Servicebot is a chatbot trained off of a custom question and answer dataset
ServiceBot works with any input or customer query
Representation of the input dataset calculated using the expert.ai relevant terms sdk.
Computing the dot product similarity between the user query and known facts/answers.
No match example
Create a customer service bot from any CSV or question/answer dataset.
Built for the Sentiment & Opinion Mining hackathon.
Powered by expert.ai for information extraction and bot dialogue.
Add your expert.ai credentials following the steps here: https://github.com/therealexpertai/nlapi-python#authorization.
Run the below commands:
pip install -r requirements.txt uvicorn main:app --reload
yarn yarn start
To get started, upload a csv on the front end that has your question database in the two column format: question,answer. See
data/example.csvas an example.
Once a csv is uploaded, the data will be saved in a file
server/db.json. To create a bot based on a new dataset, delete this file and refresh the website.
The bot uses expert.ai's relevant term resource analysis to match customer inquiry to a closest matching result from a dataset. If the bot is unable to find a remote match it will display a generic error message. The bot will work more effectively in more scenarios with a larger csv or connected dataset.