Evangelos Milios

• Dalhousie University

This research introduces the Conceptual Document Clustering Explanation Model (CDCEM), a novel explanation model for explaining unsupervised textual clustering. CDCEM explains the discovered clusters and document assignments. Furthermore, it ensures faithfulness—meaning it accurately reflects the decision-making process—using the core elements of b...

Wikification (entity annotation) is a challenging task in Natural Language Processing (NLP). It is a method to automatically enrich a text with links to Wikipedia as a knowledge base. Wikification starts from detecting ambiguous mentions in the document, and later tries to disambiguate those mentions. In the core of the Wikification task, there is...

Contextualized sense embedding has been shown to carry useful semantic information to improve the final results of various Natural Language Processing tasks. However, it is still challenging to integrate them with the information of the knowledge base, which is one lack in current state-of-the-art representations. This integration is helpful in NLP...

Word Sense Disambiguation (WSD) aims to determine the meaning of a word in context. Different approaches have been proposed in supervised and unsupervised domains. In most cases, supervised learning provides superior WSD performance. Since sense-annotated corpora can be difficult or time-consuming to obtain, which must be repeated for new domains,...

Word Sense Disambiguation (WSD) is a well-known problem in the field of Natural Language Processing (NLP) related to automatically determining the most appropriate sense of words in context. Several machine learning-based approaches have been proposed to tackle the ambiguity of language, but the lack of labeled data to train supervised models made...

Short text clustering is a challenging task due to the lack of signal contained in short texts. In this work, we propose iterative classification as a method to boost the clustering quality of short texts. The idea is to repeatedly reassign (classify) outliers to clusters until the cluster assignment stabilizes. The classifier used in each iteratio...

Harassment on social media is a hard problem to tackle since those platforms are virtual spaces in which people enjoy the liberty to express themselves with no restrictions. Furthermore, a large amount of users generating publications on online media like Twitter contributes to the hardness of controlling sexism and sexual harassment content, reque...

Short text clustering is a challenging task due to the lack of signal contained in such short texts. In this work, we propose iterative classification as a method to b o ost the clustering quality (e.g., accuracy) of short texts. Given a clustering of short texts obtained using an arbitrary clustering algorithm, iterative classification applies out...

A key challenge for document clustering consists in finding a proper similarity measure for text documents that enables the generation of cohesive groups. Measures based on the classic bag-of-words model take into account solely the presence (and frequency) of words in documents. In doing so, semantically similar documents which use different vocab...

Harassment on social media is a hard problem to tackle since those platforms are virtual spaces in which people enjoy the liberty to express themselves with no restrictions. Furthermore, a large amount of users generating publications on online media like Twitter contributes to the hardness of controlling sexism and sexual harassment content, reque...

Modern OCR engines incorporate some form of error correction, typically based on dictionaries. However, there are still residual errors that decrease performance of natural language processing algorithms applied to OCR text. In this paper, we present a statistical learning model for post-processing OCR errors, either in a fully automatic manner or...

Short text clustering is an important but challenging task. We investigate impact of similarity matrix sparsification on the performance of short text clustering. We show that two sparsification methods (the proposed Similarity Distribution based, and k-nearest neighbors) that aim to retain a prescribed number of similarity elements per text, impro...

In this paper, we propose a high recall active document retrieval system for a class of applications involving query documents, as opposed to key terms, and domain-specific document corpora. The output of the model is a list of documents retrieved based on the domain expert feedback collected during training. A modified version of Bag of Word (BoW)...

Attention guides computation to focus on important parts of the input data. For pairwise input, existing attention approaches tend to bias towards trivial repetitions (e.g. punctuations and stop words) between two texts, and thus failed to contribute reasonable guidance to model predictions. As a remedy, we suggest taking into account the corpus-le...

This article presents Tulip, an ERD system submitted to the ERD 2014: Entity Recognition and Disambiguation Challenge. The objective of the proposed system is to spot mentions of entities in a document and link the mentions to corresponding Freebase articles. To achieve it, Tulip prunes the set of entity candidates focusing on a core subset of rela...

We propose a model for assigning Wikipedia Concepts as scientific category labels to scientific documents where their terms are first grouped together using the well-known topic modelling method, Latent Dirichlet Allocation (LDA) and then assigned to Wikipedia Concepts by wikification. We wikify the terms of the topic model of a document to extract...

A typical on-line content directory contains factual information about entities (e.g., address of a company) together with entity categories (e.g., company’s industries). The categories are a salient element of the system as they allow users to browse for entities of a chosen type. Assigning categories manually can be a challenging task, considerin...

In this work, we propose an approach based on analyzing the spatio-temporal partitions of a system log, generated by supercomputers consisting of several nodes, for alert detection without employing semantic analysis. In this case, "Spatial" refers to the source of the log event and "Temporal" refers to the time the log event was reported. Our rese...

The ability to automatically discover error conditions with little human input is a feature lacking in most modern computer systems and networks. However, with the ever increasing size and complexity of modern systems, such a feature will become a necessity in the not too distant future. Our work proposes a hybrid framework that allows High Perform...

Social networks and collaborative tagging systems are rapidly gaining popularity as a primary means for storing and sharing data among friends, family, colleagues, or perfect strangers as long as they have common interests. del.icio.us3 is a social network where people store and share their personal bookmarks. Most importantly, users tag their book...

Despite all of the advantages of tags as an easy and flexible information management approach, tagging is a cumbersome task. A set of descriptive tags has to be manually entered by users whenever they post a resource. This process can be simplified by the use of tag recommendation systems. Their objective is to suggest potentially useful tags to th...

We address the problem of evaluating the robustness of machine learning based detectors for deployment in real life networks. To this end, we employ Genetic Programming for evolving classifiers and Artificial Neural Networks as our machine learning paradigms under three different Denial-of-Service attacks at the Data Link layer De-authentication, A...

Self-awareness is an important attribute for any system to have before it is capable of self-management. A system needs to have a continuous stream of real-time data to analyze to allow it be aware of its internal state. To this end, previous approaches have utilized system performance metrics and system log data to characterize system internal sta...

Recent research efforts have highlighted the capability of entropy based approaches in the automatic discovery of alerts in system logs. In this work, we extend this research to present the evaluations of three entropy based approaches on new datasets not utilized in previous papers. We also extend the approach with the introduction of a Cluster Me...

Message types are semantic groupings of the free form messages in system log events. The message types that exist in a log file, if known, can be used in several log management and analysis tasks. In this work, we explore the use of message types as a schema definition for the storage and retrieval of messages in event logs. We show how message typ...

In this paper, we investigated four approaches for scientific corpora summarization when only gold-standard keyterms available. MEAD with built-in default vocabulary, MEAD with corpus specific vocabulary extracted by Keyphrase Extraction Algorithm (KEA), LexRank (a state-of-the-art summarization algorithm based on random walk) and W3SS (summarizati...

Message type or message cluster extraction is an important task in the analysis of system logs in computer networks. Defining these message types automatically facilitates the automatic analysis of system logs. When the message types that exist in a log file are represented explicitly, they can form the basis for carrying out other automatic applic...

Purpose Summarization of an entire web site with diverse content may lead to a summary heavily biased towards the site's dominant topics. The purpose of this paper is to present a novel topic‐based framework to address this problem. Design/methodology/approach A two‐stage framework is proposed. The first stage identifies the main topics covered in...

The objective of a tag recommendation system is to propose a set of tags for a resource to ease the tagging process done manually by a user. Tag recommendation is an interesting and well defined research problem. However, while solving it, it is easy to forget about its practical implications. We discuss the practical aspects of tag recommendation...

Manual alert detection on modern high performance clusters (HPC) is cumbersome given their increasing complexity and size of their logs. The ability to automatically detect such alerts quickly and accurately with little or no human intervention is therefore desirable. The entropy-based approach of the Node- info framework, which is in production us...

Collaborative tagging systems are popular tools for organization, sharing and retrieval of web resources. Their success is due to their freedom and simplicity of use. To post a resource, the user should only define a set of tags that would position the resource in the system's data structure -- folksonomy. This data structure can serve as a rich so...

The task of alert detection in event logs is very impor-tant in preventing or recovering from downtime events. The ability to do this automatically and accurately provides sig-nificant savings in the time and cost of downtime events. The Nodeinfo algorithm, which is currently in production use at Sandia National Laboratories, is an entropy based al...

Abstract Hierarchical,clustering,of text collections,is a key,problem,in document,management,and retrieval. In partitional hierarchical clustering, which is more efficient than its agglomerative counterpart, the entire collection is split into clusters and the individual clusters are further split until a heuristically-motivated termination criteri...

Social networks and collaborative tagging systems are rapidly gaining popularity as a primary means for storing and sharing data among friends, family, colleagues, or perfect strangers as long as they have common interests. del.icio.us(3) is a social network where people store and share their personal bookmarks. Most importantly users tag their boo...

The task of alert detection in event logs, i.e. determining which events in the event log require action from an administrator, is very important in preventing or recovering from downtime events. The ability to do this automatically and accurately provides significant savings in time and cost of downtime events. In this work we combine message type...

Given a set of peers with overlapping interests where each peer wishes to keep track of new documents that are rel- evant to their interests, we propose a self-organizing peer- to-peer document-tracking network based on common inter- est profiles. The goal of a document-tracking network is to disseminate new documents as they are published. Peers c...

This work addresses issues related to the design and implementation of focused crawlers. Several variants of state-of-the-art crawlers relying on web page content and link information for estimating the relevance of web pages to a given topic are proposed. Particular emphasis is given to crawlers capable of learning not only the content of relevant...

Size and complexity of data repositories collaboratively cre- ated by Web users generate a need for new processing ap- proaches. In this paper, we study the problem of detection of fine-grained communities of users in social networks, which can be defined as clustering with a large number of clusters. The practical size of social networks makes the...

The importance of event logs, as a source of information in systems and network management cannot be overempha- sized. With the ever increasing size and complexity of to- day's event logs, the task of analyzing event logs has become cumbersome to carry out manually. For this reason recent research has focused on the automatic analysis of these log...

People's email communications can be modeled as graphs with vertices representing email accounts and edges repre- senting email communications. Email communication data usually comes in as continuous data stream. Event detec- tion aims to identify abnormal email communications that serve as analogs of real-world events imposed upon the data stream....

Collaborative tagging systems are social data repositories, in which users manage resources using descriptive keywords (tags). An im- portant element of collaborative tagging systems is the tag recommender, which proposes a set of tags to each newly posted resource. In this pa- per we discuss the potential role of three tag sources: resource conten...

To be immersed in a virtual environment, the user must be presented with plausible sensory input including auditory cues. A virtual (three-dimensional) audio display aims to allow the user to perceive the position of a sound source at an arbitrary position in three-dimensional space despite the fact that the generated sound may be emanating from a...

Communication networks, such as networks formed by phone calls and email communications, can be modeled as dynamic graphs with vertices representing agents and edges representing communications. Anomaly detection is to identify abnormal behaviour occurring in these networks. This is crucial for anti-terrorism, resource allocation and network manage...

Sonel mapping is a Monte-Carlo-based acoustical modeling technique that approximates the acoustics of an environment while accounting for diffuse and specular reflections as well as diffraction effects. Through the use of a probabilistic Russian roulette strategy to determine the type of interaction between a sound and any objects/surfaces it may e...

To be immersed in a virtual environment, the user must be presented with plausible sensory input including auditory cues. A virtual (three-dimensional) audio display aims to allow the user to perceive the position of a sound source at an arbitrary position in three-dimensional space despite the fact that the generated sound may be emanating from a...

Event logs or log files form an essential part of any network management and administration setup. While log files are invaluable to a network administrator, the vast amount of data they sometimes contain can be overwhelming and can sometimes hinder rather than facilitate the tasks of a network administrator. For this reason several event clusterin...

Image-based abstraction (or summarization) of a Web site is the process of extracting the most characteristic (or important) images from it. The criteria for measuring the importance of images in Web sites are based on their frequency of occurrence, characteristics of their content and Web link information. As a case study, this work focuses on log...

Semantic Similarity relates to computing the similarity between conceptually similar but not necessarily lexically similar terms. Typically, semantic similarity is computed by mapping terms to an ontology and by examining their relationships in that ontology. We investigate approaches to computing the semantic similarity between natural language te...

Focused web crawling traverses the Web to collect documents on a specific topic. This is not an easy task, since focused crawlers need to identify the next most promising link to follow based on the topic and the content and links of previously crawled pages. In this paper, we present a framework based on Maximum Entropy Markov Models(MEMMs) for an...

This paper presents a statistical model for discovering top ical clusters of words in unstructured text. The model uses a hierarchical Bayesian structure and it is also able to identify segments of text which are topi cally coherent. The model is able to assign each segment to a particular topic and thus categorizes the corresponding document to po...

Network administration is a task that requires experience in relating symptoms of network problems with possible causes and corrective actions. We describe the design of a system and more specifically its information retrieval component, which aims to retrieve articles relevant to a given problem case from a collection of articles describing previo...

Security and Intrusion detection in WiFi networks is currently an active area of research where WiFi specific Data Link layer attacks are an area of focus; particularly recent work has focused on producing machine learning based IDSs for these WiFi specific attacks. These proposed machine learning based IDSs come in addition to the already deployed...

Netpal is a web-based dynamic knowledge base system designed to assist network administra- tors in their troubleshooting tasks, in recall- ing and storing experience, and in identify- ing new failure cases and their symptoms. In the context of web hosting environments, Net- pal summarises network data and and supports retrieval of relevant organisa...

A great challenge of text mining arises from the increasingly large text datasets and the high di- mensionality associated with natural language. In this research, a systematic study is conducted of six Dimension Reduction Techniques (DRT) in the context of the text clustering problem using three standard benchmark datasets. The methods considered...

Social networks and collaborative tagging systems are rapidly gaining popularity as a primary means for storing and sharing data among friends, family, colleagues, or perfect strangers as long as they have common interests. del.icio.us is a social network where people store and share their personal bookmarks. Most importantly, users tag their bookm...

Web Site Summarization is the process of automatically generating a concise and informative summary for a given Web site. It has gained more and more attention in recent years as effective summarization could lead to enhanced Web information retrieval systems such as searching for Web sites. Extraction-based approaches to Web site summarization rel...

This work shows that it is possible to exploit text and im- age content characteristics of logo and trademark images in Web pages for enhancing the performance of retrievals on the Web. Searching for important (authoritative) Web pages and images is a desirable feature of many Web search en- gines and is also taken into account. State-of-the-art me...

Increasingly large text damsels and the high dimensionality associated with natural language create a great challenge in text mining, In this research, a systematic study is conducted. in which three different document representation methods for text are used, together with three Dimension Reduction Techniques (DRT), in the context of the text clus...

Shrack is a peer-to-peer framework for document sharing and tracking. Shrack peers provide support to researchers in forming direct collaboration in autonomous sharing and keeping track of newly published documents based on their interests. We propose a pull-only information dissemination protocol for peers to distribute information about new docum...

The aquatic environment is almost ideal for autonomous robot development. First, it provides a range of real tasks for autonomous systems to perform, including ongoing inspection of reef damage and renewal; tasks in the oil and gas industry; and aquaculture. Second, operating in the water requires robust solutions to mobility, sensing, navigation,...

Published scientific articles are linked together into a graph, the citation graph, through their citations. This paper explores the notion of similarity based on connectivity alone, and proposes several algorithms to quantify it. Our metrics take advantage of the local neighborhoods of the nodes in the citation graph. Two variants of link-based si...

Since the dimensions of many of the objects/surfaces encountered in our daily lives are within an order of mag- nitude of the wavelength of audible sounds, diffraction is an elementary means of sound propagation. Despite its importance in the real-world, diffraction effects are often overlooked by acoustical modeling methods lead- ing to a degredat...

Web Site Summarization is the process af automatically generating a concise and informative summary for a given Web site. It has gained more and more attention in recent Years as effective summarization could lead to enhanced Web information retrieval systems such as searching for Web sites. Extraction-based approaches to Web sife summarization rel...

We present a generative model for simultaneously clustering documents and terms. Our model is a four-level hierarchical Bayesian model, in which each document is modeled as a random mixture of document topics , where each topic is a distribution over some segments of the text. Each of these segments in the document can be modeled as a mixture of wo...

In recent years, the emergence of the Web and the dramatic increase in computing, storage and networking capacity has given rise to the concept of networked information spaces. The prime example of a networked information space is the World Wide Web itself. The Web, in its pure form, is a set of hypertext documents, with links in one document point...

For barely a decade now the Web graph (the network formed by Web pages and their hyperlinks) has been the focus of scientific study. In that short a time, this study has made a significant impact on research in physics, computer science and mathematics. It has focussed the attention of the scientific community on all the different kinds of networks...

Generative models are often used in modeling real world graphs such as the Web graph in order to better understand the processes through which these graphs are formed. In order to determine if a graph might have been generated by a given model one must compare the features of that graph with those generated by the model. We introduce the concept of...

ABSTRACT Term extraction relates to extracting the most characteristic or important terms (words or phrases) in a document. This information is commonly,used for improving,the accuracy of document,indexing and retrieval in large text collections. It also allows for faster and better understanding of the contents of a document,collection without fir...

A focused crawler is designed to traverse the Web to gather documents on a specific topic. It can be used to build domain-specific Web search portals and online personalized search tools. To estimate the relevance of a newly seen URL, it must use information gleaned from previously crawled page sequences.In this paper, we present a new approach for...

IntelliSearch is a complete and fully automated information retrieval system for the Web. It supports fast and accurate responses to queries addressing text and images in Web pages by incorporating state-of-the-art text and Web link information indexing and rertieval methods in conjunction with efficient ranking of Web pages and images by importanc...

MedSearch1 is a complete retrieval system for Medline, the premier bibliographic database of the U.S. National Library of Medicine (NLM). MedSearch implements SSRM, a novel information retrieval method for discovering similarities between documents containing se- mantically similar but not necessarily lexically similar terms.

Semantic Similarity relates to computing the similarity between conceptually similar but not nec- essarily lexically similar terms. Typically, semantic similarity is computed by mapping terms to an ontology and by examining their relationships in that ontology. We investigate approaches to computing the semantic similarity between natural language...

Modeling the acoustics of an environment is a complex and challenging task. Here we describe the sonel mapping approach to acoustical rendering. Sonel mapping is a Monte-Carlo-based approach to modeling diffuse, specular, absorption and diffraction effects in an efficient manner. The approach models many of the subtle interaction effects required f...

Relevance feedback is the state-of-the-art approach for adjusting query results to the needs of the users. This work extends the ex- isting framework of image retrieval with relevance feedback on the Web by incorporating text and image content into the search and feedback process. Some of the most powerful relevance feedback methods are implemented...

Semantic Similarity relates to computing the similarity between concepts which are not lexicographically similar. We investigate approaches to computing semantic similarity by mapping terms (concepts) to an ontology and by examining their relationships in that ontology. Some of the most popular semantic similarity methods are implemented and evalua...

Automatic key phrase extraction is a useful tool in many text related applications such as clustering and summarization. State-of-the-art methods are aimed towards extracting key phrases from traditional text such as technical papers. Application of these methods on Web documents, which often contain diverse and heterogeneous contents, is of partic...

In this paper we describe recent work toprovide a filtering service for readers interested in medically related news articles from online news sources. The first task is to filter out the nonmedical news articles. The remaining articles, the medically related ones, are then assigned MeSH headings for context and then categorized further by intended...

This paper describes the application of the Huygens-Fresnel principle to acoustical diffraction modeling. A theoretical formulation of the optics-based Huygens-Fresnel principle is presented followed by a discussion regarding the modifications necessary to apply the Huygens-Fresnel principle to acoustical diffraction modeling. Experimental results...

We propose a novel method for document clustering using character N-grams. In the traditional vector-space model, the documents are represented as vectors, in which each dimension corresponds to a word. We propose a document representation based on the most frequent character N-grams, with window size of up to 10 characters. We derive a new distanc...

We describe recent results obtained with AQUA, a mobile robot capable of swimming, walking and amphibious operation. Designed to rely primarily on visual sensors, the AQUA robot uses vision to navigate underwater using servo-based guidance, and also to obtain high-resolution range scans of its local environment. This paper describes some of the pra...

Dynamic web sites commonly return information in the form of lists and tables. Although hand crafting an extraction program for a specific template is time-consuming but straightforward, it is desirable to automatically generate template extraction programs from examples of lists and tables in html documents. Supervised approaches have been shown t...

Clinical practice guidelines (CPG) are instrumental in standardizing the quality and delivery of care across different practitioners, departments and even institutions. Health practitioners when working with CPG like to reflect on the current best evidence either to validate or to supplement their understanding of the CPG. In this paper we investig...

Passive acoustic techniques are presented to solve the localization problem of a sound source in three-dimensional space using off-the-shelf hardware. Multiple microphone arrays are employed, which operate independently, in estimating the direction of arrival of sound, or, equivalently, a direction vector from the array's geometric center towards t...

We briefly describe DalTREC 2005 Spam submission. DalTREC is the TREC research project at Dalhousie University. Four packages were submitted and they resulted in a median performance. The results are interesting and may be seen positive in the light of simplicity of our approaches.

Image retrieval on the Web requires that important (authoritative) images satisfying the query selection criteria are assigned higher ranking over other relevant images. PicASHOW (Lempel and Soffer, 2002) achieves this goal using link information alone. This work proposes WPicASHOW (weighted PicASHOW), a weighted scheme for co-citation analysis tha...

In this paper we introduce a framework to supplement and tag computerized CPG with related best-evidence automatically sourced from on-line medical literature repositories. The idea is to provide CPG users with additional published evidence pertaining to the different sections of a CPG of their interest. We present a web- enabled Best-evidence Retr...

One problem with geometric (ray) based acoustical modeling approaches is handling the potentially large number of interactions between a propagating acoustic ray and objects/surfaces it may encounter. Typical solutions to modeling these interactions include emitting several "new" rays at each interaction point. Unfortunately this type of approach i...

A Focused crawler must use information gleaned from previously crawled page sequences to estimate the relevance of a newly seen URL. Therefore, good performance depends on powerful modelling of context as well as the current observations. Probabilistic models, such as Hidden Markov Models(HMMs) and Conditional Random Fields(CRFs), can potentially c...

Acoustic modeling of even small simple environments is a complex, computationally expensive task. Sound, just as light, is itself a wave phenomenon. Although there are several key differences between light and sound there are also several similarities. Given the similarities which exist between sound and light, this work investigates the applicatio...

Citation graphs representing a body of scientific literature convey measures of schol- arly activity and productivity. In this work we present a study of the structure of the citation graph of the computer science literature. Using a web robot we built several topic-specific ci- tation graphs and their union graph from the digital library ResearchI...

Based on the RHHex hexapod robot, the AQUA robot is an aquatic robot that swims via the motion of its legs, rather than using thrusters and control surfaces for propulsion. Through an appropriate set of gaits, the AQUA vehicle is capable of five-degree-of-freedom motion in the open water, it can swim along the surface and it can walk along the seab...

availability of the right medical knowledge at the right time. This concept paper presents a knowledge management research program to (a) identify, capture and organize the tacit knowledge inherent within on-line problem-solving discussions between pediatric pain practitioners; (b) establish linkages between topic-specific pediatric pain discussion...