Differences

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

--- courses:rg:2011-report-parser [2012/09/27 10:36]
ufal
+++ courses:rg:2011-report-parser [2012/09/27 11:21]
ufal
@@ Line 32: / Line 32: @@
           * they reduced the number of generic "dep/DEP" relation
               * Stanford tags are hierarchical and "dep/DEP" is the top-most one
-          * 1.3% of arcs are non-projective (out of 8.1% of all non-projective arcs) because of the following conversion (agreement can be a motivation for this, i.e. in Czech):
+          * correcting of POS using the syntactic info + additional rules for specific word forms
-            {{:courses:rg:dependency-conversion.png|}}
+          * 1.3% of arcs are non-projective (out of 8.1% of all non-projective arcs) because of the following conversion (agreement can be a motivation for this, i.e. in Czech): {{:courses:rg:dependency-conversion.png|}}
+      - additional changes and the final conversion from the intermediate output to a dependency structure
 ==== Parser ====
-  * we illustrated a step of the parser
+  * we illustrated a step of the parser:
+{{:courses:rg:ndef_parsing.png|}}
+  * we compared time complexity of this system with other commonly used ones
+| MST parser | <latex>\mathop O(n^2)</latex> | |
+| MALT parser | <latex>\mathop O(n)</latex> | in fact slower |
+| this parser | <latex>\mathop O(n\log(n))</latex> | <latex>\mathop O(n^2)</latex> - naive implementation |
+| this parser - non-projective | <latex>\mathop O(n^2\log(n))</latex> | <latex>\mathop O(n^3)</latex> - naive implementation |
+    * implemented by a heap, it can reach <latex>\mathop O(n\log(n))</latex>; however, they don't use it because it's fast enough
+    * Algorithm1
+        * we weren't sure what the variable "index" is (best index of parent or any index of queue of unprocessed words)
+        * it again confirms that pseudocode is usually more confusing than a normal code or verbal explanation
+    * "move" operation:
+{{:courses:rg:move_operation.png|}}
+==== Features ====
+  * Brown et al. clusters - they are surprisingly used rarely
+==== Features ====
+  * not much discussed in the paper
+==== Evaluation ====
+  * they make the same transformation as they did in Section 2
+==== Shallow semantic annotation ====
+  * 4 optional modules
+      * preposition sense disambiguation
+      * noun compound interpretation
+      * possesives interpretation
+      * PropBank semantic role labelling
+          * (Hajič et al., 2009) is not in the list of references
+===== Conclusion =====
+  * non-directional easy-first parsing
+  * new features - Brown et al. clusters
+  * fast and accurate
+  * modified Penn converter
+      * changes in 9.500 POS tags
+      * labels copula, coordination
+  * semantic info

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

Institute of Formal and Applied Linguistics Wiki

Differences