Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Situated Natural Language Explanations

Zining Zhu, Haoming Jiang, Jingfeng Yang, Sreyashi Nag, Chao Zhang, Jie Huang, Yifan Gao, Frank Rudzicz, Bing Yin

TL;DR

Situated NLE provides a perspective and facilitates further research on the generation and evaluation of explanations, and describes the properties of NLEs in lexical, semantic, and pragmatic categories.

Abstract

Natural language is among the most accessible tools for explaining decisions to humans, and large pretrained language models (PLMs) have demonstrated impressive abilities to generate coherent natural language explanations (NLE). The existing NLE research perspectives do not take the audience into account. An NLE can have high textual quality, but it might not accommodate audiences' needs and preference. To address this limitation, we propose an alternative perspective, \textit{situated} NLE. On the evaluation side, we set up automated evaluation scores. These scores describe the properties of NLEs in lexical, semantic, and pragmatic categories. On the generation side, we identify three prompt engineering techniques and assess their applicability on the situations. Situated NLE provides a perspective and facilitates further research on the generation and evaluation of explanations.

Situated Natural Language Explanations

TL;DR

Situated NLE provides a perspective and facilitates further research on the generation and evaluation of explanations, and describes the properties of NLEs in lexical, semantic, and pragmatic categories.

Abstract

Natural language is among the most accessible tools for explaining decisions to humans, and large pretrained language models (PLMs) have demonstrated impressive abilities to generate coherent natural language explanations (NLE). The existing NLE research perspectives do not take the audience into account. An NLE can have high textual quality, but it might not accommodate audiences' needs and preference. To address this limitation, we propose an alternative perspective, \textit{situated} NLE. On the evaluation side, we set up automated evaluation scores. These scores describe the properties of NLEs in lexical, semantic, and pragmatic categories. On the generation side, we identify three prompt engineering techniques and assess their applicability on the situations. Situated NLE provides a perspective and facilitates further research on the generation and evaluation of explanations.
Paper Structure (52 sections, 3 equations, 1 figure, 5 tables)

This paper contains 52 sections, 3 equations, 1 figure, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Natural language explanation
  4. Controllable generation
  5. Human-centered explainable AI
  6. Pragmatics and Theory of Mind
  7. Situated Evaluation
  8. Lexical category
  9. Length
  10. Self-BLEU
  11. Type-token ratio (TTR)
  12. Semantic category
  13. BLEU score
  14. Concreteness
  15. Pragmatic category
  16. ...and 37 more sections

Figures (1)

  • Figure 1: A detailed explanation might be suitable for a careful customer (left) but is not suitable for a customer casually looking around (right) -- we need to adapt the explanations to the situations.