| Paper ID | F-1-2.3 |
| Paper Title |
OPENNLU: OPEN-SOURCE WEB-INTERFACE NLU TOOLKIT FOR DEVELOPMENT OF CONVERSATIONAL AGENT |
| Authors |
Yi Fan Ong, Maulik Madhavi, National University of Singapore, Singapore; Ken Chan, ST Engineering Land Systems Ltd, Singapore, Singapore |
| Session |
F-1-2: Natural Language and Spoken Dialogue |
| Time | Tuesday, 08 December, 15:30 - 17:00 |
| Presentation Time: | Tuesday, 08 December, 16:00 - 16:15 Check your Time Zone |
|
All times are in New Zealand Time (UTC +13) |
| Topic |
Speech, Language, and Audio (SLA): |
| Abstract |
The Natural Language Understanding (NLU) module in a conversational agent interprets and understands the user's query. As the application scenario evolves, there is a need to periodically update the knowledge database that supports the agent by adapting or re-training the NLU model. Such periodic updates could be time-consuming to the developers and system administrator since it requires manual efforts. In this paper, we present the OpenNLU toolkit, a user-friendly interface for building and updating the knowledge database, and evaluating NLU module. This paper describes in detail the architecture, important features and design of the web-based tool, as well as the backend features which are supported by popular Rasa NLU toolkit, and deep learning libraries such as Tensorflow, and PyTorch. This paper also demonstrates the training and evaluation processes on in-house datasets alongside other benchmarking datasets (ATIS and Snips) to exemplify the usage of OpenNLU toolkit as to validate proof of concepts. The open-source OpenNLU toolkit is available to the research community. |