Update README.md

README.md

@@ -7,7 +7,7 @@ inference: false
 
 We have trained a model to evaluate if a paraphrase is a semantic variation to the input query or just a surface level variation. Data augmentation by adding Surface level variations does not add much value to the NLP model training. if the approach to paraphrase generation is "OverGenerate and Rank" , Its important to have a robust model of scoring/ ranking paraphrases. NLG Metrics like bleu ,BleuRT, gleu , Meteor have not proved very effective in scoring paraphrases. 
 
-```
+```python
 from transformers import AutoTokenizer, AutoModelForSequenceClassification  
 import torch
 import pandas as pd
@@ -55,7 +55,7 @@ print("Paraphrase type detection=====", list(zip(para_pairs, labels)))
 For more robust results, filter out the paraphrases which are not semantically 
 similar using a model trained on NLI, STS task and then apply the ranker .
 
-```
+```python
 from transformers import AutoTokenizer, AutoModelWithLMHead
 from transformers import AutoModelForSequenceClassification
 from sentence_transformers import SentenceTransformer, util