Desire: Deep Semantic Understanding and Retrieval for Technical Support Services

Abhirut Gupta, Arjun R. Akula, Gargi Dasgupta, Pooja Aggarwal, Prateeti Mohapatra

Abstract: Technical support services involve enterprises providing after-sales support to users of technology products. The current support structure is labor intensive with practitioners manually consulting support documentation to troubleshoot users’ problems. We propose a cognitive technical support system as one that: (a) can understand technical problems expressed by users, (b) can automatically provide relevant resolution information and (c) can learn and improve its understanding and resolution over time. A typical technical problem description contains a combination of symptoms experienced by the user, explanation of attempts already made to resolve the problem, and sometimes, a clear expression of the requirement to solve the problem. Handling such intricate descriptions is outside the scope of current retrieval based systems and requires a deep understanding of the problem, combined with reasoning over a knowledge graph.