TelegramWordCloud

Generate word cloud from Telegram chat history.

1. Export Telegram chat history in JSON format from Telegram Desktop.

2. Clone this repo:

   > git clone https://github.com/Ray-Eldath/TelegramWordCloud

3. Install dependencies:

   > pip install numpy cython pkuseg click

4. Run scripts:

   1. main.py: Extract valid sentences from result.json. You could specify ignored sender by specifying option –i. See --help for more explanation.
   2. seg.py: Segment every sentences with pkuseg. You should provide a user dictionary file dict, in which every line is a word in your dictionary. See pkuseg # 参数说明 on how to define your own dictionary or model.
   3. cloud.py: This script will transform the words (which are segmented from sentences in the previous step) into weighted CSV file that could be visualized by WordClouds.com. There are several modes and corresponding parameters to tune, check out --help for more information.

5. Upload the generated CSV file to WordClouds.com (Word list –> Import CSV) and customize your graph.

Notice

• This repo uses pkuseg for word tokenization. There are many other options, including jieba and SaaS such as Aliyun NLP 2.0. You could try different services and models to get the desired result. Be aware: user-defined dictionary is essential.
• Cover image is generated with default parameter set (n=100, mode=group-freq, groups=6). The default group-freq mode can generate the best visualization IMHO, but you can try them out on your own.