The Fifth Elephant 2019

Gathering of 1000+ practitioners from the data ecosystem

Up next

Text Classification, Interpretability, and Summarisation at Scale


Siddhant Panda


The Freshdesk product is used by over 150,000 customers for resolving customer support tickets. Each customer configures workflows within the product that are specific to their approach to ticket resolution. Traditionally, these use a hand-tuned rule-based system that serves well when a support organisation is relatively small. However, as businesses scale and customer needs become more complex, rule-based systems are unable to keep up, resulting in increases in issue resolution times and a drop in customer satisfaction.
In order to enable our customers to meet increasing customer expectations and reduce unnecessary manual effort, we have designed an NLP system that:
1. Automatically routes & prioritises tickets to the right support agent using historical data
2. Reduces dependence on a rule-based system
3. Helps agents and administrators understand why a particular prediction was made


  1. An overview of ticketing systems and rule-based routing
  2. Various approaches we looked at to solve routing and prioritisation
  3. Scaling the data science pipeline to 70,000+ models
  4. Building interpretability models to help users understand the reason for predictions
  5. Future Directions - we would like to take this further for text summarization where we would shorten the document that would provide a brief overview of it.

Speaker bio

Data scientist at Freshworks