Automated Extraction and Time-Cost Prediction of Contractual Reporting Requirements in Construction Using Natural Language Processing and Simulation

Round 1

Reviewer 1 Report

This paper proposes a Natural Language Processing (NLP)- and simulation-based method to extract the time- and cost-related requirements from construction contracts. The proposed idea is interesting, but this reviewer found the following issues to be addressed before this paper is published in the Journal. 

• The authors explain that there are two modules in the proposed system, 'extraction module' and 'prediction module'. However, this paper appears to have much more focus on the extraction module, and in the end, the reviewer was unsure how exactly the two modules work together. The authors suggest that the stochastic simulation and prediction be made possible based on historical data, but the link between the NLP-based extracted data and the simulation process is obscure in the paper. To some degree, it looks like they are two different systems (meaning that two don't require each other for them to function on their own). To address this issue, the authors would need to explain the system from a more holistic perspective. Otherwise, it is not sure how the whole system will work.
• 'Framework Application' would need to be re-titled as 'case study' or 'illustrative example', depending on what it actually was. In the current form, the paper does not clearly show whether it was a real case study or a hypothetical case study. Also, there has to be more rigour in the case study if it was to validate the proposed framework (as it appears so). 
• The results and discussion section needs improvement. Currently, it is the result of the case study, not really the result or the discussion of the entire research work done. In other words, in-depth discussion on the overall research findings, proposed system design, and results and findings from the case study is missing. In fact, these discussion points are provided in the "Considerations, Limitations and Future Work' section. This structure of paper is confusing, and it is recommended to separate the results/discussion of the case study and the discussion on the overall research. This reviewer believes that it can be relatively easily done with some effort for reorganising the content which is mostly already there in the manuscript. 
• The conclusions section can clarify what problem can be addressed by the proposed system. Again, this is a matter of how explicitly the contributions can be explained in the paper, especially in the conclusion section.  

Author Response

Thank you very much for your feedback. Please see attachment for a detailed point-by-point response to your comments.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper starts from the premise there is a problem with reporting requirement extraction from contractual documents. Understanding reporting requirements and managing them effectively is a major task for contractors although with new automated procedures on site the reporting and certification process is changing. The paper then moves on to demonstrate how various NLP and ML techniques can be used to automate this extraction. The area that the paper does not consider is that contract documents can be written through databases. Therefore the database can be used to automatically generate reporting requirements. If the contact documents are purely issued in a text form this database functionality is lost. It maybe that the structure of the contact documents needs amending so they are both human and machine readable ideally in a more rule based way. Obviously bespoke clauses can cause problems but if written using a rule based approach they can be machine readable. The volume of work required in reporting relates to the amount of instances on site. The work quantities to some extent are available through the BIM model. Although the relationship between quantities and cost is effected by location. It would be helpful to consider the role of this data source and how it integrates into the process. If the aim is to extract reporting requirements there is the question of what form of visualization or communication be conveys this to the managers and operatives.

What I am saying is that along with the NLP and ML techniques discussed as discussion should be provide on what can be achieved as considering the creation and uses of contractual documents as databases. Perhaps the way to think of all information related to construction is as linked data. The process should not be one of retrospectively creating (linked) databases construction projects should be created as integrated informational database systems from the outset whether this be relational or alternatively semantic linking. Litigation of course happens on old contracts therefore the tools you discuss could have application here. Further development of reporting can be linked to blockchain concepts of payment.

These ideas are not meant to negate the quality of the paper or the research that has been undertaken but position what has been proposed within a wider range of possible options.

In terms of the quality of research undertaken and documented the paper warrants publication but the author may wish to reflect on my comments and amend based on that reflection.

Author Response

Thank you for taking the time to review our work and share your thoughts with us. Please see the attachment for a detailed point-by-point response to your comments.

Author Response File: Author Response.pdf