Oʻzbekiston respublikasi оliy va oʻrta maxsus

Ключевые слова: выявление субъективности, cентимент анализ, алгоритм нечеткой сис- темы управления, узбекский язык, распознавание образов.
Key words: subjectivity detection, sentiment analysis, fuzzy control system algorithm, Uzbek lan- guage, pattern recognition.
Introduction. In recent years, there has been a surge in interest in identifying and extracting sub- jective information from Web sites, such as opinions. Opinions are often subjective statements represen- ting the opinions, evaluations, or emotions of persons. The goal of subjectivity detection is to ascertain
whether a given text expresses opinions (subjective) or offers facts (objective) [1, pp. 1153-1161]. Tech- niques for automated subjectivity analysis have been implemented in a variety of text processing and na- tural language applications. In a number of natural language processing applications, subjectivity detec- tion is employed as a first filtering step to give more relevant data. Our research aims to build learning techniques for classifiers that can differentiate between subjective and objective texts. In this paper, we classify subjectivity at the sentence level using language-independent feature weighting. As a test issue, we used the “Uzbek cuisine” restaurant domain review subjectivity dataset that is stored to huggingface¹.
Main part. Two different supervised machine learning approaches: a Fuzzy Control System and Adaptive Neuro-Fuzzy Inference System and applied to sentence-level subjectivity detection in a restau- rant domain review. The presence or occurrence statistics within the corpus are used by the majority of language independent feature extraction techniques. We present one such technique that is simple to use, computationally efficient, and does not need any extra human annotation or lexical understanding. We use a subjectivity dataset, which contains 5000 subjective and objective processed phrases from restaurant domain reviews. Because our target lacks lexical expertise, we consider each word to be a code word. We do not mix verbs in various tenses, such as present and past (“qaror qilish” vs. “qaroq qilishdi”), nor nouns as single or plural in our algorithm (“gul” vs “gullar”). Instead, we regard them to be the various code terms.