Neural Network Interface for Converting Complex Russian-Language Text Commands into a Formalized Graph to Control Robotic Devices

A neural network interface for converting complex Russian-language text commands for robotic devices into the RDF-format is presented. The interface involves neural network models to recover missing verbs, split compound commands into single ones, and parse single commands. In order to train these models, training and testing samples have been formed: a data set for learning to divide compound commands into single commands and to analyze single commands has been created with the help of a specially-developed text command generator, using which 55000 simple commands and 16000 composite commands have been generated using special templates; for the task of recovering missing verbs, the corpus from the Dialog-21 conference is used, containing 16000 sentences, of which 5000 have missing verbs; the test set is assembled using crowdsourcing technology and contains 1300 examples. The methods used for text analysis are based on language models and neural networks with Transformer architecture. The accuracy of using a resourceintensive language model (MultilingualBERT) and the resource-efficient distilled version RuBERT-tiny of RuBERT has been evaluated. The results of the dependence of the command parsing accuracy on the presence of punctuation marks and capital letters in the text are presented.

1. Gray J., Srinet K., Jernite Y., Yu H., Chen Z., Guo D.,Szlam A. Craftassist: A framework for dialogue-enabled interactive agents. arXiv preprint arXiv: 1907.08584, 2019.

2. Adiwardana D., Luong M.T., So D.R., Hall J., Fiedel N.,Thoppilan R., Le Q.V. Towards a human-like open-domain chatbot. arXiv preprint arXiv: 2001.09977., 2020.

3. Dong L., Lapata M. Language to logical form withneural attention. arXiv preprint arXiv: 1601.01280, 2016.

4. Sboev A.G., Gudovskih D.V., Moloshnikov I.A., Rybka R.B. Opredelenie pola avtora teksta v kollekcii russkih mnogozhanrovyh tekstov s iskazheniyami s ispol’zovaniem modelej mashinnogo obucheniya. Vestnik NIYaU MIFI, 2018, vol. 7, no. 6, pp. 531–536. (in Russian)

5. Sboev A.G., Moloshnikov I.A., Rybka R.B., Naumov A.V. Interpretaciya rezul’tatov modeli opredeleniya tipa imitacii vozrasta v tekste. [Interpretation of the results of the model for determining the type of age simulation in the text]. Vestnik NIYaU MIFI, 2020, vol. 9, no. 2, pp. 189–196. (in Russian)

6. Moloshnikov I.A., Gryaznov A.V., Vlasov D.S.,Rybka R.B., Sboev A.G. Primenenie mul’tizadachnoj modeli dlya prakticheskih zadach generacii zagolovka, opredeleniya lemm i klyuchevyh slov [Application of the multitasking model for practical tasks of title generation, definition of lemmas and keywords]. Vestnik NIYaU MIFI, 2020, vol. 9, no. 3, pp. 236–244. (in Russian)

7. Sboev A.G., Rybka R.B., Gryaznov A.V., Moloshnikov I.A. Generativno-diskriminativnaya nejrosetevaya model' dlya zadachi avtorskogo profilirovaniya [Generative-discriminative neural network model for the problem of author’s profiling]. Vestnik NIYaU MIFI. 2020, vol. 9, no. 1, pp. 50–57. (in Russian)

8. Sboev A.G., Selivanov A.A., Rybka R.B., Moloshnikov I.A., Bogachev D.S. Model’ nejronnoj seti dlya vklyucheniya sintaksicheskoj struktury predlozheniya v zadachu klassifikacii pola avtora russkogo teksta. [A neural network model for including the syntactic structure of a sentence in the task of classifying the gender of the author of a Russian text]. Vestnik NIYaU MIFI, 2019, vol. 8, no. 6, pp. 569–576. (in Russian)

9. Smurov I.M., Ponomareva M., Shavrina T.O., Droganova K. Agrr-2019: Automatic gapping resolution for russian. Computational Linguistics and Intellectual Technologies, 2019, pp. 561–575.

10. Anisimovich K.V., Druzhkin K.Ju., Minlos F.R.,Petrova M.A., Selegey V.P., Zuev K.A. Syntactic and semantic parser based on ABBYY Compreno linguistic technologies. Komp’iuternaia Lingvistika i Intellektual’nye Tehnologii: Trudy Mezhdunarodnoj Konferentsii “Dialog”, 2012, pp. 91–103.

11. Belkin I. BERT finetuning and graph modeling for gapping resolution. Komp’juternaja Lingvistika i Intellektual’nye Tehnologii, 2019, pp. 63–71.

12. Devlin J., Chang M.W., Lee K., Toutanova K. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv: 1810.04805, 2018.

13. Dale D. Malen’kij i bystryj BERT dlya russkogo yazyka [Small and fast BERT for Russian]. Available at: https://habr.com/ru/post/562064/. (accessed 24.06.2022).

14. Kuratov Y., Arkhipov M. Adaptation of deep bidirectional multilingual transformers for russian language. arXiv preprint arXiv: 1905.07213, 2019.

15. Feng F., Yang Y., Cer D., Arivazhagan N., Wang W.Language-agnostic bert sentence embedding. arXiv preprint arXiv: 2007.01852, 2020.

16. Artetxe M., Schwenk H. Massively multilingual sentence embeddings for zero-shot cross-lingual transfer and beyond. Transactions of the Association for Computational Linguistics, 2019, vol. 7, pp. 597–610.

17. Cer D., Yang Y., Kong S.Y., Hua N., Limtiaco N., JohnR.S., Kurzweil R. Universal sentence encoder. arXiv preprint arXiv: 1803.11175, 2018.

18. Zhang B., Williams P., Titov I., Sennrich R. Improvingmassively multilingual neural machine translation and zeroshot translation. arXiv preprint arXiv: 2004.11867, 2020.

19. Tiedemann J. Parallel data, tools and interfaces inOPUS. Lrec., 2012, pp. 2214–2218.

20. Williams A., Nangia N., Bowman S.R. A broad-coveragechallenge corpus for sentence understanding through inference. arXiv preprint arXiv: 1704.05426, 2017.