English - Revision history

Marcriera: /* Bidix Guidelines */

2018-05-01T22:00:24Z

Bidix Guidelines

2017-08-28T09:37:32Z

2017-08-25T07:16:32Z

Current status

2017-08-25T07:16:17Z

Current status

2017-08-23T13:51:23Z

Bidix Guidelines

2017-08-23T13:39:12Z

Tagger

2017-08-22T11:36:39Z

Future work

2017-08-22T11:36:18Z

Apertium-eng

2017-08-16T16:52:43Z

Apertium-eng

2017-08-16T15:41:32Z

Tagger

@@ Line 162: / Line 162: @@
 '''Coverage:''' 93.55% (Wikipedia)
-===Bidix Guidelines===
+===Dictionary guidelines===
 The current English dictionary is quite big (nearly 55,000 entries), so tidiness is essential to ensure future development:

@@ Line 153: / Line 153: @@
 ===Current status===
+''Last update: 28 Aug 2017''
 '''Dix entries:''' 54,453

@@ Line 158: / Line 158: @@
 '''Dix paradigms:''' 377
-'''Coverage:''' 93,55% (Wikipedia)
+'''Coverage:''' 93.55% (Wikipedia)
 ===Bidix Guidelines===

@@ Line 158: / Line 158: @@
 '''Dix paradigms:''' 377
-'''Coverage:''' ?
+'''Coverage:''' 93,55% (Wikipedia)
 ===Bidix Guidelines===

@@ Line 167: / Line 167: @@
 * Keep entries grouped by type and tags (do not mix different types of proper nouns together).
 * Check the file with apertium-dixtools (to update the number of entries and remove duplicates).
+====Spelling variants====
+The standard spelling variant in Apertium is British English. American English spelling entries are officially supported using <code>v="eng"</code> in the original British entry and <code>v="eng_US"</code> in the American entry, which should be a subform of the first one.
 === Tagger ===

@@ Line 172: / Line 172: @@
 Apertium-eng currently uses a tagger trained in a supervised manner using a hand-tagged corpus and a perceptron. The corpus can be found [https://svn.code.sf.net/p/apertium/svn/languages/apertium-eng/texts/consensus here], and should be modified after any tag change in the monolingual dictionary to make sure it matches the current status of the language module. The MTX file for English has no rules defined for now, but it should improve disambiguation after some restrictions are added.
-The tagger can be trained with the perceptron using the following command:
+The tagger can be trained with the perceptron using the following command (see [[Perceptron tagger]]):
 <pre>

← Older revision		Revision as of 11:36, 22 August 2017
Line 189:		Line 189:

	* Improve support of British/American ortography variants.		* Improve support of British/American ortography variants.


	''For further documentation about English in Apertium, check: [[:Category:English]]''		''For further documentation about English in Apertium, check: [[:Category:English]]''

@@ Line 165: / Line 165: @@
 Before making changes to the hand-tagged corpus, make sure you have read the [[Tagging_guidelines_for_English|tagging guidelines]]. A good corpus is the key to a good tagger!
+=== Constraint Grammar ===
+Apertium-eng currently has 38 CG rules. There is a lot of room for disambiguation improvement using CG, but ideally some rules should be moved to the tagger MTX.

← Older revision		Revision as of 15:41, 16 August 2017
Line 163:		Line 163:

	''Note: in the example above, "10" is the number of iterations, "eng.tagged" is the hand-tagged corpus and "eng.untagged" is the corpus tagged by apertium-tagger.''		''Note: in the example above, "10" is the number of iterations, "eng.tagged" is the hand-tagged corpus and "eng.untagged" is the corpus tagged by apertium-tagger.''

			Before making changes to the hand-tagged corpus, make sure you have read the [[Tagging_guidelines_for_English\|tagging guidelines]]. A good corpus is the key to a good tagger!