News

Vortrag von: Normix Hillemann
Titel: NLP Techniques for Summarizing and Topic Matching in Patient-Chatbot Interactions: A Comparative Study
Zeit: 24.03.2025 - 10:00 - 11:00
Ort: WE5/05.013 und Teams
Abstract:

This paper addresses the research question of how well NLP techniques can perform in matching relevant topics from patient-chatbot interactions to help doctors prepare for consultations. For this purpose, a patient-chatbot to complement patient education in the area of MRI was used as a foundation of this thesis. A comparative approach was chosen to answer this question. First, the use of NLP and LLM's in the medical field, especially radiology, was analyzed, then a selection of various models from the field of statistics, embeddings and LLM-based AI was made and looked at in more detail.

The performance of the various models in this use case was measured using the precision, recall and F1 score metrics, in each case with a synthetic dataset of artificially generated chat logs and a dataset of chat logs from people in a real-world scenario. An editorially created list of topics based on an MRI information sheet was used as ground truth for this work. A model was selected from all the models for a prototype. 

This work was created in cooperation with the company Thieme Compliance GmbH under the supervision of Nico Benz.

Keywords: Natural Language Processing (NLP), Large Language Model (LLM), Applied Computer Science, Medical Information System, Human-Computer Interaction, Chatbot, Topic Summarizing, Topic Matching

Information to participate via Teams:

Microsoft Teams Benötigen Sie Hilfe?

Jetzt an der Besprechung teilnehmen

Besprechungs-ID: 327 087 611 349

Kennung: zd36TA6F