Disunity in Cohesion: How Purpose Affects Methods and Results When AnalyzingLexical Cohesion

PACLIC 28 !513 Disunity in Cohesion: How Purpose Affects Methods and Results When Analyzing Lexical Cohesion Stuart G. Towns Richard Watson Todd Department of Applied Linguistics King Mongkut’s University of Technology Thonburi Bangkok, Thailand sgtowns@gmail.com irictodd@kmutt.ac.th Abstract Lexical Cohesion is a commonly studied lin- guistic feature as it is easily identified from the surface of a text. However, the purposes for studying lexical cohesion are varied, and each purpose requires different methods. This study analyzes two short movie review texts for four different research purposes using lex- ical cohesion: text evaluation, text segmenta- tion, text summarization, and text criticism. The analysis shows that these four different purposes produce very different results con- cerning the lexical cohesion of the two texts, suggesting that the apparently straightforward construct of lexical cohesion is actually com- plex. 1 Introduction The purposes of text analysis research can be divided into two main categories: applications and descriptions. The difference between these two areas is that applications produce results that are useful to end users who are outside of the field of linguistics, while descriptions of language are used internally by the linguistic community (Sinclair, 2004a). Many text analysis applications created for those outside of linguistics use automated tools, and therefore they focus on features that can be identified and analyzed with computers. One lin- guistic feature that can be analyzed to varying de- grees of success using computers is lexical cohesion, since lexical cohesion can be found in the surface features of text. The analysis of lexical cohesion has been used in many text analysis ap- plications, such as discourse analysis (Morris & Hirst, 1991), automatic text summarization (Barzi- lay & Elhadad, 1999), text segmentation (Stokes, Carthy, & Smeaton, 2004), word sense disambigu- ation (Okumura & Honda, 1994), and evaluation of machine translations (Wong & Kit, 2012). Lexical cohesion was defined by Halliday and Hasan (1976, p. 274) as “the cohesive effect achieved by the selection of vocabulary” and is one of five types of cohesion (the other four being ref- erence, substitution, ellipsis, and conjunction). A cohesive text is held together by explicit relation- ships found in the lexis and grammar of the text. These lexico-grammatical relationships are called cohesive ties as they connect one sentence to an- other. Multiple ties can, in turn, be combined into longer lexical chains which can span large portions of the text. Current technology can identify lexical cohe- sion ties and lexical chains of ties with varying degrees of accuracy. Some cohesive ties are very easy to identify, such as the exact repetition of a lexical unit in an adjacent sentence, while others can be more difficult to correctly identify, such as the relationship of a pronoun to a noun in a previ- ous sentence. Hoey (1991) outlined six types of lexical cohesion which are ordered by ease of iden- tification from easiest to most difficult, along with some examples in Table 1. As stated earlier, lexical cohesion has been used in text analysis research for many different Copyright …

Content truncated to 500 words