Mentor: David Carlson and Kevin Gao, Comm100
Project Room: ORCH 4062
Natural Language Processing (NLP) is being dramatically enhanced by AI and machine learning. We believe NLP is at the tipping point, about to reach a level of performance accuracy that makes widespread commercial applications more dependable. The challenge is in quickly training bots on common lexicon, intent, and reaction. The data set provided includes thousands of online chat sessions between support and sales agents of a company and their customers or online web visitors. Each chat log includes [chat transcript, duration times, and customer rating]. We’ve identified some business issues/use cases with suggested solutions for you to solve but you are not limited to the following issues or solutions (get creative! See things we don’t see in with the data that would provide real business value):
- Cluster or Correlate chat sessions:
- Build an algorithm or methodology to cluster or correlate similar issues across chat sessions defined by you
- We have a Knowledge Base (KB) tool but it is manual work on the client side
to create Q&A’s. Is there a way to progress towards full automation in building
a Knowledge Base with what’s already there?
- Build a KB comprised of questions and their corresponding answers in an automated way.
The goal here is to see if the students can: a) figure out a mathematical model for determining intent, and b) develop a tool that can be used to repeat that process with new data sets.
Comm100 is the global provider of enterprise-level customer service and communication solutions. As our motto “100% communication, 100% success” indicates, we believe that good customer communication can make a difference for your business, helping you build stronger customer relationships and achieve a better business performance.