How Will Your Tweet Be Received? Predicting the Sentiment Polarity of Tweet Replies

Overview

• This is the official repository of the paper How Will Your Tweet Be Received? Predicting the Sentiment Polarity of Tweet Replies.

• Paper: https://ieeexplore.ieee.org/document/9364527

• Paper DOI: 10.1109/ICSC50631.2021.00068

• The paper comes with a public dataset, RETWEET: https://kaggle.com/soroosharasteh/retweet/

• Dataset DOI: 10.34740/kaggle/ds/736988

• The presentation video: https://youtu.be/YXu_BuJsoKw

• The presentation slides: https://github.com/tayebiarasteh/retweet/blob/master/Presentation_main.pdf

Introduction

What is Sentiment Analysis? It is a machine learning technique that automatically analyses and detects the sentiment of text. Here we do Message-level sentiment analysis, i.e. we classify the sentiment of the whole given message (tweet or reply).

Goal of this project

The aim of this project is to build a sentiment classification tool which, given a tweet predicts how likely it gets positive, negative or neutral replies. In other words, it is a tool to predict the nature of replies to a tweet is most likely to get.

Getting Started

Prerequisites

The software is developed in Python 3.7. For the deep learning, the PyTorch 1.3.1 (including TorchText 0.5.0) framework is used.

Main Python modules required for the software can be installed from ./requirements in three stages:

1. Create a Python3 environment by installing the conda environment.yml file:

$ conda env create -f environment.yml
$ activate SentimentAnalysis

2. Install the remaining dependencies from requirements.txt.

Note: These might take a few minutes.

3. We have also made use of spaCy to tokenize our data. To install spaCy, follow the instructions here making sure to install the English models with:

$ python -m spacy download en

Code structure

1. Everything can be ran from ./main.py.

• Set the hyper-parameters and model parameters here.
• main_train(), main_test(), and main_manual_predict() run codes related to Standard Tweet Sentiment Polarity Classification.
• The data preprocessing parameters and directories can be modified from ./configs/config.json. Note that, you have to stick to the names specified in the ./configs/config.json and name your data accordingly.
• Also, you should first choose an experiment name (if you are starting a new experiment) for training, in which all the evaluation and loss value statistics, tensorboard events, and model & checkpoints will be stored. Furthermore, a config.json file will be created for each experiment storing all the information needed.
• For testing, just load the experiment which its model you need.

2. The rest of the files:

• ./models/ directory contains all the model architectures and losses.
• ./Train_Test_Valid.py contains the training, validation, and the inference processes.

Our pre-trained models for reply prediction

• BiLSTM
• CNN

In case you use this repository, please cite the original paper:

S. Tayebi Arasteh, M. Monajem, V. Christlein, P. Heinrich, A. Nicolaou, H.N. Boldaji, M. Lotfinia, S. Evert. "How Will Your Tweet Be Received? Predicting the Sentiment Polarity of Tweet Replies". Proceedings of the 2021 IEEE 15th International Conference on Semantic Computing (ICSC), Laguna Hills, CA, USA, January 2021.

BibTex

@inproceedings{RETWEET,
  title = "How Will Your Tweet Be Received? Predicting the Sentiment Polarity of Tweet Replies",
  author = "Tayebi Arasteh, Soroosh and Monajem, Mehrpad and Christlein, Vincent and
  Heinrich, Philipp and Nicolaou, Anguelos and Naderi Boldaji, Hamidreza and Lotfinia, Mahshad and Evert, Stefan",
  booktitle = "Proceedings of the 2021 IEEE 15th International Conference on Semantic Computing (ICSC)",
  address = "Laguna Hills, CA, USA",
  pages = "356-359",
  doi = "10.1109/ICSC50631.2021.00068",
  url = "https://ieeexplore.ieee.org/document/9364527/",
  month = "01",       
  year = "2021"
  }