AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Improving Neural Question Generation Using Answer Separation

Yanghoon Kim, Hwanhee Lee, Joongbo Shin, Kyomin Jung

AAAI 2019 · Large language models › Retrieval and knowledge · approach: Explanatory and multiple-choice QA benchmarks

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Separate passage and answer information and replace the answer span with a special token (ablation)ASs2s · SQuAD Split-1BLEU-4: 14.37 $\pm$ 0.28 -> 16.20 $\pm$ 0.32Improved
Add keyword-net to extract key information from the target answer (ablation)ASs2s · SQuAD Split-1BLEU-4: 13.95 $\pm$ 0.29 -> 16.20 $\pm$ 0.32Improved
Use the answer-separated decoder rather than a general LSTM decoder (ablation)ASs2s · SQuAD Split-1BLEU-4: 12.30 $\pm$ 0.26 -> 16.20 $\pm$ 0.32Improved
Use answer-separated seq2seq instead of seq2seq+APSQuAD question generation · Generated questionsComplete target-answer inclusion: 0.8\% -> 0.6\%Improved
Use answer-separated seq2seq instead of seq2seq+APSQuAD question generation · Generated questionsPartial target-answer inclusion: 17.3\% -> 9.5\%Improved
Replace answer-position features with answer separationQuestion-type prediction · SQuADRecall: 77.3\% -> 82.0\% (what); 56.2\% -> 74.1\% (how); 19.4\% -> 43.3\% (when); 3.4\% -> 6.1\% (which); 12.1\% -> 46.3\% (wherImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy