CrossBERT: a Triplet Neural Architecture for Ranking Entity Properties

Manotumruksa, J., Dalton, J. , Meij, E. and Yilmaz, E. (2020) CrossBERT: a Triplet Neural Architecture for Ranking Entity Properties. In: 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2020), Xi'an, China, 25-30 Jul 2020, pp. 2049-2052. ISBN 9781450380164 (doi:10.1145/3397271.3401265)

Full text not currently available from Enlighten.

Abstract

Task-based Virtual Personal Assistants (VPAs) such as the Google Assistant, Alexa, and Siri are increasingly being adopted for a wide variety of tasks. These tasks are grounded in real-world entities and actions (e.g., book a hotel, organise a conference, or requesting funds). In this work we tackle the task of automatically constructing actionable knowledge graphs in response to a user query in order to support a wider variety of increasingly complex assistant tasks. We frame this as an entity property ranking task given a user query with annotated properties. We propose a new method for property ranking, CrossBERT. CrossBERT builds on the Bidirectional Encoder Representations from Transformers (BERT) and creates a new triplet network structure on cross query-property pairs that is used to rank properties. We also study the impact of using external evidence for query entities from textual entity descriptions. We perform experiments on two standard benchmark collections, the NTCIR-13 Actionable Knowledge Graph Generation (AKGG) task and Entity Property Identification (EPI) task. The results demonstrate that CrossBERT significantly outperforms the best performing runs from AKGG and EPI, as well as previous state-of-the-art BERT-based models. In particular, CrossBERT significantly improves Recall and NDCG by approximately 2-12% over the BERT models across the two used datasets.

Item Type:Conference Proceedings
Status:Published
Refereed:Yes
Glasgow Author(s) Enlighten ID:Manotumruksa, Mr Jarana and Dalton, Dr Jeff
Authors: Manotumruksa, J., Dalton, J., Meij, E., and Yilmaz, E.
College/School:College of Science and Engineering > School of Computing Science
ISBN:9781450380164
Related URLs:

University Staff: Request a correction | Enlighten Editors: Update this record