MacSaar at SemEval-2016 Task 11: Zipfian and Character Features for ComplexWord Identification

Proceedings of SemEval-2016, pages 1001–1005, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics MacSaar at SemEval-2016 Task 11: Zipfian and Character Features for Complex Word Identification Marcos Zampieri1,2, Liling Tan1, Josef van Genabith1,2 1Saarland University, Germany 2German Research Center for Artificial Intelligence, Germany {first.last}@uni-saarland.de Abstract This paper presents the MacSaar system de- veloped to identify complex words in English texts. MacSaar participated in the SemEval 2016 task 11: Complex Word Identification submitting two runs. The system is based on the assumption that complex words are likely to be less frequent and on average longer than words considered to be simple. We report re- sults of 82.5% accuracy and 27% F-Score us- ing a Random Forest Classifier. The best Mac- Saar submission was ranked 8th in terms of F- Measure among 45 entries. 1 Introduction Complex Word Identification (CWI) is the task of automatically identifying complex words in texts. It is considered a sub-task carried out in most lexical simplification pipelines (Paetzold and Specia, 2015). In this step, complex words, which are likely to be difficult words for readers and language learners, are identified so they can be substituted for simpler ones (Specia et al., 2012; Shardlow, 2013). Lexical sim- plification methods are usually integrated into text simplification systems developed for a particular tar- get population (e.g. people with reading impairment or dyslexia, language learners, etc.) (Siddharthan, 2014). Given a sentence, a CWI system is trained to iden- tify words which are considered by readers to be complex. To give an example, let us consider the fol- lowing sentence extracted from the SemEval CWI task training set: (1) Leo took an oath of purgation concerning the charges brought against him , and his oppo- nents were exiled. Taking Example 1 into account, the task of the CWI system is to assign as complex the four under- lined words, namely: oath, purgation, charges, and exiled. But what makes these words complex and not, for example, opponents or took? In the lexical simplification literature, the term complex is a synonym for difficult or complicated. For practical purposes, we consider as complex words those that were assigned by a pool of human annotators, provided by the organizers of the CWI task, as difficult to be understood due to several fac- tors that we will discuss in this paper. This is a read- ability notion that is not necessarily related to intrin- sic linguistic phenomena (e.g. word formation).1 1.1 Motivation MacSaar participated in the CWI SemEval task in- terested in two aspects of complex words. The first one is related to communication principles, or in other words, what makes words complex or simple to readers. One of our assumptions is that complex words tend to be less frequent in general language corpora than simple words. The second aspect is language learning. Lexical and text simplification methods are very important to produce simpler texts 1It is important to note that the definition of complex used here is different from that used in Morphology, …

Content truncated to 500 words