9 - Lexical anisomorphism in machine-readable dictionaries

Summary

Natural-language-processing knowledge bases are substantially different from textual dictionaries in one important regard. Owing to the lack of text–human communication, all those strategies that rely on human reasoning are impossible in NLP knowledge bases. Thus, for example, explanation, specification, exemplification, and illustration cannot be used.

To see how the treatment of CLA in this context differs from that of textual dictionaries, I now examine a coding schema intended for use in NLP knowledge bases: LMF (Lexical Markup Framework: see Francopoulo et al., 2007; Francopoulo, 2013).

In a way, the LMF schema came as a response to the inadequacies of the TEI schema (Text Encoding Initiative scheme, version 5: see Burnard and Bauman, 2013), which satisfies the needs of digitizing printed dictionaries but falls short of providing an adequate platform for NLP databases. The use of the TEI schema in digitizing does not substantially differ from the rendering in printed dictionaries, apart from the fact that the text is much better segmented and its parts identifiable.

To illustrate this, let us take a look at the TEI-5 coding of the previously presented examples of zero, multiple, and partial equivalence. In addition to the aforementioned Burnard and Bauman (2013), more information about the application of the TEI coding schema in lexicography is provided by Budin, Majewski, and Mörth (2012). The schema is based on Extensible Markup Language (XML: see www.w3.org/XML for more information) and it uses sets of matching opening tags (the word within <>) and closing tags (the same word withn </>). For example, the beginning of a dictionary entry will be marked as <entry> and the end of the same entry by </entry>.