Lexical typology studies organization of semantic fields, such as verbs of motion, verbs of location, or physical qualities. Languages tend to lexicalize these semantic zones in different ways. Lexical typologists want to know how many lexical items can be in a language to cover all the meanings of a given field, what meanings are frequently opposed (that is, covered by different lexemes) in languages, and what meanings, on the contrary, are usually expressed in one word.
Our method strongly relies on lexical distribution. In this we follow the rich lexicographic tradition of the Moscow Semantic School (cf. in particular Apresjan 1974, 2000), which has perfected the art of describing synonyms. Partial equivalents that form the material of lexical typology are, at some level, cross-linguistic near synonyms and can be compared and contrasted in the same ways as a group of synonyms within one language. Subtle differences and nuances of meaning are revealed by finding contexts in which one term cannot be replaced by another (as in the well-known example by A. Wierzbicka (2006) for wide VS. broad: wide/*broad board, but broad/*wide back). We assume that the cognitive reality behind these diagnostic contexts are “frames”, or prototypical situations that constitute a semantic domain. We find this a useful concept for lexical typology: word meanings from all languages can be compared through a common inventory of frames in the domain, to maintain compatibility of definitions and highlight the various ways in which individual languages represent reality.
Another novel aspect of our work lies in visualization of results. Firstly, we have adapted for our needs the traditional, manually constructed semantic maps in the shape in which they are used in grammatical typology (see van der Auwera, Plungian 1998, Haspelmath 1997, 2003). They allow us to illustrate the scope of lexical meanings in the domain and to compare lexical systems graphically.
Secondly, we make use of the recent advances in computational linguistics. Indeed, one of the tasks that lexical typology is facing while computational linguistics and corpora continue to grow is finding ways to automate the collection and preliminary analysis of data (on parallel corpora in cross-linguistic investigations cf. Cysouw, Wälchli (eds.) 2007). Increased use of statistics, in particular, is important, since it gives more weight to typological generalizations. In a separate section of the proposed volume we discuss distributional semantic models (cf. Baroni et al. 2013) and the ways of applying them to lexical combinations in order to verify and support the choice of diagnostic contexts.
If you would like to learn more about our approach please read the article ‘Doing lexical typology with frames and semantic maps’written by Ekaterina Rakhilina and Tatiana Reznikova.