Function Calling for Supporting Large Language Models in Question Answering

The recent advancement of function-calling capabilities in Large Language Models (LLMs) offers new opportunities for improving question answering over structured technical data. In Facility Management (FM), operational building information is often stored in Building Information Modeling (BIM) systems, databases, and Common Data Environment platforms. However, these data sources are often difficult for non-technical users to access because they require knowledge of specific data structures, software interfaces, or query logic.

Vairavipillai, Visnuciry, 2026

Type of Thesis Master Thesis
Client
Supervisor Hanne, Thomas
Views: 2
Share
This thesis investigates the design, implementation, and evaluation of a prototype function-calling LLM assistant for structured BIM-based information retrieval in Facility Management. The prototype enables users to submit natural-language questions, which are translated by the LLM into controlled function calls connected to structured building data. The study is conducted within the Smart Building Information Service (SBIS) context and focuses on operational FM information provided through the LIBAL data environment.
Following a Design Science Research methodology, a prototype artifact was developed and its performance was evaluated using predefined facility management query scenarios. The evaluation considered answer correctness, function-call correctness, parameter accuracy, response time, and error types, and compared two freely accessible LLMs, Qwen3-32B and LLaMA 3.1-8B, with respect to their function-calling capabilities.
The contribution is a clearer understanding of how function-calling LLMs can support controlled, traceable, and reliable access to structured building information.
Studyprogram: Business Information Systems (Master)
Keywords
Confidentiality: öffentlich
Type of Thesis
Master Thesis
Authors
Vairavipillai, Visnuciry
Supervisor
Hanne, Thomas
Publication Year
2026
Thesis Language
English
Confidentiality
Public
Studyprogram
Business Information Systems (Master)
Location
Olten