A generic framework for Arabic to English machine ... - Acsu Buffalo

More documents

Recommendations

Info

5.2. DESIGNING AN XML LEXICON ARCHITECTURE FOR ARABIC MT BASED ON RRG file. In order to be able to analyses Arabic by computer we must first extract the lexical properties of the Arabic words. The UniArab system uses the lexicon to construct a logical structure for Arabic input sentences, also represented in XML, which is then used for generating the target language translation. We show the structure of the UniArab lexicon, discuss how it is used in the system, and show the user interface used for adding to the lexicon. The lexicon is built from individual words at present. 5.2.1 An XML-based lexicon In order to build this system and represent the data sources, we use the XML language and Java. The most recent recommendation of the XML language has been presented by Bray et al. (2008). XML has become the default standard for data exchange among heterogeneous data sources (Arciniegas 2000). The UniArab system allows data to be stored in XML format. This data can then be queried, exported and serialized into any format the developer wishes. We choose to create our data source as XML, for optimum support or different plat- forms. It was also easier as we used Arabic letters not Unicode inside the data source, XML fully supported Arabic. We created our search engine using Java. 5.2.2 Lexical representation in UniArab Lexical frames represent the language-dependent lexicon. We use an XML data source to represent the UniArab lexicon. The lexicon creates pointers to corresponding conceptual frames or attributes of each word. These frames also have relations which link them to verb class frames, which are organized hierarchically according to the particular language. 70
5.2. DESIGNING AN XML LEXICON ARCHITECTURE FOR ARABIC MT BASED ON RRG In Phase 3 in Figure 5.1, the UniArab system tokenizes a sentence into words, then sends each word to the search engine within the Lexicon to query the category of each word and all attributes for that word. The Lexicon returns the corresponding category and its attributes as detailed below. The Morphology Parser, Phase 5, receives the word meta- data and ensures that the properties of the words are consistent. The verb attributes in particular, are of great importance in correctly extracting sentence logical structure fur- ther down the processing chain, helping to answer the basic question ‘Who does what?’ In free word order sentences, for example, yh. b qys lylā, ‘Qays loves Laila’ multiple orders are possible including verb-subject-object, verb-object-subject or subject-verb-object. The attributes of the verb agree with the gender of the subject. Given the masculine gender of the verb in this case, the Syntactic Parser will look for a masculine proper noun to make the actor for this sentence. If there is more than one masculine proper noun in such a case, then Modern Standard Arabic defines the first proper noun as the actor. The Morphology Parser will be extended so that it can deal with words that are defined in multiple categories, deciding which should be processed. Meanwhile the Syntactic Parser, so far, has only been implemented for extracting word order, though it will be extended to deal with word ambiguities in future versions. 5.2.3 Lexical properties Figure 5.2 shows the structure of the Lexicon including the properties stored for each word category. For all categories, an Arabic word is stored along with its English representation. Since word ambiguity has not been dealt with so far, there is a one to one map- ping for the simple sentences which UniArab processes up to now. However, word ambiguity is supported in the structure, with each possible case stored as a separate record. All search results will be passed to the Morphology Parser to decide which is taken. 71
Page 1 and 2:
A generic framework for Arabic to E
Page 3 and 4:
sentence and its linking algorithm
Page 5 and 6:
Declaration I herby certify that th
Page 7 and 8:
advices, but also academically and
Page 9 and 10:
CONTENTS 2.3.1.2 Indefinite nouns .
Page 11 and 12:
CONTENTS 4.4.1 Non-Roman alphabet s
Page 13 and 14:
CONTENTS 7.2.8 Limitation of UniAra
Page 15 and 16:
LIST OF FIGURES 4.4 Difference betw
Page 17 and 18:
List of Tables 2.1 Dual: merely add
Page 19 and 20:
LIST OF TABLES 5.9 Test strategy: p
Page 21 and 22:
1 Introduction The following paragr
Page 23 and 24:
high-quality translation technology
Page 25 and 26:
1.3. TECHNOLOGIES ditransitive). Th
Page 27 and 28:
1.4. THESIS ORGANIZATION 2009b). Ch
Page 29 and 30:
2.1. CHARACTERISTICS OF THE ARABIC
Page 31 and 32:
2.2 Characteristics of Arabic words
Page 33 and 34:
2.2. CHARACTERISTICS OF ARABIC WORD
Page 35 and 36:
2.3. PART OF SPEECH INVENTORY OF TH
Page 37 and 38:
2.3.4 Verbs 2.3. PART OF SPEECH INV
Page 39 and 40: 2.3.4.2 Aspect 2.3. PART OF SPEECH
Page 41 and 42: 2.3. PART OF SPEECH INVENTORY OF TH
Page 43 and 44: 2.4.1.1 Verb and noun 2.4. SENTENCE
Page 45 and 46: Table 2.27: verb(V), subject(S) and
Page 47 and 48: 3 Role and Reference Grammar (RRG)
Page 49 and 50: 3.1. ROLE AND REFERENCE GRAMMAR LIN
Page 51 and 52: 3.2. FORMAL REPRESENTATION OF LAYER
Page 57 and 58: 3.3. NOUN PHRASE STRUCTURE appropri
Page 59 and 60: 3.4. LEXICAL REPRESENTATIONS FOR VE
Page 61 and 62: 3.5. WHY WE USE RRG AS THE LINGUIST
Page 63 and 64: 3.5. WHY WE USE RRG AS THE LINGUIST
Page 65 and 66: 3.6. SUMMARY there will be more com
Page 67 and 68: 4.1. ADVANTAGES OF MACHINE TRANSLAT
Page 69 and 70: 4.2.3 Lexical databases 4.2. COMPUT
Page 71 and 72: 4.3.2 Direct translation 4.3. BASIC
Page 73 and 74: 4.3. BASIC MACHINE TRANSLATION STRA
Page 75 and 76: would also be needed for an interli
Page 77 and 78: 4.4.1 Non-Roman alphabet scripts 4.
Page 79 and 80: Figure 4.5: NP rule (NP -> det n pp
Page 81 and 82: 4.5. CHALLENGES OF ARABIC TO ENGLIS
Page 83 and 84: Figure 4.7: Direct MT system 4.6.2
Page 85 and 86: 4.7 Summary Figure 4.9: Structure t
Page 87 and 88: 5 Design of Arabic to English machi
Page 89: 5.2. DESIGNING AN XML LEXICON ARCHI
Page 93 and 94: 5.2. DESIGNING AN XML LEXICON ARCHI
Page 95 and 96: 5.3. DESIGN OF TEST STRATEGY Table
Page 97 and 98: 5.4. DESIGN OF EVALUATION CRITERIA
Page 99 and 100: 6 UniArab: a proof-of-concept Arabi
Page 101 and 102: 6.1. CONCEPTUAL STRUCTURE OF THE UN
Page 107 and 108: 6.2. UNIARAB: LEXICAL REPRESENTATIO
Page 113 and 114: 6.3. UNIARAB: GENERATION which mean
Page 115 and 116: Figure 6.11: UniArab’s GUI 1 6.4.
Page 117 and 118: 6.4.1 Lexicon interface Figure 6.13
Page 119 and 120: 6.6 Summary 6.6. SUMMARY We present
Page 121 and 122: 7.2. SENTENCE TESTS to which standa
Page 123 and 124: Table 7.2: Test : Verb-subject; agr
Page 125 and 126: Figure 7.4: Verb-subject agreement
Page 127 and 128: Table 7.5: Test : gender-ambiguous
Page 129 and 130: Table 7.7: Test : Verb ‘to be’
Page 131 and 132: Table 7.9: Test : Verb ‘to have
Page 133 and 134: 7.2. SENTENCE TESTS Table 7.11: Tes
Page 135 and 136: 7.2.6 Pro-drop Arabic Table 7.13: T
Page 137 and 138: Table 7.15: Test : Intransitive 2 A
Page 139 and 140: 7.2.7.3 Ditransitive Table 7.17: Te
Page 141 and 142:
Table 7.19: Test : Ditransitive wit
Page 143 and 144:
7.2. SENTENCE TESTS In a case where
Page 145 and 146:
7.3 System evaluation 7.3. SYSTEM E
Page 147 and 148:
7.3. SYSTEM EVALUATION search engin
Page 149 and 150:
Now is not the end. It is not the b
Page 151 and 152:
8.1. THESIS SUMMARY and so we hope
Page 153 and 154:
and proving the suitability of the
Page 155 and 156:
References Abn-Aqeal, 2007. sharah
Page 157 and 158:
REFERENCES Salem, Y., Hensman, A.,
Page 159 and 160:
Appendix 139
Page 161 and 162:
• Yasser Salem and Brian Nolan. 2
Page 163 and 164:
Arabic Letter and Phonetic Value Le
Page 165 and 166:
C List of translatable sentences I
Page 167 and 168:
she is an engineer. Zaid loves Fati
Page 169 and 170:
Jack is eating with the spoon.
Page 171 and 172:
Ayah tore the page. ֓āyh
Page 173 and 174:
Eman saw Sarah. r֓at ֓iymā
Page 175 and 176:
Carl is brushing his hair. ym
Page 177 and 178:
I want a book. ֓aryd ktāb
Page 179 and 180:
Omar gave the book to Khalid.
Page 181 and 182:
D Verbs in lexicon Verbs in Arabic
Page 183 and 184:
Arabic Example Logical Structure
Page 185 and 186:
Arabic Example Logical Structure
Page 187 and 188:
Arabic ys. rs. r Example the wheel
Page 189 and 190:
Arabic ֓adrs Example I am teach
Page 191 and 192:
Package: Name of Class Class Summar
show all

A generic framework for Arabic to English machine ... - Acsu Buffalo

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?