Sentence Embeddings - Bluebert uncased (MedNLI)

Description

This model is trained to generate contextual sentence embeddings of input sentences. It has been fine-tuned on MedNLI dataset to provide sota performance on STS and SentEval Benchmarks.

Download

How to use

Use as part of an nlp pipeline with the following stages: DocumentAssembler, SentenceDetector, BertSentenceEmbeddings. The output of this model can be used in tasks like NER, Classification, Entity Resolution etc.

sbiobert_embeddings = BertSentenceEmbeddings\
     .pretrained("sbluebert_base_uncased_mli",'en','clinical/models')\
     .setInputCols(["ner_chunk_doc"])\
     .setOutputCol("sbert_embeddings")


val sbiobert_embeddings = BertSentenceEmbeddings.pretrained("sbluebert_base_uncased_mli",'en','clinical/models')
    .setInputCols(Array("ner_chunk_doc"))
    .setOutputCol("sbert_embeddings")

Results

Gives a 768 dimensional vector representation of the sentence.

Model Information

Model Name: sbluebert_base_uncased_mli
Type: BertSentenceEmbeddings
Compatibility: Spark NLP 2.6.4 +
Edition: Official
License: Licensed
Input Labels: [ner_chunk]
Output Labels: [sentence_embeddings]
Language: [en]
Case sensitive: false

Data Source

Tuned on MedNLI dataset using Bluebert weights.