Python Instant Messenger Analyser

Analyze your instant messaging behaviour. Exported WhatsApp Chat .txt files are parsed into a pandas dataframe, which is also enriched by feature columns.

Getting started

Installation

1. Install poetry, if not already installed.
2. git clone https://github.com/fabitosh/chat-analyzer.git
3. cd chat-analyzer
4. Install project dependencies poetry install
5. Export your WhatsApp Chats.
6. Configure chat_analyzer/__init__.py to your needs. I work with a gitignored data/ folder within the project.
7. Run main.py
8. View basic chat analysis in data/visualiyed/ for each chat as html files. A dash application is wip.

Structure

main.py
chat_analyzer
├── __init__.py                     # Configuration
├── analysis
│   ├── __init__.py
│   └── analysis.py                 # Module for conducting analysis on chat data
│
├── data_processing
│   ├── __init__.py
│   └── load.py
│       ├── extract.py              # Module for parsing raw chats and merging consecutive messages
│       └── feature_engineering.py  # Define feature columns
│
├── utils
│   ├── __init__.py
│   └── data_definitions.py
│
├── visualization
│   ├── __init__.py
│   └── visualize.py
│
├── data
│   ├── raw                         # Raw chat exports
│   ├── processed                   # Parsed and enriched entire dataframes as pickle.
│   └── visualized                  # Basic chat visualization html files
│
├── notebooks                       # Explaratory notebooks can go here.
└── tests

Thoughts and Notes

Load

Every chat should be parsed into a RawChat DataFrame. If you are chatting to the same person through multiple messengers, the possibility to concat/combine two RawChat should be an option if desired.

WhatsApp

Manually export every chat into a .txt file. As of this writing, this was only possible from the phone. Dump all chats of interest into a folder and set said path as PATH_WHATSAPP_MSG .

Signal Messages

Parsing is still to be implemented. https://github.com/tbvdm/sigtop

Analysis

Analysis can happen on multiple layers:

• Analysis which every message/row has.
  • Statistics derived in the context of the chat should be included in analysis.extract_single_chat_features(). Example: Time it took to reply
  • Metrics that only need the row for context are put to analysis.add_features(). Example: Number of message symbols
• Analysis of one chat behaviour:
• Analysis of the sender's instant messengers as whole