Repository logo
  • English
  • Deutsch
Log In
or
  1. Home
  2. HSG CRIS
  3. HSG Publications
  4. Exploring Knowledge Graphs in an Interpretable Composite Approach for Text Entailment
 
  • Details

Exploring Knowledge Graphs in an Interpretable Composite Approach for Text Entailment

Journal
Thirty-Third AAAI conference on artificial intelligence
Type
conference contribution
Date Issued
2019-01-27
Author(s)
Silva, Vivian
Freitas, Andre
Handschuh, Siegfried  
Abstract
Recognizing textual entailment is a key task for many seman- tic applications, such as Question Answering, Text Summa- rization, and Information Extraction, among others. Entail- ment scenarios can range from a simple syntactic variation to more complex semantic relationships between pieces of text, but most approaches try a one-size-fits-all solution that usually favors some scenario to the detriment of another. We propose a composite approach for recognizing text entailment which analyzes the entailment pair to decide whether it must be resolved syntactically or semantically. We also make the answer interpretable: whenever an entailment is solved se- mantically, we explore a knowledge base composed of structured lexical definitions to generate natural language human- like justifications, explaining the semantic relationship hold- ing between the pieces of text. Besides outperforming well- established entailment algorithms, our composite approach gives an important step towards Explainable AI, using world knowledge to make the semantic reasoning process explicit and understandable.
Language
English
HSG Classification
contribution to scientific community
Refereed
Yes
Publisher
AAAI Press
Publisher place
Palo Alto, California USA
URL
https://www.alexandria.unisg.ch/handle/20.500.14171/98958
Subject(s)

other research area

Division(s)

ICS - Institute of Co...

Contact Email Address
siegfried.handschuh@unisg.ch
Eprints ID
255897
File(s)
Loading...
Thumbnail Image

open.access

Name

aaai2019_VSetal_camera-ready.pdf

Size

196.02 KB

Format

Adobe PDF

Checksum (MD5)

7aa3377ae44400f35514eea4847f15ac

here you can find instructions and news.

Built with DSpace-CRIS software - Extension maintained and optimized by 4Science

  • Privacy policy
  • End User Agreement
  • Send Feedback