A web application for gathering business intelligence using Twitter.
A good company is more than a good product. Technologically advanced distribution methods are critical for success.
We allow you to gather data on what people are saying about your competitor's products and reach out to their disatisfied customers.
Getting data to identify complaints about a specific entity is difficult. To solve these challenges, we use a boostrapped semi=supervised approach to training.
Tweets are filtered and initially ranked based on text sentiment. Any tweet with a negative sentiment score is identified as a possible complaint.
As part of gathering data, each tweet is provided a 768 dimentional embedding from basilica.ai. As the user gives feedback, PCA and Logistic Regression are used to create a classifier based on these embeddings.
The tweepy
library is used to search for recent tweets by keyword.
The application runs on Heroku and a Heroku Postgres instance.
The web application was written using the Flask framework. The SQLAlchemy
library is used as an ORM.