A town-hall lobby plaque is 180 words about the 1912 clock tower. A visitor asks when the tower was dedicated, and the desk wants that year taken from the plaque rather than guessed. What kind of question answering should the app run?
Select an answer to reveal the explanation.
Short Explanation
The year is already on the plaque. Extractive QA copies that span instead of inventing a dedication date. A chat guess, a historic-versus-modern label, or a rewrite-then-guess summarizer will not point at the words that are already there.
Full Explanation
Extractive question answering returns a contiguous span copied from the source passage rather than a newly generated sentence. A short civic plaque plus a fact question is the classic extractive setting: the year is already in the text. A generative chat model may fabricate a year the plaque never states. Classification and summarization are different heads and do not locate that span.