CODEWITHSUNDEEP
pythontensorflowkeras
Unterbelichtet
Unterbelichtet

Reputation: 628

Tensorflow predict likeliness of two sentences to go after each other

I use AWS Textract to extract text from documents, however, some sentences are ripped apart which I need to put back together using Tensorflow. I'm quite new to TensorFlow, therefore I don't know what model configuration I need to achieve this. Here are some examples of what I need the model to predict. This is the input:

[
 [
  {
   "part": "I think that"
  },
  {
   "part": "TensorFlow is great."
  },
  "label": 1
 ],
 [
  {
   "part": "I'm"
  },
  {
   "part": "computer in the room."
  },
  "label": 0
 ],
]

I need the model to predict the label, how likely it is for these two strings to go together.

What model architecture do I need to do this?

Upvotes: 0

Views: 56

Answers (1)

Andrey
Andrey

Reputation: 6367

Use sequence classification model, e.g. BERT from Huggingface transformers.

Upvotes: 3

Related Questions