Abstract
This paper introduces Akalex, a new online tool created to help vocabulary research in Czech academic texts. The Akalex database includes almost 60 000 n-grams - candidates for typical academic words or multiword units - and can be easily searched and filtered according to several criteria.
These n-grams were extracted from the SYN2015 corpus of written contemporary Czech based on their prominent frequency in academic texts and shared occurrence in many different academic disciplines, which distinguish them from general vocabulary on one hand and specialized terminology on the other. Each n-gram in the database is also provided with additional information, such as part-of-speech, distribution in disciplines, frequency etc., making it possible to search for e.g. specific collocations with a specific lexeme (such as adjectives combined with the word výzkum 'research' or verbs with a certain preposition).
The features of Akalex were put to test in our case study covering 2-grams to 6-grams used in all 24 academic disciplines included in the SYN2015 corpus. Out of almost 900 candidates, 236 were manually chosen by two annotators as typical for academic texts.
These were then further analysed and split into groups based on their semantic, functional and formal features. Among the most frequent were lexical bundles, collocations with content words and combinations of two verbs pointing to a frequent use of passives in academic texts etc.