[ Skip to the content ]

Institute of Formal and Applied Linguistics Wiki

[ Back to the navigation ]

You are here: start » courses » rg » unsupervised-methods-for-head-assignments

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Last revision Both sides next revision
courses:rg:unsupervised-methods-for-head-assignments [2011/05/19 15:02]
bejcek vytvořeno 		courses:rg:unsupervised-methods-for-head-assignments [2011/05/23 10:34]
popel Section 5 -> Equation 5 + more details
Line 5: Line 5:
  
 ===== Introduction ===== ===== Introduction =====
-  * The paper describes **two** methods for **unsupervised** head assignment and tests them on **two** corpora byt **two** means (against gold standard and using it in constituency parser).+  * The paper describes **two** methods for **unsupervised** head assignment and tests them on **two** corpora (English PennTB/PARC700 and German Tiger/TigerDB) by **two** means (against gold standard and using it in a constituency parser).
   * It is the **first** successful unsupervised head assignment.   * It is the **first** successful unsupervised head assignment.
   * There is a problem when converting dependency structures into head annotations: their algorithm doesn't guarantee **exactly one head** assigned among siblings.   * There is a problem when converting dependency structures into head annotations: their algorithm doesn't guarantee **exactly one head** assigned among siblings.
    
 ===== What do we dislike about the paper ===== ===== What do we dislike about the paper =====
-  * There is a wrong definition of **entropy** in Section 5. +  * Equation 5 is misleading/wrong: entropy is here computed for the distribution **p** which sums up to <latex>|\mathcal{L}|</latex>. 
-  * We found three possible interpretations of "//any//" in Section 3.3, second paragraph; their approach is therefore a bit unclear.+  * We found three possible interpretations of "//any//" in Section 3.3, second paragraph; their approach is therefore a bit unclear. The interpretations are: 
 +     - apply one fixed annotation of the heads (but it can be any annotation) 
 +     - apply the (partial) annotation which is consistent with all possible head annotations 
 +     - apply all possible head annotations
   * It is also unclear, what "transformed to their respective **spines**" means exactly; Section 3.4, first item in the itemize.   * It is also unclear, what "transformed to their respective **spines**" means exactly; Section 3.4, first item in the itemize.
   * The **greediness** of the entropy model could be its main weakness resulting in not so good numbers.   * The **greediness** of the entropy model could be its main weakness resulting in not so good numbers.

[ Back to the navigation ] [ Back to the content ]