Word Embeddings : DOc2Vec Why it fails? ( twitter Sentiment Analysis)

Why does Doc2Vec performing extremenly poor while training ??
In Twitter Sentiment Analysis ( practice hackathon).
Like, It is giving extremely low F1- score in all different models. Why is it so??
Doc2Vec gives additional feature vector as Document feature vector…
But, Still, it has failed. I am curious to know the reason bcause it failed for all models…???

@AishwaryaSingh I hope you can answer this!

Hey @atul_anand ,

It depends upon how Doc2Vec is generating document level vectors? Does it sum up the individual word2vec of each words in a document? Maybe the implementation is very poor at handling the semantic meaning of text.

Feel free to share your code/links/approach then we can figure out why is this happening?

1 Like

Thanks. for the explanatin. But , I haven’t coded this by myself.
I observed in a tutorial series. You can see their conclusion above presented in the screenshot.
If you still need the code, Its in the twitter Sentiment Analysis tutorial…
Link: https://github.com/prateekjoshi565/twitter_sentiment_analysis/blob/master/code_sentiment_analysis.ipynb

I hope you come up with a better explanation now :slight_smile:

© Copyright 2013-2019 Analytics Vidhya