Corpus Linguistics, Frequency and Explanatory Dictionaries: Interaction Vectors

Abstract

With the overall growth in the number of corpora, their volumes, and their diversity, corpora are becoming increasingly specialized depending on the composition of their targeted content. Electronic corpora of the first generation (with a volume of approximately 100 million word tokens), referred to or perceived as “national” or “state” corpora, maintain a relatively balanced structure of subcorpora and address a broad audience in the social sciences and humanities. As the size of later corpora increases, their specialization develops along two vectors: 1) content-oriented monitor (continuously updated) megacorpora of newspaper and magazine texts; the target user groups for this class of corpus content include sociologists and political scientists, economists, demographers, journalists, and others; 2) thematically unlimited (non-selective) corpora that accumulate digitized texts (printed and electronic), used in computer science as raw material for natural language processing—machine pre-training of neural networks and the creation of statistical algorithms that link words into coherent textual responses of artificial intelligence.

Two of the most significant innovative developments in corpus lexicography are identified: 1) the synthesis of explanatory and frequency dictionaries in the Macmillan dictionaries (2007), later adopted by Collins and Longman; 2) componential semantic analysis of a 100,000-word lexicon using the 2,500 most frequent lexemes in Macmillan (2007) as semantic components. The capabilities of corpora will soon lead to major advancements in diachronic linguistics.