T. V. Geetha

Concept based Focused Crawling using Ontology

The constraint of a web crawler that downloads only relevant pages is still a major challenge in the field of information retrieval systems. Rather than visiting all the web pages, a focused crawler visits only the section of the web that contains relevant pages, and at the same time, tries to skip irrelevant sections. Existing ontology based web crawlers estimate the semantic content of the URL based on a domain dependent ontology, which in turn supports the methods used for prioritizing the URL queue. The crawler maintains a queue of URLs it has seen during the crawl at each level, and then selects from this queue, the next URL to visit based on the conceptual rank of the page at that level obtained from domain ontology. However in this work we represent the topic as an overall conceptual vector, obtained by combining concept vectors of individual pages associated with seed URLs. The conceptual rank is based on comparison between conceptual vectors at each depth, across depths and between the overall topics indicating seed concept vector. General Terms Data and Web Mining.

Collaborative search using an implicitly formed academic network

Purpose – The purpose of this paper is to propose the Collaborative Search System that attempts to achieve collaboration by implicitly identifying and reflecting search behaviour of collaborators in an academic network that is automatically and dynamically formed. By using the constructed Collaborative Hit Matrix (CHM), results are obtained that are based on the search behaviour and earned preferences of specialist communities of researchers, which are relevant to the users need and reduce the time spent on bad links. Design/methodology/approach – By using the Digital Bibliography Library Project (DBLP), the research communities are formed implicitly and dynamically based on the users’ research presence in the search environment and in the publication scenario, which is also used to assign users’ roles and establish links between the users. The CHM, to store the hit count and hit list of page results for queries, is also constructed and updated after every search session to enhance the collaborative sear...

Enhanced ontology-based indexing and searching

Purpose – The purpose of this paper is to improve the conceptual-based search by incorporating structural ontological information such as concepts and relations. Generally, Semantic-based information retrieval aims to identify relevant information based on the meanings of the query terms or on the context of the terms and the performance of semantic information retrieval is carried out through standard measures-precision and recall. Higher precision leads to the (meaningful) relevant documents obtained and lower recall leads to the less coverage of the concepts. Design/methodology/approach – In this paper, the authors enhance the existing ontology-based indexing proposed by Kohler et al., by incorporating sibling information to the index. The index designed by Kohler et al., contains only super and sub-concepts from the ontology. In addition, in our approach, we focus on two tasks; query expansion and ranking of the expanded queries, to improve the efficiency of the ontology-based search. The aforemention...

Automatic Generation of Templates using Ontology

Generally, Information Extraction (IE) methods use predefined templates to determine the slot fillers to obtain relevant information. The slots define the important information necessary for the particular template. However, effective of the information is decided by the predefined template. In this paper, we mine the templates from a domain corpus to act as the predefined templates for IE. We automatically generate the templates using domain ontology for identifying the slots of the template. The performance of the proposed work is compared with the existing automatic template generation system and evaluated based on the precision metric. The automatic generation of templates using ontology produces the precision of 0.82.

Pattern Based Bootstrapping Technique for Tamil POS Tagging

Part of speech (POS) tagging is one of the basic preprocessing techniques for any text processing NLP application. It is a difficult task for morphologically rich and partially free word order languages. This paper describes a Part of Speech (POS) tagger of one such morphologically rich language, Tamil. The main issue of POS tagging is the ambiguity that arises because different POS tags can have the same inflections, and have to be disambiguated using the context. This paper presents a pattern based bootstrapping approach using only a small set of POS labeled suffix context patterns. The pattern consists of a stem and a sequence of suffixes, obtained by segmentation using a suffix list. This bootstrapping technique generates new patterns by iteratively masking suffixes with low probability of occurrences in the suffix context, and replacing them with other co-occurring suffixes. We have tested our system with a corpus containing 20,000 Tamil documents having 2,71,933 unique words. Our system achieves a precision of 87.74%.

Ontology based semantic similarly ranking of documents

Semantic relatedness measures the degree to which some words or concepts are related, considering possible semantic relationships among them. Semantic relatedness is of great interest in different areas, such as Natural Language Processing, Information Retrieval, or the Semantic Web. This paper proposes policy based ranking of documents, where policy directs which path of the ontology tree is to be considered for semantic computation. An algorithm for ranking documents according to their relevance to a query is used. The given query is expanded based on the ontology concepts and the given polices. As the policy changes, the concept set used for the ranking algorithm varies. Depending on the requirements, using the same ontology, documents can be ranked differently based on the policies which in turn depend on the structure of the ontology. A ranking algorithm calculates the similarity degree of each document with respect to the user query.