Congratulations to Assistant Professor Kim JinWoo on the award of MOE T1 (Call 2/2022) Award

Project Write-up:

The construction industry is gearing up for a digital transformation of project information, knowledge, and decision-makings. This transformation has become inevitable and prevalent because the successful delivery of construction projects relies heavily on knowledge transfer and management. However, project information and knowledge have still been stored and shared with stakeholders mainly in printed and/or scanned copies of text-written documents. This practical issue has been a major obstacle to the digital transformation of the construction knowledge and information.

To address this issue, this research explores a pioneering way of transforming a non-digitalised construction project document (e.g., printed and scanned copies) into a fully digitalised, computer-understandable format.

This research will therefore test a hypothesis that a complex construction project document can be digitalised by computer vision (CV) and natural language processing (NLP), answering following three research questions:

(i) Can we detect construction document objects and recognise their semantic layout?;
(ii) Can we detect construction-specific textual words in an unstructured document image?; and
(iii) Can we segment and reconstruct a complete sentence from a corpus of words with noisy detection results?

Answering these three questions can help unlock the full potential of CV and NLP approaches in digitalising construction project documents. The findings of this research will open up a new possibility of CV and NLP for construction document digitalisation. This can also put a valuable steppingstone for future research in the field of automated construction document analytics and knowledge discovery. Furthermore, the proposed approach can stimulate knowledge transfer and management in practice, while significantly saving time and resources required for project document retrieval and management.