Exploring Verb Frames for Sentence Simplification in Hindi

International Joint Conference on Natural Language Processing, pages 1082–1086, Nagoya, Japan, 14-18 October 2013. Exploring Verb Frames for Sentence Simplification in Hindi Ankush Soni Sambhav Jain Dipti Misra Sharma Language Technologies Research Centre IIIT Hyderabad {ankush.soni, sambhav.jain}@research.iiit.ac.in, dipti@iiit.ac.in Abstract Systems processing on natural language text encounters fatal problems due to long and complex sentences. Their perfor- mance degrades as the complexity of the sentence increases. This paper addresses the task of simplifying complex sentences in Hindi into multiple simple sentences, using a rule based approach. Our approach utilizes two linguistic resources viz. verb demand frames and conjuncts’ list. We performed automatic as well as human evaluation of our system. 1 Introduction Cognitive and psychological studies, performed on ‘human reading’ states that the effort in reading and understanding a text increases with the sen- tence complexity (Klein and Kurkowski, 1974) . The modern natural language processing applica- tions are not much different, in this respect, from humans. Processing complex sentences with high accuracy has always been a challenge in computa- tional linguistics. This calls for techniques aiming at automatic simplification of sentences (Chan- drasekar et al., 1996). The sentence complexity can be mainly classi- fied into ‘lexical complexity’ and ‘syntactic com- plexity’. In context of natural language applica- tions, lexical complexity can be handled signifi- cantly by utilizing various resources like lexicons, dictionary, thesaurus etc. and substitute infrequent words with their frequent counterparts (De Belder et al., 2010). To address syntactic complexity, one can analyse the structure of the sentence and then apply proper operations to simplify the structure. There are many applications of sentence sim- plification in NLP applications. Machine Trans- lation systems when dealing with highly diverge language pairs face difficulty in translating long and complex sentences. For Parsing, it has been shown by McDonald and Nivre(2007) that syntactic parsing of long sen- tence and Identifying long distance dependencies is still a challenging task for modern day parsers. So, it looks intuitive to break down the sentence into smaller parts and use the simplified sentences for the task of parsing and Machine translation. In case of Automatic summarization, after sim- plifying sentences, it is likely that the accuracy of sentence extraction based summarization systems improves as smaller units of information are being extracted. We present a rule based approach for sentence simplification in Hindi. Our proposed system takes a sentence and returns a set of simple sen- tences, smaller in length. We have taken care to produce sentences which keep the meaning close to the original sentence. This paper is structured as follows: In Section 2, we discuss the related works for sentence sim- plification. Section 3 talks about complex sen- tence. Section 4 describes the linguistic resources we used. In section 5, we discuss our algorithm. Section 6 outlines evaluation of the system. In sec- tion 7, results are being talked about. Section 8 gives the error analysis and in Section 9, we con- clude and talk about future works in this area. 2 Related Work Chandrasekar et …

Content truncated to 500 words