Mova0.5 pre-release⌕
☰
home / docs / corpus / open / 0.3 / Tale questions

Tale questions

12,251 short questions about the tales, one JSON object per line: story, para (paragraph the answer comes from), question, answer (at most three words, copied verbatim from the text), multi (the answer needs more than one paragraph). Files are grouped by book set: grimm, aesop, andersen, blue (Blue Fairy Book), english, indian.

How made: a large language model wrote simple questions for each paragraph with an answer that occurs literally in the text; Mova keeps a question only if its answer is found in the paragraph.

License: ours, CC BY-SA 4.0.

How Mova uses it: training and evaluation of short-answer reading in world (world read-train, world read-eval): for each question Mova finds the sentence and extracts the answer span, with an averaged perceptron over features of the sentence and the candidate.

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/corpus/open/0.3/tale-questions/