---
title: "Understanding Spoken Natural Language with Omni-Directional Information Flow"
authors:
  - "Katashi Nagao"
  - "Kôiti Hasida"
  - "Takashi Miyata"
doi: ""
url: "https://ijcai.org/proceedings/1993/175"
pdf: "https://ijcai.org/proceedings/1993/0175.pdf"
venue: "Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence"
year: 1993
pages: "1268-1274"
track: "Natural Language"
keywords:
---

# Understanding Spoken Natural Language with Omni-Directional Information Flow

Katashi Nagao, Kôiti Hasida, Takashi Miyata

*Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence*, pages 1268-1274.

## Abstract

A totally constraint- based computational architecture is employed to deal with problems in speech and natural language processing. The semantics of constraints is defined as a type of dynamics, with potential energy measuring the degree of violation and field of force towards minimal energy, so that information processing emerges as dynamical interaction. This establishes tight feedback loops among heterogeneous sorts of information, and integrates various types of computation under dynamical control. In speech understanding, the omnidirectional information flow arising in this architecture leads to natural solutions to tough problems such as recognition of ill-formed or ambiguous speech, prediction of subsequent utterances, revision of former recognition, treatment of insertion/ deletion of phonemes, detection of word boundaries, and so forth. Since the architecture is constraint-based, there is nounderlying procedure specific to domains or tasks.