Spark NLP release notes 3.13.0

3.13.0

Release date: 25-05-2022

We are glad to announce that Spark OCR 3.13.0 has been released! This release focuses around VisualDocumentNer models, adding ability to fine-tune, fixing bugs, and to leverage the Annotation Lab to generate training data.

New Features

• VisualDocumentNerV21:
  • Now you can fine tune models VisualDocumentNerV21 models on your own dataset.
• AlabReaders:
  • New class to allow training data from the Annotation Lab to be imported into Spark OCR. Currently, the reader supports Visual Ner only.

Bug Fixes

• Feature extraction on VisualDocumentNer has been improved.

New notebooks

• SparkOcrFormRecognitionFineTuning.ipynb, end to end example on Visual Document Ner Fine-Tuning.
• Databricks notebooks on Github Spark-OCR Workshop repository have been updated, and fixed.