Автоматично извличане на информация от текстове на близкородствени езици (върху славянски езиков материал)
Automatic extraction of information from texts in closely related languages (on Slavic texts)

Summary/Abstract: The paper reviews computer processing of texts in the closely related Slavic languages. The major focus falls on language resources and basic computer extraction methods for processing linguistic information in Bulgarian-Russian and Bulgarian-Macedonian contrastive research. Part of the discussion is devoted to problems of language resources, mostly the availability and nature of processing of aligned parallel corpora. The linguistic results from the application of the specific information extraction methods are discussed, namely linguistic knowledge about the two pairs of languages, the automatic extraction of translation equivalents and phrases by applying an improved Lowenstein algorithm, as well as the results of subsequent evaluation and further manual and automatic processing. Special attention is devoted to the extraction of false friends from Russian-Bulgarian parallel texts. A particular linguistic adjustment of the applied statistical methods is offered, more specifically of the scale of weighing coefficients of letter substitutions. The adjustments are based on phonological, grammatical and mean 53 ing criteria. The language resources, both textual and grammatical, have been developed at the Department for Linguistic Modeling at the Institute for Information and Communication Technologies at the Bulgarian Academy of Sciences.

• Details
• Contents