文献紹介 11月7日

Explicit Retrofitting of Distributional Word Vectors

Transcript

1. Goran Glavas, Ivan Vulic Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Long Papers), pages 34–45    
   
2. )4 2 n 8"* %Retrofitting 51' ,+:&(Word vector space specialization 0 38#-/ n 96
   96 ! $2.7
3. .& 3 n *4 #/NLP ,60    n *4 #12,)-5*4+7*4 8%!"$  
   -34('   9 
4. 5' 4 n Word vector space specialization l !$3LHGV=4 *JI:+ 3L!$M
   9C l $%$@? 1PB(3L8O  n Retrofitting l 3LD!$4 0TSE R>N9C l 0TSE2.!$@? ,QUV&6 n 2 9CK- explicit retrofitting (ER) ;A l 7)F!$/<"%#$  3L
5.  5 n  !
     n !  
        
6. "$ 6 n 3)+-* !%& 1/+-
    l Ex) car – automobile - car – drive - n Negative Sampling   l  '0 #-
   K+,2  (.
7. ") 7 n .!"  #" 2,.  !# #" ′ %1&%(#" ; ()- l (:  0 n /' 2 +*1&$(
   
8.  8 n 2"#%
   $ Mean Square Distance Objective (ER-MSD) n !": #% (&!" 0 ) n !($% , $' ): #%$ ( &,   #% )
9.  9 n )$%
   $% '! ' Contrastive Objective (ER-CNT) n !":  &( ()$%  0 ") n !′: &( ' !(%& , %( ) # n ER-MSD )$%$% 
10. '+ 10 n  .1 %#&",2$  *!3 Topological Regularization n !: ER-MSD  ER-CNT -,2( n ": *0!
     *!  )/, -,2( !# = ! + "!&'(
11. 3G 11 n &+2> $ Word2Vec, GloVe, FastText ': n F=@
    4=@5( WordNet, Roget’s Thesaurus ': l 1,023,082F=@
    380,8734=@5( l !%.@7 57,320@1 → Retrofitting /E 
    )<.@06 ,8;?  n ACB7!(#$ , #& )!%#AC': 9*-D" %(0.3:
12.  12 n     
     n  l Word Similarity (SimLex-999, SimVerb-3500) l Language Transfer n  l Lexical Text Simplification (LIGHT-LS) l Dialog State Tracking
13. 7- (Word Similarity) 13 • SimLex 4 %#< !6>'&$ +.095(;= • :23
     Attract-Repel  !6&$,"
     8*&$)/ 1  ,"  
14. /* (Lexical Text Simplification) 14 • Retrofitting (,  Attract-Repel % 2'135 • #. 59.6%  Attract-Repel %"!'   +$'1 7
    A: Accuracy (-&), C: Changes ( 640))
15. -* (Lexical Text Simplification) 15 • 1+ 
    $."!2 3 ) 
    %04&    (,villain(%#) Attract-Repel protagonist(#)/' • ER-CNT demon(5)/'
16.   16 n Retrofitting Word vector space specialization =* 3 #"%$4916 n WordNet (<.+
    ,*  :/-2 8)0>'!#7; n Word Similarity  Lexical Simplification A&
    5.!#?
    0>@B