TalkAbout: Cross Domain Opinion Mining and Sentiment Analysis

Introduction

People talk about people, service, products in reviews and comment feeds. Talk About is an opinion mining and sentiment analysis platform that focuses on cross domain sentiment analysis on the items that people commented. The project focuses on three key components in drawing insights from online reviews, i.e.
i. plafform where the reviews can be gathered, integrated, processed, analyzed and projected.
ii. libraries/models comprising of improvised solutions related to sentiment analysis.
iii. knowledge base of aspect and sentiment words for different verticals.

Project I: TalkAbout Analytics Platform for Opinionated Texts

The TalkAbout analytics platform focuses on the development opinion analysis tools and applications. The project covers development of components like data sourcing, data storage & standardization, base libraries and end user prototype for opinion text processing and analysis. The scope of this project includes analytics modules like text pre-processing pipelines, intelligent modelling and visuals to project the results of analysis. Methods like NLP, machine learning, rule-based, lexicon-based techniques will be applied in the development work.

Scope

People

Deliverables

Talk About - Text Analytics Platform with Intelligent Categorization by Chiow Hui Qin (2021)

Talk About - Chatbot for Tourism Point of Interest by Chin Wai Lun (2020)

Talk About - Large-Scale Big Data Analytic for Event Monitoring by Hoh Wen Chun (2020)

Talk About - Textual Summarization for Review Analytics by Loi Hang Lieh (2019)
Check out the prototype LIVE at @ ir.cs.usm.my/talkabout/touranalytics/


Select DEMO option to proceed.

Project II: Sentiment Analysis for Opinionated Texts

Opinion rich resources come in different sizes and complexities. Mining such resources involve challenges in handling variety of cases on how sentiments are expressed and understanding of context of whether the expressed sentiments are important and relevant to the product/service. Following this, conclusion are drawn from the mined sentiments. This project focus on a multi-dimension, e.g. sizes, complexities, domain based solution on identifying sentiments given an entity/aspect and drawing conclusion from the sentiments. Techniques for exploration ranging from state-of-the-arts to latest learning algorithms.

Scope

People

Deliverables

Vivian Lee Lay Shan, Gan Keng Hoon, Tan Tien Ping, Rosni Abdullah: Using Informative Score for Instance Selection Strategy in Semi-Supervised Sentiment Classification. Computers, Materials and Continua, 74(3):4801-4818, Tech Science Press (2023). Q2 IF:3.860 (JCR 2021)

Gan Keng Hoon, Noeurn Krol: Contrast and conditional polarity rule (CCPR) for detection of positivity and negativity in opinionated sentence with indirect polarity. International Journal of Web Information Systems, 17(2):65-83, Emerald (2021).

Vivian Lee Lay Shan, Gan Keng Hoon, Tan Tien Ping, Rosni Abdullah: Semi-supervised Learning for Sentiment Classification using Small Number of Labeled Data. 5th Information Systems International Conference 2019, ISICO 2019, 23-24 July 2019, Surabaya, Indonesia, Procedia Computer Science, Elsevier (2019).

Noor Rizvana Ahamed Kabeer, Gan Keng Hoon: A Framework for Aspect and Sentiment Extraction for Online Review. ACIS 2015, 15-17 October, 2015, Penang, Malaysia. slide

Project III: Aspect-based Opinion Mining

Aspect-based opinion mining explores both domain dependent and independent approaches for identifying explicit and implicit entities and their aspects that are strongly related to a specific product/service domain. In this context, mining and summarizing these specific entities (and their aspects) can help consumers in deciding what to purchase and businesses to better monitor their reputation and understand the needs of the market. In this project, research of better techniques for processes involved in aspect-based sentiment analysis like aspect extraction, aspect categoization and aspect ranking etc. are explored.

Scope

People

Deliverables

Saif A Ahmad Alrababah, Gan Keng Hoon, Tan Tien-Ping, Mohammad N Al-Kabi: Product aspect ranking using multi-criteria decision making. Handbook of Research on Consumer Behavior Change and Data Analytics in the Socio-Digital Era, 74-99, IGI Global (2022).

Lim Jie Chen, Gan Keng Hoon: Feature expansion using lexical ontology for opinion type detection in tourism reviews domain. International Journal of Advanced Computer Science and Applications, 11(8):628-637, SAI (2020).

Saif Addeen Alrababah, Gan Keng Hoon, Tan Tien Ping: Mining Opinionated Product Features using WordNet Lexicographer Files. Journal of Information Science, 43(6):769-785, SAGE (2017). Q2 IF:1.372

Saif Addeen Alrababah, Gan Keng Hoon, Tan Tien Ping: Comparative Analysis of MCDM Methods for Product Aspect Ranking: TOPSIS and VIKOR. 8th International Conference on Information and Communication Systems, ICICS 2017, 4-6 April, 2017, Irbid, Jordan.

Saif Addeen Alrababah, Gan Keng Hoon, Tan Tien Ping: Product Aspect Ranking using Sentiment Analysis and TOPSIS. The Third International Conference on Information Retrieval and Knowledge Management, CAMP 2016, 23-24 August, 2016, Melaka, Malaysia. [Best Paper Award]

Project IV: The SummaRev Platform

** this project has ended

The massive amount of online reviews written by users about services and products are beneficial to both consumers and business providers. For consumers, learning more about a product or service before purchasing shall assist in better bargain. For business provider, being able to capture consumer feedbacks quickly shall facilitate decision making to improve effectiveness of business. However, the differences of opinion types and domains produce many details that needs to be compiled and summarized for the ease of decision making. SummaRev is a platform proposed to solve this problem, the project includes development of data sourcing, data storage standards, base libraries (using NLP, machine learning, lexicon etc.) of review processing and a set of visuals to project the results of analysis.

The first version of SummaRev was experimented hotel reviews. The on-going development will be deployed for cross domain solutions,

Scope

People

Deliverables

Teh Xin Xi, Tan Zhiyu, Wong Kean Yi, Gan Keng Hoon: Review Summarization using Feature-based Sentence Categorization. ACIS 2015, 15-17 October, 2015, Penang, Malaysia. slide

FYP Prototype Demo Slide - The Storage *

FYP Prototype Demo Slide - The Sentiment *

FYP Prototype Demo Slide - The Summary *

FYP Final Report - The Storage *

FYP Final Report - The Sentiment *

FYP Final Report - The Summary *

* This file is password protected and reserved for internal viewing only. Please send a request to my email if you wish to use the file.