[ Skip to the content ]

Institute of Formal and Applied Linguistics Wiki

[ Back to the navigation ]

You are here: start » courses » rg » morphology-induction-with-spelling-rules

Differences

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

Link to this comparison view

Next revision 		Previous revision
courses:rg:morphology-induction-with-spelling-rules [2011/04/08 13:32]
bejcek vytvořeno 		courses:rg:morphology-induction-with-spelling-rules [2011/04/19 13:05] (current)
abzianidze
Line 5: Line 5:
  
 ===== Introduction ===== ===== Introduction =====
-  * Paper describes morphology induction using Bayesian approach+  * The paper describes morphology induction using Bayesian approach 
 +    * It is based on the Minimum Description Length (MDL) principle
   * Baseline: Goldwater et al., 2006:  [[http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.70.9842&rep=rep1&type=pdf|Interpolating between types and tokens by estimation power-law generators]]   * Baseline: Goldwater et al., 2006:  [[http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.70.9842&rep=rep1&type=pdf|Interpolating between types and tokens by estimation power-law generators]]
     * only stem & suffix     * only stem & suffix
-    * Dirichlet priors over the multinominal distributions for word class, stem and for suffix +    * Dirichlet priors over the multinominal distributions for word class, stem and suffix 
-    *    + 
-  * Improvements to baseline system +  * Improvements to the baseline system 
-    * introduces spelling rules (context/change, e.g.: "ut_i"/ε→t (in "shut.ing") or "ke_i"/e→ε (in "take.ing"))+    * introduces spelling rules (context/change, e.g.: "ut_i"/ε→t (in "shut.ing") or "ke_i"/e→ε (in "take.ing")), which are simultaneously learned with morphological analysis 
     * Dirichlet priors first set by hand: to prefer empty rules to deletion/insertion     * Dirichlet priors first set by hand: to prefer empty rules to deletion/insertion
    
    
 ===== What do we dislike about the paper ===== ===== What do we dislike about the paper =====
-  * Experiments are done olny for English and only for verbs -- that's constrained too much+  * Experiments are done only for English and only for verbs -- that's constrained too much
   * Results are not convincing enough -- F-score and underlying form accuracy outperform baseline only for stems (not for suffixes) and precision doesn't outperform baseline at all   * Results are not convincing enough -- F-score and underlying form accuracy outperform baseline only for stems (not for suffixes) and precision doesn't outperform baseline at all
 +
 ===== What do we like about the paper ===== ===== What do we like about the paper =====
   * Loganathan has the code (although he couldn't compile it, yet)   * Loganathan has the code (although he couldn't compile it, yet)
 +  * Spelling rules are also simultaneously learned along with morphological analysis 
 +  * It's unsupervised and clever: using just a couple of (hyper)parameters (some of them are learned automatically), they can describe a wide range of morphological rules - within one elegant framework.

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