SentiFootball

This library performs sentiment analysis from a graph theory perspective on fan's tweets before a football match.

Python packages

SentiFootball needs the next libraries to work:

• Emoji - Emoji library for python
• Matplotlib - 2D graphics library
• Networkx - Complex networks package
• Nltk - Natural language processing with python
• Numpy - Scientific computation package
• Scipy - Mathematics and science environment
• Seaborn - Data visualization library
• Stanza - Stanford NLP package
• WordCloud - Word cloud generator with python

Environmental Vars

export TW_KEY=""
export TW_SECRET=""

Files Structure

├── GB
│   ├── 2019
│   │   ├── 38
│   │   │   ├── 1595772340_ARSWAT_neutral.json
│   │   │   ├── 1595772552_ARSWAT_positive.json
│   │   │   ├── 1595772607_ARSWAT_negative.json
│   │   │   ├── 1595773332_BURBHA_negative.json
│   │   │   ├── 1595773344_BURBHA_positive.json
│   │   │   ├── 1595773360_BURBHA_neutral.json

Data Retrieval

Querying Twitter before a match can be performed by running.
python search_twitter.py --home_team_id --away_team_id --week

python search_twitter.py CHE LIV 2

Note: GB/GB_twitter.json contains the 3 letter code of the teams.

Dataset for Tweets

This command creates a whole dataset of all tweets in a week.
python maps_tweets.py --country --season --week

python maps_tweets.py GB 2020 2

Dataset Extras

This command creates a clean dataset and classifies tweet's polarity and support to a team.
python process_tweets.py --country --season --week

python process_tweets.py GB 2020 2

Frequencies Visualization

visualization_sentiment.ipynb contains some visualizations of the weekgame.

Graph Visualization

graph_sentiment.ipynb contains some visualizations of the weekgame from a graph theory perspective.