Multimodal Sentiment Analysis

COMP/DATA/CSEC 5703 Group-Based Capstone Project.

Getting Started

This project aims to create three multimodality sentiment analysis models and test them on three multimodal datasets with texts and images using Python.

Prerequisites

Python 3.x version

Check requirement.txt for all the packages that needs to be installed.

Please open a issue if there is any problem.

Installing

Clone the whole github respository

git clone https://github.com/beholder91/Multimodal-Sentiment-Analysis.git

Install all the required packages using

pip install -r requirements.txt

Download the following three datasets from the websites and put the zip files into the folder Capstone-5703/CS15_2_virtual/DATASET

-Memotion
https://www.kaggle.com/datasets/williamscott701/memotion-dataset-7k
Rename the zip file as memotion_dataset_7k.zip

-MVSA-single
https://www.kaggle.com/datasets/vincemarcs/mvsasingle
Rename the zip file as MVSA_single.zip

-MVSA-multi
https://www.kaggle.com/datasets/vincemarcs/mvsamultiple
Rename the zip file as MVSA.zip

Running the tests

Linux/MacOS:

Using termial to navigate to the respository folder and run the following codes

make run_script clean

Windows:

Physically unzip three datasets to folder Capstone-5703/CS15_2_virtual/DATASET:

If using terminal

tar -xf memotion_dataset_7k.zip
tar -xf MVSA.zip
tar -xf MVSA_single.zip

If using zip tool, please select 'Extract Here'

Then run

python script.py

Results

After the program is done, there will be files logging the results in the output folder. Please send this output folder back to developer team.

Authors

• May Li - Code Integration - (https://github.com/MaySsssss)
• Kehao Chen - Model establishment - (https://github.com/beholder91)
• Taoxu Zhao - Fine tuning and testing -
• Yufan Lin - Version control -
• Liye Wang - File management -
• Xucheng Zhou - Team management - (https://github.com/aden1350)