Updated version of Jennifer Foster's GenERRate.
===========================================================================
GenERRate expects the following input:
- An input file containing one SGML-encoded sentence per line. If the sentences are POS-tagged, the format is:
word1 TAG1 word2 TAG2 ...
If the sentences are not POS-tagged, the format is:
word1 word2 ...
Use the -n option to specify that the input is not tagged.
Attributes can be included within the tag. These will be included in the output.
- An output file. This will be where the error annotated corpus will be written.
Each sentence in the output file will contain a grammatical error.
The original sentence will be encoded as an attribute in the SGML mark-up.
Details of the actual error will also be encoded as an attribute.
- An error analysis file, specifying one error per line.
Frequency information is optional and is specified as a ratio at the end of each error specification, e.g.
subst word an a 0.1
means that an error involving an an/a confusion should constitute 10% of the final error corpus.
See below for the full list of errors supported by GenERRate.
- A word list to be used by GenERRate when substituting a word or adding one to a sentence. The format of this is one word per line:
token TAG token TAG
To see the GenERRate usage options, type:
java GenERRate
GenERRate can be run with the sample input files provided:
java GenERRate testInput.txt testOutput.txt testErrorAnalysis.txt testWordList.txt
Please make sure that your classpath includes the xalan jar file which is in the lib folder of GenERRate.jar.
GenERRate Error Options
delete Delete a word at random
delete POS Delete word tagged as POS.
delete POS1 POS2 true Delete a word tagged as POS1 occurring after a word tagged as POS2
delete POS1 POS2 false Delete a word tagged as POS1 occurring before a word tagged as POS2
delete POS1 POS2 POS3 Delete a word tagged as POS2 occurring between a word tagged as POS1 and a word tagged as POS3
delete start POS1 POS2 Delete a word tagged as POS1 occurring at the start of the sentence and before a word tagged as POS2
delete POS1 POS2 end Delete a word tagged as POS2 occurring after a word tagged as POS1 and at the end of the sentence
delete word word1 Delete an occurrence of word1 in the sentence.
insert Insert a word at random into the sentence. Randomly choose a word from the sentence or from the extra word list.
insert file Insert a word at random into the sentence. The word is chosen at random from the extra word list
insert sentence Insert a word at random into the sentence. The word is chosen at random from the sentence.
insert file POS Insert a word tagged as POS into the sentence at random. The word is chosen at random from the extra word list.
insert sentence POS Same as above except that the word is chosen at random from the sentence.
insert file POS1 POS2 true Insert a word tagged as POS1 into the sentence after a word tagged as POS2. The word is chosen from the extra word list.
insert sentence POS1 POS2 true Same as above except that the word is chosen at random from the sentence.
insert file POS1 POS2 false Insert a word tagged as POS1 into the sentence before a word tagged as POS2. The word is chosen from the extra word list.
insert sentence POS1 POS2 false Same as above except that the word is chosen at random from the sentence.
insert file POS1 POS2 POS3 Insert a word tagged as POS2 after a word tagged as POS1 and before a word tagged as POS3. The word is chosen from the extra word list.
insert sentence POS1 POS2 POS3 Same as above except that the word is chosen at random from the sentence.
insert file start POS1 POS2 Insert a word tagged as POS1 at the start of the sentence before a word tagged as POS2. The word is chosen from the extra word list.
insert sentence start POS1 POS2 Same as above except that the word is chosen at random from the sentence.
insert file POS1 POS2 end Insert a word tagged as POS2 at the end of the sentence after a word tagged as POS1. The word is chosen from the extra word list.
insert sentence POS1 POS2 end Same as above except that the word is chosen at random from the sentence.
move Move a word chosen at random from the sentence to another position (also randomly chosen) within the sentence.
move POS Move a word tagged as POS to another position within the sentence.
move POS true n Move a word tagged as POS n places to the left.
move POS false n Move a word tagged as POS n places to the right.
subst Substitute a word chosen at random in the sentence for a word chosen at random from the extra word list.
subst POS Substitute a word tagged as POS for another word from the extra word list tagged as POS
subst word word1 word2 Substitute an occurence of word1 in the sentence for an occurence of word2
subst POS1 POS2 Substitute a word tagged as POS1 for a word tagged as POS2 from the extra word list UNLESS POS1 and POS2 are one of the pairs below. If they are, the inflection of the word is changed.
singular noun, plural noun (WSJ: NN -> NNS) plural noun, singular noun (WSJ: NNS -> NN)
third singular verb, non third singular verb (WSJ: VBZ -> VBP) non third singular verb, third singular verb (WSJ: VBP -> VBZ)
third singular verb, present participle verb (WSJ: VBZ -> VBG) present participle verb, third singular verb (WSJ: VBG -> VBZ)
present participle verb, past participle verb (WSJ: VBG -> VBN) past participle verb, present participle verb (WSJ: VBN -> VBG)
present participle verb, non third singular verb (WSJ: VBG -> VBP) non third singular verb, present participle verb (WSJ: VBP -> VBG)
present participle verb, inf marker (WSJ: VBG -> TO) inf marker, present participle verb (WSJ: TO -> VBG)
base verb, present participle verb (WSJ: VB -> VBG) past participle verb, third singular verb (WSJ: VBN -> VBZ) base verb, third singular verb (WSJ: VB -> VBZ)
base adjective, comparative adjective (WSJ: JJ -> JJR) base adjective, superlative adjective (WSJ: JJ -> JJS)
comparative adjective, superlative adjective (WSJ: JJR -> JJS) comparative adjective, base adjective (WSJ: JJR -> JJ)
superlative adjective, comparative adjective (WSJ: JJS -> JJR) superlative adjective, base adjective (WSJ: JJS -> JJ)
adverb, adjective (WSJ: RB -> JJ)