3
$\begingroup$

In machine learning for text classification, the first step after acquiring and cleaning the data is that of feature extraction. Since computers can't understand language like humans, the language needs to be converted to a vector representation.

I'm building machine classification models on medium size dataset ~ 10k rows. One of the things with machine learning is feature extraction and the better it is the better the classifier has information and the better it performs at the task.

The common way of converting a text to a vector that I know of are:

  • TFIDF
  • POS tags
  • Entities
  • Word Embeddings (although I tend to avoid since it's trained on Google news articles and not totally works in all cases. I created my own embedding)

I am trying to find research papers that talk about more recent ways/ other ways of extracting meaningful vector representation from text. If someone knows, please share.

Thanks

$\endgroup$
  • $\begingroup$ This isn't really a meaningful question, unless you include context about what you're trying to do. There are many different NLP features, but not all are useful for a given task. So what are you trying to do exactly? $\endgroup$ – Alex R. Jan 9 '18 at 0:59
  • 2
    $\begingroup$ But aside from that, look into Word2Vec, FastText, Latent Dirichlet Allocation, and LSTM hidden layer outputs. $\endgroup$ – Alex R. Jan 9 '18 at 1:00
  • $\begingroup$ well, what I meant to ask was what all you might have tried and if you can share any post/ papers regarding that. I'm trying to build a classifier. I'll add more context to the question $\endgroup$ – nEO Jan 9 '18 at 23:21

Your Answer

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

Browse other questions tagged or ask your own question.