Research task title

Development of context-aware Deep-Symbolic Hybrid Intelligence original technology and construction of language knowledge resources

Research Overview

Development of context-aware Deep-Symbolic Hybrid Intelligence original technology and construction of language knowledge resources to support autonomous knowledge learning and question answering in expert fields

A symbolic approach is a method of understanding and learning knowledge in a symbolic system, similar to the human thinking process.
Humans acquire segmented unit knowledge, learn new knowledge through a process of connection and augmentation, and understand and accumulate new knowledge through a process of generalizing and abstracting of the knowledge.
In this detailed task, research is conducted for original technology for context-aware deep symbolic hybrid intelligence capable of processing and learning knowledge in a symbolic system, using deep learning in a process similar to that of human understanding and accumulation of knowledge.
  • Extracting and utilizing context knowledge expressed in natural language
  • Acquiring segmented unit knowledge and learning new knowledge through a process of connection and augmentation
  • Realizing opinion-type question answering on the basis of cases in expert fields through generation of case-based knowledge
Researching Korean language processing technology and Korean language knowledge to facilitate a process for extracting expert knowledge about the Korean language
  • Technology of reducing Korean-language ambiguity.
  • Technology of correcting errors in Korean-language documents.

Research Goals


(1st year) Development of design and prototype of original technology for deep-symbolic hybrid intelligence and construction of language resources to support highly-difficult question answering with short answers
  • Development of context extraction and knowledge indexing technology for highly-difficult question answering with short answers.
  • Development of deep symbolic hybrid knowledge learning and reasoning technology for highly-difficult question answering with short answers
  • Development of question answering technology based on a deep symbolic hybrid knowledge base for highly-difficult question answering with short answers
  • Construction of language resources and development of a machine learning model for document correction
  • Expansion of the Korean language WordNet for providing Korean language service in expert fields
  • Research on sense-embedding technology based on knowledge-powered deep learning.
  • Development of word sense disambiguation (WSD) technology in homonyms based on a lexical map and word embedding
  • Expansion of a lexical map including terminology
(2nd year) Development of core technology for deep-symbolic hybrid intelligence and construction of language resources to support question answering on the basis of cases in expert fields
  • Development of context extraction and knowledge indexing technology for question answering on the basis of cases in expert fields
  • Development of deep symbolic hybrid knowledge learning and reasoning technology for question answering on the basis of cases in expert fields
  • Development of question answering technology based on the deep symbolic hybrid knowledge base for question answering on the basis of cases in expert fields
  • Improvement of performance of text-dependent spelling error correction on the basis of a statistical model
  • Expansion of the Korean language WordNet for providing English service in expert fields
  • Research for distributed multi-sense word embedding (DMWE) technology
  • Development of homonym WSD technology based on word embedding
  • Expansion of a lexical map including terminology
  • Convergence between visual information and text information
(3rd year) Improvement of original technology for deep-symbolic hybrid intelligence and construction of language resources to support question answering on the basis of cases in expert fields
  • Development of context extraction and knowledge indexing technology for question answering on the basis of cases in expert fields
  • Improvement of deep symbolic hybrid knowledge learning and reasoning technology for question answering on the basis of cases in expert fields
  • Improvement of question answering technology based on a deep symbolic hybrid knowledge base for question answering on the basis of cases in expert fields
  • Development of a high-quality document correction system through a combining of the rules with the statistical model
  • Research for a plan for utilizing the Korean language WordNet to provide multilingual services in expert fields
  • Research for sense embedding and DMWE-based phrase embedding technologies
  • Development of homonym WSD technology based on word embedding
  • Expansion of a lexical map including terminology
  • Efficient mining for a large-volume graph