TexPrax - Responsible Use of intelligent Text Analysis Methods to enhance Problem-Cause-Solution Processes in Industry

This is a small demonstrator to show how the data collection could work. To save some resources, the underlying code has not much logic into it. However, the predictions are made by the fine-tuned GermanBERT models we used in our paper.

The purpose of the bot is to keep track of a conversation and log problems (Problem), causes (Ursache), and solutions (Lösung). To do so, it will try to classify each message in the chat and ask for feedback via reactions. You can either confirm (Ja) or correct the predictions which will then be stored in the database.

The dashboard demonstrates how recognized messages could be sent to and stored at a different application. To try out the demonstrator, you will have to "simulate" a conversation between multiple users and provide problems, causes, and solutions on your own. To start chatting, click the chatterbox button on the bottom right of the screen. After chatting for a while, hit the "refresh" button on the dashboard to see what the bot has logged so far (only the last five problems including linked causes and solutions will be shown).

To help you get started, here is an example:

• Ich hab ein Problem mit der Index!
• Ist der Stecker eingesteckt?
• Ah ja, das Kabel war locker. Ich hab es fixiert.

For further information please check out our website!

Note: the message client in this demonstrator uses plug-ins that can experience issues if your browser has ad-blockers installed (the client can be run separately and is solely for demonstration purposes). If you experience issues with the client, please refresh the page (e.g., by pressing F5).