What problem does it solve? Thai retail product names arrive full of noise: Thai numerals, zero-width characters, duplicate spaces, promotional phrases like "ซื้อ 1 แถม 1", and inconsistent measurement units. Feeding this raw text into classification or deduplication pipelines degrades accuracy, so it must be normalized first. ## Core Features & Use Cases - Thai Text Normalization: Converts Thai digits to Arabic, strips zero-width characters, and collapses duplicate whitespace via the project's ThaiTextProcessor. - Unit Standardization: Rewrites units like "100 g", "100g", and "100ก." into a consistent form (ก., กก., มล., ล.). - Promotion & Noise Removal: Strips marketing phrases such as "ลด 50%" and "ราคาพิเศษ" before AI processing. - Batch Processing: Cleans entire product DataFrames with pandas, drops duplicates on the cleaned name, and filters out names shorter than 3 characters. - Use Case: Before importing a CSV of 5,000 retail product names into the taxonomy pipeline, run the batch cleaning pattern to normalize every name and remove duplicates. ## Quick Start Ask the AI to clean and normalize a list of Thai product names using the ThaiTextProcessor from fresh_implementations before running classification.
