Semantic Answer Type Prediction using BERT: IAI at the ISWC SMART Task 2020

This paper summarizes our participation in the SMART Task of the ISWC 2020\nChallenge. A particular question we are interested in answering is how well\nneural methods, and specifically transformer models, such as BERT, perform on\nthe answer type prediction task compared to traditional approaches. Our main\nfinding is that coarse-grained answer types can be identified effectively with\nstandard text classification methods, with over 95% accuracy, and BERT can\nbring only marginal improvements. For fine-grained type detection, on the other\nhand, BERT clearly outperforms previous retrieval-based approaches.\n