chunkseg 0.1.0__tar.gz

chunkseg-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

chunkseg-0.1.0/PKG-INFO

@@ -0,0 +1,283 @@
+Metadata-Version: 2.4
+Name: chunkseg
+Version: 0.1.0
+Summary: Evaluate segmentation quality for audio and video content in a discretized time space
+Author-email: Fabian Retkowski <f@retkow.ski>
+License: MIT
+Project-URL: Homepage, https://github.com/retkowski/chunkseg
+Project-URL: Repository, https://github.com/retkowski/chunkseg
+Project-URL: Issues, https://github.com/retkowski/chunkseg/issues
+Keywords: segmentation,evaluation,audio,video,chapter,boundary,temporal
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Multimedia :: Sound/Audio :: Analysis
+Classifier: Topic :: Scientific/Engineering
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: numpy
+Requires-Dist: segeval
+Requires-Dist: nltk
+Provides-Extra: align
+Requires-Dist: alqalign; extra == "align"
+Dynamic: license-file
+
+# chunkseg
+
+A lightweight Python package for evaluating **segmentation quality** (also called **chaptering**) in audio and video content. Chunkseg evaluates in the time space rather text space and thus is transcript-invariant, enabling comparability between a broad set of different models. The package converts segment boundaries into **fixed-size time chunks** and computes **established and comprehensive metrics** including binary classification scores (precision, recall, F1) and segmentation-specific measures (Pk, WindowDiff, Boundary Similarity, GHD).
+
+The package supports **direct timestamp evaluation** or **automatic boundary extraction from structured transcripts via forced alignment**.
+
+## Install
+
+```bash
+pip install chunkseg
+```
+
+For transcript alignment support (requires `alqalign`):
+
+```bash
+pip install chunkseg[align]
+```
+
+## Evaluation Modes
+
+Chunkseg supports three evaluation modes depending on your input format:
+
+### 1. Timestamps Mode
+Provide boundary timestamps directly as lists of floats. Use when you already have predicted timestamps (e.g., from a timestamps-only model). **No audio or transcript needed.**
+
+### 2. Structured Transcript (Forced Alignment)
+Provide a structured transcript without timestamps (e.g., `[CSTART] Title [CEND] text...`). Use when your model produces chapter structure but no timestamps. **Requires audio file and `alqalign`.**
+
+**How it works:** Parse transcript → sentence-tokenize → align to audio → derive boundary timestamps from aligned sections.
+
+### 3. Structured Transcript with Timestamps
+Provide a structured transcript with embedded timestamps (e.g., `[CSTART] 1:23:45 - Title [CEND] text...`). Use when your model produces both structure and timestamps.
+
+**Two modes:**
+- **Use provided timestamps** (default): Fast, no alignment needed, no audio required
+- **Force alignment** (`--force-alignment`): Ignores timestamps, uses audio alignment instead (requires audio + `alqalign`)
+
+## Quick Reference
+
+| Mode | Format | Audio Required? | `force_alignment` | Example |
+|------|--------|-----------------|-------------------|---------|
+| **Timestamps** | `list[float]` | No | N/A | `hypothesis=[0.0, 125.0]` |
+| **Transcript (FA)** | `cstart`, `markdown`, etc. | Yes | N/A (always aligns) | `format="cstart", audio="..."` |
+| **Transcript + Timestamps (use provided)** | `cstart_ts`, `markdown_ts`, etc. | No | `False` (default) | `format="cstart_ts"` |
+| **Transcript + Timestamps (force FA)** | `cstart_ts`, `markdown_ts`, etc. | Yes | `True` | `format="cstart_ts", force_alignment=True` |
+
+## Usage
+
+### Python API
+
+```python
+from chunkseg import evaluate, evaluate_batch, print_results
+
+# Mode 1: Timestamps mode
+result = evaluate(
+    hypothesis=[0.0, 120.5, 300.0],
+    reference=[0.0, 125.0, 310.0],
+    duration=600.0,
+    chunk_size=6.0,
+)
+
+# Mode 2: Structured transcript (forced alignment)
+result = evaluate(
+    hypothesis="[CSTART] Intro [CEND] text... [CSTART] Main [CEND] more...",
+    reference=[0.0, 125.0],
+    audio="/path/to/audio.wav",
+    duration=600.0,
+    format="cstart",
+    lang="eng",
+)
+
+# Mode 3a: Transcript with timestamps (use provided)
+result = evaluate(
+    hypothesis="[CSTART] 0:00:00 - Intro [CEND] text... [CSTART] 2:05:30 - Main [CEND]...",
+    reference=[0.0, 125.0],
+    duration=600.0,
+    format="cstart_ts",  # timestamps embedded, use them
+)
+
+# Mode 3b: Transcript with timestamps (force alignment)
+result = evaluate(
+    hypothesis="[CSTART] 0:00:00 - Intro [CEND] text... [CSTART] 2:05:30 - Main [CEND]...",
+    reference=[0.0, 125.0],
+    audio="/path/to/audio.wav",
+    duration=600.0,
+    format="cstart_ts",
+    force_alignment=True,  # ignore timestamps, use alignment
+    lang="eng",
+)
+
+# Batch evaluation
+results = evaluate_batch(
+    samples=[
+        {"hypothesis": [0.0, 120.5], "reference": [0.0, 125.0], "duration": 600.0},
+        {"hypothesis": [0.0, 300.0], "reference": [0.0, 310.0], "duration": 500.0},
+    ],
+    chunk_size=6.0,
+)
+print_results(results)
+```
+
+### CLI
+
+```bash
+# Mode 1: Timestamps mode
+chunkseg timestamps.jsonl --chunk-size 6.0
+
+# Mode 2: Structured transcript (forced alignment)
+chunkseg transcripts.jsonl --chunk-size 6.0 --format cstart --lang eng
+
+# Mode 3a: Transcript with timestamps (use provided)
+chunkseg transcripts_with_ts.jsonl --chunk-size 6.0 --format cstart_ts
+
+# Mode 3b: Transcript with timestamps (force alignment, ignore timestamps)
+chunkseg transcripts_with_ts.jsonl --chunk-size 6.0 --format cstart_ts --force-alignment --lang eng
+
+# Output to JSON file
+chunkseg samples.jsonl --chunk-size 6.0 --output results.json
+```
+
+## Input Formats
+
+### Mode 1: Timestamps Mode
+
+```json
+{"hypothesis": [0.0, 120.5, 300.0], "reference": [0.0, 125.0, 310.0], "duration": 600.0}
+```
+
+**Fields:**
+- `hypothesis`: List of predicted boundary timestamps (seconds)
+- `reference`: List of ground truth boundary timestamps (seconds)
+- `duration`: Total audio duration (seconds)
+
+### Mode 2: Structured Transcript (Forced Alignment)
+
+```json
+{"hypothesis": "[CSTART] Intro [CEND] text...", "reference": [0.0, 125.0], "audio": "/path/to/audio.wav", "duration": 600.0}
+```
+
+**Fields:**
+- `hypothesis`: Structured transcript string (no timestamps)
+- `reference`: Ground truth boundary timestamps
+- `audio`: Full path to audio file (required for alignment)
+- `duration`: Total audio duration
+
+**Note:** Requires format like `cstart`, `newline`, `markdown`, or `custom` (no `_ts` suffix).
+
+### Mode 3: Transcript with Timestamps
+
+```json
+{"hypothesis": "[CSTART] 0:00:00 - Intro [CEND] text... [CSTART] 2:05:30 - Main [CEND]...", "reference": [0.0, 125.0], "duration": 600.0}
+```
+
+**Fields:**
+- `hypothesis`: Structured transcript with embedded timestamps
+- `reference`: Ground truth boundary timestamps
+- `duration`: Total audio duration
+- `audio`: (optional) Full path to audio - only needed if using `--force-alignment`
+
+**Format:** Use `cstart_ts`, `markdown_ts`, or `custom_ts` (with `_ts` suffix).
+
+**Behavior:**
+- Default: Uses timestamps from the text
+- With `--force-alignment`: Ignores timestamps, requires `audio` field
+
+## Parser Presets
+
+| Preset | Description | Example |
+|--------|-------------|---------|
+| `cstart` | `[CSTART] Title [CEND]` tags | `[CSTART] Intro [CEND] text...` |
+| `cstart_ts` | Tags with timestamps | `[CSTART] 1:23:45 - Title [CEND] text...` |
+| `newline` | Double newline separator | `Chapter 1 text.\n\nChapter 2 text.` |
+| `markdown` | Markdown headers | `# Title\nText\n## Subtitle\nMore text` |
+| `markdown_ts` | Headers with timestamps | `# 0:15:30 - Introduction\nText...` |
+| `custom` | User-provided regex | Split on any regex pattern |
+| `custom_ts` | Regex with timestamp capture | Named groups: `(?P<timestamp>...)`, `(?P<title>...)` |
+
+### Timestamp Formats for `markdown_ts`
+
+The markdown timestamped parser supports multiple header formats:
+
+- `# 00:15:30 - Title` (timestamp first with separator)
+- `# 00:15:30 Title` (timestamp first, no separator)
+- `# Title @ 00:15:30` (timestamp last with @)
+- `## 1:23:45 - Chapter Title` (any header level)
+
+### Custom Pattern with Timestamps (`custom_ts`)
+
+For `custom_ts`, provide a regex pattern with named groups:
+
+```python
+from chunkseg import evaluate
+
+result = evaluate(
+    hypothesis="[ Intro @ 1:23:45 ] text... [ Main @ 2:00:00 ] more...",
+    reference=[0.0, 125.0],
+    duration=600.0,
+    format="custom_ts",
+    custom_pattern=r"\[\s*(?P<title>[^@]+?)\s*@\s*(?P<timestamp>\d+:\d{2}:\d{2})\s*\]",
+    timestamp_format="H:MM:SS",
+)
+```
+
+Supported `timestamp_format` values:
+- `"HH:MM:SS"` — 2-digit hours (e.g., `01:23:45`)
+- `"H:MM:SS"` — variable hours (e.g., `1:23:45`) — **default**
+- `"MM:SS"` — 2-digit minutes (e.g., `23:45`)
+- `"M:SS"` — variable minutes (e.g., `3:45`)
+- Custom regex with named groups: `(?P<h>...)`, `(?P<m>...)`, `(?P<s>...)`
+
+## Metrics
+
+### Binary Classification
+
+| Metric | Description |
+|--------|-------------|
+| F1 | Harmonic mean of precision and recall |
+| Precision | TP / (TP + FP) |
+| Recall | TP / (TP + FN) |
+| Accuracy | (TP + TN) / Total |
+| Specificity | TN / (TN + FP) |
+
+### Segmentation-Specific
+
+| Metric | Description |
+|--------|-------------|
+| Pk | Beeferman's Pk metric (via segeval) |
+| WindowDiff | Window difference metric (via segeval) |
+| Boundary Similarity | Boundary similarity score (via segeval) |
+| GHD | Generalized Hamming Distance (via nltk) |
+
+### Statistics
+
+- Bootstrap confidence intervals for all metrics
+- Standard deviation
+- Prediction and reference segment counts
+
+## Dependencies
+
+**Required:**
+- `numpy`
+- `segeval`
+- `nltk`
+
+**Optional:**
+- `alqalign` — for transcript alignment (install with `pip install chunkseg[align]`)
+
+## License
+
+This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
+
+The MIT License is a permissive open-source license that allows you to use, modify, and distribute this software for any purpose, including commercial applications, with minimal restrictions.

chunkseg-0.1.0/README.md

@@ -0,0 +1,252 @@
+# chunkseg
+
+A lightweight Python package for evaluating **segmentation quality** (also called **chaptering**) in audio and video content. Chunkseg evaluates in the time space rather text space and thus is transcript-invariant, enabling comparability between a broad set of different models. The package converts segment boundaries into **fixed-size time chunks** and computes **established and comprehensive metrics** including binary classification scores (precision, recall, F1) and segmentation-specific measures (Pk, WindowDiff, Boundary Similarity, GHD).
+
+The package supports **direct timestamp evaluation** or **automatic boundary extraction from structured transcripts via forced alignment**.
+
+## Install
+
+```bash
+pip install chunkseg
+```
+
+For transcript alignment support (requires `alqalign`):
+
+```bash
+pip install chunkseg[align]
+```
+
+## Evaluation Modes
+
+Chunkseg supports three evaluation modes depending on your input format:
+
+### 1. Timestamps Mode
+Provide boundary timestamps directly as lists of floats. Use when you already have predicted timestamps (e.g., from a timestamps-only model). **No audio or transcript needed.**
+
+### 2. Structured Transcript (Forced Alignment)
+Provide a structured transcript without timestamps (e.g., `[CSTART] Title [CEND] text...`). Use when your model produces chapter structure but no timestamps. **Requires audio file and `alqalign`.**
+
+**How it works:** Parse transcript → sentence-tokenize → align to audio → derive boundary timestamps from aligned sections.
+
+### 3. Structured Transcript with Timestamps
+Provide a structured transcript with embedded timestamps (e.g., `[CSTART] 1:23:45 - Title [CEND] text...`). Use when your model produces both structure and timestamps.
+
+**Two modes:**
+- **Use provided timestamps** (default): Fast, no alignment needed, no audio required
+- **Force alignment** (`--force-alignment`): Ignores timestamps, uses audio alignment instead (requires audio + `alqalign`)
+
+## Quick Reference
+
+| Mode | Format | Audio Required? | `force_alignment` | Example |
+|------|--------|-----------------|-------------------|---------|
+| **Timestamps** | `list[float]` | No | N/A | `hypothesis=[0.0, 125.0]` |
+| **Transcript (FA)** | `cstart`, `markdown`, etc. | Yes | N/A (always aligns) | `format="cstart", audio="..."` |
+| **Transcript + Timestamps (use provided)** | `cstart_ts`, `markdown_ts`, etc. | No | `False` (default) | `format="cstart_ts"` |
+| **Transcript + Timestamps (force FA)** | `cstart_ts`, `markdown_ts`, etc. | Yes | `True` | `format="cstart_ts", force_alignment=True` |
+
+## Usage
+
+### Python API
+
+```python
+from chunkseg import evaluate, evaluate_batch, print_results
+
+# Mode 1: Timestamps mode
+result = evaluate(
+    hypothesis=[0.0, 120.5, 300.0],
+    reference=[0.0, 125.0, 310.0],
+    duration=600.0,
+    chunk_size=6.0,
+)
+
+# Mode 2: Structured transcript (forced alignment)
+result = evaluate(
+    hypothesis="[CSTART] Intro [CEND] text... [CSTART] Main [CEND] more...",
+    reference=[0.0, 125.0],
+    audio="/path/to/audio.wav",
+    duration=600.0,
+    format="cstart",
+    lang="eng",
+)
+
+# Mode 3a: Transcript with timestamps (use provided)
+result = evaluate(
+    hypothesis="[CSTART] 0:00:00 - Intro [CEND] text... [CSTART] 2:05:30 - Main [CEND]...",
+    reference=[0.0, 125.0],
+    duration=600.0,
+    format="cstart_ts",  # timestamps embedded, use them
+)
+
+# Mode 3b: Transcript with timestamps (force alignment)
+result = evaluate(
+    hypothesis="[CSTART] 0:00:00 - Intro [CEND] text... [CSTART] 2:05:30 - Main [CEND]...",
+    reference=[0.0, 125.0],
+    audio="/path/to/audio.wav",
+    duration=600.0,
+    format="cstart_ts",
+    force_alignment=True,  # ignore timestamps, use alignment
+    lang="eng",
+)
+
+# Batch evaluation
+results = evaluate_batch(
+    samples=[
+        {"hypothesis": [0.0, 120.5], "reference": [0.0, 125.0], "duration": 600.0},
+        {"hypothesis": [0.0, 300.0], "reference": [0.0, 310.0], "duration": 500.0},
+    ],
+    chunk_size=6.0,
+)
+print_results(results)
+```
+
+### CLI
+
+```bash
+# Mode 1: Timestamps mode
+chunkseg timestamps.jsonl --chunk-size 6.0
+
+# Mode 2: Structured transcript (forced alignment)
+chunkseg transcripts.jsonl --chunk-size 6.0 --format cstart --lang eng
+
+# Mode 3a: Transcript with timestamps (use provided)
+chunkseg transcripts_with_ts.jsonl --chunk-size 6.0 --format cstart_ts
+
+# Mode 3b: Transcript with timestamps (force alignment, ignore timestamps)
+chunkseg transcripts_with_ts.jsonl --chunk-size 6.0 --format cstart_ts --force-alignment --lang eng
+
+# Output to JSON file
+chunkseg samples.jsonl --chunk-size 6.0 --output results.json
+```
+
+## Input Formats
+
+### Mode 1: Timestamps Mode
+
+```json
+{"hypothesis": [0.0, 120.5, 300.0], "reference": [0.0, 125.0, 310.0], "duration": 600.0}
+```
+
+**Fields:**
+- `hypothesis`: List of predicted boundary timestamps (seconds)
+- `reference`: List of ground truth boundary timestamps (seconds)
+- `duration`: Total audio duration (seconds)
+
+### Mode 2: Structured Transcript (Forced Alignment)
+
+```json
+{"hypothesis": "[CSTART] Intro [CEND] text...", "reference": [0.0, 125.0], "audio": "/path/to/audio.wav", "duration": 600.0}
+```
+
+**Fields:**
+- `hypothesis`: Structured transcript string (no timestamps)
+- `reference`: Ground truth boundary timestamps
+- `audio`: Full path to audio file (required for alignment)
+- `duration`: Total audio duration
+
+**Note:** Requires format like `cstart`, `newline`, `markdown`, or `custom` (no `_ts` suffix).
+
+### Mode 3: Transcript with Timestamps
+
+```json
+{"hypothesis": "[CSTART] 0:00:00 - Intro [CEND] text... [CSTART] 2:05:30 - Main [CEND]...", "reference": [0.0, 125.0], "duration": 600.0}
+```
+
+**Fields:**
+- `hypothesis`: Structured transcript with embedded timestamps
+- `reference`: Ground truth boundary timestamps
+- `duration`: Total audio duration
+- `audio`: (optional) Full path to audio - only needed if using `--force-alignment`
+
+**Format:** Use `cstart_ts`, `markdown_ts`, or `custom_ts` (with `_ts` suffix).
+
+**Behavior:**
+- Default: Uses timestamps from the text
+- With `--force-alignment`: Ignores timestamps, requires `audio` field
+
+## Parser Presets
+
+| Preset | Description | Example |
+|--------|-------------|---------|
+| `cstart` | `[CSTART] Title [CEND]` tags | `[CSTART] Intro [CEND] text...` |
+| `cstart_ts` | Tags with timestamps | `[CSTART] 1:23:45 - Title [CEND] text...` |
+| `newline` | Double newline separator | `Chapter 1 text.\n\nChapter 2 text.` |
+| `markdown` | Markdown headers | `# Title\nText\n## Subtitle\nMore text` |
+| `markdown_ts` | Headers with timestamps | `# 0:15:30 - Introduction\nText...` |
+| `custom` | User-provided regex | Split on any regex pattern |
+| `custom_ts` | Regex with timestamp capture | Named groups: `(?P<timestamp>...)`, `(?P<title>...)` |
+
+### Timestamp Formats for `markdown_ts`
+
+The markdown timestamped parser supports multiple header formats:
+
+- `# 00:15:30 - Title` (timestamp first with separator)
+- `# 00:15:30 Title` (timestamp first, no separator)
+- `# Title @ 00:15:30` (timestamp last with @)
+- `## 1:23:45 - Chapter Title` (any header level)
+
+### Custom Pattern with Timestamps (`custom_ts`)
+
+For `custom_ts`, provide a regex pattern with named groups:
+
+```python
+from chunkseg import evaluate
+
+result = evaluate(
+    hypothesis="[ Intro @ 1:23:45 ] text... [ Main @ 2:00:00 ] more...",
+    reference=[0.0, 125.0],
+    duration=600.0,
+    format="custom_ts",
+    custom_pattern=r"\[\s*(?P<title>[^@]+?)\s*@\s*(?P<timestamp>\d+:\d{2}:\d{2})\s*\]",
+    timestamp_format="H:MM:SS",
+)
+```
+
+Supported `timestamp_format` values:
+- `"HH:MM:SS"` — 2-digit hours (e.g., `01:23:45`)
+- `"H:MM:SS"` — variable hours (e.g., `1:23:45`) — **default**
+- `"MM:SS"` — 2-digit minutes (e.g., `23:45`)
+- `"M:SS"` — variable minutes (e.g., `3:45`)
+- Custom regex with named groups: `(?P<h>...)`, `(?P<m>...)`, `(?P<s>...)`
+
+## Metrics
+
+### Binary Classification
+
+| Metric | Description |
+|--------|-------------|
+| F1 | Harmonic mean of precision and recall |
+| Precision | TP / (TP + FP) |
+| Recall | TP / (TP + FN) |
+| Accuracy | (TP + TN) / Total |
+| Specificity | TN / (TN + FP) |
+
+### Segmentation-Specific
+
+| Metric | Description |
+|--------|-------------|
+| Pk | Beeferman's Pk metric (via segeval) |
+| WindowDiff | Window difference metric (via segeval) |
+| Boundary Similarity | Boundary similarity score (via segeval) |
+| GHD | Generalized Hamming Distance (via nltk) |
+
+### Statistics
+
+- Bootstrap confidence intervals for all metrics
+- Standard deviation
+- Prediction and reference segment counts
+
+## Dependencies
+
+**Required:**
+- `numpy`
+- `segeval`
+- `nltk`
+
+**Optional:**
+- `alqalign` — for transcript alignment (install with `pip install chunkseg[align]`)
+
+## License
+
+This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
+
+The MIT License is a permissive open-source license that allows you to use, modify, and distribute this software for any purpose, including commercial applications, with minimal restrictions.

chunkseg-0.1.0/pyproject.toml

@@ -0,0 +1,47 @@
+[build-system]
+requires = ["setuptools>=68.0", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "chunkseg"
+version = "0.1.0"
+description = "Evaluate segmentation quality for audio and video content in a discretized time space"
+readme = "README.md"
+license = {text = "MIT"}
+requires-python = ">=3.9"
+authors = [
+    {name = "Fabian Retkowski", email = "f@retkow.ski"}
+]
+keywords = ["segmentation", "evaluation", "audio", "video", "chapter", "boundary", "temporal"]
+classifiers = [
+    "Development Status :: 4 - Beta",
+    "Intended Audience :: Science/Research",
+    "Intended Audience :: Developers",
+    "License :: OSI Approved :: MIT License",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.9",
+    "Programming Language :: Python :: 3.10",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Topic :: Multimedia :: Sound/Audio :: Analysis",
+    "Topic :: Scientific/Engineering",
+]
+dependencies = [
+    "numpy",
+    "segeval",
+    "nltk",
+]
+
+[project.urls]
+Homepage = "https://github.com/retkowski/chunkseg"
+Repository = "https://github.com/retkowski/chunkseg"
+Issues = "https://github.com/retkowski/chunkseg/issues"
+
+[project.optional-dependencies]
+align = ["alqalign"]
+
+[project.scripts]
+chunkseg = "chunkseg.cli:main"
+
+[tool.setuptools.packages.find]
+where = ["src"]

chunkseg-0.1.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

chunkseg-0.1.0/src/chunkseg/__init__.py

@@ -0,0 +1,8 @@
+"""chunkseg — Time-chunked segmentation evaluation."""
+
+__version__ = "0.1.0"
+
+from .evaluate import evaluate, evaluate_batch
+from .display import print_results
+
+__all__ = ["evaluate", "evaluate_batch", "print_results"]