Learning Apparatus for Dialog Intention Correction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Dialog systems often misinterpret user intentions due to unpredictable user utterances, leading to costly errors in understanding and requiring frequent corrections or relearning.

Innovation Solution

A learning apparatus that uses similarity information to infer user intentions from speech recognition results, correcting interpretations based on stored dialog history and similarity scores to reduce misinterpretation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If speech recognition results are used to infer user intentions, then dialog processing can be performed, but misinterpretation of user intentions occurs

Engineering Contradiction:

Improvedialog processing efficiencyVSAvoidintention inference accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system collects feedback from multiple sources including recognition scores, dialog history, and similarity comparisons to continuously improve intention inference. The learning apparatus updates its understanding of user intentions based on feedback from successful and unsuccessful dialog interactions, adjusting its inference model to reduce misinterpretation while maintaining processing efficiency.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system uses a unified framework that handles multiple functions: speech recognition, intention inference, similarity matching, and learning from dialog history. This multi-functional approach allows the system to process various types of input (speech, text, metadata) through a single intention inference engine, improving both efficiency and accuracy by leveraging all available information sources.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Reliability

If multiple recognition result candidates are used to reduce errors, then intention inference accuracy improves, but system complexity increases

Engineering Contradiction:

Improveintention inference accuracyVSAvoidsystem structure complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system segments the intention inference process into distinct modules: speech recognition module, scoring module, similarity matching module, and intention inference module. Each module handles a specific aspect of processing, allowing the complex task of multi-candidate evaluation to be broken down into manageable, independent components that can be processed efficiently.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system changes parameters dynamically during processing, including recognition score thresholds, similarity thresholds, and candidate selection criteria. By adjusting these parameters based on dialog context and confidence levels, the system can flexibly handle multiple recognition candidates without requiring a permanently complex structure, simplifying the overall system design.

Inventive Principle:

Principle #35Parameter changes

3Reliability

If dialog history and similarity information are stored and processed, then misinterpretation is reduced, but processing time and computational cost increase

Engineering Contradiction:

Improveintention inference accuracyVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary actions by pre-processing and storing dialog history, similarity information, and recognition patterns before they are needed for intention inference. This includes pre-computing similarity metrics and organizing dialog history in accessible formats, so that during actual dialog processing, the system can quickly retrieve and compare against stored information without time-consuming computation.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system applies partial processing by selectively using dialog history and similarity information based on confidence thresholds. When recognition scores are high and confidence is sufficient, the system may skip extensive similarity searching. When confidence is lower, it performs more thorough analysis. This partial action approach balances accuracy requirements with processing time constraints.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS10446146B2Learning apparatus and method

Publication Date: 2019.10.15 KK TOSHIBA

AI summary

According to one embodiment, a learning apparatus includes a first storage and a processing circuitry. The first storage stores a plurality of similarity information items, each including at least a second text, intention candidates and certainty degrees representing a certainty with which the intention candidates are regarded as being identical to intentions of the second text. The processing circuitry detects corresponding similarity information item including the second text identical to a first text from the plurality of similarity information items, corrects a utterance intention to an intention candidate included in the corresponding similarity information item where a certainty degree included in the corresponding similarity information item is not less than a threshold.