Back to search

Democritus University of Thrace (DUTH)

Intention and sentiment mining on the web for business intelligence

Abstract

dc:description

Websites, search engines, and social media platforms are the simplest way to get information. People use social media to share their feelings and opinions about topical subjects or personal difficulties. This user behaviour provides large volumes of original unprocessed data that, with the proper manipulation, might yield substantial research discoveries. A portion of the content shared on social media is connected to the products and services available on the market. Large amounts of publicly available information on a product or service provide an essential quantity of data that businesses and other types of organisations can use. In recent years, much effort has been placed into managing these data to derive relevant insights for enterprises. The essential tools are Business Intelligence (BI) systems to quickly process, analyse, and visualise data. The purpose of BI systems is to facilitate the decision-making process inside an organisation. As a tool for making decisions, BI systems can adopt and analyse the information provided by social media users to forecast the present and future specific needs of customers and evaluate the performance of businesses. In addition, it is helpful to recognise the sentiment and the underlying intention in the large amounts of publicly available and free to obtain microblogging data to act appropriately in different scenarios. In this thesis, our goal is to find a way to get beyond these limitations, unite the research fields of sentiment analysis and intention mining, and create a novel BI system. First, we focus on pre-processing, which is the first step in text classification. Choosing the proper pre-processing techniques can improve classification effectiveness for sentiment analysis. For example, lemmatization, removing numbers and replacing contractions improve accuracy, but removing punctuation doesn't. On the other hand, ablation and combination studies show the importance of replacing numbers and punctuation. Then, we present an ensemble text sentiment classification scheme based on empirical analysis of several classifiers and other related works. The proposed system is based on linguistic pre-processing, feature engineering, and supervised machine learning techniques. Finally, we present a Business Intelligence tool that categorises tweets by intention, sentiment, and opinion. The final model automatically recognises the intent to buy, compare, and seek information about a product. Concluding, in this thesis we created a novel BI system that uses publicly available information and by using Natural Language Understanding (NLU) approaches offers to end-users important insights about a product or service.

Degree

thesis:*
Grantor dc:publisher
Democritus University of Thrace (DUTH)
Year dc:date
2022

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Symeonidis, Symeon
  • Συμεωνίδης, Συμεών

Subjects

dc:subject × 22

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/51933

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Symeonidis, Symeon; Συμεωνίδης, Συμεών. Intention and sentiment mining on the web for business intelligence. Democritus University of Thrace (DUTH), 2022. http://hdl.handle.net/10442/hedi/51933