Using NLP to recognize the timeliness of text content

Question

NLP models can classify text content as positive or negative. Except that, we also need to know the timeliness of such text content. That is whether the text is describing something that has already happened or predicting something that might happen in the future. For example:

Text 1: Stock A has risen more than 7.5%. (positive, but happened already)

Text 2: Stock B might be exposed to downside risk due to its poor financial situation (negative prediction).

It is obvious that Text 2 contains more useful information because it is a prediction of the future while Text 1 is just a description of the past. However, current NLP models, as I understand, cannot recognize such timeliness inside the text content.

Dose anyone have any ideas about this problem? Any introduction of academic journals and surveys would be very useful.

Answer 1

Assuming your goal is to infer whether a sentence is: positive already happened, positive likely to happen, negative already happened or negative likely to happen; you end up with a 4-classes classification problem, which you need to label in advance (this would be, if feasible, the tedious-human work).

After that, you can also apply word embedding layers to your model, to capture the meaning you would like to get out of your vocabulary. If you want a reference with an example from the Keras documentation, you can have a look at it: NLP approaches to infer Processes from Text