MSc Thesis Proposal Announcement of Mayank Semwal:"Active Community Opinion Network Mining and Maximization through Social Network Posts "

Thursday, April 29, 2021 - 14:00 to 16:00

SCHOOL OF COMPUTER SCIENCE 

The School of Computer Science is pleased to present...

MSc Thesis Proposal by:  Mayank Semwal 

 
Date:  Thursday, April 29th, 2021                
Time:  2:00 pm – 4:00 pm 
Passcode: If interested in attending this event, contact the Graduate Secretary at csgradinfo@uwindsor.ca
 

Abstract:  

Opinion Mining plays a vital role in social networking sites like Facebook, where users share their opinions, and by analyzing the responses (likes, comments, and shares) of users on social posts, we can obtain product popularity. Opinion Maximization (OM) is a variant of influence maximization (IM). Unlike IM, only some seed users may like the product and share positive opinions, whereas negative opinions can discredit a product's overall popularity. In OM, a small set of potential customers is considered to maximize the overall opinion spread toward a target product in the social network that shares positive and neutral opinions about a product.       
 
Existing mining systems like CONE takes a partial historical rating matrix of users on multiple products and performs opinion estimation to fill the partial matrix that maximizes overall positive opinions using OM. However, CONE does not consider social posts where users provide opinions through comments, likes and sharing about a product. OBIN mine users' low-frequency features from comments using feature mining to create a community preference influence network, then identify the relationships among nodes to create an influence network. However, OBIN does not consider slang word inclusion for Facebook, which users often use over the internet. Also, OBIN performs only feature-level sentiment analysis (SA) and does not consider sentence-level SA with modalities. For example, "a removable battery would be nice" has a positive opinion word "nice" but expresses a negative sentiment. OBIN discovers community preference but does not perform Influence or Opinion Maximization, which is more profitable to the seller when introducing a new product in the market for opinion estimation when opinion is not present about a product.   
 
This thesis proposes an algorithm called Active Community Opinion Network Mining and Maximization (ACOMax), which performs sentence and feature-level SA jointly. First, we determine the user's opinion of the overall sentence. Second, we extract all opinion features, including their sentiments, i.e., users' emotional reactions toward the product. Next, ACOMax performs OM on the influence network created by OBIN and performs OM using Multiple Linear Threshold (MLT), a multiple cascade model unlike OBIN, allowing nodes to be disregarded during the diffusion process and combine active opinion estimation for a new product using Collaborative Filtering (CB). The proposed ACOMax creates an opinion network of users who share positive opinions on posts related to a product, then that network is utilized by the seller to introduce a new product. ACOMax is a threefold system: In the first stage, using a hashtag like #iPhone12 for mining posts related to the new product using extended OpinionMiner by OBIN, which considers relationships between users. In the second stage, opinion estimation (positive, negative, and neutral) from comments of selected users on posts using the Apriori frequent pattern mining. Finally, active selection of top k users with positive and neutral opinions on the product to construct an opinion network and perform OM using MLT and CB. 
 
Keywords: Opinion Maximization, Sentiment Classification, Opinion Mining, Frequent pattern Mining, Social Network 
 
 

MSc Thesis Committee:  

Internal Reader: Dr. Pooya Moradian Zadeh 
External Reader: Dr. Animesh Sarkar 
Advisor: Dr. Christie Ezeife 
 

 MSc Thesis Proposal Announcement     Vector Institute artificial intelligence approved topic logo

 

 5113 Lambton Tower 401 Sunset Ave. Windsor ON, N9B 3P4 (519) 253-3000 Ext. 3716 csgradinfo@uwindsor.ca