Building Tempo-HindiWordNet: A resource for effective temporal information access in Hindi

Building Tempo-HindiWordNet : A Resource for Effective Temporal Information Access in Hindi Dipawesh Pawar1, Mohammed Hasanuzzaman2, Asif Ekbal1 1Department of Computer Science and Engineering Indian Institute of Technology, Patna {dipawesh.mtcs14,asif}@iitp.ac.in 2Normandie University, GREYC UMR 6072, Caen, France Mohammed.Hasanuzzaman@unicaen.fr Abstract In this paper, we put forward a strategy that supplements Hindi WordNet entries with information on the temporality of its word senses. Each synset of Hindi WordNet is automatically annotated to one of the five dimensions: past, present, future, neutral and atemporal. We use semi-supervised learning strategy to build temporal classifiers over the glosses of manually selected initial seed synsets. The classification process is iterated based on the repetitive confidence based expansion strategy of the initial seed list until cross-validation accuracy drops. The resource is unique in its nature as, to the best of our knowledge, still no such resource is available for Hindi. Keywords: Time sense annotation, Tempo-HindiWordNet, Semi-supervised learning 1. Introduction There is considerable academic and commercial inter- est in processing time information in text, where the information is expressed either explicitly, or implicitly, or connotative. Recognizing such information and ex- ploiting it for Information Retrieval (IR) and Natural Language Processing (NLP) tasks are important fea- tures that can significantly improve the functionality of these applications. Whereas most of the prior computational linguistics and text mining temporal studies have focused on temporal expressions and events, there has been a lack of work looking at the temporal orientation of word senses. Towards this direction, (Dias et al., 2014) developed TempoWordNet (TWn), an extension of WordNet (Miller, 1995), where each synset is aug- mented with its temporal connotation (past, present, future or atemporal). It mainly relies on the quantita- tive analysis of the glosses associated to synsets, and on the use of the resulting vectorial term representations for semi-supervised synset classification. The authors show that improvements can be reached for tempo- ral sentence classification and temporal query intent classification when unigrams are temporally expanded using the constructed TWn. Being motivated from the idea of Dias et al. (Dias et al., 2014), we propose to build a temporal lexical knowledge-base (Tempo-HindiWordNet) which may contribute to the information access research in Hindi, a resource-scared language. Hindi is the national lan- guage of India and 4.46 % of world population is the native speaker of it. Semi-supervised learning strat- egy is used to automatically annotate each synset of Hindi WordNet (Bhattacharyya et al., 2010) to one of the five time-sensed classes: past, present, future, neu- tral and atemporal. One might suggests some synset do not clearly fall into one of the four classes (past, present, future, atemporal) proposed in (Dias et al., 2014). For example, does synset sunday.n.01—first day of the week have a predominant temporal class? Therefore, in addition to the four classes proposed by (Dias et al., 2014), we introduce another class as ‘neu- tral’. Such synset would be marked as ‘neutral’ in our annotation scheme. Thus, one might consider our ‘neu- tral’ class to encompass past, present and future class. Classes, …

Content truncated to 500 words