Artificial intelligence techniques to develop a predictive model using textual data: a literature review

The construction industry is experiencing a data revolution, with vast amounts of digital information now available. This shift has facilitated the widespread adoption of Artificial Intelligence (AI) applications within the sector. The capability to extract valuable insights and make precise predictions has become increasingly crucial for decision-making. A comprehensive methodology for developing AI models using textual data is fundamental in this context. Therefore, this study initially examines the AI techniques used to develop predictive models to train the textual data available from construction projects. Thereby the study aims to develop an AI methodology for predictive models using textual data in construction projects. The review was carried out under four themes: Introduction to AI technologies, Emergence of digital data in the construction industry, AI Applications using textual data in the construction industry, and Proposed AI Methodology. The review identified the AI Applications using textual data under risk management, accident management, contract and compliance management, and dispute management. The research findings indicate that Natural Language Processing (NLP) is highly effective for data preprocessing and feature extraction tasks. The study suggests that combining NLP techniques with deep learning algorithms for model development could be a promising approach for developing predictive models based on textual data. An integrated approach strengthens both NLP and deep learning. NLP transforms the unstructured to structured data which the computer can understand while deep learning algorithms learn the complex patterns and relationships from the large dataset. Thus, construction professionals can utilise more accurate predictive models, leading to enhanced decision-making.