Whatsapp Chat Analyzer

Have you ever wondered who is the nastiest in the group chat? Which member posts the most? Who takes the longest to reply?

I created this project for learning by doing. Its sole purpose is to write my own parser and to learn the basics of data mining and data analytics. I went through several experiments for optimazations, performance increase and parsing issues (god damn you invisible characters!). Interesting statistics, visualizations and insights can be gained from the analyzed chat messages.

Feedback is very welcome - I love to improve myself!

Key Features

• Parsing of whatsapp messages based on exported chat.txt files
• auto detect Android / iOS file format
• Tokenization and lemmatization of messages
• Extract URLs, cursewords and emojis
• Determine message type: questions, replys, initiations, follow-up, media
• a representation of you chat in a Pandas DataFrame for further analysis
• some example analysis functions

Installation

After pulling use conda install -n MYENV --file environment.yml to install the environment and its packages (PIP version may coming soon :)).

If you want to run NLP you need to install Spacey language packages:

• English: python -m spacy download en_core_web_md
• German: python -m spacy download de_core_news_md

Running

You need to export your chat via Whatsapp. Go to Settings > Chat > export chat. Transfer the exported .txt file to you device and adapt the path in the jupyter-notebook.