Multi-Task Neural Network for Search Intent Classification

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Search engines often fail to provide meaningful suggestions to users, leading to inefficient search sessions and increased queries, as they may offer broadly relevant but not specifically helpful guidance, causing users to ignore suggestions and resulting in higher resource consumption.

Innovation Solution

A computer-implemented technique using a suggestion-generating system with a machine-trained model, such as a classification-type or generative-type neural network, to provide tailored suggestions by identifying coherent task-related intents in search sessions, generating suggestions that advance the user's task, and reducing the number of queries needed to complete a search objective.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If search engines provide broadly relevant suggestions to users, then the suggestions cover a wide range of topics, but the suggestions fail to meaningfully assist users in conducting their search

Engineering Contradiction:

Improvebreadth of suggestionsVSAvoidusefulness of suggestions

Core Design Contradiction:

Adaptability or versatilityVSEase of operation

Solution Approach 1:

The patent segments the suggestion generation process into multiple components: an intent classification model that identifies user search intent, a suggestion generation model that creates tailored suggestions, and a multi-task training framework that integrates both functions. This segmentation allows the system to first understand user intent and then generate suggestions specifically adapted to that intent, resolving the contradiction between broad coverage and targeted usefulness

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent changes the parameters of suggestion generation by incorporating intent classification as a prerequisite step. The system adjusts suggestion parameters (relevance, specificity, timing) based on the classified intent type, transforming generic suggestion generation into intent-driven suggestion generation that maintains versatility while improving usefulness

Inventive Principle:

Principle #35Parameter changes

2Productivity

If search engines provide suggestions that do not meaningfully assist users, then the system consumes computing resources, but users ignore the suggestions and conduct more queries

Engineering Contradiction:

Improvenumber of queriesVSAvoidcomputing resource consumption

Core Design Contradiction:

ProductivityVSUse of energy by moving object

Solution Approach 1:

The patent implements feedback mechanisms where the system learns from user interactions with suggestions. The multi-task training framework uses training examples derived from actual search sessions to continuously improve intent classification accuracy and suggestion relevance. This feedback loop ensures that suggestions become increasingly useful over time, reducing both the number of queries needed and the computing resources wasted on ineffective suggestions

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs preliminary intent classification before generating suggestions, ensuring that suggestions are pre-tailored to user needs. This preliminary action prevents the generation and presentation of useless suggestions, thereby avoiding the waste of computing resources and reducing the number of additional queries users must submit

Inventive Principle:

Principle #10Preliminary action

3Loss of time

If search engines increase the number of queries users enter to accomplish search objectives, then users spend more time searching, but resource consumption increases

Engineering Contradiction:

Improvesearch session durationVSAvoidcomputing resource consumption

Core Design Contradiction:

Loss of timeVSLoss of energy

Solution Approach 1:

The patent introduces an intermediary intent classification model that mediates between user queries and suggestion generation. This intermediary component analyzes user intent and guides the suggestion generation process, ensuring that suggestions are highly relevant and reduce the number of queries needed. The multi-task training framework optimizes both the intermediary classifier and the suggestion generator jointly, minimizing both time loss and energy loss

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS12099552B2Using a multi-task-trained neural network to guide interaction with a query-processing system via useful suggestions

Publication Date: 2024.09.24 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

A computer-implemented technique is described herein for assisting a user in advancing a task objective. The technique uses a suggestion-generating system (SGS) to provide one or more suggestions to a user in response to at least a last-submitted query provided by the user. The SGS may correspond to a classification-type or generative-type neural network. The SGS uses a machine-trained model that is trained using a multi-task training framework based on plural groups of training examples, which, in turn, are produced using different respective example-generating methods. One such example-generating method constructs a training example from queries in a search session. It operates by identifying the task-related intent the queries, and then identifying at least one sequence of queries in the search session that exhibits a coherent task-related intent. A training example is constructed based on queries in such a sequence.