No full-text available
To read the full-text of this research,
you can request a copy directly from the author.
Creative Language Retrieval: A Robust Hybrid of Information Retrieval and Linguistic Creativity
Abstract
Information retrieval (IR) and figurative language processing (FLP) could scarcely be more different in their treatment of language and meaning. IR views language as an open-ended set of mostly stable signs with which texts can be indexed and retrieved, focusing more on a text's potential relevance than its potential meaning. In contrast, FLP views language as a system of unstable signs that can be used to talk about the world in creative new ways. There is another key difference: IR is practical, scalable and robust, and in daily use by millions of casual users. FLP is neither scalable nor robust, and not yet practical enough to migrate beyond the lab. This paper thus presents a mutually beneficial hybrid of IR and FLP, one that enriches IR with new operators to enable the non-literal retrieval of creative expressions, and which also transplants FLP into a robust, scalable framework in which practical applications of linguistic creativity can be implemented.
... Figurative devices such as metaphors contain rich semantics, which is challenging for computational approaches to NLP to process cross-domain structure alignment [17]. On the other hand, literal senses are relatively easy to retrieve; hence Veale et al. [17] propose a hybrid way of information retrieval and figurative language processing to increase the effectiveness of detecting metaphors. ...
... Figurative devices such as metaphors contain rich semantics, which is challenging for computational approaches to NLP to process cross-domain structure alignment [17]. On the other hand, literal senses are relatively easy to retrieve; hence Veale et al. [17] propose a hybrid way of information retrieval and figurative language processing to increase the effectiveness of detecting metaphors. In this study, we also use the patterns of literal senses as basis to predict where a sense is used in its metaphoric sense. ...
... Figurative language is pervasive in our daily life and often provides the most crucial piece of tightly packed information. Due to its versatility, creativity, and deviation from literal meaning, figurative language poses great challenges and attracts considerable attention in NLP [17], [54]. This study aims to go beyond simple metaphor detection by capturing linguistic creativity in the cultural context. ...
Metaphors are compact packages of information with rich cultural background information. As one of the most powerful linguistic forms with non-literal meaning, metaphor detection in natural language processing can be both challenging and rewarding. We propose an innovative method for metaphor detection and classification leveraging culturally grounded eventive information. This culturally grounded information is organized based on ontological structure, which in turn facilitates further semantic processing of the result of our classification. As a culturally bound ontological system, the Chinese writing system has basic concepts organized according to semantic radicals, which are symbols containing rich eventive information that represent categorical concepts. This study illustrates basic design principles of applying ontological structures in metaphor detection by taking into account radicals representing body parts, instruments, materials, and movements. Our approach to leverage the eventive information of the Chinese writing system in metaphor detection is based on the fact that such information is available as an integral part of the writing system of any text. We hypothesize that eventive information can be accessed through the "embodied" source domain information represented by the radicals without syntactic processing or annotation. In terms of the theory of metaphor, we further hypothesize that eventive types in the embodied source domain maps to, and hence can help to predict, eventive meaning in the target domain of metaphor. Our studies show that the event information encoded in lexical items can facilitate classification of metaphoric events and identification of metaphors in Chinese texts effectively.We achieved improvements in Chinese metaphor detection over state-of-the-art approaches in our first classification experiment, and our proposed approach is shown to be generalizable in a second experiment involving new sets of characters with the same radicals.
... Martin, 1990;Mason, 2004). For existing metaphors are themselves a valuable source of knowledge for the production of new metaphors, so much so that a system can mine the relevant knowledge from corpora of figurative text (see Veale, 2011;Shutova, 2010). Thus, though linguistic metaphors are most naturally viewed as the output of a language generation process, and as the input to a language understanding process, it is just as meaningful to view the conceptual metaphors that underpin these linguistic forms as an input to the generation process and an output of the understanding process. ...
... They go on to show in (2007b) how this body of stereotypes can be used in a Web-based model of metaphor generation and comprehension. Veale (2011) employs stereotypes as the basis of the Creative Information Retrieval paradigm, by introducing a variety of non-literal-matching wildcards in the vein of Mihalcea (2002). In this paradigm, @Noun matches any adjective that denotes a stereotypical property of Noun (so e.g. ...
... Of course, there are important differences between metaphors, which elicit information from other humans, and IR queries, which elicit information from search engines. For one, IR fails to discriminate literal from non-literal language (see Veale 2004Veale , 2011, and reduces any metaphoric query to literal keywords and keyphrases that are matched near-identically to texts (see Salton, 1968;Van Rijsbergen 1979). Yet everyday language shows that metaphor is an ideal form for expressing our information needs. ...
Metaphor is much more than a pyrotechnical flourish of language or a fascinating conceptual puzzle: it is a cognitive lever that allows speakers to leverage their knowledge of one domain to describe, reframe and understand another. Though NLP researchers tend to view metaphor as a problem to be solved, metaphor is perhaps more fittingly seen as a solution to be used, that is, as an important tool in the support of creative thinking and the generation of diverse linguistic outputs. Since it pays to think of metaphor as a foundational cognitive service, one that can be exploited in a wide array of creative computational tasks, we present here a view of metaphor as a public Web service that can be freely called on demand.
... The layered architecture of Fig. 1 employs Creative Information Retrieval [28] to integrate a diverse set of resources into a single middleware component. This middleware hides the complexity of language resources that vary in organization (from structured to unstructured), scale (single files to massive corpora) and content (raw data, tagged information, or conceptual knowledge), while providing an expressive means of exploiting these resources to the SOA services perched above it. ...
... Though linguistic creativity primarily involves the generation of novel texts, or texts that achieve their communicative ends in novel or non-obvious ways, generation need not be computationally achieved from first principles, using explicit grammars and other formal machinery. Veale [28] argues that much linguistic creativity arises from the purposeful reuse of existing texts or language fragments. So an innovative text is not necessarily one that uses rare or fanciful words, but one that finds fresh and surprising uses for familiar forms [36]. ...
... Given a large repository of common language fragments, such as the Google n-grams [38], one needs an especially expressive query language to retrieve "readymade" phrases, based not on their form (which cannot be known before retrieval), but on their meanings, connotations or resonances. Creative Information Retrieval [28] provides one such query language to support non-literal retrieval. ...
Creativity is a long cherished and widely studied aspect of human behavior that allows us to re-invent the familiar, and to imagine the new. Computational creativity (CC) is a newly burgeoning area of creativity research that brings together academics and practitioners from diverse disciplines, genres and modalities, to explore the potential of computers to be autonomously creative, or to collaborate as co-creators with people. We describe here an architecture for creative Web services that will act as a force magnifier for CC, both for academic research, and for the effective deployment of real CC applications in industry. For researchers, this service-oriented architecture supports the pooling of technologies in a robust interoperable framework, in which CC models are conceived, developed and migrated from lab settings to an industrial strength platform. Industry developers, for their part, will be able to exploit novel results of CC research in a robust, low-risk form, without having to re-implement algorithms from a quickly moving field. We illustrate the architecture with the first of a growing set of creative Web services that provide robust figurative language processing on demand.
... One definition of creativity, which follows the above-mentioned divergence-convergence steps, is "the ability to see a challenge or a problem in a new light and to find solutions, which were not obvious" (Weston, 2008). Veale saw creativity from a linguistic perspective as "new and creative ways of expressing a given idea" (Veale, 2011), therefore, a different way of writing (and, implicitly, thinking) about given concepts. In this sense, creativity is directly related to the usage of metaphors and analogies. ...
... In the end, the presented corpus-based application might also help Veale in his quest for retrieving creative (fresh or unusual) expressions that have the same meaning as the query (Veale, 2011) or the researchers investigating how people decide the choice of words when making up a novel noun-noun compound. ...
... One definition of creativity, which follows the above-mentioned divergence-convergence steps, is "the ability to see a challenge or a problem in a new light and to find solutions, which were not obvious" (Weston, 2008). Veale saw creativity from a linguistic perspective as "new and creative ways of expressing a given idea" (Veale, 2011), therefore, a different way of writing (and, implicitly, thinking) about given concepts. In this sense, creativity is directly related to the usage of metaphors and analogies. ...
... In the end, the presented corpus-based application might also help Veale in his quest for retrieving creative (fresh or unusual) expressions that have the same meaning as the query (Veale, 2011) or the researchers investigating how people decide the choice of words when making up a novel noun-noun compound. ...
... Though this work manually generated similes but it opened an effective way for identifying figurative language. Another important contribution that we want to mention here is from [21] by creating a method for retrieving figurative language. In this work, the author defined a list of operators (i.e., neighborhood (?X), cultural stereotype (@X), and ad-hoc category (ˆX)) and the compound rules for expressing sentences. ...
... Using this methodology is able to create a lot of meaningful queries. By integrating these two techniques together and extending one more operator (antonym (-X)) from [21], the author gave a method for categorizing texts into straight simile and ironic simile by using strategies of ironic subversion [22]. In another research [19], unsupervised methods were used to find the associate from a small set of metaphorical expressions by verb and noun clustering. ...
This paper focuses on identifying the polarity of figurative language in the very short text collected from Social Network Services. Although this topic is not new, most computer scientists have solved this issue by using natural language processing techniques. This seems difficult for non-native English speakers because they have to rely on heuristics in language. Therefore, our target in this work is to find a language-independent approach to solve the problem without using any semantic resources (e.g., dictionaries and ontologies). A statistical method based on two main features (i.e., (i) textual terms and (ii) sentimental patterns) is proposed to determine the sentiment degree of three popular types of figurative language (i.e., sarcasm, irony, and metaphor). We experimented on two Test sets with about 3,800 tweets and used Cosine similarity as the correlation measurement for evaluating the performance. The results show that our Fi-Senti model (Figurative Sentiment analysis model) well performs in determining the sentiment intensity of the figurative language with the best achievement is 0.8952 with sarcasm and 0.9011 with irony.
... Existing metaphors are themselves a valuable source of knowledge for the production of new metaphors, so much so that a system can mine the relevant knowledge from corpora of figurative text (e.g. see Veale, 2011;Shutova, 2010). ...
... An expansionist approach to metaphor meaning, in which an affective metaphor is interpreted by generating the space of related metaphors and talking points that it implies, is thus very much suited to a more creative vision of IR, as e.g. suggested by Veale (2011). To expand a metaphorical query (like "company-X is a cult" or "company-Y is a dinosaur" or "Z was a tyrant"), a system must first expand the metaphor itself, into a set of plausible construals of the metaphor (e.g. a company that is viewed as a dinosaur will likely be powerful, but also bloated, lumbering and slow). ...
Metaphors pervade our language because they are elastic enough to allow a speaker to express an affective viewpoint on a topic without committing to a specific meaning. This balance of expressiveness and indeterminism means that metaphors are just as useful for eliciting information as they are for conveying information. We explore here, via a demonstration of a system for metaphor interpretation and generation called Metaphor Magnet, the practical uses of metaphor as a basis for formulating affective information queries. We also consider the kinds of deep and shallow stereotypical knowledge that are needed for such a system, and demonstrate how they can be acquired from corpora and the web.
... According to her results, when people translated sentences to and from their native language and their second language, both the metaphoric and the literal meaning tend to be active, even though it is clear from the context which meaning is intended. From a different perspective, Veale [34] found some patterns about how people express metaphorical comparisons in an information retrieval task. In addition, Zinken [35] noted in a corpus of newspaper texts how metaphors have a regular pattern to make the comparisons. ...
Literal language is commonly defined in terms of direct meaning, i.e., any literal utterance must convey a unique meaning. Such meaning has to be the one conventionally accepted to guarantee a successful communication. Figurative language, on the other hand, could be regarded as the opposite of literal language. Thus, whereas the latter is assumed to communicate a direct and explicit meaning, figurative language is related to the communication of veiled or implicit meanings. For instance, the word pozolero (stewmaker), which literally refers to a person who cooks a traditional Mexican food, when it is used in a figurative utterance, it can refer to different concepts, which are hardly related to food. Therefore, it can work instead of hitman, murderer, drug dealer, and others, in such a way its literal meaning is intentionally deviated in favor of secondary interpretations. In this regard, we are focused on analyzing the use of figurative language in an atypical context: drug trafficking. To this end, a corpus about narco language in Spanish was built. This corpus was used to train a word embedding model to identify creative ways to name narco-related concepts. The results show that various concepts are commonly expressed through figurative devices, such as metaphor, metonymy, or mental imagery. This fact corroborates that figurative language is quite recurrent in our daily communication, regardless of the context. In addition, we show how this creativity can be recognized by applying Natural Language Processing (NLP) techniques.
... Natural language understanding often goes beyond the syntactic and semantic layers, and perhaps nowhere is this more palpable than in the use of figurative language. A better understanding of figurative language use, such as metaphors, irony, or sarcasm, can not only lead to advances in computational creativity (Veale, 2011;Kuznetsova et al., 2013), but also in understanding social media content, where users often employ such pragmatic tools as irony or sarcasm (Reyes et al., 2013;Riloff et al., 2013). This type of figurative language is difficult to identify, however, at least partly due to what the influential literary poet and critic William Empson called "ambiguities" (Empson, 1947) in the language. ...
... Understanding language requires both linguistic knowledge and commonsense knowledge (LoBue and Yates 2011). We should also keep in mind that words should not always be taken at face value (Veale 2011). However, existing QA datasets are commonly sourced from Wikipedia (Yang, Yih, and Meek 2015;Rajpurkar et al. 2016) and news articles (Trischler et al. 2017), inevitably leading to creation of superficial literal questions, which are easy to comprehend. ...
A riddle is a question or statement with double or veiled meanings, followed by an unexpected answer. Solving riddle is a challenging task for both machine and human, testing the capability of understanding figurative, creative natural language and reasoning with commonsense knowledge. We introduce BiRdQA, a bilingual multiple-choice question answering dataset with 6614 English riddles and 8751 Chinese riddles. For each riddle-answer pair, we provide four distractors with additional information from Wikipedia. The distractors are automatically generated at scale with minimal bias. Existing monolingual and multilingual QA models fail to perform well on our dataset, indicating that there is a long way to go before machine can beat human on solving tricky riddles. The dataset has been released to the community.
... Creativity has been seen as a central property of the human use of natural language (McDonald and Busa, 1994). Text should not be always taken at face value, however, higher-order use of language and figurative devices such as metaphor can communicate richer meanings and needs deeper reading and more complicated reasoning skills (Veale, 2011). Recent works on processing language with creative use focus on metaphor detection (Gao If you take off my skin, I will not cry, but you will. ...
... Our method is based on the linguistic readymade (Veale 2012) that is inspired from Marcel Duchamp's idea in art where something (a phrase in our case) is taken from its conventional context of use and used in a new context that gives it new meaning and new relevance. We utilise a technique called Creative Information Retrieval (CIR) (Veale 2011) to obtain stereotypical associations of nouns with basic colours and linguistic readymades that are used as names from Google n-grams (Brants and Franz 2006). CIR defines operators such as @Adj andˆNoun for retrieving stereotypical associations with the adjective Adj (e.g. ...
Computational creativity has received a good amount of research interest in generating creative artefacts programmatically. At the same time, research has been conducted in computational aesthetics, which essentially tries to analyse creativity exhibited in art. This thesis aims to unite these two distinct lines of research in the context of natural language generation by building, from models for interpretation and generation, a cohesive whole that can assess its own generations.
I present a novel method for interpreting one of the most difficult rhetoric devices in the figurative use of language: metaphors. The method does not rely on hand-annotated data and it is purely data-driven. It obtains the state of the art results and is comparable to the interpretations given by humans. We show how a metaphor interpretation model can be used in generating metaphors and metaphorical expressions.
Furthermore, as a creative natural language generation task, we demonstrate assigning creative names to colours using an algorithmic approach that leverages a knowledge base of stereotypical associations for colours. Colour names produced by the approach were favoured by human judges to names given by humans 70% of the time.
A genetic algorithm-based method is elaborated for slogan generation. The use of a genetic algorithm makes it possible to model the generation of text while optimising multiple fitness functions, as part of the evolutionary process, to assess the aesthetic quality of the output. Our evaluation indicates that having multiple balanced aesthetics outperforms a single maximised aesthetic.
From an interplay of neural networks and the traditional AI approach of genetic algorithms, we present a symbiotic framework. This is called the master-apprentice framework. This makes it possible for the system to produce more diverse output as the neural network can learn from both the genetic algorithm and real people.
The master-apprentice framework emphasises a strong theoretical foundation for the creative problem one seeks to solve. From this theoretical foundation, a reasoned evaluation method can be derived. This thesis presents two different evaluation practices based on two different theories on computational creativity. This research is conducted in two distinct practical tasks: pun generation in English and poetry generation in Finnish.
... Creativity has been seen as a central property of the human use of natural language (McDonald and Busa, 1994). Text should not be always taken at face value, however, higher-order use of language and figurative devices such as metaphor can communicate richer meanings and needs deeper reading and more complicated reasoning skills (Veale, 2011). Recent works on processing language with creative use focus on metaphor detection (Gao et al., 2018), pun generation (He et al., 2019;Luo et al., 2019), creative story generation, and humor detection Seppi, 2019, 2020), sarcasm generation (Chakrabarty et al., 2020), etc. ...
A riddle is a mystifying, puzzling question about everyday concepts. For example, the riddle "I have five fingers but I am not alive. What am I?" asks about the concept of a glove. Solving riddles is a challenging cognitive process for humans, in that it requires complex commonsense reasoning abilities and an understanding of figurative language. However, there are currently no commonsense reasoning datasets that test these abilities. We propose RiddleSense, a novel multiple-choice question answering challenge for benchmarking higher-order commonsense reasoning models, which is the first large dataset for riddle-style commonsense question answering, where the distractors are crowdsourced from human annotators. We systematically evaluate a wide range of reasoning models over it and point out that there is a large gap between the best-supervised model and human performance -- pointing to interesting future research for higher-order commonsense reasoning and computational creativity.
... Typicality. Corpus research has shown that people tend to express property-concept relations explicitly for cases in which a concept is a particularly good example of a property (Veale and Hao, 2007;Veale, 2011Veale, , 2013. For instance, colors tend to be described in terms of things which illustrate them particularly well (e.g. as white as snow, as red as blood, as black as ebony wood 1 ). ...
... Irony detection (ID) has gained relevance recently, due to its importance to extract information from texts. For example, to go beyond the literal matches of user queries, Veale enriched information retrieval with new operators to enable the non-literal retrieval of creative expressions [40]. Also, the performances of sentiment analysis systems drastically decrease when applied to ironic texts [5,19]. ...
This paper proposes the first multilingual (French, English and Arabic) and multicultural (Indo-European languages vs. less culturally close languages) irony detection system. We employ both feature-based models and neural architectures using monolingual word representation. We compare the performance of these systems with state-of-the-art systems to identify their capabilities. We show that these monolingual models trained separately on different languages using multilingual word representation or text-based features can open the door to irony detection in languages that lack of annotated data for irony.
... We use Veale's (2011) neighbouring properties dataset (see end of this paper) to obtain links between properties. Veale used the simile pattern "as p and * as" as in "as sweet and * as" to retrieve neighbouring properties of sweet. ...
Many linguistic creativity applications rely heavily on knowledge of nouns and their properties. However, such knowledge sources are scarce and limited. We present a graph-based approach for expanding and weighting properties of nouns with given initial, non-weighted properties. In this paper, we focus on famous characters, either real or fictional, and categories of people , such as Actor, Hero, Child etc. In our case study, we started with an average of 11 and 25 initial properties for characters and categories, for which the method found 63 and 132 additional properties, respectively. An empirical evaluation shows that the expanded properties and weights are consistent with human judgement. The resulting knowledge base can be utilized in creation of figurative language. For instance, metaphors based on famous characters can be used in various applications including story generation, creative writing, advertising and comic generation.
... Irony detection (ID) has gained relevance recently, due to its importance to extract information from texts. For example, to go beyond the literal matches of user queries, Veale enriched information retrieval with new operators to enable the non-literal retrieval of creative expressions [40]. Also, the performances of sentiment analysis systems drastically decrease when applied to ironic texts [5,19]. ...
This paper proposes the first multilingual (French, English and Arabic) and multicultural (Indo-European languages vs. less culturally close languages) irony detection system. We employ both feature-based models and neural architectures using monolingual word representation. We compare the performance of these systems with state-of-the-art systems to identify their capabilities. We show that these monolingual models trained separately on different languages using multilingual word representation or text-based features can open the door to irony detection in languages that lack of annotated data for irony.
... Thus, creativity as an esthetic category is searched in the scientific discourse [5], [6], [7], [8] and is demanded in the communicative practice of users, it becomes a condition of communicative behaviourod social networks users. This is the way the actors show their individuality, social and civil activity. ...
The article considers multimedia texts as units of a communicative act in social networks. The authors think that these units are the realization of Media Aesthetics principles which generally was formed in the epoch of social networks. It is characterized with such principles as performance, combination of verbal and nonverbal types of sign, reign of image over words, democracy of communication, priority of self-presentation, interaction and materialization of sense perception. Also the article provides with the analyses of the term "multimedia text" analogous to multimedia text, defines esthetic foundation and stylistic characteristics of multimedia texts. It is determined that creativity in creation and publishing of multimedia texts becomes a form of social activity presence of social network users. The category of creativity is realized through the deconstruction of official culture, through irony and humour on everything which refers to a norm. This helped defining the means of multimedia creativity: precedence, language play, irony, collage. The method of case study helped revealing the fact that a multimedia message makes a greater impact on the communicative behaviour of social networks users than a common verbal message, it becomes a repost more frequently.
... We adapted it to the domain of style and creative language. Detecting chiasmus is a creative manipulation of texts that has potential applications in figurative language processing (Veale, 2011), where information retrieval becomes creative text retrieval. ...
Figurative language identification is a hard problem for computers. In this paper we handle a subproblem: chiasmus detection. By chiasmus we understand a rhetorical figure that consists in repeating two el- ements in reverse order: “First shall be last, last shall be first”. Chiasmus detec- tion is a needle-in-the-haystack problem with a couple of true positives for millions of false positives. Due to a lack of anno- tated data, prior work on detecting chias- mus in running text has only considered hand-tuned systems. In this paper, we ex- plore the use of machine learning on a partially annotated corpus. With only 31 positive instances and partial annotation of negative instances, we manage to build a system that improves both precision and recall compared to a hand-tuned system using the same features. Comparing the feature weights learned by the machine to those give by the human, we discover common characteristics of chiasmus.
... The pattern 'as ADJ as *' retrieves many results but with a high percentage of ironies. In contrast, web instances of 'ADJ S such as NOUN', where S denotes a superordinate of NOUN, are rarely ironic (Veale 2011(Veale , 2012. Third, each of the obtained category 'ADJ S' is used to get more members of the same category, via the query pattern 'ADJ S such as * and NOUN'. ...
Behind every good advertisement there is a creative concept, a Big Idea. In contrast to the countless number of ads, a small number of patterns of effective communication (idea templates) have been uncovered, which are invariant across content and context. In this thesis, we investigate the formalization of idea templates and use it as the basis for building computational means of generating ideas. We present a computational approach of generating one type of ideas, pictorial metaphor. A two-stage approach is proposed and implemented: Stage 1 finds concepts that are salient in a given property; Stage 2 evaluates the aptness of the concepts found as metaphor vehicles. The ideas generated by our approach were evaluated against past successful ads.
... The application might also help Veale in his quest for retrieving creative (fresh or unusual) expressions that have the same meaning as the query (Veale, 2011) or Costello to investigate how people decide the choice of words when making up a novel noun-noun compound (Costello, 2002). ...
This article proposes a new fully automated method for identifying creativity that is manifested in a divergent task. The task is represented by chat conversations in small groups, each group having to debate on the same topics, with the purpose of better understanding the discussed concepts. The chat conversations were created by undergraduate students in computer science studying human-computer interaction in several consecutive years. From this corpus of conversations, the chats from a single year were selected for analysis. These 25 chats contained 8,798 utterance, made of 82,176 word appearances with a vocabulary size of 5,948 distinct words. By analyzing the resulted dataset of chat conversations, the creative ideas expressed by participants are automatically identified and extracted. The application is a first step in supporting creativity in online group discussions by highlighting the novel concepts present in conversations (new ideas) and also by identifying topics that could have become important, but they were forgotten during the debates (lost ideas). Once the ideas are identified, the system tries to also capture their developments, the reactions they attract and the conclusions that are drawn based on them. Because group constituency might influence the level of creative discourse within a conversation, the typology of each participant to the conversation is evaluated, starting from the analysis of the ideas already discovered, along with the utterances labeled as developments, reactions and conclusions.
... Its knowledge projection mechanisms help us to grasp new concepts and generate innovative ideas. This opens many avenues for the creation of computational tools that foster creativity (Veale 2011Veale , 2014) and support assessment in education (Burstein et al. 2013). For many years, computational work on metaphor evolved around the use of handcoded knowledge and rules to model metaphorical associations, making the systems hard to scale. ...
Highly frequent in language and communication, metaphor represents a significant challenge for Natural Language Processing (NLP) applications. Computational work on metaphor has traditionally evolved around the use of hand-coded knowledge, making the systems hard to scale. Recent years have witnessed a rise in statistical approaches to metaphor processing. However, these approaches often require extensive human annotation effort and are predominantly evaluated within a limited domain. In contrast, we experiment with weakly supervised and unsupervised techniques—with little or no annotation—to generalize higher-level mechanisms of metaphor from distributional properties of concepts. We investigate different levels and types of supervision (learning from linguistic examples vs. learning from a given set of metaphorical mappings vs. learning without annotation) in flat and hierarchical, unconstrained and constrained clustering settings. Our aim is
to identify the optimal type of supervision for a learning algorithm that discovers patterns of metaphorical association from text. In order to investigate the scalability and adaptability of our models, we applied them to data in three languages from different language groups—English, Spanish, and Russian—achieving state-of-the-art results with little supervision. Finally, we demonstrate that statistical methods can facilitate and scale up cross-linguistic research on metaphor.
... The relevance of the response depends on factors that include rhyming, vocabulary, unexpectedness, semantic coherence, and humor. Tony Veale [26] illustrates other linguistically creative uses of information retrieval, e.g., for metaphor generation. He argues that phrases extracted from large corpora can be used as "readymade" or "found" objects, like objets trouvés in arts, that can take on fresh meanings when used in a new context. ...
Writing rap lyrics requires both creativity to construct a meaningful, interesting story and lyrical skills to produce complex rhyme patterns, which form the cornerstone of good flow. We present a rap lyrics generation method that captures both of these aspects. First, we develop a prediction model to identify the next line of existing lyrics from a set of candidate next lines. This model is based on two machine-learning techniques: the RankSVM algorithm and a deep neural network model with a novel structure. Results show that the prediction model can identify the true next line among 299 randomly selected lines with an accuracy of 17%, i.e., over 50 times more likely than by random. Second, we employ the prediction model to combine lines from existing songs, producing lyrics with rhyme and a meaning. An evaluation of the produced lyrics shows that in terms of quantitative rhyme density, the method outperforms the best human rappers by 21%. The rap lyrics generator has been deployed as an online tool called DeepBeat, and the performance of the tool has been assessed by analyzing its usage logs. This analysis shows that machine-learned rankings correlate with user preferences.
... Bendersky and Smith (2012) demonstrated a method for automatically culling quotations from textual corpora, yet their method was limited to individual sentences. Taking into account the stylistic schemas of quotations could facilitate the gathering of multi-sentence quotations and assist "creative text retrieval" (Veale, 2011) more generally. In the context of social media platforms where quotes circulate, stylistic patterns could also be used to recommend users stylisticallysimilar quotations to read. ...
... Figurative language: There has been substantial work for detecting and interpreting figurative language (Shutova, 2010;Li et al., 2013;Kuznetsova et al., 2013a;Tsvetkov et al., 2014), while relatively less work on generating creative or figurative language (Veale, 2011;Ozbal and Strapparava, 2012). We probe data-driven approaches to creative language generation in the context of image captioning. ...
... The remaining chunks can then be ranked by frequency or PMI between words, depending on the type of output we want to obtain. In addition specific n-gram patterns -see for example [42] -for extracting semantically exaggerated variations are used. ...
Accurate wording is essential in persuasive verbal communication. Through it speakers can provide an affective connotation to the text and reveal their disposition or induce a similar disposition on the recipient. All this is apparent in persuasion texts par excellence, such as political speech and advertisement. Automatic sentiment variations of existing linguistic expressions open the way to promising applications, yet it is a challenging problem. In this paper we describe a system which takes up this challenge, together with a framework for evaluating the persuasiveness of the newly produced expressions.
... To build our knowledge-representation of the quotidian world, we collect all the noun values of T for which the 3gram "world without [T]" is found in the Google n-grams. To do this, we use the Creative Information Retrieval (CIR) model of Veale (2011), which allows us to find ngrams that match complex non-literal queries. Our sweep of the n-grams database reveals approx. ...
We all share the same world, but are free to formulate and argue for our own interpretations of this shared reality. For different agents will grant differing degrees of importance to the same facts and norms. We cannot experiment on human cultures the way scientists experiment on cell cultures, but we can construct thought experiments that imagine the consequences of otherwise impossible changes. Successful thought experiments do not change the world, but change the way we see the world. This paper describes Gedankenstyle reasoning in an Al system that allows a computer to understand, or at least speculate on. the surprising causal interactions between apparently unrelated concepts. This system ponders alternate worlds in which the amount of a conceptual ingredient [X] is increased or decreased, to see what unexpected and apparently incongruous effects might arise from this change. Our goal is to construct a creative generator of novel what-if scenarios that can be used in the generation of perspective-shaping stories, poems and jokes. Copyright © 2014, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.
... Such a classification is organizaed into: (i) linguistic analysis, which leverages global language properties such as morphological, lexical, syntactic and semantic word relationships; (ii) corpus-specific (global) techniques, which analyze the contents of a full database to identify features used in similar ways; (iii) query-specific (local) techniques, which take advantage of the local context provided by the query; (iv) search log analysis techniques, which are based on the idea of mining query associations that have been implicitly suggested by previous users; (v) web data techniques, which are based on anchor texts that are often succinct descriptions of the destination page and as such, can be very similar to search queries. However, such techniques are usually used in accordance with traditional information retrieval approaches, which do not distinguish between creative and conventional uses of languages, or between literal and non-literal meanings [5]. But to support a more creative search, with the ultimate objective of being surprised or inspired by the results, nonliteral relationships between queries and the texts that they match should be exploited. ...
Web search engines are equipped with query expansion facilities to reformulate a seed query and improve retrieval performance. However, such techniques are usually used in accordance with traditional Information Retrieval approaches, which do not distinguish between creative and conventional uses of languages, or between literal and non-literal meanings. But to support a more creative search, with the ultimate objective of being surprised or inspired by the results, non-literal relationships between queries and the texts that they match should be facilitated. This paper presents a query expansion method with a lateral thinking approach, by suggesting, starting from a seed term given by the user, a set of lists of terms representing conceptual paths, each of which starts from the seed term. Each term in the path is reached by traversing pre-identified relationships in a given semantic network, while the selection of a specific term is driven by the assessment of a distance metrics between terms. The paper also presents a software implementation of the method, which can be accessed as a mobile web app. I. INTRODUCTION Lateral thinking [1], the term was coined by the physician Edward De Bono, is an attitude for addressing problems through an indirect and creative approach. Lateral thinking leverage reasoning that is not immediately obvious, involving ideas that may not be obtainable by using only traditional step-by-step logic. When we search for something, we are used to follow traditional pattern-based approaches. But in any patterning system, how argued by De Bono, there is an absolute, and even logic, need for something like lateral thinking, in order to yearn for something new that can further trigger creative and innovative behaviors [2]. The most common engines for searching resources over the Web evolved a lot. Currently, they do not just search for resources that exactly match keywords representing users' criteria. Indeed, most of them are equipped with query expansion facilities [3], whose aim is to reformulate a seed query to improve retrieval performance. Common query expansion techniques involve: finding synonyms of words, finding all the various morphological forms of words by stemming each word in the search query; fixing spelling errors and automatically searching for the corrected form or suggesting it in the results; weighting the terms in the original query. In [4], a classification of automated query expansion techniques is presented. Such a classification is organizaed into: (i)
... Such a classification is organizaed into: (i) linguistic analysis, which leverages global language properties such as morphological, lexical, syntactic and semantic word relationships; (ii) corpus-specific (global) techniques, which analyze the contents of a full database to identify features used in similar ways; (iii) query-specific (local) techniques, which take advantage of the local context provided by the query; (iv) search log analysis techniques, which are based on the idea of mining query associations that have been implicitly suggested by previous users; (v) web data techniques, which are based on anchor texts that are often succinct descriptions of the destination page and as such, can be very similar to search queries. However, such techniques are usually used in accordance with traditional information retrieval approaches, which do not distinguish between creative and conventional uses of languages, or between literal and non-literal meanings [5]. But to support a more creative search, with the ultimate objective of being surprised or inspired by the results, nonliteral relationships between queries and the texts that they match should be exploited. ...
Web search engines are equipped with query expansion facilities to reformulate a seed query and improve retrieval performance. However, such techniques are usually used in accordance with traditional Information Retrieval approaches, which do not distinguish between creative and conventional uses of languages, or between literal and non-literal meanings. But to support a more creative search, with the ultimate objective of being surprised or inspired by the results, non-literal relationships between queries and the texts that they match should be facilitated. This paper presents a query expansion method with a lateral thinking approach, by suggesting, starting from a seed term given by the user, a set of lists of terms representing conceptual paths, each of which starts from the seed term. Each term in the path is reached by traversing pre-identified relationships in a given semantic network, while the selection of a specific term is driven by the assessment of a distance metrics between terms. The paper also presents a software implementation of the method, which can be accessed as a mobile web app. I. INTRODUCTION Lateral thinking [1], the term was coined by the physician Edward De Bono, is an attitude for addressing problems through an indirect and creative approach. Lateral thinking leverage reasoning that is not immediately obvious, involving ideas that may not be obtainable by using only traditional step-by-step logic. When we search for something, we are used to follow traditional pattern-based approaches. But in any patterning system, how argued by De Bono, there is an absolute, and even logic, need for something like lateral thinking, in order to yearn for something new that can further trigger creative and innovative behaviors [2]. The most common engines for searching resources over the Web evolved a lot. Currently, they do not just search for resources that exactly match keywords representing users' criteria. Indeed, most of them are equipped with query expansion facilities [3], whose aim is to reformulate a seed query to improve retrieval performance. Common query expansion techniques involve: finding synonyms of words, finding all the various morphological forms of words by stemming each word in the search query; fixing spelling errors and automatically searching for the corrected form or suggesting it in the results; weighting the terms in the original query. In [4], a classification of automated query expansion techniques is presented. Such a classification is organizaed into: (i)
... To build our knowledge-representation of the quotidian world, we collect all the noun values of T for which the 3gram "world without [T]" is found in the Google n-grams. To do this, we use the Creative Information Retrieval (CIR) model of Veale (2011), which allows us to find ngrams that match complex non-literal queries. Our sweep of the n-grams database reveals approx. ...
We all share the same world, but are free to formulate and argue for our own interpretations of this shared reality. For different agents will grant differing degrees of importance to the same facts and norms. We cannot experiment on human cultures the way scientists experiment on cell cultures, but we can construct thought experiments that imagine the consequences of otherwise impossible changes. Successful thought experiments do not change the world, but change the way we see the world. This paper describes Gedanken-style reasoning in an AI system that allows a computer to understand, or at least speculate on, the surprising causal interactions between apparently unrelated concepts. This system ponders alternate worlds in which the amount of a conceptual ingredient [X] is increased or decreased, to see what unexpected and apparently incongruous effects might arise from this change. Our goal is to construct a creative generator of novel what-if scenarios that can be used in the generation of perspective-shaping stories, poems and jokes.
... The relevance of the response depends on factors that include rhyming, vocabulary, unexpectedness, semantic coherence, and humor. Tony Veale [26] illustrates other linguistically creative uses of information retrieval, e.g., for metaphor generation. He argues that phrases extracted from large corpora can be used as "readymade" or "found" objects, like objets trouvés in arts, that can take on fresh meanings when used in a new context. ...
Writing rap lyrics requires both creativity, to construct a meaningful and an
interesting story, and lyrical skills, to produce complex rhyme patterns, which
are the cornerstone of a good flow. We present a method for capturing both of
these aspects. Our approach is based on two machine-learning techniques: the
RankSVM algorithm, and a deep neural network model with a novel structure. For
the problem of distinguishing the real next line from a randomly selected one,
we achieve an 82 % accuracy. We employ the resulting prediction method for
creating new rap lyrics by combining lines from existing songs. In terms of
quantitative rhyme density, the produced lyrics outperform best human rappers
by 21 %. The results highlight the benefit of our rhyme density metric and our
innovative predictor of next lines.
... n dispatched to Google as a phrasal query. We value quality over size, as these similes will later be used to find diverse viewpoints on the web via bootstrapping. We thus manually filter each web simile, to weed out any that are ill-formed, and those intended to be seen as ironic by their authors. This gives us a body of 12,000+ valid web similes. Veale (2011 Veale ( , 2012 Veale ( , 2013) notes that web uses of the pattern " as P as C " are rife with irony. In contrast, web instances of " P S such as C " – where S denotes a superordinate of C – are rarely ironic. Hao & Veale (2010) exploit this fact to filter ironic comparisons from web similes, by re-expressing each " as P as C " simile as ...
Just as observing is more than just seeing, comparing is far more than mere matching. It takes understanding, and even inventive-ness, to discern a useful basis for judging two ideas as similar in a particular context, especially when our perspective is shaped by an act of linguistic creativity such as metaphor, simile or analogy. Structured re-sources such as WordNet offer a conven-ient hierarchical means for converging on a common ground for comparison, but offer little support for the divergent thinking that is needed to creatively view one concept as another. We describe such a means here, by showing how the web can be used to har-vest many divergent views for many famil-iar ideas. These lateral views complement the narrow vertical view offered by Word-Net, and support a system for creative idea exploration called Thesaurus Rex. We also show how Thesaurus Rex supports a novel, generative similarity measure for WordNet.
... According to her results, when people translated sentences to and from their native language (Turkish) and their second language (English), upon encountering a metaphorical usage, both the underlying metaphor and the literal meaning are likely to be active in people's perception, even though it is clear from the context which meaning is intended. Whereas in Veale [170], author approached metaphor processing by means of an information retrieval task. In contrast, by analyzing metaphors in a corpus of newspaper texts, Zinken [185] noted that metaphors follow a regular pattern when the comparisons (analogies in his terms) are made. ...
Ph. D. thesis in Computer Science written by Antonio Reyes Pérez under the supervision of Dr. Paolo Rosso (Universitat Politècnica de València). The thesis defense was done in Valencia (Spain) on July 2nd, 2012. The doctoral committee was integrated by the following doctors: Antónia Martí Antonín (University of Barcelona), Walter Daelemans (University of Antwerp), Richard Anthony (Tony) Veale (University College Dublin), Carlo Strapparava (Fondazione Bruno Kessler FBK-IRST), and José Antonio Troyano Jiménez (University of Sevilla). The obtained grade was Cum Laude.
... In different situations different measures could eventually have better results, and we will investigate this in the future. In addition specific n-gram patterns -see for example (Veale, 2011) -for extracting semantically exaggerated variations are used. ...
The need for creativity is ubiquitous, and mobile devices connected to Web services can help us. Linguistic creativity is widely used in advertisements to surprise us, to get our attention, and to stick concepts in our memory. However, creativity can also be used as a defense. When we walk in the street, we are overwhelmed by messages that try to get our attention with any persuasive device at hand. As messages get ever more aggressive, often our basic cognitive defenses—trying not to perceive those messages—are not sufficient. One advanced defensive technique is based on transforming the perceived message into something different (for instance, making use of irony or hyperbole) from what was originally meant in the message. In this article, we describe an implemented application for smartphones, which creatively modifies the linguistic expression in a virtual copy of a poster encountered on the street. The mobile system is inspired by the subvertising practice of countercultural art.
... It thus makes computational sense to calculate the affect of a word-concept as a function of the affect of its most salient properties. Veale (2011) later built on this work to show how a property-rich stereotypical representation could be used for non-literal matching and retrieval of creative texts, such as metaphors and analogies. ...
Since we can 'spin' words and concepts to suit our affective needs, context is a major determinant of the perceived affect of a word or concept. We view this re-profiling as a selective emphasis or de-emphasis of the qualities that underpin our shared stereotype of a concept or a word meaning, and construct our model of the affective lexicon accordingly. We show how a large body of affective stereotypes can be acquired from the web, and also show how these are used to create and interpret affective metaphors.
Irony is a complex linguistic phenomenon that has been extensively studied in computational linguistics across many languages. Existing research has relied heavily on annotated corpora, which are inherently biased due to their creation process. This study focuses on the problem of bias in cross-domain and cross-language irony detection and aims to identify the extent of topic bias in benchmark corpora and how it affects the generalization of models across domains and languages (English, Spanish, and Italian). Our findings offer a first insight into this issue and showed that mitigating the topic bias in these corpora improves the generalization of models beyond their training data. These results have important implications for the development of robust models in the analysis of ironic language.KeywordsIronyBiasCross-DomainCross-LanguageBias Mitigation
This paper introduces the concept of semantic wildcards which is the
idea of generalizing full-text search in a notion that enables user to search within
additional layers of syntactic or semantic information retrieved from natural
language texts. We distinguish two approaches to use semantic wildcards in
search expressions: pre-defined wildcards which offers a straightforward and
accurate query-syntax, frequently used in corpus search engines by computational
linguists, and the concept of natural language wildcards which enables
the user to specify the wildcards as a part of a natural language query. We will
show that there are two ways to define the matching behavior in the field of
natural language wildcards and we will see that the Prolog programming language
offers a clean and declarative solution for implementing a search engine
when using dependency-based matching.
Our affective perspective on a word is heavily influenced by the context in which it is used and by the features it is typically perceived to exhibit in that context. A nuanced model of lexical affect thus requires a feature-rich representation of each word’s potential to mean different things in different contexts. To this end, we present here a two-level model of lexical affect. At the first level, words are represented as bundles of the typical properties and behaviors they are commonly shown to exhibit in everyday language. To construct these bundles, we present a semi-automatic approach to harvesting stereotypical properties and behaviors from the Web. At the second level, these properties and behaviors are related to each other in a graph structure that captures how likely one is to reinforce the meaning of another. We present an effective means of constructing such a graph from a combination of text n-grams and queries to the open Web. We calculate positive and negative potentials for each property in the graph, and show how these potentials can be used in turn to calculate an overall affective value for the higher-level terms for which they are considered stereotypical.
In this paper we propose a new method for identifying creativity that is based on analyzing a corpus of chat conversations on the same topic and extracting the new ideas expressed by participants. The application is a first step in supporting creativity in online group discussions by highlighting the novel concepts present in conversations (new ideas) and also by identifying topics that could have become important, if not forgotten during the debates (lost ideas).
One of the challenges of linguistic creativity is to use words in a way that is novel and striking and even whimsical, to convey meanings that remain stubbornly grounded in the very same world of familiar experiences as serves to anchor the most literal and unimaginative language. The challenge remains unmet by systems that merely shuttle or arrange words to achieve novel arrangements without concern as to how those arrangements are to spur the processes of meaning construction in a listener. In this paper we explore a problem of lexical invention that cannot be solved without a model – explicit or implicit – of the perceptual grounding of language: the invention of apt new names for colours. To solve this problem here we shall call upon the notion of a linguistic readymade, a phrase that is wrenched from its original context of use to be given new meaning and new resonance in new settings. To ensure that our linguistic readymades – which owe a great deal to Marcel Duchamp's notion of found art – are anchored in a consensus model of perception, we introduce the notion of a lexicalised colour stereotype.
System design and evaluation methodologies receive significant attention in natural language processing (NLP), with the systems typically being evaluated on a common task and against shared data sets. This enables direct system comparison and facilitates progress in the field. However, computational work on metaphor is considerably more fragmented than similar research efforts in other areas of NLP and semantics. Recent years have seen a growing interest in computational modeling of metaphor, with many new statistical techniques opening routes for improving system accuracy and robustness. However, the lack of a common task definition, shared data set, and evaluation strategy makes the methods hard to compare, and thus hampers our progress as a community in this area. The goal of this article is to review the system features and evaluation strategies that have been proposed for the metaphor processing task, and to analyze their benefits and downsides, with the aim of identifying the desired properties of metaphor processing systems and a set of requirements for their evaluation.
Opposition theory suggests that binary oppositions (e.g., high vs. low) underlie basic cognitive and linguistic processes. However, opposition theory has never been implemented in a computational cognitive-semiotics model. In this paper, we present a simple model of metaphor identification that relies on opposition theory. An algorithm instantiating the model has been tested on a data set of 100 phrases comprising adjective-noun pairs in which approximately a half represent metaphorical language-use (e.g., dark thoughts) and the rest literal language-use (e.g., dark hair). The algorithm achieved 89% accuracy in metaphor identification and illustrates the relevance of opposition theory for modelling metaphor processing.
Divergence is a multifaceted capability of multifaceted creative individuals. It may be exhibited to different degrees, and along different dimensions, from one individual to another. The same may be true of computational creative agents: Such systems may do more than exhibit differing levels of divergence: They may also implement the mechanics of divergence in very different ways. We argue that creative capabilities such as divergence are best viewed as cognitive services that may be called upon by cognitive agents to complete tasks in ways that may be deemed “original” or to generate products that may be deemed “creative.” We further argue that in a computational embodiment of such an agent, cognitive services are best realized as modular, distributed Web services which hide the complexities of their particular implementations and which can be discovered, reused and composed as desired by other Web-aware systems with diverse creative needs of their own. We describe the workings of one such reusable service for generating divergent categorizations on demand and show how this service can be composed with others to support the generation and rendering of novel metaphors in an autonomous Twitterbot system.
In this paper, we propose a computational approach to generate neologisms consisting of homophonic puns and metaphors based on the category of the service to be named and the properties to be underlined. We describe all the linguistic resources and natural language processing techniques that we have exploited for this task. Then, we analyze the performance of the system that we have developed. The empirical results show that our approach is generally effective and it constitutes a solid starting point for the automation of the naming process.
In this paper, we introduce a system that supports the naming process by exploiting natural language processing and linguistic creativity techniques in a completely unsupervised fashion. The system generates two types of neologisms based on the category of the service to be named and the properties to be underlined. While the first type consists of homophonic puns and metaphors, the second consists of neologisms that are produced by adding Latin suffixes to English words or homophonic puns. During this process, both semantic appropriateness and sound pleasantness of the generated names are taken into account.
E-Science relies heavily on manipulating massive amounts of data for research purposes. Researchers should be able to contribute their own data and methods, thus making their results accessible and reproducible by others worldwide. They need an environment which they can use anytime and anywhere to perform data-intensive computations. Virtual observatories serve this purpose. With the advance of the Semantic Web, more and more data is available in Resource Description Framework based databases. It is often desirable to have the ability to link data from local sources to these public data sets. We present a prototype system, which satisfies the requirements of a virtual observatory over semantic databases, such as user roles, data import, query execution, visualization, exporting result, etc. The system has special features which facilitate working with semantic data: visual query editor, use of ontologies, knowledge inference, querying remote endpoints, linking remote data with local data, extracting data from web pages.
Sense based query expansion never proved its effectiveness except for the so-called "open domain question answering" task. The present work is still inconclusive at this regard, due to some experimental limitations, but we provide interesting evidence suggesting new guidelines for future research. Word sense disambiguation is in fact only one of the problems involved with sense based query expansion. The second is how to use sense information (and ontologies in general) to expand the query. We show that expanding with synonyms or hyperonyms has a limited effect on web information retrieval performance, while other types of semantic information derivable from an ontology are much more effective at improving search results.
The relationship between metaphor and literal meaning is often discussed in terms that imply that the distinction is absolute: a statement either is a metaphor or it is not. This paper adduces evidence in support of analysis of metaphors by reference to stereotypical usage, and concludes that some metaphors are more metaphorical than others. At the present time, sharply defined boundaries of categories in linguistics are being questioned in the light of empirical evidence, and metaphor is no exception. Theories that invoke partial or full matching to 'best examples' of categories – norms or prototypes – seem to explain linguistic phenomena more adequately than theories that invoke necessary conditions and sharp distinctions. The question then arises, how to handle fuzzy sets, to which an empirically well-founded theory of metaphor can itself offer useful answers. Two detailed case studies are offered as a contribution to the study of metaphor in this context. Against those who argue that metaphor is merely a diachronic phenomenon, the paper shows that metaphor is a useful synchronic, empirical semantic classification, although its boundaries are fuzzy and a distinction must be made between dynamic metaphors (ad-hoc coinages) and conventional metaphors.
In previous work, we found that a great deal of information about noun attributes can be extracted from the Web using simple text patterns, and that enriching vector-based models of concepts with this information about attributes led to drastic improvements in noun categorization. We extend this previous work in two ways: (i) by comparing concept descriptions extracted using patterns with descriptions extracted with a parser, and (ii) by developing an improved dataset balanced with respect to ambiguity, frequency, and WordNet unique beginners.
Applications such as office automation, news filtering, help facilities in complex systems, and the like require the ability to retrieve documents from full-text databases where vocabulary problems can be particularly severe. Experiments performed on small collections with single-domain thesauri suggest that expanding query vectors with words that are lexically related to the original query words can ameliorate some of the problems of mismatched vocabularies. This paper examines the utility of lexical query expansion in the large, diverse TREC collection. Concepts are represented by WordNet synonym sets and are expanded by following the typed links included in WordNet. Experimental results show this query expansion technique makes little difference in retrieval effectiveness if the original queries are relatively complete descriptions of the information being sought even when the concepts to be expanded are selected by hand. Less well developed queries can be significantly improved by expansion of hand-chosen concepts. However, an automatic procedure that can approximate the set of hand picked synonym sets has yet to be devised, and expanding by the synonym sets that are automatically generated can degrade retrieval performance.
In most research on concept acquisition from corpora, concepts are modeled as vectors of relations extracted from syntactic structures. In the case of modifiers, these relations often specify values of attributes, as in (attr red); this is unlike what typically proposed in theories of knowledge representation, where concepts are typically defined in terms of their attributes (e.g., color). We compared models of concepts based on values with models based on attributes, using lexical clustering as the basis for comparison. We find that attribute-based models work better than value-based ones, and result in shorter descriptions; but that mixed models including both the best attributes and the best values work best of all.
Human categorization is neither a binary nor a context-free process. Rather, some con- cepts are better examples of a category than others, while the criteria for category mem- bership may be satisfied to different degrees by different concepts in different contexts. In light of these empirical facts, WordNet's static category structure appears both exces- sively rigid and unduly fragile for process- ing real texts. In this paper we describe a syntagmatic, corpus-based approach to re- defining WordNet's categories in a func- tional, gradable and context-sensitive fash- ion. We describe how the diagnostic prop- erties for these definitions are automati- cally acquired from the web, and how the increased flexibility in categorization that arises from these redefinitions offers a ro- bust account of metaphor comprehension in the mold of Glucksberg's (2001) the- ory of category-inclusion. Furthermore, we demonstrate how this competence with figu- rative categorization can effectively be gov- erned by automatically-generated ontologi- cal constraints, also acquired from the web.
Irony is an effective but challenging mode of communication that allows a speaker to express sentiment-rich viewpoints with concision, sharpness and humour. Irony is especially common in online documents that express subjective and deeply-felt opinions, and thus represents a significant obstacle to the accurate analysis of sentiment in web texts. In this paper we look at one commonly used framing device for linguistic irony --the simile --to show how irony is often marked in ways that make it computationally feasible to detect. We conduct a very large corpus analysis of web-harvested similes to identify the most interesting characteristics of ironic comparisons, and provide an empirical evaluation of a new algorithm for separating ironic from non-ironic similes.
Techniques that exploit knowledge of distributional similarity between words have been proposed in many areas of Natural Language Processing. For example, in language modeling, the sparse data problem can be alleviated by estimating the probabilities of unseen co-occurrences of events from the probabilities of seen co-occurrences of similar events. In other applications, distributional similarity is taken to be an approximation to semantic similarity. However, due to the wide range of potential applications and the lack of a strict definition of the concept of distributional similarity, many methods of calculating distributional similarity have been proposed or adopted.
In this work, a flexible, parameterized framework for calculating distributional similarity is proposed. Within this framework, the problem of finding distributionally similar words is cast as one of co-occurrence retrieval (CR) for which precision and recall can be measured by analogy with the way they are measured in document retrieval. As will be shown, a number of popular existing measures of distributional similarity are simulated with parameter settings within the CR framework. In this article, the CR framework is then used to systematically investigate three fundamental questions concerning distributional similarity. First, is the relationship of lexical similarity necessarily symmetric, or are there advantages to be gained from considering it as an asymmetric relationship? Second, are some co-occurrences inherently more salient than others in the calculation of distributional similarity? Third, is it necessary to consider the difference in the extent to which each word occurs in each co-occurrence type?
Two application-based tasks are used for evaluation: automatic thesaurus generation and pseudo-disambiguation. It is possible to achieve significantly better results on both these tasks by varying the parameters within the CR framework rather than using other existing distributional similarity measures; it will also be shown that any single unparameterized measure is unlikely to be able to do better on both tasks. This is due to an inherent asymmetry in lexical substitutability and therefore also in lexical distributional similarity.
The quantification of lexical semantic relatedness has many applications in NLP, and many different measures have been proposed. We evaluate five of these measures, all of which use WordNet as their central resource, by comparing their performance in detecting and correcting real-word spelling errors. An information-content-based measure proposed by Jiang and Conrath is found superior to those proposed by Hirst and St-Onge, Leacock and Chodorow, Lin, and Resnik. In addition, we explain why distributional similarity is not an adequate proxy for lexical semantic relatedness.
The paper discusses the incorporation of richer semantic structures into the Preference Semantics system: they are called pseudo-texts and capture something of the information expressed in one type of frame proposed by Minsky (q.v.). However, they are in a format, and subject to rules of inference, consistent with earlier accounts of this system of language analysis and understanding. Their use is discussed in connection with the phenomenon of extended use: sentences where the semantic preferences are broken. It is argued that such situations are the norm and not the exception in normal language use, and that a language understanding system must give some general treatment of them. A notion of sense projection is proposed, leading on to an alteration of semantic formulas (word sense representations) in the face of unexpected context by drawing information from the pseudo texts. A possible implementation is described, based on a new semantic parser for the Preference Semantics system, which would cope with extended use by the methods suggested and answer questions about the process of analysis itself. It is argued that this would be a good context in which to place a language understander (rather than that of question-answering about a limited area of the real world, as is normal) and, moreover, that the sense projection mechanisms suggested would provide a test-bed on which the usefulness of frames for language understanding could be realistically assessed
The met* method distinguishes selected examples of metonymy from metaphor and from literalness and anomaly in short English sentences. In the met* method, literalness is distinguished because it satisfies contextual constraints that the nonliteral others all violate. Metonymy is discriminated from metaphor and anomaly in a way that [1] supports Lakoff and Johnson's (1980) view that in metonymy one entity stands for another whereas in metaphor one entity is viewed as another, [2] permits chains of metonymies (Reddy 1979), and [3] allows metonymies to co-occur with instances of either literalness, metaphor, or anomaly. Metaphor is distinguished from anomaly because the former contains a relevant analogy, unlike the latter. The met* method is part of Collative Semantics, a semantics for natural language processing, and has been implemented in a computer program called meta5. Some examples of meta5's analysis of metaphor and metonymy are given. The met* method is compared with approaches from artificial intelligence, linguistics, philosophy, and psychology.
The paper discusses the incorporation of richer semantic structures into the Preference Semantics system: they are called pseudo-texts and capture something of the information expressed in one type of frame proposed by Minsky (q.v.). However, they are in a format, and subject to rules of inference, consistent with earlier accounts of this system of language analysis and understanding. Their use is discussed in connection with the phenomenon of extended use: sentences where the semantic preferences are broken. It is argued that such situations are the norm and not the exception in normal language use, and that a language understanding system must give some general treatment of them. A notion of sense projection is proposed, leading on to an alteration of semantic formulas (word sense representations) in the face of unexpected context by drawing information from the pseudo texts. A possible implementation is described, based on a new semantic parser for the Preference Semantics system, which would cope with extended use by the methods suggested and answer questions about the process of analysis itself. It is argued that this would be a good context in which to place a language understander (rather than that of question-answering about a limited area of the real world, as is normal) and, moreover, that the sense projection mechanisms suggested would provide a test-bed on which the usefulness of frames for language understanding could be realistically assessed.
This paper describes the structure-mapping engine (SME), a program for studying analogical processing. SME has been built to explore Gentner's structure-mapping theory of analogy, and provides a “tool kit” for constructing matching algorithms consistent with this theory. Its flexibility enhances cognitive simulation studies by simplifying experimentation. Furthermore, SME is very efficient, making it a useful component in machine learning systems as well. We review the structure-mapping theory and describe the design of the engine. We analyze the complexity of the algorithm, and demonstrate that most of the steps are polynomial, typically bounded by O(N2). Next we demonstrate some examples of its operation taken from our cognitive simulation studies and work in machine learning. Finally, we compare SME to other analogy programs and discuss several areas for future work.
A theory of analogy must describe how the meaning of an analogy is derived from the meanings of its parts. In the structure‐mapping theory, the interpretation rules are characterized as implicit rules for mapping knowledge about a base domain into a target domain. Two important features of the theory are (a) the rules depend only on syntactic properties of the knowledge representation, and not on the specific content of the domains; and (b) the theoretical framework allows analogies to be distinguished cleanly from literal similarity statements, applications of abstractions, and other kinds of comparisons.
Two mapping principles are described: (a) Relations between objects, rather than attributes of objects, are mapped from base to target; and (b) The particular relations mapped are determined by systematicity, as defined by the existence of higher‐order relations.
A system is described for the automatic adjustment of queries addressed to information retrieval systems employing a structurised thesaurus for the coordinate indexing of an average of at least five or six descriptors per document. Starting with at least two documents considered by the user as relevant to his inquiry, the system formulates different queries using descriptors occuring in the relevant documents. Results from these queries are presented to the user for relevance assessment as a result of which the most efficient queries are automatically selected and loosened (broadened). The new documents retrieved are again checked for relevance by the user; and with new relevant documents the loop starts again.The result of the automatic procedure is independent of the point of departure. The automatic procedure is superior to traditional searching procedures in terms of both recall and precision. The automatic procedure requires more computing, but probably for more than 80% of the inquiries the need for a documentalist as an intermediary between the user and the system can be avoided.
Information retrieval (IR) is an effective mechanism for text management that has received widespread adoption in the world at large. But it is not a particularly creative mechanism, in the sense of creating new conceptual structures or reorganizing existing ones to pull in documents that describe, in novel and inventive ways, a user's information needs. Since language is a dynamic and highly creative medium of expression, the concepts that one seeks will therefore represent a moving target for IR systems. We argue that only by thinking creatively can an IR system effectively retrieve documents that express themselves creatively.
Analogy is a powerful boundary-transcending process that exploits a conceptual system’s ability to perform controlled generalization in one domain and re-specialization into another. The result of this semantic leap is the transference of meaning from one concept to another from which metaphor derives its name (literally: to carry over). Such generalization and re-specialization can be achieved using a variety of re-representation techniques, most notably abstraction via a taxonomic backbone, or selective projection via structure-mapping over propositional content. In this paper we explore both the extent to which a bilingual lexical ontology for English and Chinese, called HowNet, can support each technique, and the extent to which both are, ultimately, variations of the same process of creative rerepresentation.
CorMet is a corpus-based system for discovering metaphorical mappings between concepts. It does this by finding systematic variations in domain-specific selectional preferences, which are inferred from large, dynamically mined Internet corpora.
Metaphors transfer structure from a source domain to a target domain, making some concepts in the target domain metaphorically equivalent to concepts in the source domain. The verbs that select for a concept in the source domain tend to select for its metaphorical equivalent in the target domain. This regularity, detectable with a shallow linguistic analysis, is used to find the metaphorical interconcept mappings, which can then be used to infer the existence of higher-level conventional metaphors.
Most other computational metaphor systems use small, hand-coded semantic knowledge bases and work on a few examples. Although Cor Met's only knowledge base is Word Net (Fellbaum 1998) it can find the mappings constituting many conventional metaphors and in some cases recognize sentences instantiating those mappings. CorMet is tested on its ability to find a subset of the Master Metaphor List (Lakoff, Espenson, and Schwartz 1991).
People construct ad hoc categories to achieve goals. For example, constructing the category of “things to sell at a garage sale” can be instrumental to achieving the goal of selling unwanted possessions. These categories differ from common categories (e.g., “fruit,” “furniture”) in that ad hoc categories violate the correlational structure of the environment and are not well established in memory. Regarding the latter property, the category concepts, concept-to-instance associations, and instance-to-concept associations structuring ad hoc categories are shown to be much less established in memory than those of common categories. Regardless of these differences, however, ad hoc categories possess graded structures (i.e., typicality gradients) as salient as those structuring common categories. This appears to be the result of a similarity comparison process that imposes graded structure on any category regardless of type.
The IRSLO (Information Retrieval using Semantic and Lexical Operators) project aims at integrating semantic and lexical information into the retrieval process, in order to overcome some of the impediments currently encountered with today's information retrieval systems. This paper introduces the semantic wildcard, one of the most powerful operators implemented in IRSLO, which allows for searches along general-specific lines. The semantic wildcard, denoted with #, acts in a manner similar with the lexical wildcard, but at semantic levels, enabling the retrieval of subsumed concepts. For instance, a search for animal# will match any concept that is of type animal, including dog, goat and so forth, thereby going beyond the explicit knowledge stated in texts. This operator, together with a lexical locality operator that enables the retrieval of paragraphs rather than entire documents, have been both implemented in the IRSLO system and tested on requests of information run against an index of 130,000 documents. Significant improvement was observed over classic keyword-based retrieval systems in terms of precision, recall and success rate.
greatest challenges in natural language learning. We first define a word similarity measure based on the distributional pattern of words. The similarity measure allows us to construct a thesaurus using a parsed corpus. We then present a new evaluation methodology for the automatically constructed the- saurus. The evaluation results show that the the- saurus is significantly closer to WordNet than Roget Thesaurus is.
Automatic query expansion has long been suggested as a technique for dealing with the fundamental issue of word mismatch in information retrieval. A number of approaches to expansion have been studied and, more recently, attention has focused on techniques that analyze the corpus to discover word relationships (global techniques) and those that analyze documents retrieved by the initial query ( local feedback). In this paper, we compare the effectiveness of these approaches and show that, although global analysis has some advantages, local analysis is generally more effective. We also show that using global analysis techniques, such as word context and phrase structure, on the local set of documents produces results that are both more effective and more predictable than simple local feedback. 1 Introduction The problem of word mismatch is fundamental to information retrieval. Simply stated, it means that people often use different words to describe concepts in their queries than auth...
We describe a method for the automatic acquisition of the hyponymy lexical relation from unrestricted text. Two goals motivate the approach: (i) avoidance of the need for pre-encoded knowledge and (ii) applicability across a wide range of text. We identify a set of lexicosyntactic patterns that are easily recognizable, that occur frequently and across text genre boundaries, and that indisputably indicate the lexical relation of interest. We describe a method for discovering these patterns and suggest that other lexical relations will also be acquirable in this way. A subset of the acquisition algorithm is implemented and the results are used to augment and critique the structure of a large hand-built thesaurus. Extensions and applications to areas such as information retrieval are suggested. 1 Introduction Currently there is much interest in the automatic acquisition of lexical syntax and semantics, with the goal of building up large lexicons for natural language processing. Projects...
Evaluating Word-Net-based Measures of Lexical Semantic Related-ness Structure-Mapping Engine: Algorithm and Exam-ples Met*: a method for discriminating metonymy and metaphor by computer Processing Metonymy and Metaphor
- A Budanitsky
- G Hirst
- B Falkenhainer
- K Forbus
- D Gentner
Budanitsky, A. and Hirst, G. (2006). Evaluating Word-Net-based Measures of Lexical Semantic Related-ness. Computational Linguistics, 32(1):13-47. Falkenhainer, B., Forbus, K. and Gentner, D. (1989). Structure-Mapping Engine: Algorithm and Exam-ples. Artificial Intelligence, 41:1-63. Fass, D. (1991). Met*: a method for discriminating metonymy and metaphor by computer. Computa-tional Linguistics, 17(1):49-90. Fass, D. (1997). Processing Metonymy and Metaphor
- J P Guilford
Guilford, J.P. (1950) Creativity, American Psychologist
5(9):444–454.
Metaphor and Cognition: Studies in Cognitive Systems
- B Indurkhya
Indurkhya, B. (1992). Metaphor and Cognition: Studies
in Cognitive Systems. Kluwer Academic Publishers,
Dordrecht: The Netherlands.
Metaphoricity is gradable Corpus- Based Approaches to Metaphor and Metonymy
- P Hanks
Hanks, P. (2006). Metaphoricity is gradable. In: Anatol
Stefanowitsch and Stefan Th. Gries (Eds.), Corpus-
Based Approaches to Metaphor and Metonymy,. 17-
35. Berlin: Mouton de Gruyter.
Knowledge Representation and Metaphor. Studies in Cognitive systems
- E C Way
Way, E. C. (1991). Knowledge Representation and
Metaphor. Studies in Cognitive systems. Holland:
Kluwer.
Web 1T 5-gram Ver. 1. Linguistic Data Consortium
- T Brants
- A Franz
Brants, T. and Franz, A. (2006). Web 1T 5-gram Ver. 1.
Linguistic Data Consortium.
The Challenge of Creative Information Retrieval. Computational Linguistics and Intelligent Text Processing: Lecture Notes in Computer Science
- T Veale
Veale, T. (2004). The Challenge of Creative Information Retrieval. Computational Linguistics and Intelligent Text Processing: Lecture Notes in Computer
Science, Volume 2945/2004, 457-467.
Poetic and Non-Poetic Simile: Structure, Semantics, Rhetoric
- D Fishlov
Fishlov, D. (1992). Poetic and Non-Poetic Simile:
Structure, Semantics, Rhetoric. Poetics Today, 14(1),
1-23.
Similes and Sets: The English Preposition 'like
- P Hanks
Hanks, P. (2005). Similes and Sets: The English Preposition 'like'. In: Blatná, R. and Petkevic, V. (Eds.),
Languages and Linguistics: Festschrift for Fr. Cermak. Charles University, Prague.