Dear Josimar,

The idea of transforming text into image is not clear to me. What is the purpose? Is there any applicability of this transformation?

If you have pdf you can use an OCR for text regognition, then you can esaily perform the feature extraction. Anyway, it is better if you can work with txt, csv, doc, or similar files.

Thank you. I was talking about transform text to image, i.e. a paper to one matrix of pixels to extract features.

Dear Josimar,

The idea of transforming text into image is not clear to me. What is the purpose? Is there any applicability of this transformation?

I could get features from a text using NLP algorithms, then I was wondering if there is a transformation function to get an image from a text. If I have an image I can use Image Processing algorithms to represent text.

I agree with Sobhan Sarkar

++

Text is a 1 dimensional analysis domain, which is very easy compared to analyzing 2-dimensional image objects.

Really, it makes no sense to convert text to images. What extra features do you get? The text font maybe?

It just was an idea because there very well-know algorithms to get features from images and I was trying to explore new ways to get features from text. The next idea was to use it for steganography and hide text in images.

There are no extra features to be found, unless the way the text is printed conveys extra meaning, like a calligram:

I think from text you can extract variety of features such as orthographic, lexico,syntactic and semantic and you can apply deep learning based word embedding approaches so that 100's to 1000's of features can be extracted from text. Even surrounding text can improve the performance in TM. If that is not the case you can use vector representation of the text in which matrix can easily be generated using which kernel based or graph based models with 100's of features can be generated. Using feature selection methods such as MOO or PSO is an import in TM because 100's of feature can easily be generated from which we have to identify the best set for higher performance of our classifier.

Thank you for the answers.