This article analyzes current problems in the development of artificial intelligence and natural language processing (NLP) models for low-resource languages, in particular, the Karakalpak language. The main goal of the research is to develop a methodology for data refinement and normalization to improve model accuracy under conditions of data scarcity. In the methods section, special normalization algorithms and Byte Pair Encoding (BPE) tokenization were used for specific letters of the Karakalpak language. The results showed that the proposed preprocessing methods made it possible to significantly reduce the model error (Loss) and increase the translation and search accuracy (BLEU score) from 12.4 to 24.8. The research findings serve as a theoretical and practical basis for creating digital dictionaries and intelligent systems for low-resource languages.
| Mualliflar | Aimbetova , Gulara, Sarsenbaeva, Hu’rlixa, Djumabaev , Alpamis |
|---|---|
| Jurnal | Techscience.uz - техника фанлари долзарб масалалри |
| Nashr sanasi | 2026-05-14 |
| Jild | 4 |
| Son | 5 |
| Betlar | 43-48 |
| Til | O‘zbek |
| DOI | 10.47390/ts-v4i5y2026n07 |
DOI: 10.47390/ts-v4i5y2026n07 · Maqolaning asl sahifasi
NLP, low-resource languages, Karakalpak language, preprocessing, normalization, artificial intelligence, machine translation, NLP, kam resursli tillar, qoraqalpoq tili, preprocessing, normallashtirish, sun’iy intellekt, mashinali tarjima.
This article presents the results of using deep learning methods to determine the level of land salinity and predict its dynamics in the Shortanbay massif of the Nukus district of the Republic of Karakalpakstan based on…
The paper presents an automated monitoring system for agricultural lands in the Nukus district (Republic of Karakalpakstan, Uzbekistan). The dataset comprises 1,282 observations spanning January 6, 2023 through February…
This article considers RFID-based parking access logging together with short-term occupancy estimation from the collected data. In Uzbekistan, the growth of car ownership and the increasing pressure on parking spaces…
This article presents ideas about the system of improving the moral culture of students through fiction. Through them, the further development of the process of reading fiction, the socio-cultural activities of the…
This article examines the mathematical foundations of Wiener-Hopf filter construction algorithms for digital image processing. The filter construction algorithm is described, and the efficiency of noise suppression in…
The article discusses the application of the Model Predictive Control (MPC) method to solid material drying processes. Bulk materials. A control system was synthesized using the MPC algorithm, and the generation of…
This research is aimed at studying the effectiveness of reinforcing brick walls using ECC, steel mesh, steel tapes, and zinc sheets to increase their seismic resistance. As a result of experimental and numerical…
This article discusses modern 3D modeling of the pishtaq element of historical monuments and presents a comparative analysis of calculation results. Within the scope of the study, various strengthening methods are…
This article presents the theoretical foundations of plow aggregation. Efficient aggregation of plows with tractors in agricultural operations plays an important role in improving soil tillage quality, reducing fuel…
The rapid growth of electric vehicle adoption creates both promising opportunities and serious challenges for contemporary power distribution networks. Uncoordinated EV charging can impose substantial stress on the…
Techscience.uz - техника фанлари долзарб масалалри — barcha maqolalar