Semantic Network Analysis as a Knowledge Representation and Retrieval Approach Applied to Unstructured Documents of Construction Projects

Reusing knowledge from past projects is a critical task in construction, given the increasing complexity in such projects: numerous stakeholders, a multi-disciplinary domain, and multi-objectives besides the traditional ones such as cost and schedule. Unstructured data, such as progress reports and minutes, is a rich source of knowledge that can be revisited in projects as the contextual nature of documents permits describing the nuances of the interrelations and uniqueness of each project. However, texts are difficult to formalize in a way that the process of retrieval and analysis of relevant knowledge be automated using computers. In this paper, we present an innovative approach that encompasses formalization, retrieval, analysis, and reuse of knowledge from case studies of past construction projects. Assuming that energy is an objective, the cases are represented as a network of common concepts found in every project. The nodes are the concepts, and the links between them are established whenever there is an association between two concepts that affects the energy use in construction. Conversely, the comments of team members of a current project can also be captured and represented using the same standardized set of concepts. Using network analysis, we can retrieve the most relevant cases, which are similar to the current project, study the most important concepts, extract clusters of concepts, and capture the nuances of the cases in a more objective way. A concept map based on the literature and three case studies of past oil and gas projects are developed to undertake this approach. We evaluate the method by simulating the collaborative environment of one of the cases through the participation of ten volunteers in Green 2.0, an online media to discuss construction projects. At the end of the test, we perform a correlation between the networks of the test and the case study.