whereismyai
Cjkfold
Cjkfold: Python library that folds look-alike CJK glyphs from PDFs so text comparisons and searches match again.
Cjkfold repairs look-alike CJK glyphs extracted from PDFs so that downstream text comparisons, keyword searches, and vector retrievers work correctly. Standard Unicode normalization leaves many characters in the CJK Radicals Supplement block unfixed, but cjkfold closes this gap using the EquivalentUnifiedIdeograph property from the Unicode Character Database. It provides pure Python functions with zero runtime dependencies and supports Python 3.9+. The library also includes exactboost for dense retrieval to boost verbatim n-gram matches, along with scanning tools to inspect invisible characters and damage. It is designed for developers building knowledge-base pipelines from PDF extractions.
Cjkfold at a glance
- Best for: Software engineer, Researcher
- Runs on: Python, CLI
- Pricing: Free
- Released: October 2026
- Category: AI research tool
- Website: github.com/cdcailf/cjkfold
Key features
- Fold look-alike CJK glyphs
- Zero runtime dependencies
- Pure Python functions
- Exactboost for dense retrieval
- Scan invisible characters
- Report unreadable records
- Command line interface
Use it for
- Fixing broken text extracted from PDFs
- Improving keyword searches on PDF text
- Boosting vector retrieval with verbatim matches
- Checking text for invisible characters and damage
- Cleaning data before ingestion into knowledge bases
Alternatives to Cjkfold
- MiniCPM OCR — Efficient document parsing with correctness assessment and high throughput.
- PaperFold Gallery — Explore landmark research papers from eight fields as expandable maps.
- EasyRead，让你更舒服地读论文 — It runs locally, stays open source, and uses your own AI models for background translation.
- LEANN — Efficiently index millions of documents locally with 97% storage savings.
- PageIndex — It provides 98.7% accuracy on FinanceBench without vector databases.
- MTEB — Evaluates embedding models across over 1000 tasks and 1000 languages.
- Pagesift — Extracts web data with high memory efficiency and speed for AI agents.
- Paper Xray — Recovers hidden failure modes, author bets, and concrete symbol examples.
- Morphik — Provides native Model Context Protocol support and granular multi-user access.
- Lighteval — Evaluates LLMs across multiple backends with thousands of built-in tasks.
- RAG-Anything — It unifies text and multimodal document retrieval within a single framework.
- PaperLeet — Practice implementing 89 Python exercises derived directly from AI research papers.
All Cjkfold alternatives