Query Country Identification via ML Geocoder Routing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing geocoding systems face challenges in accurately identifying the intended country for ambiguous queries, leading to decreased user experience and processing efficiency due to incorrect or unnecessary country identifications.

Innovation Solution

A machine learning-based approach that utilizes various location signals such as place entities, location labels, location candidates, and user location to predict the intended country associated with a query, reducing errors and improving processing efficiency by providing a list of probable countries with associated probabilities.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If country-specific geocoders are used to improve geocoding accuracy, then geocoding precision is improved, but system complexity increases due to multiple geocoders

Engineering Contradiction:

Improvegeocoding accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs preliminary country identification using a machine learning model before routing queries to country-specific geocoders. This advance classification step enables the system to select the appropriate geocoder in advance, achieving high geocoding accuracy through specialized geocoders while managing complexity through automated country detection.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

A machine learning-based country identification model serves as an intermediary between the user query and multiple country-specific geocoders. This intermediary automatically determines the intended country and routes queries to the appropriate geocoder, eliminating the need for manual country selection and reducing system complexity.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If multiple country-specific geocoders are maintained to handle ambiguous queries, then geocoding reliability is improved, but processing time increases

Engineering Contradiction:

Improvegeocoding reliabilityVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The geocoding system is segmented into distinct components: a machine learning-based country identification model and multiple country-specific geocoders. This segmentation allows each component to specialize in its function, with the ML model handling ambiguous queries and routing to the appropriate geocoder, improving reliability without requiring all geocoders to process all queries simultaneously.

Inventive Principle:

Principle #1Segmentation

3Productivity

If country identification is performed using simple keyword matching, then processing speed is improved, but identification accuracy deteriorates due to ambiguous queries

Engineering Contradiction:

Improveprocessing speedVSAvoidcountry identification accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The system replaces traditional mechanical keyword-matching methods with a machine learning model for country identification. The ML model analyzes multiple features including place entities, location labels, location candidates, and user location data to accurately determine the intended country, significantly improving identification accuracy while maintaining processing speed through efficient feature-based classification.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system changes the parameters used for country identification from simple keyword presence to a multi-feature analysis including place entities, location labels, location candidates, and user location. This parameter transformation enables the system to distinguish between ambiguous location references and accurately identify the intended country.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS11347820B2Facilitating identification of an intended country associated with a query

Publication Date: 2022.05.31 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

Aspects of the technology described herein are directed towards, among other things, identifying a country(s) likely intended in association with a query. In embodiments, a query is analyzed to identify a set of place entities that each indicate a specific location and a set of location labels that each indicate an aspect of a location. Additionally, map view data is analyzed to identify a set of location candidates that each indicate a country presented within a map view. The map view data indicates at least a portion of a boundary of the map view. Thereafter, a machine learned model is used to identify a set of countries likely to be intended in association with the query based on the set of place entities, the set of location labels, and the set of location candidates. The set of countries can be provided for use in identifying information related to the query.