v0.6.3

What's Changed

• Show failing sentences for quality test by @tushuhei in #453
• Japanese model improvement by @tushuhei in #454
• Bundle budoux-th Web Components by @tushuhei in #452
• Support new CSS display syntax by @kojiishi in #483
• Remove Node.js 16 support and add Node.js 22 support by @tushuhei in #484
• Replace WBR with ZWSP in demo page by @tushuhei in #494
• Sort ICU format output by key by @tushuhei in #495
• Update JS README about Web Workers by @tushuhei in #509
• [security] Include DOMPurify in the demo bundle by @tushuhei in #658
• [nodejs] Override with [email protected] to suppress punycode deprecation warning by @tushuhei in #657
• [demo] Bind the input content and the query param by @tushuhei in #656
• Migrate to eslint flat config using @eslint/migrate-config by @tushuhei in #673
• Mention Korean support in README.md by @tushuhei in #701
• Correct a small typo/missing word in README.md by @adamsilverstein in #746
• [Java] Handle comment nodes by @tushuhei in #764
• [Java] Skip node at the end of input by @tushuhei in #765

New Contributors

• @adamsilverstein made their first contribution in #746

Full Changelog: v0.6.2...v0.6.3

v0.6.2

Thai is now supported! 🎉

What's Changed

• Add the scale argument to encode_data.py by @tushuhei in #408
• Nit fix for an ignored test by @tushuhei in #407
• Ja model improvement by @tushuhei in #410
• Add granularity option to prepare_knbc.py by @tushuhei in #417
• Add Thai language support by @tushuhei in #421
• Improve typing by @amitmarkel in #426
• Update README for Thai support by @tushuhei in #429
• Rename @returns to @return by @tushuhei in #415

New Contributors

• @amitmarkel made their first contribution in #426

Full Changelog: v0.6.1...v0.6.2

v0.6.1

What's Changed

• Bump @typescript-eslint/eslint-plugin from 6.9.1 to 6.10.0 in /javascript by @dependabot in #353
• Bump org.apache.maven.plugins:maven-surefire-plugin from 3.2.1 to 3.2.2 in /java by @dependabot in #354
• Bump actions/dependency-review-action from 3.1.1 to 3.1.2 by @dependabot in #357
• Bump @types/node from 20.8.3 to 20.9.0 in /javascript by @dependabot in #356
• Support weighted samples by @tushuhei in #358
• Fix unpaired close tags and self-closing tags by @kojiishi in #360
• [Java] Stop emitting close tags if self-closing by @kojiishi in #362
• Update Google Java Format action by @tushuhei in #363
• Bump actions/dependency-review-action from 3.1.2 to 3.1.3 by @dependabot in #364
• Bump @typescript-eslint/eslint-plugin from 6.10.0 to 6.11.0 in /javascript by @dependabot in #365
• [java] Fix errors by collapsed white spaces and <br> by @kojiishi in #367
• Bump github/codeql-action from 2.22.5 to 2.22.6 by @dependabot in #368
• [java] Replace wholeText() with NodeVisitor by @kojiishi in #369
• Implement tail for node visitor by @tushuhei in #370
• Update jsoup to 1.16.2 by @tushuhei in #371
• Version up to 0.6.1 by @tushuhei in #372

Full Changelog: v0.6.0...v0.6.1

v0.6.0

Noteworthy changes

• BudouX Web Components don't use Shadow DOM anymore. The segmentation results will be reflected in their Light DOM, where the global styles can apply. #291
• Phrases are segmented by ZWSP (U+200B) not <wbr> for a better screen reader experience. #346
• You can insert non-breaking markup (<nobr and white-space: nowrap) when you have a phrase you don't want to break. #240

What's Changed

• Remove dependency to gts by @tushuhei in #187
• Add Parser.parseBoundaries for JavaScript by @kojiishi in #234
• Replace slice with substring by @kojiishi in #241
• Support non-breaking content (<nobr> and white-space: nowrap) by @kojiishi in #240
• Make scripts run without install by @tushuhei in #239
• Add permissions to style check action by @tushuhei in #246
• Specify maxsplit to handle colon symbols properly by @tushuhei in #247
• Support non-breaking content in java by @kojiishi in #248
• Support non-breaking content in Python by @kojiishi in #251
• Nit: use get_nowait instead of get by @tushuhei in #253
• Remove utils from JavaScript module by @tushuhei in #262
• Move hasChildTextNode to HTML Processor by @tushuhei in #274
• Fix mypy issues by @tushuhei in #308
• Fix Python dependency issues by @tushuhei in #316
• Avoid inserting separators to where the source has one by @kojiishi in #342
• [Web Components] Use Light DOM instead of Shadow DOM by @tushuhei in #291
• Use ZWSP instead of WBR by @tushuhei in #346
• [Java] Use ArrayDeque instead of Stack by @tushuhei in #349
• Rename applyElement to applyToElement by @tushuhei in #348
• Update README to use ZWSP by @tushuhei in #347
• Version up to 0.6.0 by @tushuhei in #343

Full Changelog: v0.5.2...v0.6.0

v0.5.2

What's Changed

• Use overflow-wrap: anywhere; instead of overflow-wrap: break-word; by @tamanyan in #144
• Add a script to finetune models. by @tushuhei in #145
• Add quality regression test by @tushuhei in #146
• Release finetuned model by @tushuhei in #147 #154 #161
• Add validation data arg to train.py by @tushuhei in #148
• Remove direct dependency to NumPy by @tushuhei in #149
• Add a README for BudouX Scripts by @tushuhei in #155
• Add score scale arg to build_model.py by @tushuhei in #156
• Separate HTML processing as a mixin by @tushuhei in #159

New Contributors

• @step-security-bot made their first contribution in #163

Full Changelog: v0.5.1...v0.5.2

v0.5.1

What's Changed

• Add Java module by @tushuhei in #124
• Separate HTML processing as html_processor.py by @tushuhei in #126
• Fix bug with nodes to skip by @tushuhei in #127
• Rename test_utils.py to utils.py by @tushuhei in #129
• Remove test utils by @tushuhei in #130
• Universal unit testing by @tushuhei in #125
• Replace textarea with another skip node by @tushuhei in #131
• Java style fix by @tushuhei in #132
• Java code improvement by @tushuhei in #133
• Java style fix by @tushuhei in #134
• Fix mypy issue by @tushuhei in #135
• [Java] Inherit from sonatype oss parent by @tushuhei in #136
• Improve KNBC HTML Parser by @tushuhei in #137

Full Changelog: v0.5.0...v0.5.1

v0.5.0

Highlights

• No major change in using default parsers.
• If you're using a custom model, you need to update it. Read on the "Updating Models" section.
• The defineClassAs method in javascript/src/html_processor.ts is removed.

Updating Models

As described in #112, the model file structure has been updated for performance improvement and file size reduction. The change is simple; it just adds one layer depth by grouping features as the following example shows.

Before:

{"UW1:a": 123, "UW3:b": 271}

After:

{"UW1": {"a": 123}, "UW3": {"b": 271}}

You can update your custom model to the latest by running scripts/translate_model.py.

$ python translate_model.py --format=json old-model.json > new-model.json

What's Changed

• Nit fix on some test descriptions by @tushuhei in #109
• Delete unused tsconfig by @tushuhei in #110
• Add unit test for Web Components by @tushuhei in #111
• Update the model structure for faster processing by @tushuhei in #112
• Refactor feature_extractor by @tushuhei in #113
• Use tempfile for unit test by @tushuhei in #114
• Add model translator for ICU by @tushuhei in #115
• Add a model format updater by @tushuhei in #117
• Remove defineClassAs function by @tushuhei in #119
• Remove unnecessary assertion by @tushuhei in #118
• Remove skip nodes data from JS by @tushuhei in #120
• Update the Prepare KNBC script to break chunks by specified sequences by @tushuhei in #121
• Update JA model by @tushuhei in #122
• Version Bump to 0.5.0 by @tushuhei in #123

Full Changelog: v0.4.1...v0.5.0

v0.4.1

⚠️ Breaking Change ⚠️

We added a significant change to the model training script scripts/train.py.

• The --chunk-size option is removed because the bottleneck of memory consumption has shifted due to the overhaul.
• The script does not shuffle the input data any more. You need to shuffle the data by yourself using tools such as shuf if needed.

What's Changed

• Faster training with sparse matrix by @tushuhei in #103
• Add lang option to JS CLI by @tushuhei in #102
• Bump json5 from 2.2.1 to 2.2.3 in /javascript by @dependabot in #104
• Cleanup the training script by @tushuhei in #105
• More accurate Japanese model by @tushuhei in #106

Full Changelog: v0.4.0...v0.4.1

v0.4.0

What's Changed

• Traditional Chinese support by @tushuhei in #101

Full Changelog: v0.3.0...v0.4.0

v0.3.0

What's Changed

Faster model training

We made model training faster by applying JAX's JIT compilation, pooling file writes, etc.

• Faster training data encoding by @tushuhei in #89
• Add out_span option for better GPU utilization by @tushuhei in #90
• Apply JAX JIT compiling for faster training by @tushuhei in #95
• Check in updated Simplified Chinese model by @tushuhei in #99

Smaller models

We made models smaller by removing less important features, disabling ASCII encoding, etc.

• Remove Unicode Block features by @tushuhei in #86
• Disable ASCII encoding when building the model file by @tushuhei in #98
• Output compact model by @tushuhei in #100

Misc

• encode_data: write without break line join by @tushuhei in #91
• Update unit tests for the encoding script by @tushuhei in #92
• Add more granularity in weight outputs by @tushuhei in #93
• Remove tar module dependency by @tushuhei in #96

Full Changelog: v0.2.1...v0.3.0