Hello,
I am planning to build a textual entailment model using a traditional machine learning approach (SVM). I am using SNLI dataset for this purpose. However, I am a bit confused as to how to give the input to the model as there are 2 sentences (premise, hypothesis) and 1 label for the classification.
What I understand is, I can build all the required features(tokens, POS, NER etc.) of both premise and hypothesis and make a one-hot encoding out of these features for both premise and hypothesis. Then concatenate the one-hot encoding of both premise and hypothesis and put this concatenated list as input to the model.
An example of textual entailment looks like following.
Premise: Two men on bicycles competing in a race.
Hypothesis: People are riding bikes.
Label: entailment
Is this right? Can someone please help me in this regards? Or, maybe some code examples will be also good.
Thank you