Ranks graphical visualizations of a dataset by attribute fit. The visualization-ranking primitive for Trends-style dashboards — per dataset, the best chart form emerges from data-attribute scoring.
Patent Overview
- Inventor
- Yossi Matias, others
- Assignee
- Google LLC
- Filed
- 2011
- Granted
- 2014-08-19
The Challenge
The Challenge
Per dataset, many visualization forms are possible (line, bar, pie, map, heatmap, scatter). Only some are appropriate. The system needs to rank visualization candidates by how well they fit the data attributes, automating chart-form selection.
- Wrong Charts Mislead — Pie chart for time-series data misleads. Line chart for categorical data misleads. Selection matters.
- Data Attributes Determine Fit — Time dimension favors lines; categorical comparison favors bars; geographic data favors maps. Per attribute, chart fit varies.
- Multiple Charts May Be Valid — Some datasets support multiple valid visualizations. Ranking picks the best.
- User Context Shapes Selection — Per user task (exploration, comparison, summary), preferred chart forms differ.
- Ranking Must Scale — Per dataset, ranking runs in real time for interactive dashboards.
Innovation
How The System Works
The system extracts dataset attributes, scores each candidate visualization by attribute fit and user-task context, ranks visualizations, and presents top candidate or surfaces choices for user selection.
- Extract Dataset Attributes — Per dataset, extract attribute types (temporal, categorical, numerical, geographic).
- Enumerate Visualization Candidates — Per attribute mix, enumerate candidate visualization forms.
- Score Per-Candidate Fit — Per candidate, score fit to dataset attributes.
- Apply User Context — Per user task, candidates aligned with task earn boost.
- Rank Candidates — Combined scoring ranks visualizations.
- Present Top Or Surface Choices — Above-threshold top candidate presented; tied candidates surface as user choices.
- Learn From User Selection — Per user, selection feedback feeds back into ranking.
Attribute Fit Drives Chart Selection
The patent's load-bearing idea is that visualization selection is a ranking problem. Per dataset attributes plus user context, candidates rank, top wins. The architectural primitive is the ranking itself.
Attribute-Driven Plus Context-Adjusted
Per dataset attributes, baseline fit. Per user task, context adjustment. Together, ranking selects appropriate chart.
- Attribute Extraction — Per dataset, attribute types extracted.
- Per-Candidate Fit Scoring — Per visualization, fit to attributes scored.
- User-Task Context — Per user task, candidates aligned with task earn boost.
Technical Foundation
Technical Foundation
The patent specifies the attribute extractor, candidate enumerator, fit scorer, context adjuster, ranker, presenter, and learning loop.
- Attribute Extractor — Per dataset, extracts attribute types.
- Candidate Enumerator — Per attribute mix, enumerates candidates.
- Fit Scorer — Per candidate, scores fit.
- Context Adjuster — Per user task, adjusts scores.
- Ranker — Combined scoring ranks candidates.
- Presenter — Top candidate presented or choices surfaced.
The Process
The Process
Per dataset render, the visualization ranking pipeline runs in real time.
- Receive Dataset — Dataset arrives for visualization.
- Extract Attributes — Attributes extracted.
- Enumerate Candidates — Candidates enumerated.
- Score Fit — Per candidate, fit scored.
- Apply Context — User-task context applied.
- Rank — Candidates ranked.
- Present — Top candidate rendered.
Quality Control
Quality Control
Wrong visualization selection misleads. The patent specifies safeguards.
- Fit-Scoring Validation — Per candidate, fit scoring validated against labeled data.
- User-Task Calibration — Per user task, context adjustment calibrated.
- Misleading-Chart Filtering — Known misleading-chart patterns filtered.
- User-Selection Feedback — User selections feed back into ranking.
- Continuous Recalibration — Scoring and ranking recalibrate against fresh data.
Real-World Application
Visualization ranking underpins Google Trends visualizations, Looker Studio defaults, and Data Studio chart suggestions. The attribute-fit plus context pattern is the architectural template for automated chart selection.
- Attribute-driven Selection Basis — Per dataset, attribute types drive candidate fit scoring.
- Context-aware Task Adjustment — Per user task, scores adjust.
- Ranking-based Output Form — Per dataset, candidates ranked; top presented or choices surfaced.
Why Clean Dataset Structure Helps Visualization Quality
Datasets with clean attribute types (clearly temporal, clearly categorical, clearly geographic) produce strong fit signals. Mixed-attribute datasets produce ambiguous fit and weaker visualization selection.
Why Task-Appropriate Charts Compound Comprehension
Per user task, the right chart form aids comprehension. Selection that matches task improves understanding; mismatch produces friction.
<\/section>What This Means for SEO
What This Means for SEO
This patent ranks candidate chart forms for a dataset by how well they fit the data attributes and the user's task. SEO implication: clean, clearly-typed datasets and task-appropriate chart choices produce stronger visualizations and better comprehension in data-driven content.
- Clean Dataset Structure Produces Strong Fit — Datasets with clearly-typed attributes, distinctly temporal, categorical, or geographic, produce strong fit signals and confident chart selection. Mixed or ambiguously-typed data yields weak fit and poor visualization choices.
- Match The Chart To The Task — Per user task, whether exploration, comparison, or summary, the right chart form aids comprehension. Choosing the chart that fits the question your data answers improves understanding; a mismatch creates friction.
- Wrong Charts Mislead And Get Filtered — Known misleading-chart patterns, like a pie chart for time series, are filtered. Presenting data in the form that honestly fits it aligns with the system and serves readers.
- Attribute Types Drive Candidate Selection — Time dimensions favor lines, categorical comparisons favor bars, and geographic data favors maps. Structuring your data so its type is unambiguous lets the right chart form be selected.
- Clear Data Powers Data-Driven Content — The pattern underpins Trends, Looker Studio, and chart-suggestion tools. Publishing well-structured, clearly-typed data makes your data-driven content render in strong, comprehensible visualizations.
- Multiple Valid Charts Get Ranked — When several visualizations are valid, the system ranks them and picks the best. Providing data that supports a clearly best representation produces a confident choice rather than an ambiguous one.
- Comprehension Is The End Goal — Task-appropriate charts compound comprehension. Designing visualizations around what the reader is trying to understand, not around visual novelty, is what makes data content effective.