Sentiment Analyser

A platform that receives SMS customer queries and responds to them accordingly in order of priority.

Comment

DurHack 2022: Sentiment Analyser

How to use

Send your query to +447480802870 via SMS, and you will get an answer through our AI chatbot, or a 'Connecting you to agent...' message. If you received a 'Connecting you to agent...' message, it means that the AI chatbot is still learning how to answer that query, and your message is passed over to an agent, which you could see on the demo website below.

What is this project?

  • A platform that receives SMS customer queries and responds to them accordingly in order of priority.

What inspired us to come up with Sentiment Analyser?

  • The problem introduced by Atom Bank on ordering customer queries
  • The initial demonstration of building an automated SMS messaging app using Twilio

Main features of Sentiment Analyser

  • Twilio
  • Twilio (AI) Autopilot
  • Firebase
  • MonkeyLearn
  • Fetch API
  • Rest API

What does Sentiment Analyser do?

  • Receives SMS messages sent to a phone number attached to a Twilio account
  • Send SMS replies using a chatbot
  • Save the SMS query to Firebase
  • Provide a REST API that's queried from a frontend website using the fetch API
  • Display customer queries on the website
  • Order the queries based on ML semantic analysis, keyword frequency detection and date created

How is Sentiment Analyser linked to the theme Unity?

  • We unify companies and customers who have specific queries through the use of Twilio
  • We unify companies with customer information
  • The text, and Semantic Machine Learning model (from the website) unifies the APIs and websites

How did we develop Sentiment Analyser?

  • Interactions via SMS (SMS chats) are done using Twilio
  • The REST API was built with NodeJS
  • The fetch API and frontend website was built with Javascript, HTML and CSS
  • ML semantic analysis was done using a third party API called MonkeyLearn

What challenges did we run into when building Sentiment Analyser?

  • We were briefly only able to receive SMS messages from one phone number
  • We ran into issues querying MonkeyLearn with the fetch API due to a lack of documentation
  • We decided to use a cloud hosting platform as we ran into issues running the Twilio endpoints locally
  • Initially we used a JSON file and the cloud hosting platform deleted it every time the server went down. To overcome this, instead of storing the data locally, we stored it in Firebase
  • We used ngrok to solve an error that we were facing in the Terminal. ngrok allowed us to expose a web server running on our local machine to the Internet.
  • We had to import the module dotenv in order to protect the authentication token and account SID

What accomplishments are we proud of as a result of Sentiment Analyser?

  • Understanding how to use and get Twilio to work without any prior knowledge
  • Developing a usable platform that provides useful functionalities
  • Learning Firebase without prior knowledge

What did we learn from building Sentiment Analyser?

  • How to implement automated SMS messaging and a chatbot with Twilio
  • How to store and fetch a JSON object in Firebase

Further Improvements

  • Expanding the program into a multi-messaging platform, e.g. email
  • A more creative design to make the platform visually more appealing

Built With

Share this project:

Updates