Population Language Model Clustering for Adaptive Vocabulary Recognition

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Traditional language recognition systems fail to accurately recognize user-specific vocabulary and context, leading to misrecognition of important words like proper names, as they do not dynamically adapt to individual user preferences and usage patterns.

Innovation Solution

A population language processing system that collects and analyzes user events from local language models to identify clusters of users with similar characteristics, generating and updating population language models to improve local language model accuracy by incorporating commonly used words and context-specific vocabulary.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a local language model is trained individually for each user based on their inputs, then the model can adapt to user-specific vocabulary, but the system complexity and training time increase significantly

Engineering Contradiction:

Improveuser-specific vocabulary recognitionVSAvoidlanguage model training system

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges individual user language models with a population-level language model. User-specific vocabulary and patterns are integrated into a shared population model that serves multiple users, combining the benefits of personalization with the efficiency of shared learning across the user base.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The population language model serves multiple functions: it provides baseline language understanding for all users, accumulates vocabulary from multiple users, and can be selectively applied to individual users based on their needs, reducing the need for extensive individual training for each user.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Reliability

If a comprehensive language model includes all possible words, then recognition accuracy improves, but the model size and processing time increase

Engineering Contradiction:

Improveword recognition accuracyVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system applies local quality by customizing the language model for each user based on their specific vocabulary needs and usage patterns. Rather than using a uniform comprehensive model for all users, the model is tailored to include only the words and patterns relevant to each user's communication style, improving accuracy while reducing unnecessary processing overhead.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS10176803B2Updating population language models based on changes made by user clusters

Publication Date: 2019.01.08 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

Technology for improving the predictive accuracy of input word recognition on a device by dynamically updating the lexicon of recognized words based on the word choices made by similar users. The technology collects users' vocabulary choices (e.g., words that each user uses, or adds to or removes from a word recognition dictionary), associates users who make similar choices, aggregates related vocabulary choices, filters the words, and sends words identified as likely choices for that user to the user's device. Clusters may include, for example, users in a particular location (e.g., sets of people who use words such as “Puyallup,”“Gloucester,” or “Waiheke”), users with a particular professional or hobby vocabulary, or application-specific vocabulary (e.g., word choices in map searches or email messages).