Lexical feature transfer - Second report - Revision history

Fpetkovski: /* Second model */

2012-07-27T12:21:10Z

Second model

Fpetkovski: /* Review */

2012-07-27T12:19:45Z

Review

Fpetkovski: /* Corpora, sets and alignment */

2012-07-27T12:18:48Z

Corpora, sets and alignment

Fpetkovski: /* Corpora, sets and alignment */

2012-07-27T12:18:30Z

Corpora, sets and alignment

Fpetkovski: /* Second model */

2012-07-27T11:50:57Z

Second model

Fpetkovski: /* Second model */

2012-07-27T11:50:33Z

Second model

Fpetkovski: /* Corpora, sets and alignment */

2012-07-27T00:38:53Z

Corpora, sets and alignment

Fpetkovski: /* Corpora, sets and alignment */

2012-07-26T17:14:15Z

Corpora, sets and alignment

Fpetkovski: /* First Model */

2012-07-26T17:13:59Z

First Model

Fpetkovski: /* Second model */

2012-07-26T16:59:32Z

Second model

@@ Line 47: / Line 47: @@
 == Second model ==
-The second model is a simpler one, where instead of one trigram, two bigrams are used as features. The source-language nouns / verbs are merged with the pronoun to form a set in the following format:
+The second model is a simpler one, where instead of one trigram, two bigrams are used as features. The source-language nouns / verbs are merged with the preposition to form a set in the following format:
 <pre>

@@ Line 1: / Line 1: @@
 == Review ==
-In the first attempt at trying to solve the problem of corpus-based preposition selection, both a Naive Bayes and and SVM classifier were tried out. The lemmas and some of the tags of the surrounding words were extracted as features for the classifier. The source-language corpus was used to extract training examples from <n1> <pr> <n2> -> <n1> <pr> <n2> patterns, and the target-language corpus was used to label the extracted training examples. <br />
+In the first attempt at trying to solve the problem of corpus-based preposition selection, both a Naive Bayes and and SVM classifier were tried out. The lemmas and some of the tags of the surrounding words were extracted as features for the classifier. The source-language corpus was used to extract training examples from (n1) (pr) (n2) -> (n1) (pr) (n2) patterns, and the target-language corpus was used to label the extracted training examples. <br />
 Around 12.000 of the extracted examples were aligned to their target-language translations and labeled. There was some improvement in the translation quality, however, there were many wrong predictions as a result of the small training set and formatting errors in the training set.

@@ Line 13: / Line 13: @@
 <pre>
   (n | v) (pr) (adj | adv)* (n | v)  ->    (n | v) (pr) (adj | adv | det)* (n | v)
-                                         | (n)     (n)
+                                         | (n) (n)
 </pre>

@@ Line 12: / Line 12: @@
 This time, the aligner was extended to match the pattern: <br />
 <pre>
-  <n | v> <pr> <adj | adv>* <n | v>  ->    <n | v> <pr> <adj | adv | det>* <n | v>
+  (n | v) (pr) (adj | adv)* (n | v)  ->    (n | v) (pr) (adj | adv | det)* (n | v)
-                                         | <n>     <n>
+                                         | (n)     (n)
 </pre>

@@ Line 75: / Line 75: @@
 Number of position-independent correct words: 720714
 Position-independent word error rate (PER): 51.67 %
 after:

@@ Line 71: / Line 71: @@
 ----------------------
 before:
-Edit distance: 842588
+Edit distance: 804749
-Word error rate (WER): 68.33 %
+Word error rate (WER): 66.18 %
-Number of position-independent correct words: 692408
+Number of position-independent correct words: 720714
-Position-independent word error rate (PER): 54.52 %
+Position-independent word error rate (PER): 51.67 %
 after:
-Edit distance: 842418
+Edit distance: 795336
-Word error rate (WER): 68.32 %
+Word error rate (WER): 65.40 %
-Number of position-independent correct words: 692811
+Number of position-independent correct words: 720913
-Position-independent word error rate (PER): 54.49 %
+Position-independent word error rate (PER): 50.71 %
 ===================================================
 Tested on the affected lines only:
 ----------------------------------
 before:
-Edit distance: 241251
+Edit distance: 312138
-Word error rate (WER): 72.12 %
+Word error rate (WER): 69.77 %
-Number of position-independent correct words: 191790
+Number of position-independent correct words: 266954
-Position-independent word error rate (PER): 56.96 %
+Position-independent word error rate (PER): 54.26 %
 after:
-Edit distance: 233708
+Edit distance: 303829
-Word error rate (WER): 69.86 %
+Word error rate (WER): 67.92 %
-Number of position-independent correct words: 192187
+Number of position-independent correct words: 266111
-Position-independent word error rate (PER): 54.21 %
+Position-independent word error rate (PER): 51.50 %
 </pre>

@@ Line 26: / Line 26: @@
 </pre>
-A total of 41.631 training and 62.676 testing examples were extracted.
+A total of 41.631 training and 62.676 testing examples were extracted. The reason why there are more testing examples extracted, even though the testing set is smaller, is because the translation from apertium is much closer to the original text in terms of synonym use, grammatical structure etc.
 == First Model ==

@@ Line 28: / Line 28: @@
 == First Model ==
-In the first model, the lemmas from the extracted nouns / verbs and preposition were used as one feature, and a NB classifier was used.
+In the first model, the lemmas from the extracted nouns / verbs and preposition were used as one feature, and a NB classifier was used. <br />
 <pre>

@@ Line 55: / Line 55: @@
 на--удел      | на--профит         | of
 за--данок     | за--поединец       | for
-за--профит    | за--буџет | to
+за--профит    | за--буџет          | to
 </pre>