Natural language can describe events and relations that are likely or unlikely, possible or impossible, etc. We need to be able to handle these extra-propositional aspects of meaning, sometimes known as modality and negation.

What is modality?

It refers to the speaker’s attitude towards its own statements. How certain, reliable, and subjective the speaker is. Where does the sources of information comes from and what’s the perspective?

You can categorise modality in several ways including:

  • Future
  • Interrogative
  • Imperative
  • Conditional
  • Subjective

Information extraction is concerned with negation and certainty. You can link negation with modal calculus of certainty, likelihood, and possibility, leading to a 2D schema. This form the basis for the FactBank corpus, with annotations of the factuality (fact-checking) of all sentences in 208 documents of news articles.

The concept hedging, referring to situation when speakers limit their commitment to a proposition as shown below:

In sentence a, the word “suggest” and “might” sends off the signal of uncertainty. In sentence b, the word “technically” tells us that the evidence for the proposition might not be what the reader’s think it is. Hedging is popular within scientific texts.

Another concept related to modality is evidentiality, whereby the speaker marks the information source.



Data Scientist

Leave a Reply