MRPC-Paraphrasing

Using MRPC dataset identify whether 2 sentences are similar in meaning or not.

I've implemented ensemble model using xgboost and LSTM (Siamese LSTM). The LSTM implementation is inspired from this post. I have also created it as an API to directly query for similarity. The docker image can be found here. This is the first time I've worked with docker so please ignore naming conventions, it was just a try :)

The service expects a form JSON with key "injson". A sample input:

{"Sentence1": "PCCW 's chief operating officer , Mike Butcher , and Alex Arena , the chief financial officer , will report directly to Mr So .",
"Sentence2": "Current Chief Operating Officer Mike Butcher and Group Chief Financial Officer Alex Arena will report to So ."}

Curl request:

curl --location --request POST 'http://localhost:4002/get_sentence_similarity' \
--form 'injson={"Sentence1": "PCCW '\''s chief operating officer , Mike Butcher , and Alex Arena , the chief financial officer , will report directly to Mr So .",
"Sentence2": "Current Chief Operating Officer Mike Butcher and Group Chief Financial Officer Alex Arena will report to So ."}'

Response:

[
  { 
    "Sentence1": "PCCW 's chief operating officer , Mike Butcher , and Alex Arena , the chief financial officer , will report directly to Mr So .",
    "Sentence2": "Current Chief Operating Officer Mike Butcher and Group Chief Financial Officer Alex Arena will report to So .",
    "isSimilar": 1
  }
]