Interaction and Natural Language Processing Based Sentiment Analysis of Bangla Food Review in Social Media

Author: Ataf Fazledin Ahamed (1705066 at ugrad.cse.buet.ac.bd)
Supervisor: Dr. Sadia Sharmin (sadia at cse.buet.ac.bd)

Acknowledgement

We are thankful to our advisor, Dr. Sadia Sharmin for her continuous support in our related research. Her guidance and motivation during our research have been an inspiration as well as a factor of courage. We are grateful to her for her enthusiasm, guidance, and most importantly, her confidence in us. We would also like to thank all the board members of our B.Sc. thesis defense for their valuable feedback and profound observations of our work shared during the thesis defense presentation. Our heartful thanks go to our classmates and seniors who have helped us during our research. Whether it may be answering a question or giving suggestions based on their experience, we found their support with us in this journey.

Abstract

With the increasing availability of the internet and mobile devices, more and more people are connecting to social media. Social media has evolved from a simple way of connecting with friends and family to a global village. It now presents itself as a medium of entertainment, a news platform, a marketplace, and a place for advertisement. Users are nowadays attracted by social media content. In this study, we have focused on the food review-related contents of social media and performed sentiment analysis. During this study, we collected Bangla food reviews and related comments from popular Facebook pages. We have used CrowdTangle, a social media analytics tool to collect data from social media pages and used a scoring system based on reactions. Later on, we used Natural Language Processing to classify the collected comments and compared the results to the scoring system. Our study shows a high tendency for positive sentiments in food reviews

Content Tree

├───banglabert-files
│   └───outputs
├───datasets
│   ├───facebook
│   │   ├───khudalagse
│   │   │   ├───csv
│   │   │   └───pred
│   │   ├───metroman
│   │   │   ├───csv
│   │   │   └───pred
│   │   ├───petuk-couple
│   │   │   ├───csv
│   │   │   └───pred
│   │   ├───rafsan
│   │   │   ├───csv
│   │   │   └───pred
│   │   └───zoltanbd
│   │       ├───csv
│   │       └───pred
│   ├───preprocess
│   │   └───SentNoB Dataset
│   └───training
├───src
│   ├───notebooks
│   └───scripts
└───thesis-book
    ├───presentation (thesis defense)
    └───source (latex files)