Souza, Douglas Lopes de (2024) Natural language processing of technical standard for the construction sector the study of the nbr 15.575. PhD thesis, Universidade Estadual de Campinas, Brazil.
Abstract
ABNT NBR 15.575 is one of the main mechanisms available to encoraje the quality of housing in Brazil and has been implemented since 2013. Currently, the national construction industry is mediated by a large number of regulations, complicating the alignment between them and causing possible conflicts in their application. In addition, regulations are complex documents that require interpretation by experts because they present various parameters that must be verified throughout the building's lifecycle. The regulations are prepared in natural language requiring manual analysis, prone to errors, and demanding a high level of expertise for programming the rules on platforms dedicated to automated compliance checking. National research related to the theme is limited to isolated attempts at manual conversion of the standard's objective guidelines into rules for specialized conformity checking platforms. This research aims to create a rule-based information extraction method for the automatic translation of the Performance Standard (PS) into an open, semi-structured format that can be computed or incorporated into automatic rule verification platforms. The solution is classified in the field of Natural Language Processing of Artificial Intelligence. The research adopts the Design Science Research method to guide the production of a sequence of contributions. The result includes a rule-based information extraction algorithm generating a semantic model representation of the PS. The algorithm comprises various Natural Language Processing techniques based on portuguese linguistic patterns. The semantic model of Part 1 of the Performance Standard is expressed in formal, semi-structured language in Extensible Markup Language, which can be processed in applications related to conformity verification and translated into Information Delivery Specification. The accuracy of the algorithm indicates the need for incorporating new methods based on artificial intelligence, highlighting the Retrieval-Augmented Generation technique in large language models. The research dismissed the adoption of an ontology to increase the performance of text analysis and mapping algorithms given the absence of an ontology for the construction context in portuguese or the inadequacy of adopting generic or specific ontologies in english. Thus, a brazilian gap was precisely identified. The contribution and novelty of the research lie in the algorithmic formalization of auditable NLP of requirements in portuguese in neutral and open language.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Thesis advisor: | Ruschel, Regina Coeli |
| Uncontrolled Keywords: | natural language processing; information extraction; regulations; standards |
| Index terms: | large language model, Brazil, accuracy, text analysis, ontology, compliance, design science research, specification, construction sector, housing, mapping, regulation, documents, programming, conversion, platform, information delivery, construction industry, artificial intelligence, lifecycle |
| Subjects: | project delivery, data science, artificial intelligence, information systems, political science, industry analysis, professional development, contractual condition, education and knowledge transfer, design practice, analytical methods, programming, construction type, digital design, Geography, health safety and environment, spatial and geospatial analysis, manufacturing engineering |
| Topics: | Engineering Principles, Project Management, Geographical Context, Health and Safety, Governance, Construction Technology, Information Management, Research Practice, Contract Administration, Digital Applications, Design Practice |
| Descriptive scope: | 4 PCTA |
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