What problem does it solve? Bibliometric datasets retrieved from databases like Web of Science often contain irrelevant records caused by wildcard expansion, cross-language homonyms, and cross-discipline term collisions. This Skill analyzes the search strategy before cleaning begins, building a noise expectation model so downstream cleaning is targeted rather than blind. ## Core Features & Use Cases - Search Query Decomposition: Breaks a full search string into concept groups, search terms, and actual matched word forms, annotating search method (exact phrase, wildcard, Boolean) and field (TS, TI, AB, DE). - Wildcard Risk Assessment: Evaluates each wildcard term across five dimensions (root expansion, cross-language homographs, cross-discipline homographs, abbreviation conflicts, naming conflicts) and assigns a three-level risk rating with expected noise proportion. - Three-Layer Noise Prediction: Forecasts word-form noise, topic noise, and boundary noise per concept group, plus cross-group overlap analysis to prioritize records for review. - Use Case: After building a concept-group mapping table for a literature review on a term like "cell", run this Skill to discover that the wildcard query also matches biological, battery, and prison-related records, then receive a tailored cleaning strategy before batch semantic judgment. ## Quick Start Analyze the noise risk structure of my current Web of Science search strategy and predict the noise types for each concept group.
