Master's Thesis Michael Legenc

Using Natural Language Processing and Machine Learning to Assist First-Level Customer Support for Contract Management

Check24 is a comparison portal, whose customer support is crucial to its business model. At peak times 1.800 incoming emails are recorded per day, which entails a corresponding supply of manpower to process them. 

These emails have recurring topics of which one can probably draw a profit from by automation. Especially, questions regarding bonus payments, duration and terms and conditions of available or concluded contracts repeat themselves in the domain of check24´s energy department. A bot with artificial intelligence using text mining approaches could automatically answer a part of these mails by enriching boilerplates for the answers with personal data and information from the email´s context.

Thus, the purpose of this research is to find an appropriate approach to cluster emails by their semantics, to estimate the correctness of their assignment to a cluster and then to collect enough data for the answer to be generated. These clusters and their emails´ estimations are required for selecting emails, that are suitable for automatic processing without failing the well-known Turing test. This test checks whether a human is capable of telling if an answer has been written by a computer or a human.

All of this will be applied in the context of check24´s energy department to gather knowledge about applicability and in how far both, customers and economic viability, can benefit from it.