[PREVIEW]
Patent Lens is a comprehensive patent analytics and visualization platform that transforms raw patent metadata into actionable competitive intelligence. While many tools simply display patent search results, Patent Lens interprets the intricate relationships between patents, inventors, assignees, and citation networks to reveal technology trends that would otherwise remain hidden in dense legal documents.
This repository serves as the central documentation hub and development roadmap for the Patent Lens project—a modular system designed for IP attorneys, R&D directors, innovation strategists, and academic researchers who need to understand why patents matter, not just what they contain.
Think of patent data as a vast ocean. Most search tools give you a fishing line—you cast it, catch a few documents, and hope for the best. Patent Lens is a sonar system that maps the entire underwater terrain, showing you where the fish are concentrated, how they travel in schools, and which underwater currents (citation pathways) connect them.
The system ingests raw patent records from multiple international patent offices, normalizes the data into a unified schema, and applies graph-based algorithms to surface non-obvious connections. The result is a searchable, filterable, and visual dashboard that turns thousands of legal documents into a handful of strategic insights.
Key differentiators of Patent Lens:
- Citation Proximity Mapping - Visualizes how closely related any two patents are through shared citations, co-authorship, or assignee relationships
- Temporal Trend Analysis - Identifies which technology classes are accelerating, plateauing, or declining across filing years
- Inventor Mobility Tracking - Traces when key inventors change employers and how that correlates with patent filing spikes
- Jurisdiction Comparison - Overlays filing patterns across USPTO, EPO, WIPO, and JPO to spot regional innovation clusters
- Claims Breadth Scoring - Calculates a proprietary "protection radius" metric based on claim language complexity and dependent claim structure
Traditional keyword search misses synonyms, alternative phrasing, and technical jargon variations. Patent Lens uses an embedding-based retrieval system that understands conceptual similarity. Search for "electric vehicle battery cooling" and the system will also surface patents describing "thermal management for EV power storage" even if no keywords overlap.
Each patent becomes a node in an interactive graph. Edges represent forward citations, backward citations, and examiner-cited references. The visualization engine supports force-directed layouts, circular dendrograms, and timeline-based animations that show how knowledge flows between inventors and companies over time.
Patenting activity follows predictable S-curve patterns across technology lifecycles. Patent Lens fits statistical models to filing data and automatically classifies each technology subclass into emerging, growth, maturity, or saturation phases. This allows portfolio managers to allocate resources based on where the market is heading, not where it has been.
Patent claims are written in dense legal language that varies dramatically between jurisdictions. The comparator module translates claims into a normalized structured representation using a hybrid of rule-based parsing and neural machine translation. You can then see side-by-side linguistic transforms that amplify or restrict the scope of protection across different patent offices.
For corporate users, Patent Lens analyzes an entire patent portfolio and calculates a diversification score based on technology class distribution, geographic coverage, and citation independence. This metric helps identify overconcentration risks and untapped expansion opportunities.
Configure watchlists for competitors, technology classes, or specific inventors. When new patents are published that match your criteria, the system sends tailored email digests or webhook notifications with AI-generated executive summaries that explain the strategic significance of each new filing.
[DOWNLOAD]
Patent Lens is designed as a cloud-native application that can also run on-premises for security-conscious organizations. The core engine requires Python 3.11+ and PostgreSQL 15+ with the PostGIS extension. The web dashboard is built as a React SPA with WebSocket support for real-time visualizations.
The system consists of five loosely coupled modules:
- Ingestion Service - Periodically fetches patent records from national offices via their open data APIs or bulk XML/JSON dumps
- Normalization Engine - Converts heterogeneous patent schemas into a canonical model with standardized inventor names, assignee hierarchies, and IPC/CPC classification mappings
- Graph Builder - Constructs a property graph database where patents, persons, organizations, and classifications are nodes; citations, employments, and filings are edges
- Analytics API - Exposes RESTful endpoints for querying the graph, computing metrics, and generating time-series aggregations
- Visualization Frontend - Interactive web interface that consumes the API and renders network graphs, heatmaps, and drill-down tables
The configuration file uses YAML syntax and supports environment variable interpolation. Key settings include:
patent_sources: List of patent office identifiers with their API credentialsgraph_depth: How many citation hops to traverse when building the proximity maplookback_years: Rolling window for trend analysis (default is 15 years)language_models: Which multilingual models to load for claim comparison
After ensuring the database is initialized and migration scripts have been executed, the system can be launched in development mode with live reload enabled. The API will listen on port 8080 by default, while the frontend dev server runs on port 3000 with proxy configuration to forward API requests.
- Documentation Hub - Comprehensive API references, data model diagrams, and configuration guides are maintained in the
/docsdirectory - Issue Tracker - Bug reports, feature requests, and usage questions are welcome via the GitHub issues interface
- Discussion Forum - Monthly virtual meetups are organized to discuss patent analytics techniques and share tips
- Response Time - The core maintainers aim to respond to all support queries within 24 hours on weekdays and 48 hours on weekends (24/7 support tier available for enterprise licensees)
Patent databases are powerful search engines, but they operate like library card catalogs—they tell you that a document exists, but they don't tell you why that document matters. Researchers end up manually reading dozens of patents, extracting references, and building their own mental map of the technology landscape. This process is slow, error-prone, and impossible to scale across thousands of documents.
Traditional patent search is like looking up names in a phone book—alphabetical, isolated, and lacking context. Patent Lens treats patents as members of a social network. It shows you who influences whom, which collaborations produce the most impactful inventions, and how ideas migrate between fields. This social-network vantage point is the difference between listing residents and understanding a city's dynamics.
Patenting decisions are often made under time pressure. Companies file patents to secure funding, block competitors, or satisfy licensing requirements. But the strategic value of a patent depends on how well it fits into the broader innovation ecosystem. Patent Lens helps organizations avoid the "innovation cliff"—the phenomenon where patents are filed but never cited, never licensed, and never asserted, effectively becoming sunk costs.
When conducting freedom-to-operate analyses, attorneys can use Patent Lens to quickly identify the closest prior art and assess whether a client's proposed claims would infringe on existing patents. The citation proximity map reveals not just direct references, but second-order connections that might otherwise be missed.
Before allocating budget to new research initiatives, R&D leaders can query Patent Lens to see whether competitors are already actively patenting in that space. The temporal trend analysis highlights whether a technology is still fertile ground or already saturated with filings that would make differentiation difficult.
Universities generate diverse inventions across many departments. Patent Lens helps technology transfer offices see which of their inventions have the highest citation growth potential, allowing them to prioritize patent filing and licensing efforts toward the technologies most likely to gain traction.
When evaluating startups, VC analysts can assess the quality of a startup's patent portfolio beyond simple counts. The claims breadth scoring and citation patterns reveal whether the startup has obtained meaningful protection or merely filed defensive publications.
Organizations that share patents through licensing platforms can use Patent Lens to match their available technologies with potential licensees. By analyzing which technology classes are growing fastest in specific industries, they can identify companies that might be seeking exactly the type of protection they own.
Patent Lens supports a plugin system for adding new patent office connectors, custom analytics algorithms, and specialized visualization types. Plugins are distributed as Python packages that conform to a standard entry-point interface.
Users can define their own composite metrics by combining base metrics with arithmetic operations, logical conditions, and temporal aggregations. For example, you might create a "Blocking Power" metric that combines forward citation counts, duration of remaining validity, and geographic reach.
For consulting firms or patent analytics services, Patent Lens can be white-labeled with custom branding and domain names. The frontend theming system supports complete visual rebranding through CSS variable overrides.
For large-scale academic research, Patent Lens supports batch export of analysis results in CSV, JSON, and GraphML formats. This allows external tools like R, Python notebooks, or Gephi to further process the graph data.
Patent Lens aggregates data from the following public sources:
- USPTO (United States Patent and Trademark Office) - Full text and bibliographic data from 1976 onward
- EPO (European Patent Office) - European patent applications and granted patents via PATSTAT global database
- WIPO (World Intellectual Property Organization) - PCT international applications
- JPO (Japan Patent Office) - Japanese patent abstracts and full-text documents since 1993
- KIPO (Korean Intellectual Property Office) - Korean patent data with English abstract translations
The data ingestion layer performs deduplication using family identifiers (docdb, inpadoc) so that the same invention filed in multiple jurisdictions is treated as a single family with multiple members.
The graph database is designed to handle portfolios containing up to 10 million patents and 50 million citation relationships. Query performance is optimized through:
- Materialized Views - Precomputed aggregations for common analytical queries
- Vectorized Execution - Pandas-based in-memory analytics for metric calculations
- Columnar Storage - Apache Parquet files for archival and batch analysis
- Horizontal Scaling - The analytics API is stateless and can be deployed behind a load balancer
Benchmarks conducted on a mid-range server (8 cores, 32GB RAM) show that a typical proximity search across a 2-million-patent subset returns results in under 2 seconds, while trend calculations over 15 years of data complete in under 8 seconds.
Patent data is public information, but the analytics derived from it can reveal proprietary strategy. Patent Lens supports role-based access controls that restrict which patent families and which analytics modules certain user groups can access.
For European users, the system provides data deletion workflows, anonymization options for inventor names, and full audit trails of all user queries. No personal data beyond login credentials is stored.
For organizations with strict security requirements, Patent Lens can be deployed in an air-gapped environment where no external network connections are permitted. The system only requires outbound access to patent office APIs during the initial data ingestion phase.
- Transformer-Based Claim Summarization - Automatically generate plain-language summaries of patent claims anchored by legal term fidelity
- Cross-Industry Synonym Dictionary - Better understand how the same technical concept appears in different industries with different terminology
- AR Visualization Mode - View citation networks overlaid on physical patent wallboards using augmented reality headsets
- Voice Query Interface - Ask "What did Tesla patent last quarter?" in natural language and get a spoken executive briefing
We welcome contributions in the form of bug fixes, documentation improvements, new patent office connectors, and analytics algorithms. Please review the contributing guidelines in the repository before submitting pull requests.
Q: How is Patent Lens different from using Google Patents directly? A: Google Patents is an excellent search engine for individual patent lookup. Patent Lens is an analytics engine that combines patent data with network science to answer strategic questions about technology landscapes. It complements rather than replaces direct search tools.
Q: Can Patent Lens predict the future? A: No statistical model can predict the future with certainty. Patent Lens identifies historical patterns and extrapolates them with confidence intervals. The output should be treated as informed decision support, not as prophecy.
Q: Is Patent Lens suitable for academic research? A: Yes, academic researchers use Patent Lens for studies on innovation diffusion, knowledge spillovers, collaboration networks, and technology evolution. The batch export functionality is particularly useful for creating datasets for further analysis.
Q: What is the learning curve? A: The web dashboard can be mastered in about an hour for basic searches and visualizations. The analytics API and custom metric definition require familiarity with graph concepts and basic Python syntax, which typically takes 2-3 days of focused effort.
This project is released under the MIT License. You are free to use, modify, and distribute the software, provided that attribution to the original authors is retained.
Patent Lens is an analytical tool and does not constitute legal advice. While we strive for data accuracy, patent data from public sources may contain errors, omissions, or discrepancies. Users should always verify critical findings with the original patent documents and consult with qualified legal professionals before making decisions based on this tool's output. The authors bear no liability for actions taken based on the analysis results. Patent laws and filing requirements vary by jurisdiction and change over time—this tool reflects the state of data as ingested but does not interpret legal regulations.
The technology trends, scoring metrics, and network analyses presented by Patent Lens are algorithmic inferences, not legal determinations. A high citation count does not guarantee an enforceable patent, and a low citation count does not indicate invalidity. Always engage professional counsel for substantive patent prosecution, litigation, or licensing matters.
This project builds upon the foundational work of the open-source graph database community, the patent informatics research published in scholarly journals, and the efforts of patent offices worldwide in making public data accessible. We are grateful for the ecosystem that makes tools like this possible.
For general inquiries, partnership proposals, or enterprise licensing discussions, the maintainers are reachable through the GitHub repository discussions tab. We aim to respond within one business day.
Thank you for exploring Patent Lens. We believe that clarity in patent intelligence empowers better decisions, faster innovation, and a more efficient marketplace for ideas.
[DOWNLOAD]