RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2

Schema-Guided Dialogue (SGD) a Set of dialogs for training of voice assistants

Product
Developers: Google
Date of the premiere of the system: October, 2019
Branches: Internet services
Technology: Speech technologies

Content

2019: Announcement

At the end of October, 2019 it became known that Google released in open access the world's largest set of dialogs for training of voice assistants. The project of the American corporation received the name Schema-Guided Dialogue (SGD).

File:Aquote1.png
Modern virtual assistants help users to execute a broad spectrum of tasks, including search of runs, search of the next actions and movies, reservation something, information search on the Internet, etc — the Abkhinav Rastogi software engineer (Abhinav Rastogi) and the technical lead of department of Google Research Pranav Khaitan say. Despite huge progress, adaptivity problems often are overlooked in modern models. Partially it is connected with lack of suitable data sets which correspond to the scale and complexity which such virtual assistants face.
File:Aquote2.png

Google released in open access the world's largest set of dialogs for training of voice assistants. The new project of the American corporation received the name Schema-Guided Dialogue (SGD)

The SGD base includes 18 thousand problem-oriented marked dialogs between people and the virtual assistant. Dialogs can are broken on 17 different subjects, including media, banks, actions, the calendar, travel and weather. For the majority that provides several different API, many of which have the crossed functions, but the different interfaces reflecting typical scenarios of the real world.

According to creators of SGD, it is the first set given, covering such number of areas and having on some API for each of them.

Developers also state that the unified model which is the cornerstone of open model of Google for tracking of a status of dialogs facilitates representation of the general knowledge between similar concepts in different services. It allows developers to address new services which did not study at a basis of data from SGD earlier.[1]

You look also (voice assistants)



Notes