The impact of using pre-trained word embeddings in Sinhala chatbots

Autor:	Bimsara Gamage, Randil Pushpananda, Ruvan Weerasinghe
Rok vydání:	2020
Předmět:	Open source business.industry Computer science First language The Internet Artificial intelligence computer.software_genre business Chatbot computer Natural language processing Natural language Word (computer architecture)
Zdroj:	2020 20th International Conference on Advances in ICT for Emerging Regions (ICTer).
DOI:	10.1109/icter51097.2020.9325440
Popis:	Providing conversational interfaces to IT services enable more people to access them conveniently. For maximum benefit, such interfaces need to be in the native language of the users in a community. Popularly known as chatbots, one of their critical tasks is to accurately identify the user's intention from their natural language input. This paper describes an attempt at improving this accuracy in an implementation based on the open source RASA stack, by using Sinhala word embeddings. This approach shows improvement in both intent detection as well as the confidence in the detected intents.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::c3e50546a85dd4f245b4b3465f4cca08 https://doi.org/10.1109/icter51097.2020.9325440 Zobrazit plný text záznamu