3 ms·
Teaching Machines to Read and Comprehend [pdf]
- deepnet 11y agonon pdf links to arxiv abstracts: Teaching Machines to Read and Comprehend by Hermann, Kočiský, Grefenstette, Espeholt, Kay, Suleyman, & Blunsom http://arxiv.org/abs/1506.03340 http://arxiv.org/abs/1506.03340 and (same group) Reasoning about Entailment with Neural Attention http://arxiv.org/abs/1509.06664 http://arxiv.org/abs/1509.06664 Learning to Transduce with Unbounded Memory http://arxiv.org/abs/1506.02516 http://arxiv.org/abs/1506.02516
- netinstructions 11y agoWhat a neat way to amass a corpus of reading comprehension data. They collected about a million news stories from CNN and Daily Mail. In addition to the news story, CNN and Daily Mail both provide a "key points" bullet list of summary details which allowed them to construct a Cloze[1] styled question like the following (from Wikipedia): Today, I went to the ________ and bought some milk and eggs. I knew it was going to rain, but I forgot to take my ________, and ended up getting wet on the way. [1] https://en.wikipedia.org/wiki/Cloze_test https://en.wikipedia.org/wiki/Cloze_test
- 100ideas 11y agoThanks for posting this, fascinating and promising work! Is "Knowledge graph" the hip new way of saying "semantic web"? or is it fundamentally different?