A new computing method for extracting contiguous phraseological sequences from academic text corpora

This study aims to develop a new computing method for extracting contiguous phraseological sequences (PSs) of various lengths from academic text corpora by measuring internal associations of n-grams. We construct a new normalizing algorithm of probability-weighted average for refining the MI measure and enhancing precision in extracting PSs from corpora. This computing method is applied to the data in a medium-sized text corpus of academic English. Results indicate that the resultant new MI measure can provide statistics which better reveal internal associations within an n-gram, regardless of size. Lexico-grammatical sequences extracted with this method are more complete and less arbitrary in terms of grammar and semantics. The method can be applied to treating a variety of linguistic phenomenon, ranging from well-established phrases to likely phrasal entities, thus having potentially practical applications in corpus-based studies of phraseology and natural language processing.