Formal Methods in Czech Theory of Verse

The paper deals with possibilities of using exact methods and computer technology in verse theory. First we present the system for encoding syllable properties proposed by Květa Sgallová (1960s). Then we describe an algorithm for metre recognition of Czech accentual-syllabic verse, which we have been developing from 2010 to 2013 at the Institute of Czech Literature, Czech Academy of Sciences. Compared to model of Sgallová, which was a combination of manual and computational processing, the new model relies entirely on automatic processing done by the computer program. It consists of several components: phonetic, morphological, metrical. After the phonetic transcription, POS tagging, and lemmatization are done, particular lines are segmented into syllables and these are further annotated according to (1) prominence (stressed/unstressed), (2) position within a word (initial/ other), (3) nucleus length (long/short), (4) whether it is a monosyllabic preposition proper, (5) whether it is preceded by a syntactic break. Next all the metrical patterns consisting of S-positions (strong) and W-positions (weak) are generated according to following conditions: (1) each S/W symbol corresponds to one syllable of a line, (2) no S is followed by another S-position, (3) W at the beginning of a pattern is not followed by another W-position, (3) no more than two W-positions occur in a row. Meter recognition is based on assigning each position of each of the generated patterns a numerical weights corresponding to the degree of metricality of its realization by a syllable with given features. The degree of metricality of particular metres is then calculated basing on these values.