Text Message Categorization
Tolstoy helps Bitesize, a messaging startup, categorize thousands of SMS messages every month.
Bitesize is a San Francisco-based startup that helps users SMS text and respond to thousands of people at once. This is used by realtors, car dealerships, and politicians to chase leads and respond to customers right on their phone.
One of their core features is allowing users to quickly filter the most interested leads, as well as “not now, but later” leads and uninterested prospects. Since SMS messages are highly irregular—with nuanced content, misspellings, emojis, and the full spectrum of human expression—Bitesize did not have a good way to apply a set of rules to tag the messages.
Tolstoy created a model for Bitesize to categorize their messages with high accuracy, using the latest natural language processing techniques. The first deployed model performed with 90-92% accuracy. Bitesize staff provided feedback to the model to ultimately boost the accuracy to 95%+.
The model operates in real-time via API, helping Bitesize route thousands of messages each month for its users. This, in turns, helps the users land more deals and process their pipeline more efficiently.