Sajadfar, Narges (2022) Application of machine learning to automate classification and information extraction in industrial construction documents. PhD thesis, University of Alberta, Canada.
Abstract
Abstract: Industrial construction projects are usually mega-projects that involve millions of labour person-hours and generate hundreds of thousands of documents. Construction documents represent a vital source of information and knowledge regarding the project scope. Documents come in different types and include structured information such as data tables and unstructured information such as text, images, and drawings. The documents may consist of contract forms, drawings that define the quantities and qualities of materials, standards, and specifications required to carry out the project. Documents usually involve multi-versions and address different systems in a project, such as architectural, structural, electrical, and mechanical systems. The ability to extract and organize structured and unstructured information from these documents is a time-consuming process that is critical for effective project control and decision making. This task is more challenging and labour-intensive when documents are provided in image formats requiring human intervention to extract the required information. The objective of this research is to address this challenge by introducing an automated approach for managing and extracting information from construction documents. This research describes the development of automatic classification and information extraction based on both the text and images in industrial construction documents. The development of the proposed method includes the testing of various deep learning classification algorithms, to identify suitable models for construction documents. The results of the research confirmed the effectiveness of machine learning algorithms for classifying and extracting information from unstructured construction documents with limited text. This dissertation makes a major contribution by presenting a high-precision classification approach for construction documents that incorporates scanned images, with different sizes and resolutions. Furthermore, the method of automatic title block detection was demonstrated for unstructured construction documents in this research.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Thesis advisor: | Yasser, Mohamed |
| Uncontrolled Keywords: | information extraction; construction documents; classification |
| Index terms: | industrial construction, project scope, testing, mechanical system, documents, decision-making, project control, machine learning, drawing, resolution, specification, effectiveness, deep learning, dissertation |
| Subjects: | control systems, research dissemination and communication, building construction, decision analysis, conflict resolution, technical documentation, performance management, artificial intelligence, professional practice, mechanical systems, contractual condition, professional development, scope management |
| Topics: | Contract Administration, Research Practice, Stakeholder Management, Project Management, Quality Management, Risk Management, Digital Applications, Design Practice, Engineering Principles, Information Management |
| Descriptive scope: | 2 PC |
N.B. Descriptive scope is a count of how many of the five facets of empirical research are indicated by the words used in title, abstract and keywords. It is not intended as a judgement on the research; merely a count of the kind of word we would expect to indicate Phenomenon, Concepts, Theoretical framing, Empirical techniques, Analytical techniques. If all five are present, then a code of “5 PCTEA” will indicate this. If you feel the coding for this record is questionable, we welcome discussion around the terms we matched or the way we categorized them. The facet you would expect may not be coded, or a facet may be coded inappropriately. This can also bear on a larger question, of which facets should be treated as defining in construction management research. Please get in touch, and we will look at it. More details here