文献紹介 4月18日

Transcript

1. ) ( Towards a Seamless Integration of Word Senses into Downstream NLP Applications
2. 
   ▪  ▪ Mohammad Taher Pilehvar, Jose Camacho-Collados, Roberto Navigli and Nigel Collier ▪ Towards a Seamless Integration of Word Senses into Downstream NLP Applications ▪ Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics ▪ pp. 1857--1869 ▪  ▪  
   ,  2
3. 
    ▪ ) N P P i P ▪ topic categorization polarity detection P i L P ▪ S P ( ) NA i NT 3
4. 
    ▪ L P P P ▪ P N ▪ L N P N ▪ it can hamper their efficiency in handling words that are not encountered frequently during training, such as multiwords, inflections and derivations. ▪ N N ▪ it can restrict their semantic understanding to the level of words, with all their ambiguities, and thereby prevent accurate capture of the intended meanings. 4
5. 
    ▪ ) ( ( D S C S ▪ S 2 ( ) - ( 2 W S 5
6. 
    ▪ RaceHamilton, Formula      Entity"% ▪ $!#
      6
7. 
    ▪ b d ( ( ) ae ng i ▪ S n o mn m o a d ur eW oD ▪ multiple topic categorization ▪ polarity detection ▪ tu o n s ui o 7
8. 
    ▪ '#%*   )(!& ▪  '#,  *" 
    8
9. 
    ▪ '#%*   )(!& ▪  '#,  *" 
    9
10. 
     ▪ 10
11. 
     ▪ Embedding Layer! 
     ! "3*  ▪ 46/3 ▪ DeConf $)(!&) "2,05 (Pilehvar and Collier, 2016) ▪ &'""! %"*) "2,05 ▪ event, animal, and quantity "44,4-. 11
12. 
    12 ▪Classification model ▪ Embedding Layer 300 dimention ▪ 10-fold cross-validation ▪Semantic Network ▪ network of Wikipedia hyperlinks ▪ WordNet via the mapping provided by BabelNet ▪Pre-trained word and sense embeddings ▪ Word2vec
13. 
    
14. 
    
15. 
    
16. 
     ▪ &D@BE5-= CBA435'&0$# )! 5-=+2>,/ ▪ 1! 57.6$#6! :% 6" )54=+2>,/ ▪ :;< 4> ,0$#6 >;<90&*/& 16