What problem does it solve? During bibliometric data cleaning, irrelevant records (noise) keep appearing in recurring forms, but without a structured archive each cleaning round rediscovers the same patterns from scratch. This Skill archives every noise pattern with a code, identification features, semantic judgment rules, boundary conditions, and real cases, so cleaning rules accumulate and improve across rounds and projects. ## Core Features & Use Cases - Structured Pattern Records: Each noise pattern is documented with a code (N1, N2...), status (active/pending/deprecated), layer (word-form, topic, boundary), identification features, boundary conditions, example titles, and revision history. - Three-Layer Classification Framework: Organizes noise into word-form noise (wildcard overflow, cross-language homographs, OCR errors, abbreviation conflicts), topic noise (cross-disciplinary homonyms, brand/place names), and boundary noise (passing mentions, review coverage, weak relevance). - Lifecycle Management Operations: Supports adding new patterns (validated with 3+ cases), revising patterns when manual review finds misjudgments, and deprecating patterns without deleting them for historical reference. - Use Case: While cleaning a meme-related bibliographic dataset, you discover that "memetic algorithm" papers keep slipping through. Archive this as pattern N1 with its journal/keyword features, then reuse it in every subsequent cleaning round and export it to the cleaning log. ## Quick Start Load the existing noise patterns from the cleaning log and archive the newly discovered noise pattern from this cleaning round using the standard template.
