Rogelio Nazar’s research while affiliated with Pontifical Catholic University of Valparaíso and other places

This paper presents a statistical model of the opinion column, a discourse genre typical of the press. The model was derived from a relatively small sample (ca. 4000 texts), and it takes into account discourse variables such as text and paragraph length, discourse markers, deixis and modalization. In order to test the accuracy of the model, it was evaluated against a different corpus of mixed column and non-column documents. The idea was to test whether the model is able to identify those texts pertaining to the target genre. Results show that it is indeed accurate, with results ranging from 85 to 77% precision and 40–61% recall, depending on how restrictive the application of the model is.

We propose a method for the task of automatic terminology extraction in the context of a larger project devoted to the automation of part of the tasks involved in the production of terminological databases. Terminology extraction is the key to drafting the macrostructure of a terminological resource (i.e., the list of entries), to which information can be later added at the microstructural level with grammatical or semantic information. To this end, we developed a statistical method that is conceptually simple compared to modern neural network approaches. It is a lightweight method because it is based on term dispersion and co-occurrence statistics that can be computed with basic hardware. For the evaluation, we experimented with corpora of lexicography and linguistics in English and Spanish of ca. 66 million tokens. Results improve baselines in almost 20%.

Se describe Estilector, un sistema de retroalimentación automática de escritura académica en castellano. El software está orientado principalmente a la detección de problemas textuales, aunque también incorpora, en menor medida, correcciones gramaticales y ortográficas. Pretende apoyar el aprendizaje de la escritura aportando feedback formativo, y proporciona referencias bibliográficas y sugerencias de revisión más extensas que las de los procesadores de texto. El programa funciona actualmente en línea y es de acceso gratuito. En sus años de existencia, ha tenido gran número y variedad de usuarios, y se ha ido adaptando a un público que requiere apoyo básico y transversal a la hora de escribir. En este artículo ofrecemos una descripción del estado actual del proyecto con sus métodos de detección y revisión de errores además de algunas reflexiones finales, incluyendo sus limitaciones y posibilidades de desarrollo futuro.

We propose a method for the automatic induction of categories of Spanish discourse markers using parallel corpora, based on a quantitative and empirical approach that minimises explicit linguistic knowledge. We conducted the analysis the using a large Spanish-English parallel corpus. First, we used this corpus to obtain a list of parenthetical discourse markers in each language. Then, we used it as a “semantic mirror”, inspecting the English equivalences and assessing which Spanish discourse markers fulfil a similar function in discourse and vice versa. The result of this procedure is an emerging categorisation of discourse markers. The main contribution is to offer empirical evidence for the adequacy of existing manually-compiled taxonomies and the potential for discovery of new, unaccounted categories. In this article we focus on units pertaining to the Spanish language but, since the method is purely quantitative, it is possible to apply it to different languages as well.

Digital Humanities, Corpus and Language Technology: A look from diverse case studies is an outstanding collection of research contributions that explores the intersection of technology and the humanities. The authors provide a comprehensive overview of how these technologies can enhance research across various disciplines, from literature to history to anthropology. This book is a mustread for anyone interested in future research in the humanities. Digital Humanities, Corpus, and Language Technologies are rapidly growing fields that have the potential to revolutionize research across various disciplines. New technologies have opened up new perspectives for research, allowing scientists to analyze data in previously impossible ways. The interdisciplinary approach and practical applications make it an invaluable resource for researchers, students, and anyone interested in the intersection of technology and the humanities.

A partir de la consideración de los medios de comunicación como espacios propicios para la estandarización de la norma lingüística, la presente investigación analiza el grado de aceptación de algunas normas que aparecieron en la Ortografía de la lengua española (RAE y ASALE 2010): a) la tilde de guión/guion y otras palabras similares, b) sólo/solo y los pronombres demostrativos y c) algunos casos de extranjerismos y latinismos adaptados (como whisky/wiski). Se analizaron dos diarios de cinco países hispanohablantes (Argentina, Chile, Colombia, España y México). Se utilizó GoogleApi para recopilar 179.238 contextos de aparición en total, divididos por años y diarios, durante los primeros diez años de vigencia de la obra (2010-2019). Los resultados indican que, al final del periodo estudiado, permanecen aún, en conjunto, numerosos casos de formas con la ortografía antigua en los tres grupos.

This paper presents a series of methods for automatically determining the gender of proper names, based on their co-occurrence with words and grammatical features in a large corpus. Although the results obtained were for Spanish given names, the method presented here can be easily replicated and used for names in other languages. Most methods reported in the literature use pre-existing lists of first names that require costly manual processing and tend to become quickly outdated. Instead, we propose using corpora. Doing so offers the possibility of obtaining real and up-to-date name-gender links. To test the effectiveness of our method, we explored various machine-learning methods as well as another method based on simple frequency of co-occurrence. The latter produced the best results: 93% precision and 88% recall on a database of ca. 10,000 mixed names. Our method can be applied to a variety of natural language processing tasks such as information extraction, machine translation, anaphora resolution or large-scale delivery or email correspondence, among others.

The appearance of new verbs can be observed regularly, but verbs are not frequently investigated in neology, and they are difficult to detect automatically. In this study, a corpus-based method is proposed to detect Spanish verbs with a series of algorithms that analyse the morphology of regular verbs. The vocabulary was drawn from a large corpus and contrasted with a major dictionary of Spanish. Then, a series of filters were applied to distinguish between valid neologism candidates and spelling mistakes. Around 88% of the neologisms proposed by the method were correct and we estimate that the system detected 76% of the neologisms present in the corpus. This procedure can be included in the workflow of a lexicographic project as a regular part of the task, as a systematic way of collecting new verbs from the data and avoiding under-representation or bias.

In this paper we present the problem of a noisy lexical taxonomy and suggest two tasks as potential remedies. The first task is to identify and eliminate incorrect hypernymy links, and the second is to repopulate the taxonomy with new relations. The first task consists of revising the entire taxonomy and returning a Boolean for each assertion of hypernymy between two nouns (e.g. brie is a kind of cheese ). The second task consists of recursively producing a chain of hypernyms for a given noun, until the most general node in the taxonomy is reached (e.g. brie → cheese → food → etc.). In order to achieve these goals, we implemented a hybrid hypernym-detection algorithm that incorporates various intuitions, such as syntagmatic, paradigmatic and morphological association measures as well as lexical patterns. We evaluate these algorithms individually and collectively and report findings in Spanish, English and French.

This paper presents a series of methods for automatically determining the gender of proper names, based on their co-occurrence with words and grammatical features in a large corpus. Although the results obtained were for Spanish given names, the method presented here can be easily replicated and used for names in other languages as well. Most methods reported in the literature use pre-existing lists of first names that require costly manual processing and tend to become quickly outdated. Instead, we propose using corpora. Doing so offers the possibility of obtaining real and up-to-date name-gender links. To test the effectiveness of our method, we explored various machine learning methods as well as another method based on simple frequency of co-occurrence. The latter produced the best results: 93% precision and 88% recall on a database of ca. 10,000 mixed names. Our method can be applied to a variety of natural language processing tasks such as information extraction, machine translation, anaphora resolution or large-scale delivery or email correspondence, among others.