Can Tensor Flow be used for unsupervised learning for solving nlp classification?

Question

Is it possible to use Tensorflow for text classification problems for unlabeled data?

score 1 · Accepted Answer · answered Mar 15 '20 at 19:34

TensorFlow isn't a complete solution by itself, it's a set of tools that come together to help solve problems relying on high computational workloads (like unsupervised deep learning). These tools are optimized to help solve some of the complex mathematics that deep learning requires.

Unsupervised NLP learning problems typically comprise clustering (sorting based on unique attributes), anomaly detection, association mining, or feature reduction. If these are what you meant in your question, then deep learning via TensorFlow tools can certainly help you with your problem. If you meant 'classification,' as in the mapping of inputs to a limited set of outputs, then this is a supervised (i.e. labeled) learning problem.

Tools like TensorFlow and PyTorch can still help in supervised learning scenarios, the objective is just different.

I find q&a like this useful. I wish there was a site dedicated to these. Trying to learn about tensorflow is like opening an encyclopaedia, every page is a barrage of technical information which seems to bear no relation to my problem, yet pre-trained tools seem to solve a very similar problem. But how did they get from here to there? — silicontrip, Aug 24 '22 at 12:17

score 0 · Answer 2 · answered Mar 15 '20 at 15:42

First, text classification problem is supervised. Thus, you will need labeled data, it can't be done supervised. If you want to make it unsupervised, you need to do clustering instead, but you won't be able to make sure that each cluster will be classified according to your label.

In case you have very limited access to label data, try exploring on semi-supervised or supervised methods.

Can Tensor Flow be used for unsupervised learning for solving nlp classification?

2 Answers2