Asif Ekbal

The CHEMDNER corpus of chemicals and drugs and its annotation principles

The automatic extraction of chemical information from text requires the recognition of chemical entity mentions as one of its key steps. When developing supervised named entity recognition (NER) systems, the availability of a large, manually annotated text corpus is desirable. Furthermore, large corpora permit the robust evaluation and comparison of different approaches that detect chemicals in documents. We present the CHEMDNER corpus, a collection of 10,000 PubMed abstracts that contain a total of 84,355 chemical entity mentions labeled manually by expert chemistry literature curators, following annotation guidelines specifically defined for this task. The abstracts of the CHEMDNER corpus were selected to be representative for all major chemical disciplines. Each of the chemical entity mentions was manually labeled according to its structure-associated chemical entity mention (SACEM) class: abbreviation, family, formula, identifier, multiple, systematic and trivial. The difficulty and consistency of tagging chemicals in text was measured using an agreement study between annotators, obtaining a percentage agreement of 91. For a subset of the CHEMDNER corpus (the test set of 3,000 abstracts) we provide not only the Gold Standard manual annotations, but also mentions automatically detected by the 26 teams that participated in the BioCreative IV CHEMDNER chemical mention recognition task. In addition, we release the CHEMDNER silver standard corpus of automatically extracted mentions from 17,000 randomly selected PubMed abstracts. A version of the CHEMDNER corpus in the BioC format has been generated as well. We propose a standard for required minimum information about entity annotations for the construction of domain specific corpora on chemical and drug entities. The CHEMDNER corpus and annotation guidelines are available at: http://www.biocreative.org/resources/biocreative-iv/chemdner-corpus/

Combining multiple classifiers using vote based classifier ensemble technique for named entity recognition

In this paper, we pose the classifier ensemble problem under single and multiobjective optimization frameworks, and evaluate it for Named Entity Recognition (NER), an important step in almost all Natural Language Processing (NLP) application areas. We propose the solutions to two different versions of the ensemble problem for each of the optimization frameworks. We hypothesize that the reliability of predictions of each classifier differs among the various output classes. Thus, in an ensemble system it is necessary to find out either the eligible classes for which a classifier is most suitable to vote (i.e., binary vote based ensemble) or to quantify the amount of voting for each class in a particular classifier (i.e., real vote based ensemble). We use seven diverse classifiers, namely Naive Bayes, Decision Tree (DT), Memory Based Learner (MBL), Hidden Markov Model (HMM), Maximum Entropy (ME), Conditional Random Field (CRF) and Support Vector Machine (SVM) to build a number of models depending upon the various representations of the available features that are identified and selected mostly without using any domain knowledge and/or language specific resources. The proposed technique is evaluated for three resource-constrained languages, namely Bengali, Hindi and Telugu. Results using multiobjective optimization (MOO) based technique yield the overall recall, precision and F-measure values of 94.21%, 94.72% and 94.74%, respectively for Bengali, 99.07%, 90.63% and 94.66%, respectively for Hindi and 82.79%, 95.18% and 88.55%, respectively for Telugu. Results for all the languages show that the proposed MOO based classifier ensemble with real voting attains the performance level which is superior to all the individual classifiers, three baseline ensembles and the corresponding single objective based ensemble.

A Modified Joint Source-Channel Model for Transliteration

Most machine transliteration systems transliterate out of vocabulary (OOV) words through intermediate phonemic mapping. A framework has been presented that allows direct orthographical mapping between two languages that are of different origins employing different alphabet sets. A modified joint source-channel model along with a number of alternatives have been proposed. Aligned transliteration units along with their context are automatically derived from a bilingual training corpus to generate the collocational statistics. The transliteration units in Bengali words take the pattern C+M where C represents a vowel or a consonant or a conjunct and M represents the vowel modifier or matra. The English transliteration units are of the form C*V* where C represents a consonant and V represents a vowel. A Bengali-English machine transliteration system has been developed based on the proposed models. The system has been trained to transliterate person names from Bengali to English. It uses the linguistic knowledge of possible conjuncts and diphthongs in Bengali and their equivalents in English. The system has been evaluated and it has been observed that the modified joint source-channel model performs best with a Word Agreement Ratio of 69.3% and a Transliteration Unit Agreement Ratio of 89.8%.

Part of Speech Tagging in Bengali Using Support Vector Machine

Part of speech (POS) tagging is the task of labeling each word in a sentence with its appropriate syntactic category called part of speech. POS tagging is a very important preprocessing task for language processing activities. This paper reports about task of POS tagging for Bengali using support vector machine (SVM). The POS tagger has been developed using a tagset of 26 POS tags, defined for the Indian languages. The system makes use of the different contextual information of the words along with the variety of features that are helpful in predicting the various POS classes. The POS tagger has been trained, and tested with the 72,341, and 20 K wordforms, respectively. Experimental results show the effectiveness of the proposed SVM based POS tagger with an accuracy of 86.84%. Results show that the lexicon, named entity recognizer and different word suffixes are effective in handling the unknown word problems and improve the accuracy of the POS tagger significantly. Comparative evaluation results have demonstrated that this SVM based system outperforms the three existing systems based on the hidden markov model (HMM), maximum entropy (ME) and conditional random field (CRF).

Weighted Vote-Based Classifier Ensemble for Named Entity Recognition: A Genetic Algorithm-Based Approach

In this article, we report the search capability of Genetic Algorithm (GA) to construct a weighted vote-based classifier ensemble for Named Entity Recognition (NER). Our underlying assumption is that the reliability of predictions of each classifier differs among the various named entity (NE) classes. Thus, it is necessary to quantify the amount of voting of a particular classifier for a particular output class. Here, an attempt is made to determine the appropriate weights of voting for each class in each classifier using GA. The proposed technique is evaluated for four leading Indian languages, namely Bengali, Hindi, Telugu, and Oriya, which are all resource-poor in nature. Evaluation results yield the recall, precision and F-measure values of 92.08%, 92.22%, and 92.15%, respectively for Bengali; 96.07%, 88.63%, and 92.20%, respectively for Hindi; 78.82%, 91.26%, and 84.59%, respectively for Telugu; and 88.56%, 89.98%, and 89.26%, respectively for Oriya. Finally, we evaluate our proposed approach with the benchmark dataset of CoNLL-2003 shared task that yields the overall recall, precision, and F-measure values of 88.72%, 88.64%, and 88.68%, respectively. Results also show that the vote based classifier ensemble identified by the GA-based approach outperforms all the individual classifiers, three conventional baseline ensembles, and some other existing ensemble techniques. In a part of the article, we formulate the problem of feature selection in any classifier under the single objective optimization framework and show that our proposed classifier ensemble attains superior performance to it.

Web-based Bengali News Corpus for Lexicon Development and POS Tagging

Lexicon development and Part of Speech (POS) tagging are very important for almost all Natural Language Processing (NLP) applications. The rapid development of these resources and tools using machine learning techniques for less computerized languages requires appropriately tagged corpus. We have used a Bengali news corpus, developed from the web archive of a widely read Bengali newspaper. The corpus contains approximately 34 million wordforms. This corpus is used for lexicon development without employing extensive knowledge of the language. We have developed the POS taggers using Hidden Markov Model (HMM) and Support Vector Machine (SVM). The lexicon contains around 128 thousand entries and a manual check yields the accuracy of 79.6%. Initially, the POS taggers have been developed for Bengali and shown the accuracies of 85.56%, and 91.23% for HMM, and SVM, respectively. Based on the Bengali news corpus, we identify various word-level orthographic features to use in the POS taggers. The lexicon and a Named Entity Recognition (NER) system, developed using this corpus, are also used in POS tagging. The POS taggers are then evaluated with Hindi and Telugu data. Evaluation results demonstrates the fact that SVM performs better than HMM for all the three Indian languages.

Voted NER System using Appropriate Unlabeled Data

This paper reports a voted Named Entity Recognition (NER) system with the use of appropriate unlabeled data. The proposed method is based on the classifiers such as Maximum Entropy (ME), Conditional Random Field (CRF) and Support Vector Machine (SVM) and has been tested for Bengali. The system makes use of the language independent features in the form of different contextual and orthographic word level features along with the language dependent features extracted from the Part of Speech (POS) tagger and gazetteers. Context patterns generated from the unlabeled data using an active learning method have been used as the features in each of the classifiers. A semi-supervised method has been used to describe the measures to automatically select effective documents and sentences from unlabeled data. Finally, the models have been combined together into a final system by weighted voting technique. Experimental results show the effectiveness of the proposed approach with the overall Recall, Precision, and F-Score values of 93.81%, 92.18% and 92.98%, respectively. We have shown how the language dependent features can improve the system performance.

A hidden Markov model based named entity recognition system: Bengali and Hindi as case studies

Named Entity Recognition (NER) has an important role in almost all Natural Language Processing (NLP) application areas including information retrieval, machine translation, question-answering system, automatic summarization etc. This paper reports about the development of a statistical Hidden Markov Model (HMM) based NER system. The system is initially developed for Bengali using a tagged Bengali news corpus, developed from the archive of a leading Bengali newspaper available in the web. The system is trained with a training corpus of 150,000 wordforms, initially tagged with a HMM based part of speech (POS) tagger. Evaluation results of the 10-fold cross validation test yield an average Recall, Precision and F-Score values of 90.2%, 79.48% and 84.5%, respectively. This HMM based NER system is then trained and tested on the Hindi data to show its effectiveness towards the language independent abilities. Experimental results of the 10-fold cross validation test has demonstrated the average Recall, Precision and F-Score values of 82.5%, 74.6% and 78.35%, respectively with 27,151 Hindi wordforms.

Gene expression data clustering using a multiobjective symmetry based clustering technique

The invention of microarrays has rapidly changed the state of biological and biomedical research. Clustering algorithms play an important role in clustering microarray data sets where identifying groups of co-expressed genes are a very difficult task. Here we have posed the problem of clustering the microarray data as a multiobjective clustering problem. A new symmetry based fuzzy clustering technique is developed to solve this problem. The effectiveness of the proposed technique is demonstrated on five publicly available benchmark data sets. Results are compared with some widely used microarray clustering techniques. Statistical and biological significance tests have also been carried out.

Temporal analysis of sentiment events: a visual realization and tracking

In recent years, extraction of temporal relations for events that express sentiments has drawn great attention of the Natural Language Processing (NLP) research communities. In this work, we propose a method that involves the association and contribution of sentiments in determining the event-event relations from texts. Firstly, we employ a machine learning approach based on Conditional Random Field (CRF) for solving the problem of Task C (identification of event-event relations) of TempEval-2007 within TimeML framework by considering sentiment as a feature of an event. Incorporating sentiment property, our system achieves the performance that is better than all the participated state-of-the-art systems of TempEval 2007. Evaluation results on the Task C test set yield the F-score values of 57.2% under the strict evaluation scheme and 58.6% under the relaxed evaluation scheme. The positive or negative coarse grained sentiments as well as the Ekmans six basic universal emotions (or, fine grained sentiments) are assigned to the events. Thereafter, we analyze the temporal relations between events in order to track the sentiment events. Representation of the temporal relations in a graph format shows the shallow visual realization path for tracking the sentiments over events. Manual evaluation of temporal relations of sentiment events identified in 20 documents sounds satisfactory from the purview of event-sentiment tracking.