1
$\begingroup$

I actually have a basic doubt in NLP,

When we consider traditional models like Decision trees, The feature column order is important, Like first column is fixed with some particular attribute. So If, I have Tf-Idf Each word will have some fixed index and the model can learn.

But in the case of LSTM, Sentences can be jumbled. For eg: "There is heavy rain", "Heavy rain is there"

In the above 2 sentences, The word heavy occurs in different places. So in order for the model to understand that we have passed the word "There", We would require some unique representations for the word "there". Either a One-Hot or Word2vec. Is my understanding so far right?

My final doubt is, If I use tfidf for the above, How will it work? How will the model understand that "heavy" word is passed? This doubt has been bugging me for long. Kindly clarify this! Thanks a ton!

$\endgroup$

1 Answer 1

1
$\begingroup$

First of all in BOW model, the order is not represented. Decision Tree does not care if "heavy" is the first feature or last feature and it works the same for both as BOW just models the "existence" of words in documents. So your final doubt is actually nothing to worry about. In both sentences you have the word "heavy" and in that column you get a $1$ for both sentences (or TF, or TF-IDF or any other count you use).

LSTM sees the order as it has a memorising behaviour. That means word "heavy" has an index and during long training, your model learns the probability of usage of word heavy. That means it models your text. so your understanding is right, there is a representation of words for last too.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.