ABOUT THE PROJECT

In this project, I designed an intelligent chat bot as part of a proof of concept for the New Jersey Courts Innovation Team. It uses the Azure Dispatcher design to incorporate multiple LUIS / QNA Maker apps for a cleaner and more robust architecture. This allows for a parent domain to predict and send the user to a child app for a more specific prediction. It also incorporates Azure Language Understanding Service (LUIS) to detect user utterances and ranks the response which will select the best intent. It also categorizes keywords and phrases into entities using regular expression which helps to provide a correct response. The bot service also uses Azure Search to search specific indexes that are related to the intent. This was done by using Azure Functions to created custom documents with specific key value pair and were pushed into the index through the Azure Search API. The bot also utilizes Azure's QNA Service to answer basic questions that do not require additional dialog. In the case of a NONE intent, the application will use our Elastic Search engine to fetch a response throughout all of our websites.

FEATURES

• Azure Bot Framework with Node.js / JavaScript SDK
• Language Understanding Service (LUIS) to detect users utterance and intents
• Uses Azure Search with content specific indexes to return relevant information
• Azure Functions to push documents into index on a daily schedule
• Elastic Search returns response when intent rankings are too low
• Storing data into Azure Blob Storage for easy integration into Azure Search
• Queries multiple cross domain JSON files in certain intent use cases and returns structured data responses

MADE FOR

Innovation Team