paragrafs

A lightweight TypeScript library designed to reconstruct paragraphs from AI transcriptions. It helps format unstructured text with appropriate paragraph breaks, handles timestamps for transcripts, and optimizes for readability.

Features

• Segment Recognition: Intelligently groups text into logical paragraphs
• Filler Removal: Identifies and removes common speech fillers (uh, umm, etc.)
• Gap Detection: Detects significant pauses to identify paragraph breaks
• Timestamp Formatting: Converts seconds to readable timestamps (HH:MM:SS)
• Punctuation Awareness: Uses punctuation to identify natural segment breaks
• Customizable Parameters: Configure minimum words per segment, max segment length, etc.
• Arabic Support: Handles Arabic question marks and other non-Latin punctuation
• Transcript Formatting: Converts raw token streams into readable text with appropriate line breaks
• Ground-Truth Token Mapping: Aligns AI-generated word timestamps to human-edited transcript text using an LCS-based algorithm with intelligent interpolation

Installation

npm install paragrafs

or

pnpm install paragrafs

or

yarn add paragrafs

or

bun add paragrafs

Usage

Basic Example

import { estimateSegmentFromToken, markAndCombineSegments, mapSegmentsIntoFormattedSegments } from 'paragrafs';

// Example token from transcription
const token = {
    start: 0,
    end: 5,
    text: 'This is a sample text. It should be properly segmented.',
};

// Estimate segment with word-level tokens
const segment = estimateSegmentFromToken(token);

// Combine and format segments
const formattedSegments = mapSegmentsIntoFormattedSegments([segment]);

console.log(formattedSegments[0].text);
// Output: "This is a sample text. It should be properly segmented."

Working with Transcriptions

import {
    markAndCombineSegments,
    mapSegmentsIntoFormattedSegments,
    formatSegmentsToTimestampedTranscript,
} from 'paragrafs';

// Example transcription segments
const segments = [
    {
        start: 0,
        end: 6.5,
        text: 'The quick brown fox!',
        tokens: [
            { start: 0, end: 1, text: 'The' },
            { start: 1, end: 2, text: 'quick' },
            { start: 2, end: 3, text: 'brown' },
            { start: 3, end: 6.5, text: 'fox!' },
        ],
    },
    {
        start: 8,
        end: 13,
        text: 'Jumps right over the',
        tokens: [
            { start: 8, end: 9, text: 'Jumps' },
            { start: 9, end: 10, text: 'right' },
            { start: 10, end: 11, text: 'over' },
            { start: 12, end: 13, text: 'the' },
        ],
    },
];

// Options for segment formatting
const options = {
    fillers: ['uh', 'umm', 'hmmm'],
    gapThreshold: 3,
    maxSecondsPerSegment: 12,
    minWordsPerSegment: 3,
};

// Process the segments
const combinedSegments = markAndCombineSegments(segments, options);
const formattedSegments = mapSegmentsIntoFormattedSegments(combinedSegments);

// Get timestamped transcript
const transcript = formatSegmentsToTimestampedTranscript(combinedSegments, 10);

console.log(transcript);
// Output:
// 0:00: The quick brown fox!
// 0:08: Jumps right over the

Aligning AI Tokens to Human-Edited Text

import { updateSegmentWithGroundTruth } from 'paragrafs';

const rawSegment = {
    start: 0,
    end: 10,
    text: 'The Buick crown flock jumps right over the crazy dog.',
    tokens: [
        /* AI-generated word timestamps */
    ],
};

const aligned = updateSegmentWithGroundTruth(rawSegment, 'The quick brown fox jumps right over the lazy dog.');
console.log(aligned.tokens);
// Each token now matches the ground-truth words exactly,
// with missing words interpolated where needed.

API Reference

Core Functions

estimateSegmentFromToken(token: Token): Segment

Splits a single token into word-level tokens and estimates timing for each word.

markTokensWithDividers(tokens: Token[], options): MarkedToken[]

Marks tokens with segment breaks based on fillers, gaps, and punctuation.

groupMarkedTokensIntoSegments(markedTokens: MarkedToken[], maxSecondsPerSegment: number): MarkedSegment[]

Groups marked tokens into logical segments based on maximum segment length.

mergeShortSegmentsWithPrevious(segments: MarkedSegment[], minWordsPerSegment: number): MarkedSegment[]

Merges segments with too few words into the previous segment.

mapSegmentsIntoFormattedSegments(segments: MarkedSegment[], maxSecondsPerLine?: number): Segment[]

Converts marked segments into clean, formatted segments with proper text representation.

formatSegmentsToTimestampedTranscript(segments: MarkedSegment[], maxSecondsPerLine: number): string

Formats segments into a human-readable transcript with timestamps.

markAndCombineSegments(segments: Segment[], options): MarkedSegment[]

Combined utility that processes segments through all the necessary steps.

mapTokensToGroundTruth(segment: Segment): Segment

Synchronizes AI-generated word timestamps with the human-edited transcript (segment.text):

• Uses a longest-common-subsequence (LCS) to find matching words and preserve their original timing.
• Evenly interpolates timestamps for runs of missing words (only when two or more are missing).
• Falls back to estimateSegmentFromToken if no matches are found.

Types

type Token = {
    start: number; // Start time in seconds
    end: number; // End time in seconds
    text: string; // The transcribed text
};

type Segment = Token & {
    tokens: Token[]; // Word-by-word breakdown with timings
};

type MarkedToken = 'SEGMENT_BREAK' | Token;

type MarkedSegment = {
    start: number;
    end: number;
    tokens: MarkedToken[];
};

Utility Functions

isEndingWithPunctuation(text: string): boolean

Checks if the text ends with punctuation (including Arabic punctuation).

formatSecondsToTimestamp(seconds: number): string

Formats seconds into a human-readable timestamp (H:MM:SS).

Use Cases

• Transcript Formatting: Convert raw transcriptions into readable text
• Subtitle Generation: Create properly formatted subtitles from audio transcriptions
• Document Reconstruction: Rebuild properly formatted documents from extracted text

Contributing

Contributions are welcome! Please make sure your contributions adhere to the coding standards and are accompanied by relevant tests.

To get started:

1. Fork the repository
2. Install dependencies: bun install (requires Bun)
3. Make your changes
4. Run tests: bun test
5. Submit a pull request

License

paragrafs is released under the MIT License. See the LICENSE.MD file for more details.

Author

Ragaeeb Haq

Built with TypeScript and Bun. Uses ESM module format.