The following package consists of a rule-based verb inflector in Persian developed by Mohammad Sadegh Rasooli. The code was mainly used for preprocessing the Persian dependency treebank.

• If you use this software in your research work, please cite to the following paper:

  Mohammad Rasooli, Heshaam Faili, and Behrouz Minaei-Bidgoli. "Unsupervised Identification of Persian Compound Verbs", in Proceedings of the 10th Mexican international conference on Advances in Artificial Intelligence - Volume Part I (LNCS 7094), Pages 394-406, Puebla, Mexico, 2011.

  In ACM library: http://dl.acm.org/citation.cfm?id=2178197.2178234&coll=DL&dl=GUIDE&CFID=97422849&CFTOKEN=39528935

  In Springer library: http://www.springerlink.com/content/n3r0181wu2h6p337/

• Please send the bugs and your questions to <rasooli.ms{AT}gmail.com>

The code is compatible with C# 3.5 or upper versions.

There are two options for getting a verb analyzed sentence:

1. Without part of speech tags (without disambiguation, considering all the words as potential verbs). In SentenceAnalyzer.cs:

    public static VerbBasedSentence MakeVerbBasedSentence(string sentence)
   

   or

    public static VerbBasedSentence MakeVerbBasedSentence(string[] sentence)
   

2. With part of speech and morphosyntactic tags (with a good accuracy): the pos tags are the same as Bijankhan corpus:

    public static VerbBasedSentence MakeVerbBasedSentence(string[] sentence, string[] posSentence, string[] lemmas, MorphoSyntacticFeatures[] morphoSyntacticFeatureses)
   

In the program.cs file there is a test output of a Persian sentence that can be used as a starting point.

var analyzer = new SentenceAnalyzer("../../../Data/VerbList.txt");
var sentence = "من دارم به شما می‌گویم که این صحبت‌ها به راحتی گفته نخواهد شد و من با شما صحبت زیاد خواهم کرد.";
var result = SentenceAnalyzer.MakeVerbBasedSentence(sentence);
var output = new StringBuilder();
foreach (var dependencyBasedToken in result.SentenceTokens)
{
    output.AppendLine(dependencyBasedToken.WordForm + "\t" + dependencyBasedToken.Lemma + "\t" +
                      dependencyBasedToken.CPOSTag
                      + "\t" + (dependencyBasedToken.HeadNumber+1).ToString() + "\t" +
                      dependencyBasedToken.DependencyRelation);
}
File.WriteAllText("../../../testOutPut.txt",output.ToString());

Output in testOutPut.txt:

من	_	_	0	_
دارم	داشت#دار	V	5	PROG
به	_	_	0	_
شما	_	_	0	_
می‌گویم	گفت#گو	V	0	_
که	_	_	0	_
این	_	_	0	_
صحبت‌ها	_	_	0	_
به	_	_	0	_
راحتی	_	_	0	_
گفته نخواهد شد	گفت#گو	V	0	_
و	_	_	0	_
من	_	_	0	_
با	_	_	0	_
شما	_	_	0	_
صحبت	_	_	18	NVE
زیاد	_	_	0	_
خواهم کرد	کرد#کن	V	0	_

The file is tab-separated with the following fields:

• verbType: integer

  1: simple, 2: prefix verb, 3: compound verb, 4: compound prefix verb , 5: prepositional compound prefix verb, 6: enclitic verb, 7: prepositional verb

• transitivity: integer

  0: intransitive, 1: transitive, x 2: bitransitive

• past tense root: string

  "-" if not present

• present tense root: string

  "-" if not present

• Non-verbal element: string

  "-" if not present

• Prefix: string

  "-" if not present

• Preposition: string

  "-" if not present

• amrShodani: string

  "-" =true, *: false

• vowelEnd: string

  End of present root vowel: U: ends with u, I: ends with ei, A: ends with a, ?: else

• maziVowel: string

  Start vowel type of past tense root A: starts with "a" or "\ae", @: else

• mozarehVowel: string

  Start vowel type of present tense root bU: starts with "bu", ba: start with "b\ae", bA: starts with "ba", A: starts with "a" or "\ae", !: else

I assumed the character set is being refined when you pass array argument to the methods. As shown in the follwoing code, I used Virastyar library for refining characters and tokenizing strings.

public static VerbBasedSentence MakeVerbBasedSentence(string sentence)
{
    sentence = StringUtil.RefineAndFilterPersianWord(sentence); // using the refiner of Virastyar software
    var tokenized = PersianWordTokenizer.Tokenize(sentence,true); // using the tokenizer of Virastyar software
    return MakeVerbBasedSentence(tokenized.ToArray());
}

You can go to Virastyar official site in order to know more about its options http://virastyar.ir.

If you do not want to use it for your purposes you can clean the mentioned lines from the code

You can find a morphological-based POS tagger that can be used in your code. You can also use the tagger to help improve learner POS taggers such as HMM tagger.

I assumed that the writers use semi-space for verb inflections. In Bijankhan corpus, you can replace space with semi-space in words with verb tag.