Synset Ranking of Hindi WordNet

Synset Ranking of Hindi WordNet Sudha Bhingardive, Rajita Shukla, Jaya Saraswati, Laxmi Kashyap, Dhirendra Singh and Pushpak Bhattacharyya Department of Computer Science and Engineering, Indian Institute of Technology Bombay, India. {sudha,rajita,jayas,yupu,dhirendra,pb}@cse.iitb.ac.in Abstract Word Sense Disambiguation (WSD) is one of the open problems in the area of natural language processing. Various supervised, unsupervised and knowledge based approaches have been proposed for automatically determining the sense of a word in a particular context. It has been observed that such approaches often find it difficult to beat the WordNet First Sense (WFS) baseline which assigns the sense irrespective of context. In this paper, we present our work on creating the WFS baseline for Hindi language by manually ranking the synsets of Hindi WordNet. A ranking tool is developed where human experts can see the frequency of the word senses in the sense-tagged corpora and have been asked to rank the senses of a word by using this information and also his/her intuition. The accuracy of WFS baseline is tested on several standard datasets. F-score is found to be 60%, 65% and 55% on Health, Tourism and News datasets respectively. The created rank- ings can also be used in other NLP applications viz., Machine Translation, Information Retrieval, Text Summarization, etc. Keywords: Synset Ranking, WordNet, Hindi WordNet, Word Sense Disambiguation, WordNet First Sense, NLP 1. Introduction Word Sense Disambiguation (WSD) is the ability to identify the meaning of words in context in a computa- tional manner (Navigli, 2009). It is one of the toughest areas in natural language processing (NLP). Recently, a lot of research has been done for making powerful WSD systems with supervised, semi-supervised and unsupervised techniques. In WSD, the heuristics of choosing the most frequent sense is often found to be very hard for any WSD system. The WordNet First Sense (WFS) baseline is the most powerful baseline in WSD, even though it does not consider the context while assigning the senses. This baseline can be cre- ated by considering the sense-annotated statistics. For English, WFS baseline is created by using the frequen- cies of word senses from the sense-annotated SemCor corpus. Senses that have not occurred in SemCor are ordered arbitrarily. This WFS baseline is a very stong baseline in English WSD. Considering both precision and recall, only 5 of 26 systems in the Senseval-3 En- glish all-words task were able to beat this baseline. Our goal is to create a WFS baseline for Indian lan- guage WordNets. We focus on Hindi language as the synsets of Hindi WordNet are not ranked according to the actual usage. This is because Hindi Wordnet was built using a dictionary where words were picked up according to the alphabetical order. The rest of the paper is organized as follows. Section 2 gives a detailed description of Hindi WordNet. Hindi WordNet synset ranking methodology is explained in section 3. Section 4 gives the statistics of the ranked synsets. Section 5 highlights the performance of WFS baseline on various domains. Discussion is given in section 6, …

Content truncated to 500 words