Automated customer feedback analyzer
An automated feedback analyzer that allowed a prominent travel agency in my city to automatically segregate positive and negative feedback, bypassing the time and resources spent doing the same manually,
CLIENT
My Travel Junction
team members
2
duration
1.5
months
tech stack
Flask, Python, Firebase real-time database and Firebase messaging
THE FEEDBACK PROBLEM
Our client - My Travel Junction had a problem with handling customer feedback. With the limited people they could employ, they found themselves often spending a little too much time going through customer feedback, extracting useful information from them and then designing strategies to address them. They were looking for a solution that could help them save time and effort and help them understand their customer feedback considering that in customer-centric business feedback is an immensely important thing.
Solution objectives
Based on the requirements of our client, the following solution objectives were identified:
-
Segregate feedback based on sentiments
-
Dashboard to visualize data
-
Extract keywords from all feedbacks
Control flow of our solution
The components
1. The mobile application:
The users submitted their feedback on the trip through the client's official mobile application. The feedback was mostly a point-based rating of each aspect of the trip and also had a text-box for general comments. It is this feedback that our system analyzed.
2. The Database:
The feedback was stored in a Firebase realtime-database, our system accessed this data to pull feedback and then analyzed it accordingly.
3. Feedback collector and classifier:
This was a Flask server that pulled data from Firebase, passed it through a Naive Bayes classifier (to classify it as positive or negative) and exposed REST APIs for front-end use.
4. The Dashboard:
This was the front-end that invoked the REST APIs to get the necessary analysis results. The data collected was presented to the user and also visualized with the help of charts.
The dashboard
Pick the city for which the feedback has to be seen
The feedback is pulled by the server, is classified as positive or negative and displayed on the dashboard by invoking the necessary REST APIs.
In addition to displaying the text of the positive and negative feedback, there were also charts that visualized the distribution of the sentiments of the feedback.