The Model of Coreference Resolution in Uzbek Texts: A Review

Аннотация

Coreference resolution plays a crucial role in natural language processing by enabling accurate understanding of a text and identifying its semantic structure. While effective Coreference resolution systems have been developed for resource-rich languages such as English, German, and Chinese, research and practical systems in this area remain insufficient for the Uzbek language. Uzbek differs significantly from other languages due to its agglutinative structure, flexible word order, and rich morphology. These linguistic features necessitate unique approaches and models for Coreference resolution. This article discusses the Uzbek-language Coreference resolution system – UzCoref – highlighting its functional capabilities, system architecture, data flow, underlying model, testing process, comparative analysis with other systems, and the advantages of UzCoref. Keywords: Coreference resolution, Uzbek language, UzCoref, text, model, system, corpus, architecture. for the Uzbek language, such large-scale annotated Introduction: The phenomenon of coreference refers corpora and ready-made CR systems are currently to a linguistic event where different expressions in a unavailable. Therefore, the task of CR for Uzbek texts text (such as pronouns, nouns, etc.) refer to the same remains both relevant and novel. person, object, or event. In the field of Natural Language Processing (NLP), resolving and linking such The UzCoref system we present is specifically designed expressions is a significant task. For example, in the for CR in Uzbek language texts. The system has been sentence “Zulfiya wrote a book. Her work...”, developed using a scientific and practical approach identifying that the pronoun “her” refers to “Zulfiya” is aligned with the field of computational linguistics. an example of Coreference resolution (CR). UzCoref enables the identification of words and expressions in a given text that refer to the same For major languages like English and Russian, CR referent, clusters them accordingly, and allows for systems and models are well-developed. Large deep semantic analysis of the text. The system not only annotated corpora (e.g., the OntoNotes corpus) have provides a core algorithmic solution but also offers been used to train modern neural models. It is known broad functional capabilities, ensuring ease of use for that the latest methods, particularly those based on users through various methods. The system is specialized machine learning approaches, have accessible at https://uzcoref.uz/ (Figure 1): achieved around 83% F1 score accuracy [1]. However, American Journal Of Philological Sciences 84 https://theusajournals.com/index.php/ajps American Journal Of Philological Sciences (ISSN – 2771-2273) Figure 1. Interface of the UzCoref System system for Uzbek texts. The UzCoref system serves as a foundational tool for various NLP tasks in the Uzbek language. RESULTS METHODOLOGY The UzCoref system provides several usage modes and interfaces tailored to user needs. Specifically, the The issue of CR has been widely studied since the 1970 system offers the following functional features: s. Notably, R.Mitkov provided a comprehensive description of anaphora and coreference phenomena, – Web Interface (Front-End): A user-friendly graphical defined their terminology, proposed methods for web interface is provided. Users can input text into a automatic detection, and introduced a multi-stage special field (or upload a file) and click the “Detect” integrated model [2, 3]. Vincent Ng explored button to view the result directly in the browser. The unsupervised learning models for the CR problem [4], web interface displays coreference clusters with color while Altaf Rahman and Vincent Ng analyzed highlights in the text, facilitating visual understanding supervised learning models for addressing CR [5]. A of relationships within the text. research team led by R.Sukthanker theoretically – Command Line Interface (CLI): A dedicated CLI tool is examined the differences between coreference and available to run the system via console or terminal. anaphora, their types, evaluation metrics, the Users can input a text file using the uzcoref command evolution of CR techniques in NLP, as well as the and output results either to the console or export them strengths and weaknesses of these approaches [6]. to a file. This mode is particularly useful for developers Martha Recasens Potau published a doctoral and advanced users, allowing for integration of UzCoref dissertation focused on the coreference issue using the functions into scripts and automated workflows. Spanish language as a case study. She proposed the – RESTful API Service: The system can also be used as a CISTELL model as a CR, and used AnCora-Co, a web service via REST-style API endpoints, enabling coreference-annotated corpus for Spanish and Catalan, remote access. This allows integration of UzCoref into for testing [7]. V.Prokopenya and T.Chernigovskaya external applications. For