Qualification: 
MCA, M.Tech
veenag@am.amrita.edu

Veena G. currently serves as the Assistant Professor at the Department of Computer Science Applications at Amrita School of Engineering, Amritapuri. She has completed her M. Tech. in Computer Science.

Publications

Publication Type: Journal Article

Year of Publication Publication Type Title

2018

Journal Article

G. Veena, Pillai, L. R., and Dr. Deepa Gupta, “An Extended Model for Semantic Role Labeling Using Word Sense Disambiguation and Dependency Parsing”, Journal of Engineering and Applied Sciences, vol. 12, 2018.[Abstract]


In NLP, the main two fundamental tasks are Word Sense Disambiguation (WSD) and Semantic Role Labeling (SRL). Semantic role labeling is used to label the arguments in the sentence with the help of predicates. Word sense disambiguation is the procedure of predicting the correct definition of a word in a given context. In the research, we improved SRL using WSD and dependency parsing. The dependency parser helps to improve the semantic relationship between the predicates and its arguments. A modified Conditional Random Field (CRF) is used to bind dependency parser with SRL. We have used SVM classifier for WSD and PractNLP tool is used for dependency parser. The model is evaluated and compared with an online WSD with SRL tool. From the results obtained with the aid of our proposals, the labeling performs much better than a tool.

More »»

2018

Journal Article

G. Veena, Dr. Deepa Gupta, Lakshmi S, and Jacob, J. T., “Named entity recognition in text documents using a modified conditional random field”, Advances in Intelligent Systems and Computing, vol. 709, pp. 31-41, 2018.[Abstract]


The Named Entity Recognition in documents is an active and challenging research topic in text mining. The major objective of our work is to extract a phrase from the sentence and classify this phrase to one of the predefined named entities. The proposed system works in two layers, in the first phase each and every word in the phrase is tagged using word feature extraction approaches. In the second phase the model recognizes named entities in the phrase level using Modified Conditional Random Field. This work identifies four classes of entities such as Person, Organization, Location and Other. Our algorithm first parses the text document and identifies the sentence structure. From this sentence structure concepts are extracted. In this work the feature extraction module make use of the yahoo Geoplanet Web service for identifying the location. We have created person ontology of all available Indian names to check whether a word is name or not. Inorder to check whether the word is organization or not we have used a database with company name indicators. Finally, our MCRF assign a label to the tagged phrase. © Springer Nature Singapore Pte Ltd. 2018

More »»

Publication Type: Conference Paper

Year of Publication Publication Type Title

2018

Conference Paper

L. R. Pillai, Veena, G., and Gupta, D., “A Combined Approach Using Semantic Role Labelling and Word Sense Disambiguation for Question Generation and Answer Extraction”, in Second International Conference on Advances in Electronics, Computers and Communications (ICAECC), Bangalore, India, 2018.[Abstract]


Most question answering systems are used to predict an expected answer type given a question. In this work, we present a Question Answering System based on the combined approach of Word Sense Disambiguation (WSD) and Semantic Role Labeling (SRL). Our motivation is to generate reasonable questions and solve co-referencing problem extracted from the answer. The proposed model of work is factoid sense based question generation system. We have used Lesk algorithm for WSD and Senna tool for SRL. Based on the sense associated with the sentence, the system generates questions of semantically resolvable. Using deep syntax and semantics analysis, we have extracted an answer from the given question. Hobbs algorithm resolved co-reference problem generated in answer extraction. The experimental results show promising results for the proposed approach.

More »»

2017

Conference Paper

G. Veena, Dr. Deepa Gupta, Daniel, A. N., and Roshny, S., “A learning method for coreference resolution using semantic role labeling features”, in 2017 International Conference on Advances in Computing, Communications and Informatics (ICACCI), Udupi, India, 2017.[Abstract]


Coreference resolution plays a significant role in natural language processing systems. It is the method of figuring out all the noun phrases that refer back to the identical real world entity. Several researches have been done in noun phrase coreference resolution by using certain machine learning techniques. Our paper proposes a machine learning approach using support vector machines (SVM) towards coreference resolution at document level. In this research work, 17 well-defined syntactic and semantic features including the 13 baseline features with Semantic Role labeling (SRL) have been used. The use of SVM classifier leads to a better outcome when compared to other machine learning models. The system was evaluated using a machine learning technique and a non-machine learning technique. Experiments show that addition of SRL improves the performance of the system when compared to the Decision tree model and a non machine learning approach.

More »»

2017

Conference Paper

G. Veena, Athulya, S., Shaji, S., and Dr. Deepa Gupta, “A graph-based relation extraction method for question answering system”, in 2017 International Conference on Advances in Computing, Communications and Informatics (ICACCI), Udupi, India, 2017, vol. 2017-January, pp. 944-949.[Abstract]


Question Answering (QA) is the method of automatically answering a question asked by human in natural language using either a pre-structured database or a collection of documents. It is a rising new information service following the popularization of search engines. In this paper we introduce a graph-based QA system for reading comprehension tests that pick out the sentence in the passage that best answers a given question by extracting the relations. In order to improve the accuracy of the system, we do a gender analysis, morphological analysis and synonyms check along with coreference resolution. We tested our system with 60 comprehension passages each having five questions thereby a total of 300 questions and attained an accuracy of 79.67%. We achieved best results in terms of accuracy compared to the existing system having only 40%.

More »»

2016

Conference Paper

G. Veena and Krishnan, S., “A Concept Based Graph Model for Document Representation Using Coreference Resolution”, in Intelligent Systems Technologies and Applications, Cham, 2016.[Abstract]


Graph representation is an efficient way of representing text and it is used for document similarity analysis. A lot of research has been done in document similarity analysis but all of them are keyword based methods like Vector Space Model and Bag of Words. These methods do not preserve the semantics of the document. Our paper proposes a concept based graph model which follows a Triplet Representation with coreference resolution which extract the concepts in both sentence and document level. The extracted concepts are clustered using a modified DB Scan algorithm which then forms a belief network. In this paper we also propose a modified algorithm for Triplet Generation.

More »»

2016

Conference Paper

G. Veena, Peter, A. S., Rajkumari, K. A., and Ramanan, N., “A concept-based model for query management in service desks”, in International Conference on Innovations in Computer Science & Engineering (ICICSE-2016) Proceedings in Springer Advances in Intelligent Systems and Computing, Guru Nanak Institutions, Hyderabad, India, 2016, vol. 413, pp. 255-265.[Abstract]


Thousands of email queries are often received by help desks of large organizations nowadays. It is a cumbersome and time-consuming task to manage these emails manually. Also, the support staff who initially answers the query may not always be technically sound to do this themselves. In that case, they forward the queries to higher authorities, unnecessarily wasting their precious time. A large amount of time and human effort is being wasted for this manual classification and query management process. So, in this paper, we propose a new concept-based semantic classification technique to automatically classify the help desk queries into multiple categories. Our system also proposes an approach for retrieving powerful information related to the queries. In our work, the dataset is represented using a graph model and the concept of ontology is used for representing semantics of data.

More »»

2015

Conference Paper

G. Veena and B. U, U. Sree Veni, “Improving the Accuracy of Document Similarity Approach using Word Sense Disambiguation”, in WCI '15 Proceedings of the Third International Symposium on Women in Computing and Informatics, Kochi, India, 2015.[Abstract]


The aspects of Artificial Intelligence and statistics such as Text mining, Data Mining can provide solutions to the area of concept mining. It provides powerful insights into the meaning and documents similarity without exploiting the semantics of the terms or phrases in the document. Our work determines the similarity of documents using semantic processing namely Word Sense Disambiguation by annotating the senses of the words in the documents and then performs traditional PageRank algorithm over it. The Algorithm ranks the possible senses and finds the correct sense according to the context. Our paper proposes the method of disambiguating the ambiguous words in order to find the document similarity. Moreover it is compared with the cosine similarity approach, which is frequently used to determine similarity between two documents to prove the accuracy of our work.

More »»

2015

Conference Paper

G. Veena and K., L. N., “An Extended Chameleon Algorithm for Document Clustering”, in Advances in Intelligent Informatics, Cham, 2015.[Abstract]


A lot of research work has been done in the area of concept mining and document similarity in past few years. But all these works were based on the statistical analysis of keywords. The major challenge in this area involves the preservation of semantics of the terms or phrases. Our paper proposes a graph model to represent the concept in the sentence level. The concept follows a triplet representation. A modified DB scan algorithm is used to cluster the extracted concepts. This cluster forms a belief network or probabilistic network. We use this network for extracting the most probable concepts in the document. In this paper we also proposes a new algorithm for document similarity. For the belief network comparison an extended chameleon Algorithm is also proposed here.

More »»

2014

Conference Paper

G. Veena and Lekha, N. K., “A concept based clustering model for document similarity”, in 2014 International Conference on Data Science Engineering (ICDSE), 2014, pp. 118-123.[Abstract]


A lot of research work has been done in the area of concept mining and document similarity in past few years. But all these works were based on the statistical analysis of keywords. The major challenge in this area involves the preservation of semantics of the terms or phrases. Our paper proposes a graph model to represent the concept in the sentence level. The concept follows a triplet representation. A modified DB scan algorithm is used to cluster the extracted concepts. This cluster forms a belief network or probabilistic network. We use this network for extracting the most probable concepts in the document. In this paper we also proposes a new algorithm for document similarity.

More »»

Faculty Research Interest: