JU_NLP at SemEval-2016 Task 11: Identifying Complex Words in a Sentence

Proceedings of SemEval-2016, pages 986–990, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics JU NLP at SemEval-2016 Task 11: Identifying Complex Words in a Sentence Niloy Mukherjee, Braja Gopal Patra, Dipankar Das, and Sivaji Bandyopadhyay Department of Computer Science and Engineering, Jadavpur University Kolkata, India {niloymukherjee93,brajagopal.cse,dipankar.dipnil2005}@gmail.com sivaji cse ju@yahoo.com Abstract The complex word identification task refers to the process of identifying difficult words in a sentence from the perspective of readers be- longing to a specific target audience. This task has immense importance in the field of lexi- cal simplification. Lexical simplification helps in improving the readability of texts consist- ing of challenging words. As a participant of the SemEval-2016: Task 11 shared task, we developed two systems using various lexi- cal and semantic features to identify complex words, one using Na¨ıve Bayes and another based on Random Forest Classifiers. The Na¨ıve Bayes classifier based system achieves the maximum G-score of 76.7% after incorpo- rating rule based post-processing techniques. 1 Introduction Extensive research has been performed in the field of lexical simplification (Specia et al., 2012; Rello et al., 2013; Paetzold, 2015). Lexical simplification refers to identifying complex words and replacing them with lexically simple substitutes (Specia et al., 2012). The English lexical simplification task1 was organized in the year 2012, in which the complex words were provided by the organizers. The complex word (CW) identification is the first step towards the lexical simplification task. Under- standing words which are not frequently used in any language is very difficult for non-native speakers. It may be challenging for a reader to interpret a par- ticular word because it might be absent in his vo- cabulary. Also it may so happen that he knows the 1https://www.cs.york.ac.uk/semeval-2012/task1.html word but cannot comprehend it as he fails to capture the context it is used in. Generally, it is observed that the frequent use of complex words decreases the readability of the document. Thus, the complex word identification (CWI) task aims to classify those challenging words in a sentence with respect to a particular target audience. For example, in the following sentence, the words in italics are complex words. These words are re- lated to biology and are rarely used in our daily life. e.g. “The first amniotes, such as Casineria, resem- bled small lizards and evolved from amphibian rep- tiliomorphs about 340 million years ago.” Some research has been performed in CWI task in comparison to the lexical simplification (Shard- low, 2013; Paetzold, 2015). The important fea- tures which have been used previously in the CWI task are frequency thresholding and lexical match- ing etc. (Shardlow, 2013). Some motivations of the CWI task are to under- stand the defining characteristics of the words which are challenging for non-native speakers to interpret. Another is assessing an individual’s vocabulary lim- itations from the group he is a part of. We have participated in the SemEval 2016-Task 11: Complex Word Identification2 (Paetzold and Specia, 2016). The main goal of this task is to iden- tify …

Content truncated to 500 words