Railroad Trespassing and Safety: A Systematic Analysis of Twitter Data

https://doi.org/10.1016/j.cstp.2024.101154

Overview

This is the completed version of codes for the paper "Railroad Trespassing and Safety: A Systematic Analysis of Twitter Data"

Data

This folder contains code data_collection.ipynb is used in data collection and saved in csv file format.

Data Processing

The code preprocessing.py is used to preprocess data and save it in csv file format.

User analysis

Simple Python codes for data analysis.

Topic modeling, Sentiment and Emotion prediction

We used some existing models for topic modeling, sentiment, and emotion prediction.

Topic modeling: "https://github.com/MaartenGr/BERTopic"

Sentiment prediction: "https://huggingface.co/cardiffnlp/twitter-roberta-base-sentiment"

Emotion prediction: "https://github.com/nikicc/twitter-emotion-recognition"

Results

All numerical results are created in csv format

Description of steps

The railroad-related twitter data analysis is divided into six parts:

1. Data cleaning process: Remove tweets that are not related to railroad, rail, or safety.
2. User analysis: Identify users and the type of users involved in posting rail safety-related tweets worldwide. Group the users in different subgroups.
3. Topic modeling: Extract topics of rail safety-related tweets.
4. Sentiment analysis: Identify the sentiments(pos,neg,neu) of the tweets.
5. Emotion Analysis: Identify the emotional (joy,..) of tweets.
6. Hashtags and Mention analysis: Extract hashtags and mentions from tweets and utilize that information for extracting organizational and geographical information.