Unified Speech Indexing via Text Metadata Probabilities

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The challenge lies in efficiently searching and indexing untranscribed speech data, as manual transcription is costly and raises privacy concerns, while existing methods struggle to make speech data readily searchable.

Innovation Solution

Creating an index for spoken documents by generating probabilities of occurrence and positional information of words in both speech data and text metadata, allowing them to be processed similarly and combined into a single index for efficient querying and relevance ranking.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Difficulty of detecting and measuring

If manual transcription of speech data is performed to make it searchable, then searchability is improved, but cost and time consumption increase significantly

Engineering Contradiction:

ImprovesearchabilityVSAvoidtranscription time

Core Design Contradiction:

Difficulty of detecting and measuringVSLoss of time

Solution Approach 1:

The patent creates a textual copy or representation of speech data through automatic speech recognition, generating searchable text metadata without requiring full manual transcription. This allows the system to work with synthesized text versions of speech content, enabling search functionality while avoiding the time-consuming manual transcription process.

Inventive Principle:

Principle #26Copying

2Difficulty of detecting and measuring

If manual transcription of speech data is performed, then searchability is improved, but privacy concerns are raised

Engineering Contradiction:

ImprovesearchabilityVSAvoidprivacy concerns

Core Design Contradiction:

Difficulty of detecting and measuringVSObject-affected harmful factors

Solution Approach 1:

The system creates searchable text metadata as a copy or representation of speech content rather than requiring complete transcription. This approach enables search functionality while potentially reducing privacy risks by working with processed text representations rather than full transcriptions of sensitive speech data.

Inventive Principle:

Principle #26Copying

3Measurement precision

If separate indexing is created for speech data and text metadata, then processing accuracy is improved, but system complexity increases

Engineering Contradiction:

Improveprocessing accuracyVSAvoidindexing system complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent merges the indexing processes for speech data and text metadata into a unified indexing system. By treating both types of data through a common indexing framework, the system reduces complexity while maintaining processing accuracy through consistent handling of both speech-derived text and original text metadata.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The indexing system is designed to handle both speech data and text metadata using the same indexing mechanisms and data structures. This universal approach allows a single indexing process to serve multiple functions - indexing both automatically recognized speech text and manually created text metadata - thereby simplifying the overall system architecture.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS7809568B2Indexing and searching speech with text meta-data

Publication Date: 2010.10.05 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

An index for searching spoken documents having speech data and text meta-data is created by obtaining probabilities of occurrence of words and positional information of the words of the speech data and combining it with at least positional information of the words in the text meta-data. A single index can be created because the speech data and the text meta-data are treated the same and considered only different categories.