Project: Dialogue Act Modeling
Description
At KPN we collect the transcriptions of our customer contacts with our Customer Services centers executed via the chat and call channels. The structure of such dialogues is made by a number of classifiable parts some of which always occur, for example greetings, customer identification, problem description, follow procedure, plan follow-up, etc.
The goal of this thesis is to build a classification model to recognize such structural parts for turn within each dialogue of the Consumer Market customer contacts (~20K per week).
Such information is used to enable Process Mining, by analyzing bottlenecks and loophole across the dialogue structural parts, and to feed back relevant information to the CRM system thru a real-time streaming event platform (Kafka).
The training dataset is built via a weak-supervision approach (see snorkel.org) and the classification model is built by fine tuning BERTje with available chats and calls transcriptions in Dutch.
Details
- Supervisor
-
Mykola Pechenizkiy
- Secondary supervisor
-
KBKPN: Gianluigi Bardelloni
- Interested?
- Get in contact