Open IkokObi opened 5 years ago
比較モデルとして,Semantic Similarity Model (SSM), BM25を用いている
BM25からの相対的な精度改善率
BM25とのEnsembleモデルを学習し,精度改善率を見て,改善率が小さいほど元のモデルはBM25の重要な性質を組み込めているという斬新な発想(自分が知らないだけ?)による性能比較
比較したモデルでは上手くいかなかったが,Match-Tensorでは上手くいった検索結果
納得できるミスの仕方をしていて興味深い
ざっくり言うと
キーワード
1. 情報
論文リンク
著者
Aaron Jaech, Hetunandan Kamisetty, Eric Ringger, Charlie Clarke (Facebook)
投稿日付
2. 先行研究と比べてどこがすごい?
3. 技術や手法のキモはどこ?
4. どうやって有効だと検証した?
5. 議論はある?
6. 次に読むべき論文は?