This is a small dataset created to evaluate the performance of Question Answering (QA) for large-scale language models in the field of civil engineering. The dataset is targeted at the field of bridge design, and was created using the document (Survey on the common issues of bridge design (2018-2019) (Technical Note of NILIM No.1162)) used in bridge design projects in Japan. The dataset consists of 50 pairs of QAs, where each pair consists of a question asking about the content of a document and an answer extracted from the document associated with the question.
Each column of the csv file shows the following data.
Column 1: ID of the QA
Column 2: Referenced page (page number of the full pdf)