User:Darshak/GSoC 2014 Report - Revision history

Darshak at 18:11, 21 August 2014

2014-08-21T18:11:54Z

Darshak: /* Constraint Grammar */

2014-08-21T15:37:12Z

Constraint Grammar

Darshak: /* Constraint Grammar */

2014-08-21T15:35:08Z

Constraint Grammar

Hectoralos: /* Translation quality */

2014-08-20T05:19:36Z

Translation quality

Hectoralos: /* Coverage */

2014-08-20T05:18:10Z

Coverage

Hectoralos: /* Coverage */

2014-08-20T05:17:20Z

Coverage

Hectoralos: /* Translation quality */

2014-08-20T03:51:56Z

Translation quality

Hectoralos: /* Translation quality */

2014-08-20T03:44:43Z

Translation quality

Hectoralos: Results

2014-08-20T03:33:50Z

Results

Hectoralos: /* Vocabulary */

2014-08-18T19:36:57Z

Vocabulary

@@ Line 11: / Line 11: @@
 <pre>$ echo 'Darkness falls across the land.' | apertium -d . en-eo
 Mallumaj faloj trans la lando.</pre>
 === After ===
@@ Line 37: / Line 38: @@
 === Coverage ===
 With a Wikipedia corpus of 49,759,540 words, the coverage with the previous (trunk) version of the translator was 91.7% (4,123,893 unknown words). With the new version it improved to 92.0% (3,958,640 unknown words).
 === Translation quality ===
 To measure translation quality I got Wikipedia's featured articles of the last two days (August, 18 and 19 2014), as a random sample of Wipipedia texts, which are the main translator's target ( https://en.wikipedia.org/wiki/Episode_2_%28Twin_Peaks%29 and https://en.wikipedia.org/wiki/Leslie_Groves ). The first text has 267 words, the second one 501.
@@ Line 51: / Line 50: @@
 So translation quality seemly has improved between 2 and 4 percentage points.
+==Future Work==
+Although there has been improvement, there's still quite some work to do before we can call it state-of-the-art.
+* Better interchunk rules, because a lot of problematic sentences at [[English_and_Esperanto/Outstanding_tests]] can be solved only by that.
+* While the new proper names have significantly expanded the dix, they have brought with them some ambiguities. These need to be solved.
+==Thanks==
+Jacob and Hector have guided me through the thick and thin of the trimester. Also, many others from Apertium community have helped me when I was confused about something. So I sincerely thank all of them.

@@ Line 26: / Line 26: @@
 == Constraint Grammar ==
-I used the English CG rules from English-Kazakh language pair as a base. It contained 124 rules, 16 of which I modified. 15 new rules were added and 1 was removed.
+I used the English CG rules from English-Kazakh language pair as a base. It contained 124 rules, 16 of which I modified. 15 new rules were added and 1 was removed (more like, commented).
 The accuracy was calculated on a part of the corpus used for tagger training. The accuracy on this corpus increased from 73.67% to 76.19%.

@@ Line 27: / Line 27: @@
 I used the English CG rules from English-Kazakh language pair as a base. It contained 124 rules, 16 of which I modified. 15 new rules were added and 1 was removed.
+The accuracy was calculated on a part of the corpus used for tagger training. The accuracy on this corpus increased from 73.67% to 76.19%.
 == Structural Transfer ==

@@ Line 42: / Line 42: @@
 To measure translation quality I got Wikipedia's featured articles of the last two days (August, 18 and 19 2014), as a random sample of Wipipedia texts, which are the main translator's target ( https://en.wikipedia.org/wiki/Episode_2_%28Twin_Peaks%29 and https://en.wikipedia.org/wiki/Leslie_Groves ). The first text has 267 words, the second one 501.
-Word error rate (WER) for the second text was, in the initial version, 29.59%, and in the final version  27.72%.<br/>
+The word error rate (WER) for the second text was, in the initial version, 29.59%, and in the final version  27.72%.<br/>
-Position-independent word error rate (PER) was, respectively, 25.47% and 22.47%.
+The position-independent word error rate (PER) was, respectively, 25.47% and 22.47%.
-Word error rate (WER) for the second text was, in the initial version, 38.52%, and in the final version  34.93%.<br/>
+The word error rate (WER) for the second text was, in the initial version, 38.52%, and in the final version  34.93%.<br/>
-Position-independent word error rate (PER) was, respectively, 30.14% and 27.15%.
+The position-independent word error rate (PER) was, respectively, 30.14% and 27.15%.
-So translation quality seemly has improved some 2 to 4 percentage points.
+So translation quality seemly has improved between 2 and 4 percentage points.

@@ Line 36: / Line 36: @@
 === Coverage ===
-From a Wikipedia corpus of 49,759,540 words the coverage with the previous (trunk) version of the translator was 91.7% (4,123,893). With the new version it improved to 92.0% (3,958,640 unknown words).
+With a Wikipedia corpus of 49,759,540 words, the coverage with the previous (trunk) version of the translator was 91.7% (4,123,893 unknown words). With the new version it improved to 92.0% (3,958,640 unknown words).
 === Translation quality ===

@@ Line 31: / Line 31: @@
 Structural transfer rules were added for almost all possible date formats used in English. Moreover, some rules were added for cases where correctly tagging a token would not suffice.
+== Results ==
+=== Coverage ===
+From a Wikipedia corpus of 49,759,540 the coverage with the previous (trunk) version of the translator was 91.7% (4,123,893). With the new version it improved to 92.0% (3,958,640 unknown words).
+=== Translation quality ===
+To measure translation quality I got Wikipedia's featured articles of the last two days (August, 18 and 19 2014), as a random sample of Wipipedia texts, which are the main translator's target ( https://en.wikipedia.org/wiki/Leslie_Groves and )

@@ Line 22: / Line 22: @@
 * 933 female given names
 * 2000+ surnames
-* and the names of a few companies and productsm, most of which were likely to be mistranslated due to being dictionary words
+* and the names of a few companies and products, most of which were likely to be mistranslated due to being dictionary words
 == Constraint Grammar ==