A Study of Lexical Ambiguity in Large Forum Discussions for Multidisciplinary Knowledge Engineering

Lexical ambiguity is a challenging issue in multidisciplinary knowledge engineering due to the tendency that lexical terms can be used among different domains with different specific meanings. Particularly in large forum discussions, such ambiguous cross-disciplinary terms hard to be identified and detected by the discussion participants because domain expertise from several relevant fields is required to detect those terms and discover the actual divergence of interpretation. Having many ambiguous terms in the discussion context will result in gradual misunderstanding and delayed knowledge construction. We studied the effects of data sizes and morphological analysis in discovering ambiguous cross-disciplinary terms in large forum discussions. Our findings are twofold. First, it is more likely to discover cross-disciplinary terms as forum discussions deepen. This correlates with domain experts’ tendency to use general terms in metaphorically describing domain-specific concepts, therefore causing lexical ambiguity. Second, we found that lemmatization outperforms stemming in forming more understandable key terms. This is because lemmatization eliminates only inflectional affixes and keeps derivational affixes. On the other hand, stemming eliminates both types of affixes, causing semantic bleaching.