Sentence Embeddings - sbert medium (tuned)

Description

This model is trained to generate contextual sentence embeddings of input sentences.

Live Demo Open in Colab Download Copy S3 URI

How to use

sbiobert_embeddings = BertSentenceEmbeddings\
.pretrained("sbert_jsl_medium_umls_uncased","en","clinical/models")\
.setInputCols(["sentence"])\
.setOutputCol("sbert_embeddings")

val sbiobert_embeddings = BertSentenceEmbeddings
.pretrained("sbert_jsl_medium_umls_uncased","en","clinical/models")
.setInputCols("sentence")
.setOutputCol("sbert_embeddings")

import nlu
nlu.load("en.embed_sentence.bert.jsl_medium_umls_uncased").predict("""Put your text here.""")

Results

Gives a 768 dimensional vector representation of the sentence.

Model Information

Data Source

Tuned on MedNLI and UMLS dataset

Benchmarking

MedNLI   Score
Acc      0.744 
STS(cos) 0.695