Goal detection from natural language queries

This paper aims to identify the communication goal(s) of a user's information-seeking query out of a finite set of within-domain goals in natural language queries. It proposes using Tree-Augmented Naive Bayes networks (TANs) for goal detection. The problem is formulated as N binary decisions, and each is performed by a TAN. Comparative study has been carried out to compare the performance with Naive Bayes, fully-connected TANs, and multi-layer neural networks. Experimental results show that TANs consistently give better results when tested on the ATIS and DARPA Communicator corpora.

A comparative study of bayesian models for unsupervised sentiment detection

This paper presents a comparative study of three closely related Bayesian models for unsupervised document level sentiment classification, namely, the latent sentiment model (LSM), the joint sentiment-topic (JST) model, and the Reverse-JST model. Extensive experiments have been conducted on two corpora, the movie review dataset and the multi-domain sentiment dataset. It has been found that while all the three models achieve either better or comparable performance on these two corpora when compared to the existing unsupervised sentiment classification approaches, both JST and Reverse-JST are able to extract sentiment-oriented topics. In addition, Reverse-JST always performs worse than JST suggesting that the JST model is more appropriate for joint sentiment topic detection.

Issues in creating a corpus for EAP pedagogy and research

UK universities are accepting increasing numbers of students whose L1 is not English on a wide range of programmes at all levels. These students require additional support and training in English, focussing on their academic disciplines. Corpora have been used in EAP since the 1980s, mainly for research, but a growing number of researchers and practitioners have been advocating the use of corpora in EAP pedagogy, and such use is gradually increasing. This paper outlines the processes and factors to be considered in the design and compilation of an EAP corpus (e.g., the selection and acquisition of texts, metadata, data annotation, software tools and outputs, web interface, and screen displays), especially one intended to be used for teaching. Such a corpus would also facilitate EAP research in terms of longitudinal studies, student progression and development, and course and materials design. The paper has been informed by the preparatory work on the EAP subcorpus of the ACORN corpus project at Aston University. © 2007 Elsevier Ltd. All rights reserved.

Joint sentiment/topic model for sentiment analysis

Sentiment analysis or opinion mining aims to use automated tools to detect subjective information such as opinions, attitudes, and feelings expressed in text. This paper proposes a novel probabilistic modeling framework based on Latent Dirichlet Allocation (LDA), called joint sentiment/topic model (JST), which detects sentiment and topic simultaneously from text. Unlike other machine learning approaches to sentiment classification which often require labeled corpora for classifier training, the proposed JST model is fully unsupervised. The model has been evaluated on the movie review dataset to classify the review sentiment polarity and minimum prior information have also been explored to further improve the sentiment classification accuracy. Preliminary experiments have shown promising results achieved by JST.

LRD : Latent Relation Discovery for vector space expansion and information retrieval

In this paper, we propose a text mining method called LRD (latent relation discovery), which extends the traditional vector space model of document representation in order to improve information retrieval (IR) on documents and document clustering. Our LRD method extracts terms and entities, such as person, organization, or project names, and discovers relationships between them by taking into account their co-occurrence in textual corpora. Given a target entity, LRD discovers other entities closely related to the target effectively and efficiently. With respect to such relatedness, a measure of relation strength between entities is defined. LRD uses relation strength to enhance the vector space model, and uses the enhanced vector space model for query based IR on documents and clustering documents in order to discover complex relationships among terms and entities. Our experiments on a standard dataset for query based IR shows that our LRD method performed significantly better than traditional vector space model and other five standard statistical methods for vector expansion.

A weakly-supervised Bayesian model for violence detection from social media

Social streams have proven to be the mostup-to-date and inclusive information on cur-rent events. In this paper we propose a novelprobabilistic modelling framework, called violence detection model (VDM), which enables the identification of text containing violent content and extraction of violence-related topics over social media data. The proposed VDM model does not require any labeled corpora for training, instead, it only needs the in-corporation of word prior knowledge which captures whether a word indicates violence or not. We propose a novel approach of deriving word prior knowledge using the relative entropy measurement of words based on the in-tuition that low entropy words are indicative of semantically coherent topics and therefore more informative, while high entropy words indicates words whose usage is more topical diverse and therefore less informative. Our proposed VDM model has been evaluated on the TREC Microblog 2011 dataset to identify topics related to violence. Experimental results show that deriving word priors using our proposed relative entropy method is more effective than the widely-used information gain method. Moreover, VDM gives higher violence classification results and produces more coherent violence-related topics compared toa few competitive baselines.

Corpus linguistics

Corpora—large collections of written and/or spoken text stored and accessed electronically—provide the means of investigating language that is of growing importance academically and professionally. Corpora are now routinely used in the following fields: The production of dictionaries and other reference materials; The development of aids to translation; Language teaching materials; The investigation of ideologies and cultural assumptions; Natural language processing; and The investigation of all aspects of linguistic behaviour, including vocabulary, grammar and pragmatics.

Creation of a Digital Corpus of Bulgarian Dialects

The paper presents our considerations related to the creation of a digital corpus of Bulgarian dialects. The dialectological archive of Bulgarian language consists of more than 250 audio tapes. All tapes were recorded between 1955 and 1965 in the course of regular dialectological expeditions throughout the country. The records typically contain interviews with inhabitants of small villages in Bulgaria. The topics covered are usually related to such issues as birth, everyday life, marriage, family relationship, death, etc. Only a few tapes contain folk songs from different regions of the country. Taking into account the progressive deterioration of the magnetic media and the realistic prospects of data loss, the Institute for Bulgarian Language at the Academy of Sciences launched in 1997 a project aiming at restoration and digital preservation of the dialectological archive. Within the framework of this project more than the half of the records was digitized, de-noised and stored on digital recording media. Since then restoration and digitization activities are done in the Institute on a regular basis. As a result a large collection of sound files has been gathered. Our further efforts are aimed at the creation of a digital corpus of Bulgarian dialects, which will be made available for phonological and linguistic research. Such corpora typically include besides the sound files two basic elements: a transcription, aligned with the sound file, and a set of standardized metadata that defines the corpus. In our work we will present considerations on how these tasks could be realized in the case of the corpus of Bulgarian dialects. Our suggestions will be based on a comparative analysis of existing methods and techniques to build such corpora, and by selecting the ones that fit closer to the particular needs. Our experience can be used in similar institutions storing folklore archives, history related spoken records etc.

The Latest Prague Contributions to Written Cultural Heritage Processing

* The following text has been originally published in the Proceedings of the Language Recourses and Evaluation Conference held in Lisbon, Portugal, 2004, under the title of "Towards Intelligent Written Cultural Heritage Processing - Lexical processing". I present here a revised contribution of the aforementioned paper and I add here the latest efforts done in the Center for Computational Linguistic in Prague in the field under discussion.

Institute for Bulgarian Language, BAS

The paper presents the history, structure and ongoing activities of the Institute for Bulgarian Language of Bulgarian Academy of Sciences.

Language Resources – a Part of World Cultural Heritage

This article briefly reviews multilingual language resources for Bulgarian, developed in the frame of some international projects: the first-ever annotated Bulgarian MTE digital lexical resources, Bulgarian-Polish corpus, Bulgarian-Slovak parallel and aligned corpus, and Bulgarian-Polish-Lithuanian corpus. These resources are valuable multilingual dataset for language engineering research and development for Bulgarian language. The multilingual corpora are large repositories of language data with an important role in preserving and supporting the world's cultural heritage, because the natural language is an outstanding part of the human cultural values and collective memory, and a bridge between cultures.

Bilingual Corpus - Digital Repository for Preservation of Language Heritage

The article briefly reviews bilingual Slovak-Bulgarian/Bulgarian-Slovak parallel and aligned corpus. The corpus is collected and developed as results of the collaboration in the frameworks of the joint research project between Institute of Mathematics and Informatics, Bulgarian Academy of Sciences, and Ľ. Štúr Institute of Linguistics, Slovak Academy of Sciences. The multilingual corpora are large repositories of language data with an important role in preserving and supporting the world's cultural heritage, because the natural language is an outstanding part of the human cultural values and collective memory, and a bridge between cultures. This bilingual corpus will be widely applicable to the contrastive studies of the both Slavic languages, will also be useful resource for language engineering research and development, especially in machine translation.

Dialect variation

Relatively little research on dialect variation has been based on corpora of naturally occurring language. Instead, dialect variation has been studied based primarily on language elicited through questionnaires and interviews. Eliciting dialect data has several advantages, including allowing for dialectologists to select individual informants, control the communicative situation in which language is collected, elicit rare forms directly, and make high-quality audio recordings. Although far less common, a corpus-based approach to data collection also has several advantages, including allowing for dialectologists to collect large amounts of data from a large number of informants, observe dialect variation across a range of communicative situations, and analyze quantitative linguistic variation in large samples of natural language. Although both approaches allow for dialect variation to be observed, they provide different perspectives on language variation and change. The corpus- based approach to dialectology has therefore produced a number of new findings, many of which challenge traditional assumptions about the nature of dialect variation. Most important, this research has shown that dialect variation involves a wider range of linguistic variables and exists across a wider range of language varieties than has previously been assumed. The goal of this chapter is to introduce this emerging approach to dialectology. The first part of this chapter reviews the growing body of research that analyzes dialect variation in corpora, including research on variation across nations, regions, genders, ages, and classes, in both speech and writing, and from both a synchronic and diachronic perspective, with a focus on dialect variation in the English language. Although collections of language data elicited through interviews and questionnaires are now commonly referred to as corpora in sociolinguistics and dialectology (e.g. see Bauer 2002; Tagliamonte 2006; Kretzschmar et al. 2006; D'Arcy 2011), this review focuses on corpora of naturally occurring texts and discourse. The second part of this chapter presents the results of an analysis of variation in not contraction across region, gender, and time in a corpus of American English letters to the editor in order to exemplify a corpus-based approach to dialectology.

A Stochastic Approach for Finding of Semantically Related Words

2000 Mathematics Subject Classification: 62P99, 68T50