Named Entity Recognition Algorithm for iBISDS Using Neural Network: Fid-Bau Portal

Named Entity Recognition Algorithm for iBISDS Using Neural Network

Wang, Ning / Issa, Raja R. A. / Anumba, Chimay J.

Conversational Artificial Intelligence (AI) systems have become more and more popular to provide information support for human daily life. However, the construction industry still lags other industries in developing a conversational AI system to support construction activities. The developed intelligent Building Information Spoken Dialogue System (iBISDS) is a conversational AI system that provides a speech-based virtual assistant for construction personnel with considerable building information to support construction activities. The iBISDS enables construction personnel to use flexible spoken natural language queries instead of detecting exact keywords. To build an iBISDS, it is necessary to understand the intents of natural language queries for building information. This research aims to develop a named entity recognition (NER) algorithm for iBISDS to recognize and classify keywords within natural language queries. A dataset with 2,008 building information-related natural language queries was developed and manually annotated for training and testing. A Neural Network (NN) deep learning method was trained to recognize named entities within natural language queries. After training, the developed NER algorithm was applied to the testing dataset which achieved a precision of 99.74, a recall of 99.87, and an F1-score of 99.81. The preliminary result indicated that the developed NER algorithm can recognize named entities within the natural language queries accurately. This research will facilitate the further development of conversational AI systems in the construction industry.