Browsing by Author "Pulkkinen, Tapio"

Filter results by typing the first few letters
Now showing 1 - 3 of 3
  • Generative AI for identifying conflicts in construction industry documents
    (2024-08-19) Pulkkinen, Tapio
     Insinööritieteiden korkeakoulu | Master's thesis
    The construction industry has the potential for significant advancements by utilizing knowledge from textual data such as contracts and meeting minutes in construction projects. This could be transformative as it could enhance the knowledge of industry personnel cost-effectively, streamline project management, facilitate better knowledge sharing, and boost both productivity and decision-making. However, more than 80% of this data is unstructured and typically re-quires extensive manual analysis. The emergence of large language models (LLMs) in the 2020s has transformed this process by reducing the need for manual labour. Therefore, this thesis evaluates the use of LLMs to identify conflicts in construction documents, an essential factor in improving project outcomes. Conducted under Aalto University's Building 2030 project in spring 2024, the study investigates the capabilities of LLMs in detecting conflicts in unstructured text. The study utilised authentic data from Finnish construction projects, which was preprocessed through pseudonymisation to protect sensitive information. The research examined GPT-4 to evaluate its conflict detection abilities using various approaches: standalone use, semantic search, and knowledge graph techniques. The findings reveal that although LLMs can identify certain conflicts, they frequently misclassify non-conflictual data as conflicts, demonstrating a lack of industry-specific knowledge. Neither semantic search nor knowledge graph techniques offered improvements, as they were found to be more complex and expensive. The results indicate that LLMs currently lack the reliability needed for conflict detection in construction documents without further refinement. Nonetheless, they hold potential for extracting useful information. Future work should aim to enhance LLMs' industry-specific accuracy by fine-tuning and developing better methods for managing extensive textual data. In summary, while LLMs perform well in straightforward tasks, they struggle with more complex applications like conflict identification in construction texts. This study lays groundwork for future efforts to enhance LLMs capabilities in the construction sector, which could lead to more effective and precise information management.
  • Kokeita sorvipöllien koneellisella keskittäjällä
    (1953) Pulkkinen, Tapio
     Helsinki University of Technology | Master's thesis
  • Tekoälyn hyödyntäminen rakennusprosessissa
    (2022-04-29) Pulkkinen, Tapio
     Insinööritieteiden korkeakoulu | Bachelor's thesis