PatSeek Patent Search and Technical Intelligence

Challenges in Patent Search

In patent analysis, users often face tasks like novelty search, FTO (Freedom to Operate) analysis, and technology research. Traditional methods rely on manual searches, which can miss key documents or lead to misjudgments due to inaccurate terminology. For example, searching patents with only keywords may overlook synonyms or cross-language variants. In FTO analysis, it's crucial to cover all relevant patents in target markets to avoid infringement risks. Patent data is scattered across different databases with varying field structures, making manual retrieval time-consuming and results difficult to verify. Additionally, handling multi-language and multi-market data requires adapting search strategies, adding complexity.

How PatSeek Works

PatSeek is a patent search and technical intelligence tool designed for efficient patent document retrieval and structured analysis support. Its core capabilities include:

• Task Routing: Automatically selects workflows based on user goals (e.g., novelty, FTO, research), avoiding the use of patentability searches as substitutes for FTO.
• Search Modes: Offers quick reconnaissance and full-depth modes. Quick reconnaissance is for preliminary screening, while full-depth is for formal decisions. Users must be presented with this choice before starting.
• Bool Search: Uses precise Boolean logic to construct search queries, supporting fields like AP (applicant), IPC (classification), CC (country code). Queries must include object, mechanism, and distinguishing feature modules, following the Condition Observability Gate to ensure new conditions are effective.
• Semantic Search: In full-depth mode with detailed user-provided descriptions, a semantic search is executed first to confirm IPC distribution, expand keywords, and discover new candidates.
• Exploratory Search: Before systematic search, 1-2 exploratory searches are performed to validate term effectiveness, assess result volume, and calibrate subsequent searches.
• Round Guidance: Full-depth search is divided into multiple rounds; after each round, samples are annotated as H/M/N/D and user adjustments are awaited to avoid mechanical repetition.
• Candidate Pool Management: Search results are annotated for relevance, aggregated by family, and evidence locations are recorded. Key patents require figure display and quick links (e.g., https://xinxin.chat/patent?pid=).
• API Key Management: Keys are passed via environment variables; if invalid, the process stops immediately and prompts the user, prohibiting redirection to internet engines.

Key steps include: confirming the task type, selecting the search mode, constructing search queries, performing exploratory searches, iterating in rounds, verifying candidate pools, and generating reports. Reports must include xinxin.chat quick links and figure displays for enhanced readability.

Applicable Boundaries and Considerations

PatSeek is suitable for patent novelty search, FTO analysis, technology research, and invalidity search, but note:

• Quick Mode Limitations: Quick reconnaissance does not provide formal novelty or creativity conclusions and is not for FTO or formal application decisions.
• Non-direct Legal Conclusions: The tool is for document recall; similarity scores or hit counts are not direct legal conclusions and require manual review.
• Data Coverage: Bool search syntax differs between CN and world markets; for non-Chinese patents, specialized guides must be consulted.
• Confidentiality Handling: Unpublished technologies require confidentiality prompts; --no-call-log disables local logs, but cache may save responses.
• Cost Transparency: Costs are based on API-returned credits_charged, not estimates.
• Goal Loop Controller: Invalidity search uses a dedicated script to drive continuous loops, ensuring saturation gate judgment.
  In summary, PatSeek emphasizes verifiability and user control, avoiding automation that substitutes professional judgment.