• DialoQ is a research project to develop a dialog system that can capture user intent which is oriented towards search, particularly from science data.
• It is supported by the Grants-in-Aid for Scientific Research program (Category C, No. 19K12132) of Japan Society for the Promotion of Science (JSPS).

Introduction

Background 1

• While data science has shown a great potential as a new paradigm of science, scientists are often unaware of how to access science data for their research, and
• there are not nearly enough data scientists available to help them, which hinders many scientists from benefiting from the opportunity of data science.

Background 2

• Meanwhile, there has been increasing interest in the technology of conversational agent, which was greatly contributed by recent breakthroughs made in AI.
• Search intent of scientists will not be expressed in just one query. It will be rather expressed by a series of related queries.
• Scientists often may not know what is their exact queries. Rather, they will often develop their queries during attempt of multiple related queries.

Goal

• To develop a dialog system for search-oriented dialog, to help scientists search science data

Example

1. gives an initial question
2. modifies the question by adding a new condition ('human')
3. changes the condition (from 'human' to 'mouse')
4. adds a new condition as an alternative

Implementation Environment

Platform (open)

• LODQA - an open source project for natural language query interface for searching SPARQL endpoints.

Subject domains (open)

• Glyco Biology

Roadmap

2019

• To develop a benchmark dataset

2020

• To develop a baseline system
• to develop an evaluation framework

2021

• To develop a full system (with a focus on user intent detection and its translation into SPARQL)