fuzzy-context-finder 0.1.2

search for keywords and their context

Keyword Context Finder

Searching for context doesn't have to be a chore!

A Python utility for finding keywords and their surrounding context in MangoCR markdown files. This tool supports fuzzy matching and provides flexible context extraction around matched terms.

Features

• Fuzzy string matching for approximate keyword finding
• Customizable context window sizes (before, after, and around matches)
• Page number tracking for MangoCR formatted documents
• Adjustable similarity threshold for matches
• Returns results in a pandas DataFrame for easy analysis

Installation

pip install pandas rapidfuzz

Dependencies

• Python 3.6+
• pandas
• rapidfuzz
• regex

Usage

from fuzzy_context_finder import keyword_context_finder

# Example usage
content = """
## document_page_1
This is the content of page one with some keywords.
## document_page_2
More content on page two with different keywords.
"""

search_terms = ["keyword", "content"]
file_name = "example_document.pdf"

results = keyword_context_finder(
    content=content,
    terms=search_terms,
    file_name=file_name,
    words_before=250,
    words_after=250,
    words_around=50,
    match_threshold=80
)

Parameters

• content (str): Document content with pages separated by MangoCR markers (## filename_page_number)
• terms (list): List of search terms to find in the document
• file_name (str): Name of the file being processed
• words_before (int, default=250): Number of words to capture before the term
• words_after (int, default=250): Number of words to capture after the term
• words_around (int, default=50): Number of words to capture around the term
• match_threshold (int, default=80): Minimum similarity score (0-100) for fuzzy matching

Return Value

Returns a pandas DataFrame with the following columns:

• File Name
• Page Marker
• Page Number
• Matched Term
• Original Term
• Similarity Score
• Search Term with Context (configurable width)
• Previous Words Context
• Next Words Context

Returns None if no matches are found.

Example Output

>>> results.head()
   File Name    Page Marker  Page Number  Matched Term  Original Term  Similarity Score  ...
0  example.md  document_p_1          1      keyword       keyword              100      ...

Document Format Requirements

The tool expects documents to follow the MangoCR format with page markers:

## filename_page_1
Content for page 1
## filename_page_2
Content for page 2

Error Handling

• Empty pages are automatically skipped
• Returns None if no matches are found
• Handles out-of-bounds context windows gracefully

Contributing

Feel free to open issues or submit pull requests with improvements.

License

MIT License