Research data supporting 'Comparing Natural Language Processing Methods to Cluster Construction Schedules'

The names of construction activities are the only unstructured data attribute in construction schedules and often guide construction execution. Activity names are devised to communicate between stakeholders, and therefore are often written using inconsistent terminologies across repetitive activities with omitted contextual information. This the challenge that machine learning systems face to learn patterns from construction schedules. This paper compares the performance of state-of-the-art text-related clustering methods in identifying repetitive activities. This was achieved via creating a ground truth dataset on the basis of the construction work classification in the Standard Method of Measurement, and then compared the precision, recall and F1 score of Latent Semantic Analysis (LSA), Latent Dirichlet Allocation (LDA), word2vec, and FastText to group activity names in 27 construction schedules. Results indicate that the F1 score of LSA outperforms LDA (0.84% over 0.88%), while the results of language-models-based clustering depend on the quality of word embedding and the paired clustering method. The results of this study provide an insight into how to pre-process activity names of construction schedules for further AI-based quantitative analysis. This is the python executable file to execute the clustering method described in the manuscript. See the main manuscript for more details.