Visualizing Text Classification

Question

Suppose I have a sentence of words, where each word (or character) has an associated numeric value, or color. As an example, this could be coming from an RNN sentiment classifier, which would produce something like:

I'm looking for a lightweight way of visualizing words/characters in sentences in Jupyter with python. Is there an elegant way of doing this inline in a notebook? I've mostly seen this done with additional javascript, in a separate html file. Note that I'm fine with just a static visualization. I've seen you can change the color of the font of each letter, but I'd prefer to just manipulate the background color (fill?), while keeping the text black. I'm just not sure what this is referred to.

Answer 1

Not sure if there is a preferred way to achieve this; here is a quick approach that uses a tiny html template and IPython.display.display_html:

from IPython.display import display_html

def to_html(text, r, g, b):
    return "<var style='background-color:rgb({}, {}, {});'>{} </var>".format(
        r, g, b, text
    )

example = "A quick brown fox jumps over the lazy dog.".split()
res = ''.join(to_html(word, *np.random.randint(0,256,size=3)) for word in example)
display_html(res, raw=True)

Resulting in something like this:

<var style='background-color:rgb(144, 237, 221);'>A </var><var style='background-color:rgb(28, 208, 84);'>quick </var><var style='background-color:rgb(142, 241, 214);'>brown </var><var style='background-color:rgb(67, 199, 115);'>fox </var><var style='background-color:rgb(121, 120, 116);'>jumps </var><var style='background-color:rgb(251, 46, 48);'>over </var><var style='background-color:rgb(128, 147, 44);'>the </var><var style='background-color:rgb(48, 215, 5);'>lazy </var><var style='background-color:rgb(239, 90, 48);'>dog. </var>