1.2.0-BETA2

🚀 MultiClass Logistic Regression
🚀 OneHotEncoder

1.2.0-BETA

🚀 TransfomersPipeline

✅ PyTorch through JIT-saved TorchScript models
✅ ONNX Models directly through the hub, e.g. TokenClassificationPipeline.create("optimum/bert-base-NER")
Where optimum/bert-base-NER is a model on the HuggingFace Hub
✅ Load both PyTorch (TorchScript) & ONNX model through local path
ClassificationPipeline and TokenClassificationPipeline exists

See the following test for some examples on how to use it

1.1.1

Bumping DJL to 1.5.0 as it's now released for real.

Full Changelog: v1.1.0...v1.1.1

1.1.0

What's Changed

• feat: BagOfWords, TfIdf & BM-25 by @Lundez in #42
• feat: Adding new CLF by @Lundez in #46
• feat: Cooccurence keywords by @Lundez in #77
• perf: LightWordEmbeddings with more efficient caching by @Lundez in #81
• docs: Adding doc-generation and deployment by @Lundez in #82
• chore: Bump multiple dependencies by @Lundez

Full Changelog: v1.0.0...v1.1.0

1.1.0-BETA

This is the initial BETA for 1.1.0

🚀 Vectorizers (BagOfWord through CountVectorizer)
🚀 Transformers (TF-IDF, BM25 which also exists as Vectorizers using BagOfWord as input)
🚀 Regression (SimpleLinearRegression)
🚀 Classifier (LogisticRegression without intercept & Naïve Bayes)
🚀 Sequence Classifier (Hidden Markov Model)

✅ Moved majority of code to multik
✅ Started adding DJL PyTorch Tensor support, ramping up for neural networks
✅ Added some Metrics

🙄 ...And some extra!

1.0.0

🎉 1.0.0! 🎉

It's finally here.
For this update some dependencies has been pruned & tests has been added. Following we'll add ourself to kotlin-jupyter.

1.0-beta

First beta 🎉

API is stabilizing.

🚀 BytePieceEmbeddings (https://nlp.h-its.org/bpemb/) -- Supporting 275 (!) languages out of the box with a lot of customability of sizes.
🚀 SentencePiece Tokenizer -- Supporting 275 (!) languages out of the box with a lot of customability of sizes. (OBS: JNI-based)
🚀 FastText (non-ngram) support -- Supporting 175 languages out of the box.
🎉 Documentation now in a Kotlin Notebook (README.ipynb). This means you can run the code yourself simply locally
... And some minor bugg-fixes in DownloadHelper where it'd redownload some WordFrequencies etc.

1.0-SNAPSHOT

This is a SNAPSHOT of the 1.0 release.
Stability of API should be ok for "completed" segments.

✅ Stopwords
✅ WordFrequencies
✅ Tokenizer + CharTokenizer & SimpleTokenizer
✅ Basic Trie Structure (no 'merge node' function yet)
✅ Stemmer
✅ Embeddings (Word Embeddings & Light Word Embeddings)
❓ Sentence Embeddings (AvgSentence & USif should be good to go!)