Processes ambiguous search requests in Google Maps. The Maps-search infrastructure patent — handles fuzzy queries, partial addresses, landmark references, and ambiguous business names against the geographic information system.
Patent Overview
- Inventor
- Lawrence Greenfield, Daniel Egnor, Francois Bailly, John Hanke
- Assignee
- Google LLC
- Filed
- 2014
- Granted
- 2019-02-05
The Challenge
The Challenge
Maps queries are messy. Partial addresses, fuzzy landmark references, ambiguous business names, multi-word place phrases. The GIS must process these reliably and return the right map view, the right business, or the right place — without forcing users to type fully-qualified inputs.
- Maps Queries Are Often Partial — Users type 'pizza near me', 'starbucks downtown', 'AT&T near 5th'. None are fully qualified.
- Multiple Resolution Strategies Needed — Address resolution, business search, landmark matching, ZIP-code lookup — each strategy handles different query shapes.
- Fuzzy Matching Required — 'Starbuks' typo should still match Starbucks. 'AT&T near 5th' should match locations near 5th Avenue. Fuzzy matching is structural.
- Result Disambiguation Matters — When multiple plausible results exist, surfacing the right disambiguation choice is critical.
- User Location Anchors Interpretation — Without user location, 'pizza' is ambiguous. With user location, it resolves to local pizza places.
Innovation
How The System Works
The system parses queries through multiple resolution strategies, runs each strategy against the GIS, scores candidate results by confidence, applies user-location anchoring, and either returns the top candidate or surfaces disambiguation when multiple are plausible.
- Parse Query — Per query, parse to identify possible resolution strategies (address, business, landmark, ZIP, fuzzy phrase).
- Run Resolution Strategies In Parallel — Per strategy, run against the GIS. Each produces candidate results with confidence.
- Apply Fuzzy Matching — Per strategy, fuzzy matching tolerates typos, partial matches, abbreviation expansions.
- Anchor To User Location — Per query, user location anchors candidate scoring. Closer candidates earn proximity weight when query is local-intent.
- Score Candidates — Per candidate, combined confidence score from strategy match, fuzzy distance, proximity.
- Return Top Or Disambiguate — Above-threshold single candidate returned directly. Multi-candidate cases surface disambiguation UI.
- Learn From User Choice — Per query-resolution pair, user click choice feeds back into resolution scoring.
Multi-Strategy GIS Search
The patent's load-bearing idea is that Maps search needs multiple resolution strategies running in parallel. No single strategy covers all query shapes; the union of strategies plus user-location anchoring produces robust resolution.
Parallel Strategies, Combined Scoring
Per query, parallel strategies each produce candidates. Combined scoring across strategies plus user-location anchoring selects the winner. The parallelism is the architectural primitive.
- Multi-Strategy Parallel Execution — Address, business, landmark, ZIP, fuzzy phrase strategies run in parallel.
- Fuzzy Matching — Per strategy, fuzzy matching tolerates typos and partial matches.
- User-Location Anchoring — Per query, user location anchors candidate scoring with proximity weight.
Technical Foundation
Technical Foundation
The patent specifies the query parser, strategy runners, fuzzy matcher, user-location anchorer, candidate scorer, resolver, and learning loop.
- Query Parser — Per query, identifies applicable resolution strategies.
- Strategy Runners — Per strategy (address, business, landmark, ZIP, fuzzy), runs against GIS in parallel.
- Fuzzy Matcher — Per strategy, fuzzy matching tolerates typos and partial matches.
- User-Location Anchorer — Per query, anchors candidate scoring with user location proximity.
- Candidate Scorer — Per candidate, combined confidence score.
- Resolver — Above-threshold returns top; multi-candidate cases surface disambiguation.
The Process
The Process
Per Maps query, the multi-strategy pipeline runs in real time. User location captured per session.
- Receive Query — Maps query arrives.
- Parse And Identify Strategies — Query parsed; applicable strategies identified.
- Run Strategies In Parallel — Each strategy executes against GIS.
- Apply Fuzzy Matching — Fuzzy matching tolerates input variations.
- Anchor To User Location — Proximity weight applied.
- Score And Select — Combined confidence selects winner or triggers disambiguation.
- Learn From Click — User choice feeds back into resolution scoring.
Quality Control
Quality Control
Resolution quality determines Maps usability. The patent specifies safeguards.
- Per-Strategy Calibration — Each strategy calibrated against labeled query-resolution pairs.
- Fuzzy-Matching Bounds — Fuzzy matching tuned to avoid over-matching.
- Disambiguation Threshold — Multi-candidate threshold calibrated to balance direct-return vs disambiguation UI.
- Privacy Preservation For User Location — User location captured with privacy preservation.
- Continuous Recalibration — Strategies and scoring recalibrate against fresh data.
Real-World Application
Multi-strategy GIS search powers Google Maps query handling at scale. The parallel-strategy plus user-location anchoring pattern is the architectural template for any modern map search.
- Multi-strategy Resolution Method — Address, business, landmark, ZIP, fuzzy phrase strategies run in parallel.
- Fuzzy Match Tolerance — Typos, partial matches, abbreviations tolerated.
- Location-anchored Proximity Weighting — User location anchors candidate scoring with proximity weight.
Why Complete NAP Citations Matter For Maps
Address resolution depends on consistent, complete NAP (Name, Address, Phone) citations across the web. Sites with complete, consistent NAP earn cleaner resolution in Maps queries.
Why Landmark And Cross-Street References Help Discovery
Pages mentioning nearby landmarks, cross-streets, and neighborhood context enable the landmark strategy to surface them on landmark-style queries that pure address strategies miss.
<\/section>What This Means for SEO
What This Means for SEO
This Google Maps patent resolves messy queries through parallel strategies (address, business, landmark, ZIP, fuzzy match) anchored to user location. SEO implication: complete consistent NAP citations and landmark context make your business resolvable across the many ways people search Maps.
- Complete Consistent NAP Wins Address Resolution — The address strategy depends on consistent, complete Name-Address-Phone citations across the web. Sites with clean, matching NAP everywhere resolve cleanly; inconsistent citations fragment your resolution.
- Landmark And Cross-Street Context Adds Reach — A dedicated landmark strategy surfaces businesses on landmark-style queries. Mentioning nearby landmarks, cross-streets, and neighborhood context makes you findable on searches that pure address matching misses.
- Fuzzy Matching Tolerates Typos And Abbreviations — The system matches misspellings, partial inputs, and abbreviation expansions. A distinctive, consistent business name resolves even when typed imperfectly, so name clarity and consistency pay off.
- User Location Anchors Local Intent — Proximity weight is applied for local-intent queries. Being genuinely present and accurately located in your service area is what lets you win the near-me style searches anchored to a user nearby.
- Multiple Strategies Run In Parallel — Address, business, landmark, ZIP, and fuzzy strategies all execute at once. Covering several of these signals, accurate address plus clear name plus landmark context, gives you more paths to surface.
- Disambiguation UI Splits Weak Matches — When several candidates are plausible, the system shows a disambiguation choice instead of picking one. Strong, distinctive signals help you be the confident single result rather than one option among many.
- User Clicks Train Resolution — Click choices feed back into resolution scoring. Being the result users actually pick for a query reinforces your association with it over time, which rewards genuinely matching the intent.