ServiceBot

• 

  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

ServiceBot

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.

See slides

https://docs.google.com/presentation/d/1fLPORsXEtYSHxpIQdB-LL4HICs6gL5utpdlWhnvobbw/edit?usp=sharing

Testing

Backend server

1. Add your expert.ai credentials following the steps here: https://github.com/therealexpertai/nlapi-python#authorization.

2. Run the below commands:

   pip install -r requirements.txt
   uvicorn main:app --reload
   

Frontend service-bot

yarn
yarn start

1. 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.csv as an example.

2. 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.

3. 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.

Built With

• expertai
• python
• react