Low Accuracy in Odd one out challenge

#1

I am getting low accuracy in this odd-one out challenge with 55% accuracy. Not sure what might be the issue here, or is the difference is because of using different model or different similarity index? In case of different mode can the model link be provided as well?. The code can be checked at https://ide.codingblocks.com/s/298283

#2

hey @tisandas2011,
I guess the way or method you are performing to calculate similarity is correct.
The issue can be with embeddings.
You can try other embeddings to check if the score improves or not. Like currently you are using embeddings with 50 dimensions , give a try to 100,200,300 and there are separate embeddings provided by google news vectors and another is FastText. Give them a try and experiment how they perform.

I hope this will help you increase your results.
Thank You :slightly_smiling_face:.