Text Aligner

A Python library for aligning texts based on edit distance algorithms. This tool is particularly useful for comparing and aligning reference texts with hypothesis texts, such as in speech recognition evaluation.

Features

• Edit Distance Alignment: Uses sequence matching algorithms to align texts based on similarity
• Space Agnostic Mode: Option to ignore differences in spacing during alignment
• Punctuation Agnostic Mode: Option to ignore differences in punctuation during alignment

Installation

pip install text-aligner

Usage

Command Line Interface

# Align two individual text strings
align-text "AI cannot replace human creativity" "AI can not replace human creativity."
# Output: AI cannot replace human creativity

align-text "AI can not replace human creativity." "AI cannot replace human creativity"
# Output: AI can not replace human creativity.

align-text "He is a well-known writer." "She is a well known writer"
# Output: She is a well-known writer.

align-text "She is a well known writer" "He is a well-known writer."
# Output: He is a well known writer

# Align texts from scp format files (e.g., `utt_id text`)
align-text ref.txt hyp.txt output.txt

Python API

from text_aligner import align

# Preserve reference formatting with hypothesis content
align("AI cannot replace human creativity", "AI can not replace human creativity.")
# Output: "AI cannot replace human creativity"

align("He is a well-known writer.", "She is a well known writer")
# Output: "She is a well-known writer."

Options

• -s, --space-agnostic: Ignore differences in spacing (default: True)
• -p, --punctuation-agnostic: Ignore differences in punctuation (default: True)
• output-file: Optional output file path (default: stdout)

License

Apache License 2.0