DEVELOPMENT OF A NEURO-SYMBOLIC ALGORITHM FOR IDENTIFYING THE SUBJECT AND ATTRIBUTE IN UZBEK SENTENCES
Sharipov M. S., Avezmatov I. D., Bakhtiyorova M. N. & Alimbaeva Y. A.
Annotatsiya
This article details the development of an efficient and reliable neuro-symbolic algorithm and software pipeline architecture designed for the automatic identification of the subject and attribute, which are considered the main syntactic parts of sentences in Uzbek texts. The proposed computational linguistics solution consists of three main sequential stages: a specialized lexical tokenization system directly adapted to the complex nature of the Uzbek language, characteristic of the agglutinative language family; a Part-of-Speech (POS) tagging architecture using a additionally trained and fine-tuned BERT model; and a syntactic role extraction engine based on precise and complex linguistic rules. To ensure the stability of error-free subject identification within the text, the system relies on the predicate identified in the previous stage. Furthermore, to identify the attribute, it refers to morphological-logical chains such as the genitive case and adjectives. Once the predicate is found based on morphological analysis, specific syntactic rules defining the subject and attribute are triggered by the system. Evaluation was carried out on a large test set consisting of 1000 sentences collected from various news reports and scientific articles in Uzbekistan, manually annotated by experts. As a result, the proposed system recorded an overall average F1-score of 85.4% for subject identification and 87.2% for attribute identification.
Adabiyotlar
- Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2018). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of NAACL-HLT 2019 (pp. 4171-4186).
- Kuriyozov, E., Matlatipov, S., Alonso, H. M., & Gómez-Rodríguez, C. (2020). Construction and evaluation of BPE embeddings for the Uzbek language. In Proceedings of the 7th Workshop on NLP for Similar Languages, Varieties and Dialects.
- Sharipov M., Avezmatov I., Adinaev H. (2025). Development of a rule-based model and algorithm for predicate identification in Uzbek language texts. 10th International Conference on Computer Science and Engineering (UBMK). IEEE. pp. 594-598.
- D’Avila Garcez, A., & Lamb, L. C. (2020). Neurosymbolic AI: The 3rd wave. Artificial Intelligence Review, 56(11), 12387-12406.
- Sharipov M., Kuriyozov E., Yuldashev O., Sobirov O. (2024). UzbekVerbDetection: Rule-based detection of verbs in Uzbek texts. Proceedings of LREC-COLING 2024. pp. 17343-17347.