Open ymym3412 opened 5 years ago
A Structured Variational Autoencoder for Contextual Morphological Inflection Lawrence Wolf-Sonkin∗ Jason Naradowsky∗ Sebastian J. Mielke∗ Ryan Cotterell∗
POS tagとlemmaからInflectされたsurfaceを予測するContextual Morphological InflectionをSVAEを使って半教師ありで解く手法
従来は教師ありがメインだったが半教師ありで解いている
POS tagとlemmaの系列が与えられたときに、どういった表層系を持つ単語の系列になるかを予測するContextual Morphological Inflectionを生成モデルで解く手法。 tag, lemma, formの同時確率を分解し、個々のモデルでモデリング。 またラベルなしデータを学習に活用するために、Structured VAEを使った変分推論による学習を導入した。
従来のFST, NNを使った手法と比べると高いAccuracyで安定している
FSTではprefix, suffixなどのmorph間の結合に引っ張られがちだが、提案手法ではその影響は小さい
The CoNLL-SIGMORPHON 2017 shared task: Universal morphological reinflection in 52 languages Ryan Cotterell, Christo Kirov, John Sylak-Glassman, Géraldine Walther, Ekaterina Vylomova, Patrick Xia, Manaal Faruqui, Sandra Kübler, David Yarowsky, Jason Eisner, and Mans Hulden
The SIGMORPHON 2016 shared task— morphological reinflection Ryan Cotterell, Christo Kirov, John Sylak-Glassman, David Yarowsky, Jason Eisner, and Mans Hulden.
0. 論文
A Structured Variational Autoencoder for Contextual Morphological Inflection Lawrence Wolf-Sonkin∗ Jason Naradowsky∗ Sebastian J. Mielke∗ Ryan Cotterell∗
1. どんなもの?
POS tagとlemmaからInflectされたsurfaceを予測するContextual Morphological InflectionをSVAEを使って半教師ありで解く手法
2. 先行研究と比べてどこがすごい?
従来は教師ありがメインだったが半教師ありで解いている
3. 技術や手法のキモはどこ?
POS tagとlemmaの系列が与えられたときに、どういった表層系を持つ単語の系列になるかを予測するContextual Morphological Inflectionを生成モデルで解く手法。 tag, lemma, formの同時確率を分解し、個々のモデルでモデリング。 またラベルなしデータを学習に活用するために、Structured VAEを使った変分推論による学習を導入した。
4. どうやって有効だと検証した?
従来のFST, NNを使った手法と比べると高いAccuracyで安定している
5. 議論はある?
FSTではprefix, suffixなどのmorph間の結合に引っ張られがちだが、提案手法ではその影響は小さい
6. 次に読むべき論文は?
The CoNLL-SIGMORPHON 2017 shared task: Universal morphological reinflection in 52 languages Ryan Cotterell, Christo Kirov, John Sylak-Glassman, Géraldine Walther, Ekaterina Vylomova, Patrick Xia, Manaal Faruqui, Sandra Kübler, David Yarowsky, Jason Eisner, and Mans Hulden
The SIGMORPHON 2016 shared task— morphological reinflection Ryan Cotterell, Christo Kirov, John Sylak-Glassman, David Yarowsky, Jason Eisner, and Mans Hulden.