Incremental Natural Language Understanding for Low Latency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current Spoken Language Understanding (SLU) systems face a bottleneck in latency due to the sequential processing of automatic speech recognition (ASR) and natural language understanding (NLU) modules, where NLU has to wait for ASR to finish processing, leading to delayed responses in real-time applications.

Innovation Solution

Implementing incremental natural language understanding by generating speech recognition outputs for portions of an utterance, allowing intent determination and resource retrieval before completing the utterance, and incorporating these resources into media content in real-time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If sequential processing of ASR and NLU modules is used, then processing accuracy is maintained, but latency increases

Engineering Contradiction:
ImprovelatencyVSAvoidprocessing architecture
Core Design Contradiction:
Loss of timeVSDevice complexity

Solution Approach 1:

The patent segments the utterance into multiple portions and processes them incrementally. The ASR module generates recognition results for portions of the utterance as they become available, and the NLU module processes these incremental results to determine intent. This segmentation allows both modules to operate in parallel rather than sequentially, reducing latency while maintaining processing accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements preliminary action by determining intent based on incremental speech recognition results before the utterance is completely processed. The NLU module can start determining intent and retrieving resources based on partial recognition results, allowing the system to respond faster without waiting for complete utterance processing.

Inventive Principle:
Principle #10Preliminary action

2Loss of time

If NLU waits for ASR to finish processing, then processing completeness is ensured, but response time increases

Engineering Contradiction:
Improveresponse timeVSAvoidprocessing completeness
Core Design Contradiction:
Loss of timeVSReliability

Solution Approach 1:

The patent introduces dynamics by making the NLU processing adaptive to the availability of ASR results. Instead of waiting for complete ASR processing, the NLU module dynamically adjusts its processing based on incremental results becoming available. This allows the system to respond as soon as sufficient information is detected while maintaining the ability to refine processing as more utterance data becomes available.

Inventive Principle:
Principle #15Dynamics

3Loss of time

If incremental speech recognition is implemented, then latency is reduced, but computational complexity increases

Engineering Contradiction:
Improveprocessing latencyVSAvoidcomputational complexity
Core Design Contradiction:
Loss of timeVSUse of energy by moving object

Solution Approach 1:

The patent applies partial action by processing only the necessary portions of the utterance as they become available. Instead of waiting for the complete utterance and then processing everything at once, the system processes portions incrementally as they are detected, reducing the total computational load and latency. The system processes exactly what is needed at each step rather than preparing for all possible future processing requirements.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3886088B1System and methods for incremental natural language understanding
Publication Date: 2024.03.27 DISNEY ENTERPRISES INC
  • EP3886088B1 patent drawingFigure 1
  • EP3886088B1 patent drawingFigure 2
  • EP3886088B1 patent drawingFigure 3

AI summary

A system for incremental natural language understanding includes a media module, a memory storing a software code, and a hardware processor communicatively coupled to the media module. The hardware processor is configured to execute the software code to receive an audio stream including a first utterance, and generate a first and second incremental speech recognition outputs based on first and second portions of the first utterance. In addition, the hardware processor is configured to execute the software code to determine, prior to generating the second incremental speech recognition output, a first intent of the first utterance based on the first incremental speech recognition output. The hardware processor is further configured to execute the software code to retrieve a first resource based on the determined first intent, and incorporate the first resource in the media content to be played by the media module.