SOCIAL MEDIA SENTIMENT ANALYSIS FOR SERVICE CLASSIFICATION

TALAKSEW, KASSA YAZEW

dc.contributor.author	TALAKSEW, KASSA YAZEW
dc.date.accessioned	2024-04-19T08:19:13Z
dc.date.available	2024-04-19T08:19:13Z
dc.date.issued	2023-07
dc.identifier.uri	http://ir.bdu.edu.et/handle/123456789/15765
dc.description.abstract	Social media analysis is an essential mechanism for organizations to grasp new knowledge and information from social media data. In social media analysis, sentiment analysis and service classification are important research areas. These two tasks help organizations to know how their customers are using their services and identify areas for improvement. Sentiment analysis is the task of categorizing feedback as positive, negative, or neutral. Service classification is the task of identifying the service of the organization that a text refers to. Organizations use social media for many purposes, and social media analysis can help them to get the public's opinion. In the organization, public relations is the administration sector, having a responsibility to build strategic relationships between organizations and the public. This task is highly connected with social media analysis in today's day and age. Many research studies have been done on sentiment analysis and service classification. Most of these studies were conducted in the business sector, and many research studies have been done on English and Amharic languages separately. Rather in this study, both sentiment analysis and service classification are performed on the higher education using both Amharic and English datasets. In this study, multi-label text classification (MLTC) is utilized to handle sentiment analysis and service classification tasks. MLTC is for converting to multiple single-label binary classifications. Also used a deep learning approach, specifically LSTM, to train our model. Our study area is specifically on the Bahir Dar University Facebook page. The dataset used has Amharic and English languages because the university uses and the public responds in both languages. The results of the analysis showed an accuracy of 45% of the accuracy of the merged dataset result of both language datasets. This shows the study attempts to get a specific view and provide more information in one study, and more relevant to real-world inquiries or more realistic. Keywords: Public Relations, Social Media Analytics, Semantic analysis, Service classification, Multi-label text classification	en_US
dc.language.iso	en_US	en_US
dc.subject	Software Engineering	en_US
dc.title	SOCIAL MEDIA SENTIMENT ANALYSIS FOR SERVICE CLASSIFICATION	en_US
dc.type	Thesis	en_US