ÐÑи ÑеÑении Ð¼Ð½Ð¾Ð³Ð¸Ñ Ð·Ð°Ð´Ð°Ñ Ð¾Ð±ÑабоÑки еÑÑеÑÑвеннÑÑ ÑзÑков ÑаÑÑо нÑжно пÑоизводиÑÑ ÑинÑакÑиÑеÑкий анализ ÑекÑÑа на оÑнове гÑаммаÑики ÑзÑка. СÑеди ÑпоÑобов опиÑÐ°Ð½Ð¸Ñ Ð³ÑаммаÑик болÑÑой инÑеÑÐµÑ Ð¿ÑедÑÑавлÑÑÑ ÐºÐ¾Ð¼Ð±Ð¸Ð½Ð°ÑоÑнÑе каÑегоÑиалÑнÑе гÑаммаÑики (CCG) â лингвиÑÑиÑеÑки обоÑнованнÑй ÑоÑмализм, коÑоÑÑй позволÑÐµÑ Ð¾Ð¿Ð¸ÑÑваÑÑ ÑиÑокий клаÑÑ ÑзÑковÑÑ Ñвлений и в Ñо же вÑÐµÐ¼Ñ Ð´Ð¾Ð¿ÑÑÐºÐ°ÐµÑ ÑÑÑекÑивнÑй ÑинÑакÑиÑеÑкий анализ, а Ñакже пÑедоÑÑавлÑÐµÑ Ð´Ð¾ÑÑаÑоÑно пÑоÑÑой ÑпоÑоб ÑиÑÑемаÑиÑеÑки ÑÑÑоиÑÑ ÑеманÑиÑеÑкие пÑедÑÑÐ°Ð²Ð»ÐµÐ½Ð¸Ñ Ð¿Ñедложений на оÑнове Ð¸Ñ Ð´ÐµÑевÑев вÑвода. ÐÐ»Ñ Ð¿Ð¾ÑÑÑÐ¾ÐµÐ½Ð¸Ñ Ð³ÑаммаÑик, Ð¾Ñ Ð²Ð°ÑÑваÑÑÐ¸Ñ ÑиÑокий клаÑÑ ÑеалÑнÑÑ ÑекÑÑов, ÑаÑе вÑего иÑполÑзÑеÑÑÑ ÑÑаÑиÑÑиÑеÑкий вÑвод. ÐвÑомаÑиÑеÑÐºÐ¾Ð¼Ñ Ð²ÑÐ²Ð¾Ð´Ñ Ð³ÑаммаÑик на оÑнове CCG поÑвÑÑено болÑÑое колиÑеÑÑво ÑабоÑ, однако в Ð½Ð¸Ñ Ð² оÑновном ÑаÑÑмаÑÑиваÑÑÑÑ ÑолÑко гÑаммаÑики английÑкого ÑзÑка. ÐÐ»Ñ Ð¼Ð¾ÑÑологиÑеÑки богаÑÑÑ ÑзÑков, в Ñом ÑиÑле Ð´Ð»Ñ ÑÑÑÑкого ÑзÑка, авÑомаÑиÑеÑкий вÑвод гÑаммаÑик CCG и пÑименение поÑÑÑоеннÑÑ Ð³ÑаммаÑик Ð´Ð»Ñ ÑеÑÐµÐ½Ð¸Ñ Ð¿ÑикладнÑÑ Ð·Ð°Ð´Ð°Ñ Ð¾Ð±ÑабоÑки ÑеÑÑа оÑÑаеÑÑÑ Ð½ÐµÐ´Ð¾ÑÑаÑоÑно иÑÑледованнÑм. ÐÐ°Ð½Ð½Ð°Ñ ÑабоÑа поÑвÑÑена ÑазÑабоÑке и ÑеализаÑии алгоÑиÑма авÑомаÑиÑеÑкого вÑвода гÑаммаÑики CCG Ð´Ð»Ñ ÑÑÑÑкого ÑзÑка на оÑнове неÑазмеÑенного коÑпÑÑа ÑекÑÑов. РоÑнове ÑабоÑÑ Ð»ÐµÐ¶Ð°Ñ Ð°Ð»Ð³Ð¾ÑиÑмÑ, ÑазÑабоÑаннÑе Ð. ÐиÑком и Ðж. Ð¥Ð¾ÐºÐµÐ½Ð¼Ð°Ð¹ÐµÑ Ð´Ð»Ñ Ð°Ð½Ð³Ð»Ð¸Ð¹Ñкого ÑзÑка, коÑоÑÑе бÑли адапÑиÑÐ¾Ð²Ð°Ð½Ñ Ð´Ð»Ñ ÑÑÑÑкого ÑзÑка, в пеÑвÑÑ Ð¾ÑеÑÐµÐ´Ñ Ñ ÑÑеÑом его моÑÑологии. ÐепоÑÑедÑÑвенное иÑполÑзование доÑÑÑпной ÑеализаÑии ÑÑого Ð¿Ð¾Ð´Ñ Ð¾Ð´Ð° Ð´Ð»Ñ ÑÑÑÑкого ÑзÑка невозможно: в ÑÑÑÑком ÑзÑке ÐºÐ°Ð¶Ð´Ð¾Ð¼Ñ ÑÐ»Ð¾Ð²Ñ ÑооÑвеÑÑÑвÑÐµÑ Ð³Ð¾Ñаздо болÑÑее колиÑеÑÑво ÑоÑм, ÑоглаÑование коÑоÑÑÑ Ð½ÐµÐ¾Ð±Ñ Ð¾Ð´Ð¸Ð¼Ð¾ ÑÑиÑÑваÑÑ Ð¿Ñи ÑоÑмиÑовании деÑева вÑвода, пÑи ÑÑом поÑÑдок Ñлов ÑвлÑеÑÑÑ ÑвободнÑм. ÐÐ»Ñ Ð°Ð´Ð°Ð¿ÑаÑии алгоÑиÑма к оÑобенноÑÑÑм ÑÑÑÑкого ÑзÑка в данной ÑабоÑе ÑÑоÑмÑлиÑÐ¾Ð²Ð°Ð½Ñ Ð¿Ñавила вÑвода новÑÑ ÐºÐ°ÑегоÑий и пÑедÑÑмоÑÑÐµÐ½Ñ ÑпоÑÐ¾Ð±Ñ Ñ ÑÐ°Ð½ÐµÐ½Ð¸Ñ ÑазлиÑнÑÑ ÑоÑм Ñлов и Ð¸Ñ ÑвÑзей. ÐÑÑ Ð¾Ð´Ð½Ñми даннÑми Ð´Ð»Ñ Ð¿Ð¾ÑÑÑÐ¾ÐµÐ½Ð¸Ñ Ð¼Ð¾Ð´ÐµÐ»Ð¸ ÑвлÑеÑÑÑ Ð½ÐµÑазмеÑеннÑй коÑпÑÑ ÑекÑÑа, Ð´Ð»Ñ ÐºÐ°Ð¶Ð´Ð¾Ð³Ð¾ пÑÐµÐ´Ð»Ð¾Ð¶ÐµÐ½Ð¸Ñ Ð² коÑоÑом ÑнаÑала пÑоизводиÑÑÑ Ð¼Ð¾ÑÑологиÑеÑкий анализ. ÐÐ°Ð¶Ð´Ð¾Ð¼Ñ ÑÐ»Ð¾Ð²Ñ Ð¿ÑипиÑÑваеÑÑÑ Ð½Ð°Ð±Ð¾Ñ ÑÑандаÑÑнÑÑ ÐºÐ°ÑегоÑий. Ðалее пÑоизводиÑÑÑ ÑинÑакÑиÑеÑкий ÑÐ°Ð·Ð±Ð¾Ñ Ð¿ÑÐµÐ´Ð»Ð¾Ð¶ÐµÐ½Ð¸Ñ Ñ ÑÑеÑом ÑоглаÑÐ¾Ð²Ð°Ð½Ð¸Ñ Ð¼Ð¾ÑÑологиÑеÑÐºÐ¸Ñ Ð¿Ñизнаков Ñлов. Ð ÑловаÑÑ Ð·Ð°Ð½Ð¾ÑÑÑÑÑ Ñе каÑегоÑии, Ñ ÐºÐ¾ÑоÑÑми пÑÐµÐ´Ð»Ð¾Ð¶ÐµÐ½Ð¸Ñ ÑдалоÑÑ ÑазобÑаÑÑ, Ñ Ð´Ð¾Ð±Ð°Ð²Ð»ÐµÐ½Ð¸ÐµÐ¼ ÑооÑвеÑÑÑвÑÑÑÐ¸Ñ Ð¾Ð³ÑаниÑений на моÑÑологиÑеÑкие пÑизнаки. ÐоÑле ÑÑого ÑÑÑоиÑÑÑ Ð²ÐµÑоÑÑноÑÑÐ½Ð°Ñ Ð¼Ð¾Ð´ÐµÐ»Ñ. ÐÐ»Ñ ÑеализаÑии данного пÑогÑаммного комплекÑа вÑбÑан ÑзÑк Python, Ð´Ð»Ñ Ð¿ÑедваÑиÑелÑной обÑабоÑки и ÑаÑÑеÑеÑной ÑазмеÑки ÑекÑÑов на ÑÑÑÑком ÑзÑке иÑполÑзÑеÑÑÑ Ð±Ð¸Ð±Ð»Ð¸Ð¾Ñека pymorphy2, Ð´Ð»Ñ ÑинÑакÑиÑеÑкого анализа на ÑÑапе ÑоÑмиÑÐ¾Ð²Ð°Ð½Ð¸Ñ ÑловаÑÑ Ð¸ÑполÑзÑеÑÑÑ Ð¿Ð°ÑÑеÑ, Ð²Ñ Ð¾Ð´ÑÑий в библиоÑÐµÐºÑ NLTK. РкаÑеÑÑве Ð±Ð°Ð·Ñ Ð´Ð°Ð½Ð½ÑÑ Ð² пÑоекÑе иÑполÑзÑеÑÑÑ ÑелÑÑÐ¸Ð¾Ð½Ð½Ð°Ñ Ð¡Ð£ÐÐ sqlite., In many natural language processing tasks, it is often necessary to parse the text based on the language grammar. Among the methods for describing grammars, combinatorial categorical grammars (CCG) are of great interest. This linguistically grounded formalism allows the description of a broad class of linguistic phenomena and, at the same time, allows for efficient syntactic analysis and also provides a systematic way to construct semantic representations. Most wide-coverage grammars are probabilistic and rely on syntactically annotated corpora, which are expensive to create. An alternative approach is to apply unsupervised or semisupervised algorithms to infer grammars from unlabeled texts. There are several papers on the automatic inference of CCG, but they usually consider only grammars for English. For morphologically rich languages, including Russian, the automatic derivation of CCG grammars and the use of the constructed grammars for solving applied problems of test processing remain insufficiently studied. This work aims to develop and implement an algorithm for the automatic derivation of the CCG grammar for the Russian language based on an unmarked corpus of texts. It is based on algorithms developed by Y. Bisk and J. Hockenmaier for the English language. These algorithms were adapted for the Russian language, primarily taking into account its morphology. The adaptation of the algorithm was necessary: in Russian, each word corresponds to a much larger number of forms, the coordination of which must be taken into account when forming the inference tree, while the word order is free. In this work, we developed a set of rules for the derivation of new categories and implemented methods for storing various forms of words and connections between them. The input data for building a model is an unlabeled corpus of texts. For each sentence in the corpus, morphological analysis is performed first. Each word is assigned a set of standard categories. Next, the syntactic analysis of the sentence is performed, taking into account the agreement of the morphological features of the words. The dictionary contains those categories with which the sentences can be parsed with given restrictions on morphological features. Finally, a probabilistic model is built. The software is implemented using the Python programming language. We used the pymorphy2 library for text preprocessing and part-of-speech marking of texts in Russian. The stage of ditionary formation uses the CCG parser included in the NLTK library. The relational database SQLite is used as a database in the project.