Alisher Navoiy nomidagi Toshkent davlat o‘zbek tili va adabiyoti universiteti

Alisher Navoiy nomidagi Toshkent 
davlat o‘zbek tili va adabiyoti 
universiteti 
“O‘ZBEK MILLIY VA TA’LIMIY 
KORPUSLARINI YARATISHNING NAZARIY 
HAMDA AMALIY MASALALARI”
Xalqaro ilmiy-amaliy konferensiya
Vol. 1
№. 01 (2021) 
289 
 
A PROBLEM OF PART-OF-SPEECH TAGGING IN THE UZBEK LANGUAGE CORPUS 
O‘ZBEK TILI KORPUSIDA SO‘Z TURKUMLARINI TEGLASH MASALASI 
 Rabbimov Ilyos Mehriddinovich
*
Kobilov Sami Saliyevich
**
102
 Qurdoshev Zarifjon Mansur o‘g‘li
***
103
 
Annotation.
Feature extraction is important in opinion classification using machine learning 
algorithms. In this paper, the methods of feature extraction used in automatic opinion classification are 
analyzed, and the issue of opinion classification of Uzbek text is performed. 
Keywords:
 opinion classification, machine learning, feature extraction, stylistic features, statistical 
features, part-of-speech features, semantic features. 
Annotatsiya. 
Mashinali o‘qitish algoritlaridan foydalanib fikrlarni tasniflashda informativ 
belgilarni ajratish muhim hisoblanadi. Ushbu maqolada fikrlarni avtomatik tasniflashda qo‘llaniladigan 
informativ belgilarni ajratish usullari tahlil qilinadi va oʻzbek tilida yozilgan matnlardagi fikrlarni 
tasniflash masalasi bajariladi.
Kalit so‘zlar: 
fikrlarni tasniflash, mashinali o‘qitish, informativ belgilarni ajratish, uslubiy 
xususiyatlar, statistik xususiyatlar, so‘z turkumlariga asoslangan xususiyatlar, semantik xususiyatlar.
 
As the use of the Internet increases, so does the large amount of text created on social networks, 
blogs, and e-commerce platforms. The demand for intellectual data analysis methods and algorithms, 
which are used to extract high-quality data from such texts, is growing. One of the scientific directions of 
intellectual analysis of textual data is the problem of automatic opinion classification. Opinion mining is 
the study of people’s moods, viewpoints, opinions, attitudes, and feelings toward subjects, such as 
services, products, research, political issues, organizations, and other topics. The purpose of opinion 
classification is to determine whether the opinion in the text is positive, negative or neutral. Automatic 
opinion classification can be used to determine customers’ opinions of products or services and adapting 
them to their needs; in the analysis of public opinion on political events or new laws; when organizations 
seek feedback on their employees; in the automatic analysis of opinions of famous people and 
organizations about themselves and their brand. In general, the opinion classification is done at the 
document level, sentence level, and aspect level. Approaches to automatic opinion classification can be 
divided into rules-based methods, machine-based learning methods, and hybrid methods. The opinion 
classification includes the following main steps: 

collection of textual information; 

pre-processing of textual data; 

feature extraction; 

designing classification algorithms. 
In this paper, the methods of feature extraction used in automatic opinion classification are 
analyzed, and the issue of opinion classification of Uzbek text is performed. 
In the process of opinion classification using machine learning, deep learning algorithms, there is a 
need to express texts in the form of numerical vectors. There is a lot of scientific research on the problem 
of obtaining feature vectors that represents the statistical, lexical, stylistic, and semantic features of the 
text. Stylistic, syntactic, part-of-speech (POS), lexicon-based features, as well as one-hot encoding, term 
frequency-inverse document frequency (TF-IDF), Word2Vec, GloVe, and fastText models, are used for 
feature extraction from text.