Fig 1: Deconversion process
International Journal of Scientific & Engineering Research, Volume 4, Issue 12, December-2013
ISSN 2229-5518
198
Generation Rules to Deconvert UNL Expressions to Bangla Sentences
Mohammad Zakir Hossain Sarker, Md Nawab Yousuf Ali, Jugal Krishna Das
Abstract— This paper discusses the design and implementation of Generation Rules for deconverting Universal Networking Language or UNL expressions to the Bangla Language. The Universal Networking Language (UNL) facilitates translation between Natural Languages across the world. The paper also focuses on the linguistic aspects of Bangla required for the deconversion process. We have developed a set of generation rules for converting UNL expression to Bangla sentences. Our experiment shows that these rules successfully generate correct Bangla sentences from UNL expressions. These rules can currently produce simple Bangla sentences. It is being enhanced to handle more types of sentences. The present system is a small achievement towards attaining a complete translation facility for the Bangla language.
Index Terms— Generation Rules, UNL, UNL Expression, Deconverter, Converter, Generation Window, Condition Window
—————————— ——————————
he UNL system allows people to communicate with peoples of different languages in their mother tongue. The UNL is a common language to exchange information through computers which can deal with natural languages. The UNL system basically consists of language servers, UNL
editors and UNL viewers [1].
descriptions of Generation rules.
Co-occurrence relation-based word selection for natural collocation can also be carried out synchronously. For this purpose, a co-occurrence dictionary of the target language is
necessary. The UNL Ontology is also helpful when no corres-
IJSER
A conversion system from native languages into UNL is
called "enconverter", and the one that deconverts from UNL
into native languages is called "deconverter". Information "en-
converted," from any language is exchanged in UNL format
via networks. Information represented in UNL is "decon- verted" into each native language on the terminal network [1].
In this paper, we have emphasized on “Deconverter” espe- cially on the Generation/Deconversion rules that are required to deconvert UNL expressions to Bangla sentences.
The DeConverter is a language independent generator, which provides a framework for syntactic and morphological genera- tion synchronously. It can convert UNL Expressions into a variety of natural languages using respective word dictiona- ries and sets of grammar rules of deconversion of the languag- es. A word dictionary contains the information of words that correspond to UWs included in the input of UNL Expressions and grammatical attributes (features) that describe the beha- viors of the words. Deconversion/Generation rules (grammar rules of deconversion) describe how to construct a sentence using the information from the input of UNL Expressions and defined in a word dictionary. The DeConverter converts UNL Expressions into sentences of a target language following the
————————————————
Co-Author Md. Nawab Yousuf Aki is currently working as the Chairman, CSE Department of East West University, Bangladesh. PH-
+8801xxxxxxxxx. E-mail: nawab@ewubd.com
Co-Author Jugal Krishna Das is a professor of CSE Department, Jahangir-
nagar University, Bangladesh. E-mail: drdas64@gmail.com
pondent word for a particular UW exist in a language. In this
case, the DeConverter consults to the UNL Ontology to try to
find a more general (upper) UW of which a correspondent
word exists in its word dictionary and use the word of the
upper UW to generate the target sentence instead.
The DeConverter works in the following way. It first trans- forms the input of a UNL expression – a set of binary relations
- into a directed graph structure with hyper-nodes called node-net. The root node of a node-net is called entry node and represents the head (e.g. the main verb) of a sentence. Decon- version of a UNL Expression is carried out by applying Gen- eration Rules to the nodes of node-net. It starts from the entry node, to find an appropriate word for each node and generate a word sequence (a list of words in grammatical order) of a target language. In this process, the syntactic structure is de- termined by applying syntactic rules, and morphemes are si- milarly generated by applying morphological rules. The de- conversion process ends when all words for all nodes are found and a word sequence of target sentence is completed.
Fig 1. shows the structure of the DeConveter. “G” indicates generation windows, and “C” indicates condition windows of the DeConverter. The DeConverter operates on the node-list through generation windows. Condition windows are used to check conditions when applying a rule. In the initial stage, in opposite to the EnConveter, the entry node of a UNL Expres- sion exists in the node-list. At the end of deconversion, the node-list is the list of all morphemes, with each as a node, that are converted from the node-net and constitute the target sen- tence.
IJSER © 2013
International Journal of Scientific & Engineering Research Volume 4, Issue 12, December-2013
ISSN 2229-5518
199
Fig 1: Deconversion process
second person and 3P for third person respectively and p is the temporary attribute for person to prevent recursive opera- tions. Some examples of rules for not alternative roots are as follows:
:"HPRON,1P,SUBJ::agt:"{ROOT,VEND,^ALT,#AGT,^1p:1p:
:}P9;
:"HPRON,2P,SUBJ::agt:"{ROOT,VEND,^ALT,#AGT,^2p:2p:
:}P9;
:"HPRON,2P,SUBJ,@respect,^@contempt,HON,^NGL::agt:"
{ROOT,VEND,^ALT,#AGT,^2p:2p::}P9;
:"HPRON,2P,SUBJ,^@respect,@contempt,^HON,NGL::agt:"
{ROOT,VEND,^ALT,#AGT,^2p:2p::}P9;
:"HPRON,3P,SUBJ::agt:"{ROOT,VEND,^ALT,#AGT,^3p:3p:
:}P9;
:"HPRON,(x)P,SUBJ, [^] @respect, | [^]@contempt, | [^]HON,[^]NGL::agt:"
{ROOT,VEND,[^]@present|@progress|@complete,VEG(y),ALT,
#AGT,^(x)p:(x)p::}P10;
where, UNL attributes @respect, @contempt, @present, @past,
@future, @progress, and @complete denotes for respected person,
neglected person, present tense, past tense, future tense, conti-
nuous tense and perfect tense respectively. ALT indicates alterna-
IJSER
For deconverting UNL expressions to Bangla sentence firstly,
we have gone through Universal Networking Language
(UNL) [2, 3, 4, 5, 6] where we have learnt about UNL expres-
sion, Relations, Attributes, Universal Words, UNL Knowledge
Base, Knowledge Representation in UNL, Logical Expression in UNL, UNL systems and specifications of Enconverter. All
these are key factors for preparing Bangla word dictionary, deconversion rules in order to deconvert UNL expressions to a natural language sentence (here Bangla sentence). Secondly, we have rigorously gone through the Bangla grammar [7, 8, 9,
10], Verb and roots (Vowel and Consonant Ended), Morpho- logical Analysis [11, 12, 13, 14, 19], suffixes [15, 16, 17, 18], con- struction of Bangla sentence based on semantic structure. Us- ing above references we extort ideas about Bangla grammar for morphological and semantic analysis in order to prepare Bangla word dictionary (for verb root, verbal inflections, etc) in the format of UNL provided by the UNL center of the UNDL Foundation.
In this section, we have proposed the formats of some gen- eration rules that are to be used for converting UNL expres- sions to Bangla sentence.
:"HPRON,(x)P,SUBJ::agt:"{ROOT,VEND,^ALT,#AGT,^(x)p: (x)p::}P9;
where, grammatical attributes HPRON denotes human pronoun, SUBJ for subject of a sentence, agt for agent relation, ROOT for verb root, VEND for vowel ended root, ^ALT for not alternative root, #AGT indicates that the corresponding root involves with agent relation. P indicates person and when the value of x=1, 2 and 3 then 1P denotes first person, 2P for
tive roots and when the value of y=1 then VEG1 is to be used for vowel ended group 1 and when y=2 then VEG2 for vowel ended group 2 and so on. ‘^’ expresses negation of the following value. Some examples of rules for not alternative roots are as follows:
:"HPRON,1P,SUBJ::agt:"{ROOT,VEND,@present,^@progress,
@complete,VEG1,ALT,#AGT,^1p:1p::}P10;
:"HPRON,1P,SUBJ::agt:"{ROOT,VEND,@past,^@progress,^@c
omplete,VEG1,ALT,#AGT,^1p:1p::}P10;
:"HPRON,2P,SUBJ,^@respect,^@contempt,^HON,^NGL,^PL:
:agt:"{ROOT,VEND,@present,^@progress,^@complete,VEG4,AL
T,#AGT,^2p:2p::}P10;
:"HPRON,2P,SUBJ,@respect,HON::agt:"{ROOT,VEND,@prese
nt,^@progress,^@complete,VEG4,ALT,#AGT,^2p,^hon:2p,hon::}
P10;
:"HPRON,2P,SUBJ,@contempt,NGL::agt:"{ROOT,VEND,@pres
ent,^@progress,^@complete,VEG4,ALT,#AGT,^2p,^ngl:2p,ngl::}
P10;
:"HPRON,3P,SUBJ::agt:"{ROOT,VEND,@present,^@progress,^
@complete,VEG2,ALT,#AGT,^3p:3p::}P10;
:"HPRON,3P,SUBJ::agt:"{ROOT,VEND,@past,^@progress,^@c omplete,VEG1,ALT,#AGT,^3p:3p::}P10;
:"HPRON,(x)P,SUBJ::aoj:"{ROOT,VEND,^ALT,#AOJ,^(x)p:(x)
p::}P9; Examples:
:"HPRON,1P,SUBJ::aoj:"{ROOT,VEND,^ALT,#AOJ,^1p:1p::}P
9;
:"HPRON,2P,SUBJ::aoj:"{ROOT,VEND,^ALT,#AOJ,^2p:2p::}P
9;
:"HPRON,3P,SUBJ::aoj:"{ROOT,VEND,^ALT,#AOJ,^3p:3p::}P
IJSER © 2013
International Journal of Scientific & Engineering Research Volume 4, Issue 12, December-2013
ISSN 2229-5518
200
9;
:"HPRON,(x)P,SUBJ, [^] @respect, [^]@contempt,[^]HON,[^]NGL::aoj:"
{ROOT,VEND,[^]@present,@progress,@complete,VEG(y),ALT,# AOJ,^(x)p:(x)p::}P10;
Examples:
:"HPRON,1P,SUBJ::aoj:"{ROOT,VEND,@present,^@progress,
@complete,VEG1,ALT,#AOJ,^1p:1p::}P10;
:"HPRON,1P,SUBJ::aoj:"{ROOT,VEND,@past,^@progress,^@c
omplete,VEG1,ALT,#AOJ,^1p:1p::}P10;
:"HPRON,2P,SUBJ,^@respect,^@contempt,^HON,^NGL,^PL:
:aoj:"{ROOT,VEND,@present,^@progress,^@complete,VEG4,AL
T,#AOJ,^2p:2p::}P10;
:"HPRON,2P,SUBJ,@respect,HON::aoj:"{ROOT,VEND,@prese
nt,^@progress,^@complete,VEG4,ALT,#AOJ,^2p,^hon:2p,hon::}
P10;
:"HPRON,2P,SUBJ::aoj:"{ROOT,VEND,@past,^@progress,@co
mplete,VEG1,ALT,#AOJ,^2p:2p::}P10;
Rule 5.1
?{:::}{PRON,SG,@pl:::}P8;
N,NGL:::"P10;
:{ROOT,VEND,(x)p,[^]@present,[^]@progress,[^]@complete,^ hon,|^ngl,^kbiv:kbiv::}"[[KBIV]],KBIV,VEND,(x)P,PRS|PST|FU T,[^]PRGR,[^]CMPL,^HON,^NGL:::"P10;
Examples:
:{ROOT,VEND,3p,@present,^@progress,^@complete,^hon,^k
biv:kbiv::}"[[KBIV]],KBIV,VEND,3P,PRS,^PRGR,^CMPL,^HON::
:"P10;
:{ROOT,VEND,3p,@present,@progress,^@complete,^ngl,^kbi
v:kbiv::}"[[KBIV]],KBIV,VEND,3P,PRS,PRGR,^CMPL,^NGL:::"P1
0;
:"N,[^]@pl,^SUBJ:SUBJ:agt:"{ROOT,VEND,#AGT,^3p,[^]sg|p l:3p,sg|pl::}P10;
Examples:
:"N,^@pl,^SUBJ:SUBJ:agt:"{ROOT,VEND,#AGT,^3p,^sg:3p,sg
::}P10;
:"N,@pl,^SUBJ:SUBJ:agt:"{ROOT,VEND,#AGT,^3p,^pl:3p,pl::}
P10;
Rule 10.1 (For singular)
Rule 5.2
?{:::}{PRON,PL,^@pl:::}P8;
Rule 5.3
IJSE:{N,^@pl,@defR,^boch:boch::}"[[BIV]],BIV,BOCH,DEF,^PL:::"P1
?{:::}{HPRON,SUBJ,HON,^@respect:::}P8;
Rule 5.4
?{:::}{HPRON,SUBJ,^HON,@respect:::}P8;
Rule 5.5
?{:::}{HPRON,SUBJ,NGL,^@contempt:::}P8;
Rule 5.6
?{:::}{HPRON,SUBJ,^NGL,@contempt:::}P8;
:{ROOT,VEND,(x)p,[^]@present,[^]@progress,[^]@complete,^
kbiv:kbiv::}"[[KBIV]],KBIV,VEND,(x)P,PRS|PST|FUT,[^]PRGR,[
^]CMPL:::"P10;
Examples:
:{ROOT,VEND,1p,@present,^@progress,^@complete,^kbiv:kb
iv::}"[[KBIV]],KBIV,VEND,1P,PRS,^PRGR,^CMPL:::"P10;
:{ROOT,VEND,1p,@present,@progress,^@complete,^kbiv:kbi
v::}"[[KBIV]],KBIV,VEND,1P,PRS,PRGR,^CMPL:::"P10;
:{ROOT,VEND,(x)p,[^]@present,[^]@progress,[^]@complete,[
^]hon,[^]ngl,^kbiv:kbiv::}"[[KBIV]],KBIV,VEND,(x)P,PRS|PST|F
UT,[^]PRGR,[^]CMPL,[^]HON,[^]NGL:::"P10; Examples:
:{ROOT,VEND,2p,@present,^@progress,^@complete,^hon,^n gl,^kbiv:kbiv::}"[[KBIV]],KBIV,VEND,2P,PRS,^PRGR,^CMPL,^ HON,^NGL:::"P10;
:{ROOT,VEND,2p,@present,^@progress,^@complete,^hon,ngl
,^kbiv:kbiv::}"[[KBIV]],KBIV,VEND,2P,PRS,^PRGR,^CMPL,^HO
Rule 10.2 (For plural)
:{N,[^]HUMN,@pl,@def,^boch:boch::}"[[BIV]],BIV,BOCH,DEF
,PL,[^]HUMN:::"P10;
Examples:
:{N,^HUMN,@pl,@def,^boch:boch::}"[[BIV]],BIV,BOCH,DEF,P
L,^HUMN:::"P10;
:{N,HUMN,VEND,@pl,@def,^boch:boch::}"[[BIV]],BIV,BOCH,DE
F,PL,HUMN,VEND:::"P10;
In this section, we have converted UNL expression into Bangla sentence. For this conversion process, we need Dictio- nary, which we have developed but not discussed in this pa- per and a set of Generation rules given below.
:”HPRON,1P,SUBJ::agt:”{ROOT,VEND,^ALT,#AGT,^1p:1p
::}P9
R{:::}{SUBJ:::}
:{SUBJ,^blk:blk::}”[],BLK:::”P10;
R{:::}{SUBJ:::}
:”N,^obj:obj:obj:”{ROOT,VEND,^#AGT,#OBJ:::}P9;
R{:::}{N:::}
IJSER © 2013
International Journal of Scientific & Engineering Research Volume 4, Issue 12, December-2013
ISSN 2229-5518
201
:{N,^blk:blk::}”[ ],BLK:::”P10; Rule 8: (Right Shift Rule) R{N:::}{BLK:::}
R{BLK:::}{ROOT:::}
:{ROOT,VEND,1p,#AGT,^@progress,^@complete,^kbiv:kb iv::}”[[KBIV]],KBIV,VEND,1P,PRS,^PRGR,^CMPL:::”P10;}
R{V:::}{:::}
These generation rules will be applied to the nodes in the
node-list for operation on them and/or inserting nodes from
the Node-net into the Node-list.
Rule 1 is pronoun insertion rule that describes that when root “যা” (ja)is in the RGW, the pronoun “আিম” (aami) to be in- serted in the LGW. Rule 2 is the right shift rule which is ap- plied to shift the windows of DeConverter to right. The blank insertion rule (rule 3) is applied to insert a blank between pro- noun “আিম” (aami) and root “যা” (ja). Again right shift rule (rule
4) is applied to shift the windows to right. Now, rule 5 to be
used to insert noun “বািড়” (bari) in the RGW. And if noun “বািড়”
(bari) is in the RGW then windows will be shifted to the right
by applying rule 6. Now, rule 7 will be applied to insert a blank between noun “বািড়” (bari) and root “যা” (ja). Rule 8 and
9 both are right shift rules to be applied to shift the windows
UNL expressions to respective native language.
SCREEN 2: DECONVERTER FILE (DECOL27.EXE)
Screen 3 shows the way of selecting files to be used for decon- version
two steps to right. After that verbal inflexion “ই” (i) to be in- serted into the node-list by applying rule 10 . Finally, rule 11 to be applied to place the sentence tail in the RGW to complete the conversion process to get the target sentence - আিম বািড় যাই
To convert the UNL expression into Bangla language sentence we have used following three files (created by us) and 2 more files (see Screen 1)
1. Input file (unl expressions)
2. Dictionary file (WD.txt)
3. Rules File (DeRules.txt), contains the Generation Rules, described in Section 5
4. Decoder (DeCoL27.exe)
5. A dictionary builder file (DicBldL.exe) provided by the
UNDL Foundation of UNL center, which have been downloaded
SCREEN 1: SHOWS THE FILES FOR DECONVERSOION OF UNL EX- PRESSION TO BANGLA NATIVE SENTENCE
Screen 2 shows the DeConverter (DeCoL27) that can convert
SCREEN 3: FILES TO BE USED FOR DECONVERSION
Screen 4 shows the contents of output file i.e. the Bangla sen- tence, “আিম বািড় যাই“
SCREEN 4: BANGLA SENTENCE “আিম বািড় যাই” FROM GIVEN UNL EX- PRESSION
IJSER © 2013
International Journal of Scientific & Engineering Research Volume 4, Issue 12, December-2013
ISSN 2229-5518
202
In this paper, we have developed few Generation Rules. These rules can currently produce simple Bangla sentences. It is be- ing enhanced to handle more types of sentences. The present system is a small achievement towards attaining a complete translation facility for the Bangla language.
[1] http://www.unl.ru/system.html last accessed on October 7, 2013 [2] http://www.undl.org last accessed on October 10, 2013
[3] H. Uchida, M Zhu. T.G. Della Senta. Universal Networking Language,
2005/6-UNDL Foundation, International Environment House.
[4] H. Uchida, M. Zhu. The Universal Networking Language (UNL) Specifica- tion Version 3.0 Edition 3 , Technical Report, UNU, Tokyo:, 2005/6- UNDL Foundation, International Environment House, 2004
[5] EnConverter Specification, Version 3.3, UNL Center/UNDL Foundation,
Tokyo 150-8304, Japan 2002
[6] DeConverter Specification, Version 2.7, UNL Center, UNDL Foundation, Tokyo 150-8304, Japan 2002
[7] D.M. Shahidullah. Bangla Baykaron, Dhaka: Ahmed Mahmudul Haque of Mowla Brothers prokashani, 2003
[8] D. C. Shuniti Kumar. Bhasha-Prakash Bangala Vyakaran, Calcutta : Rupa
and Company Prokashoni, July 1999, pp.170-175
[9] Humayun Azad. Bakkotottyo - Second edition, Dhaka: Bangla Academy
gla Sentence into Universal Networking Language Expression (IJCSI),
International Journal of Computer Science Issues, Vol. 8, Issue 2, March
2011, pp. 64-73
[19] M.N.Y. Ali, M. Z. H. Sarker, G. F. Ahmed, J.K. Das, “Rules for Morpho- logical Analysis of Bangla Verbs for Universal Networking Language”, International Conference on Asian Language Processing (IALP), Harbin, China, December 2010, pp. 31-34
Publishers, 1994
IJSER
[10] D. S. Rameswar. Shadharan Vasha Biggan and Bangla Vasha, Pustok Bi-
poni Prokashoni, November 1996, pp.358-377
[11] M.N.Y. Ali, J.K. Das, S. M. Abdullah Al Mamun, M. E. H. Choudhury. “Specific Features of a Converter of Web Documents from Bengali to Universal Networking Language”, International Conference on Computer and Communication Engineering 2008 (ICCCE’08), Kuala Lumpur, Ma- laysia. pp. 726-731
[12] M.N.Y. Ali, J.K. Das, S.M. Abdullah Al Mamun, A. M. Nurannabi. “Morphological Analysis of Bangla words for Universal Networking Language”, International Conference on Digital Information Management, icdim, 2008, London, England, pp. 532-537
[13] M.N.Y.Ali, A. M. Nurannabi, G. F. Ahmed, J.K. Das. “Conversion of Bangla Sentence for Universal Networking Language”, International Conference on Computer and Information Technology (ICCIT), Dhaka, 2010 pp. 108-113
[14] M. Z. H. Sarker, M.N.Y. Ali, J.K. Das, Dictionary Entries for Bangla Consonant Ended Roots in Universal Networking Language, Interna- tional Journal of Computational Linguistics (IJCL), Volume (3) : Issue (1) :
2012 pp 79-87
[15] M. Z. H. Sarker, M.N.Y. Ali, J.K. Das, Outlining Bangla Word Dictio- nary for Universal Networking Language, IOSR Journal of Computer Engineering (IOSRJCE), ISSN: 2278-0661 Sep-Oct. 2012
[16] M. Z. H. Sarker, M.N.Y. Ali, J.K. Das, Development of Dictionary En- tries for the Bangla Vowel Ended Roots for Universal Networking Language, International Journal of Computer Applications 52(19), August
2012. Published by Foundation of Computer Science, New York, USA. pp 38-45
[17] M.N.Y. Ali, M. Z. H. Sarker, J.K. Das, Analysis and Generation of Ben- gali Case Structure Constructs for Universal Networking Language, In- ternational Journal of Computer Applications (IJCA), Volume 18, No. 7, March 2011, pp. 34-41
[18] M.N.Y. Ali, M. Z. H. Sarker, G. F. Ahmed, J.K. Das, Conversion of Ban-
IJSER © 2013