Chapter 13

Knowledge patterns in corpora

Knowledge patterns (KPs) have evolved as tools for detecting, extracting, and classifying occurrences of conceptual and semantic relations expressed in texts. KPs offer many strengths, including clear identification of relations present. The extraction of relation-containing contexts, often referred to as knowledge-rich, allows users to observe the items participating in the relation and nuances conveyed by other textual elements. Nevertheless, designing and implementing pattern-based approaches is challenging, in part because of patterns’ linguistic (i.e., lexical, grammatical) or paralinguistic nature. KPs must be selected and represented with care for a given use, and their performance may vary. In this chapter, we will explore some benefits and challenges of pattern-based approaches to corpus analysis, and some prospects for future applications.