Context-Sensitive Phrase Structure Rule for Structural

Representation of Bangla Natural Language Sentences.

Mohammed Moshiul Hoque Muhammad Masroor Ali

Dept. of Computer Science and Engineering Institute of Information and Communication Technology
Chittagong University of Engineering and Technology, Bangladesh University of Engineering and Technology,
Chittagong-4349, Bangladesh Dhaka-1000, Bangladesh
Email: [email protected] Email: [email protected]

and how these phrases can themselves be aggregated

finally into sentences. A grammar captures the legal
ABSTRACT: structure in a language and thus allows a sentence to be
Phrase-structure rules are used in context-sensitive analyzed [2]. By using the Context-free grammar
grammars to describe a given language /’s syntax. This (CFG) rules, it cannot be possible to generate all the
is accomplished by attempting to break language down sentence patterns in Bangla languages. To generate all
into its constituent parts. Phrasal categories include the sentences of Bangla, we have to use CSG’s. The
the noun phrase, verb phrase, and adjective phrase; CSG PS rules says that how parts of speech can put
lexical categories include noun, verb, adjective, adverb together to make grammatical sentences.
and many others. This paper proposes a set of context- A parser module can used the CSG’s phrase-structure
sensitive phrase structure rules to parse the all kinds of rules and also can find out how words in sentence
Bangla natural language sentences including simple, related to each other. Therefore, parsing a sentence by
complex and compound sentences. The proposed rules using a set of PS rules produces underlying structure
can parse the all types of Bangla sentences. This paper for sentence. The result is usually a tree or structural
also describes a technique to decompose a complex representation (SR) that can be used as input for a
sentence into a dependent and independent clause and Machine Translation Engine. Machine translation (MT)
a compound sentence into a simple sentence [1] [2] is an emerging paradigm for processing natural
respectively. The inflection of Bangla verb phrase languages. Now a day, MT and NLP are being used by
called auxiliary can have different forms depending on language industries for translation one language to
the tense, the person and the class of the subject of the another and allow people to interact with computers in
verb. In this paper, we also presented a decomposition a natural human language that will save a lot of time
procedure for Bangla verb phrase into several and cost of money.
subparts and then extracting necessary information
from the auxiliary part. 1.1 Previous work and Contribution of this
Keywords: Phrase structure, SR, Parsing, Paper:
Decomposition, Clause. Parsing of Bangla sentences by using CSG’s rules is in
rudimentary stage still now. Very few research work
INTRODUCTION: have been conducted regarding parsing of Bangla
Natural language processing (NLP) are developed both sentences but a significant number of research
to explore general theories of human language activities have been conducted on the recognition of
processing tasks such as providing natural language Bangla characters. Among them [4] describe the
interfaces or front ends to application system. The parsing of Bangla sentences for simple sentence, while
subject of NLP is an immense field with many [5] describes detail analysis of Bangla phrases and
potential applications, including translation from one different types of sentences by using the
language into another, retrieval of information from Transformational Generative Grammar’s (TGG’s).
databases, human/computer interaction, and automatic Syntax analysis for Bangla sentences is presented in
dictation. A language understanding program must [7]. Besides these, [6] focuses on designing a Bangla
have considerable knowledge about structure of the conversion processor using natural language
language including what words are and how they processing. Parsing methodology for Bangla simple
combine into phrases and sentences. It must also know sentences by using context-free grammars (CFG’s) is
the meaning of the words and how they contribute to presented in [8]. This paper is an extension to scheme
the meaning of the sentence and to the context within presented in [8]. The extensions and refinements
which they are being used. In this paper, we represent a presented in this paper over previous work are as
set of context-sensitive grammar (CSG) rules to parse follows:
the all kinds of Bangla sentences. To interpret any
• A set of context-sensitive PS rules for simple
language requires determining the sentence structure of
sentence.
that language. Phrase structure (PS) grammars define
• A sets of context-sensitive PS rules for
how the basic components of symbol strings, the
complex sentences.
symbols themselves, can be aggregated into phrases,
 • A sets of context-sensitive PS rules for Rule CSG’s Rule
compound sentences. No.
• Decomposition procedures for auxiliary of 01. SS NP VP
02. VP (NP) (AP) VF
verb phrase.
03. NP N
• Decomposition technique for complex and 04. NP N (DET)
compound sentences into simple sentence and 05. NP NP (Biv) (NP)
parse these sentences as simple sentences. 06. NP (DEMO) (SPR) (AP) N
• SR has been used to represent the complex 07. NP N PM
and compound sentences respectively. 08. N PER
09. AP AD/AD’S
2. CONTEXT-SENSITIVE GRAMMAR (CSG’s) 10. DEMO (DD) (DO)
11. SPR QFR (PP)
FOR BANGLA:
12. VF VR AUX
Contex – sensitive grammars (CSGs) are the collection 13. PER FP / SP/ TP
of a set of PS rules in which, at least one PS rule is 14. FP ami / amara (Avwg/Avgiv)
context- sensitive [3]. Applications of PS rules are 15. SP SPH / SPNH / SPP
restrictive by the context that is if certain conditions 16. SPH apani / apanara (Avcwb/Avcbviv)
regarding the context are fulfilled then they will be 17. SPNH tumi/tomara (Zzwg /†Zvgiv)
activated. In CSGs, auxiliaries are depended on 18. SPP tui/tora (ZzB/†Zviv)
subjects. The category of person + class for both the 19. TP TPH / TPNH
auxiliary and subject should be the same. Now, 20. TPH tini/tara/shikak (wZwb/Zviv/wk¶K)
21. TPNH se/tara/pakhi/boi (‡m/Zviv/cvwL)
according to the above discussion, we can represent a
22. DET ti (wU)
set of context – sensitive phrase grammar rules for 23. QFR ek/dui/bahu/anek (GK/`yB../eû/A‡bK)
simple sentence in table 1. Form the table 1, we find 24. PP ti/ta/khana/jon (wU/Uv /Lvbv/ Rb)
the structure of the verb phrase (VP) and verb form 25. AD khub/bhalo/sundor (Lye/fvj/ my›`i)
(VF). 26. PM era/ra/der/samuha (Giv/iv/†`i/mg~n)
3. STRUCURE OF BANGLA VERB PHRASE 27. DD ai/sei/oi (GB/†mB/H)
(VP): 28. DO prothom/ ditia (cÖ_g/wØZxq.)
Every sentence in Bangla must have a verb phrase. The 29. Biv ke/der/re/kar (‡K/†`i/†i/Kvi)
compulsory part of a VP is the verb Form (VF). 30. VR likh/por/chal/ja (wjL/co/Pj/ hv)
31. AUX e/che/bhe/bo/chhen (G/†Q/†e/e/†Qb)
The original part may contain a NP or an AP or a NP
and AP. As a rule VP can be expressed as follows-
auxiliary should be “en”( He) or for “Tumi”(a¤¢j), it
VP (NP) (AP) VF should be “o”(A) or “e”(H) are accepted and so on.
Such as- “Apani Boi Porchen” (Bf¢e hC fs-Re) [Apani
3.1. Verb Form (VF): (Bf¢e)+ Boi (hC) + Por (fs) + chh (R) + en (He)]. But
Verb Form is the most important word in a sentence. It it is not possible to expressed – “Apani Boi
is the word used for stating something about a person Porcho”(Bf¢e hC fs-R¡) [Apani (Bf¢e) + Boi (hC) +
or thing. The Bangla verb form can be segmented into Por (fs) + chh (R) + o (A)], Because “o”(A) does not
two parts: These are- i) Verb Root (VR) ii) Auxiliary agree with the subject “Apani”(Bf¢e). Now, we will try
(AUX) to shown in table 2, 3, 4, 5 how the auxiliary part of
verb form (VF) can relate or agree with the various
4. AGREEMENT: person and classes of the subject and with the various
In every language, there is an agreement between the tenses of the verb.
subject and verb form. Similarly, there is also an
agreement of person + class between verb form and 4.1. Subcategorization and updating the
subject in Bangla. The auxiliary changes for tense, Grammar Rules:
aspect and person + class. For example- Subcategorization expresses the constraints that a
“Tini Boi Porchen”(¢a¢e hC fs-Re). Here, -------------- predicate (verb for now) places on the number and type
VP Boi Porchen (eB co‡Qb) of the argument it wants to take. Therefore, from
VF Porchen (co‡Qb) the above analysis, we can concluded that the inflection
VR Por (co) of Bangla verb called AUX can have different forms
AUX Chen (‡Qb) and hence subcategorized depending on the tense, the
AUX Aspect + Tense + Concord person and the class of the verb. The forms are
Che = Ch + Φ+ en described in table 2, 3, 4 and 5. After a thorough
= Continuous + Present + Concord of Third Person observation of table 2, 3, 4 and 5, we proposed a set of
Honorific for Present Tense updated Context-sensitive grammar rules to further
That means, in Bangla, the person + class of AUX decompose and extract information from the verb of a
follows the person + class of subject NP. For example, sentence, with the following PS rules of table 6.
if subject is “Apani”( Bf¢e) then person + class of
Table 1. CSG’s PS rules for Simple Sentence of Bangla
Natural language.
 Table 2: Agreement of Person + class with the verb auxiliary and subject for Past Tense:
Person and Class
FP SPH SPNH SPP TPH TPNH
Subject ami/amara apani/apanara tumi/tomara tui/tora tini/tara se/tara
(Avwg/Avgiv) (Avcwb/Avcbviv) (Zzwg/‡Zvgiv) (ZzB/‡Zviv) (wZwb/Zviv) (‡m/Zviv)
Iam =l o+am len = lo + en le = lo + e li = lo + i len = 1o + en lo (j)
Indefinite (jvg=j+Avg) (†jb=j+Gb) (†jb=j+G) (wj=j+B) (†jb=j+Gb) kor + lo
kor+lam kor + len kor + le kor + li + kor + len = = korlo
= korlam (Kijvg) = korlen (Ki‡jb) = korle (Ki‡j) = korli (Kiwj) korlen (Ki‡jv)
(Ki‡jb)
chholam = chh + i chhilen = chh + i chhile = chh + i chhili = chh + i chhilen = chh chilo = chh + i
+ lo + am (wQjvg + lo + en + lo + e + lo + i + i + lo + en + lo
Continuous = Q+B+j+Avg) (wQ‡jb = wQ‡j = (wQwj=Q+B+j (wQ‡jb=Q+B (wQj=Q+B+j
kor + hhi Q+B+j+Gb) Q+B=j+G +B) +j+Gb) +A)
- alm = kor + chhilen = kor + chhile = kor + chhili = kor + chhilen kor + chilo
Aspect

korchhialm korchhilen korchhile korchhili = korchhilen = korchilo

(KiwQjvg) (KiwQ‡jb) (KiwQ‡j) (KiwQwj) (KiwQ‡jb) (KiwQj)
echhilam =e + chh echhilen = e + chh echhile = e + echhili= e + echhilen=e+ echilo=e+chh
+ i + lo + am + i + lo + en chh + i + lo + e echh+i+lo+i chh+i+lo+en +i+lo
Perfect (GwQjvg= (GwQ‡jb=G+Q+ (GwQ‡j) (GwQwj) = (GwQj =
G+Q+B+j + Avg) B+j+Gb) kor + echhile = =kor + echhili (GwQ‡jb) G+Q+B+j+A)
kor + chhilam kor + echhilen = korechhile = korchhili =kor + echhilen kor+echilo
= korechhilam korechhilen (K‡iwQ‡j) (K‡iwQwj) = korechhilen = korechilo
(K‡iwQjvg) (K‡iwQ‡jb) (K‡iwQ‡jb) (K‡iwQj)

Imperative Not applicable

Table 3: Agreement of Person + class with the verb auxiliary and subject for Future Tense:
Person and Class
FP SPH SPNH SPP TPH TPNH
Subject ami/amara apani/apanara tumi/tomara tui/tora tini/tara se/tara
(Avwg/Avgiv) (Avcwb/Avcbvi (Zzwg/‡Zvgiv) (ZzB/‡Zviv) (wZwb/Zviv) (‡m/Zviv)
v)
bo (e) ben (†eb) be (†e) bi (we) ben (†eb) be (†e)
Indefinite kor + bo = kor + ben = kor + be = kor + bi = kor + ben kor + be
(Ki+e = (Ki+†eb=Ki‡eb (Ki+†e=Ki‡e) (Ki+we=Kiwe) (Ki+†eb= (Ki+†e
korbo (Ki‡ev) ) korbe korbi Ki‡eb) = korbe
korben = korben Ki‡e)
te thakbo (†Z _vKe) te thakben te thakbe te thakbi te thakben te thakbe
(†Z _vK‡eb) (†Z _vK‡e) (†Z _vKwe) (†Z _vK‡eb) (†Z _vK‡e)
Aspect

Continuous
e thakbo (G _vK‡e) e thakben e thakbe e thakbi e thakben e thakbe
(G _vK‡eb) (G _vK‡e) (G _vKwe) (G _vK‡eb) G _vK‡e)
Perfect
Imperative Not applicable ben (†eb) o (A) ish (Bm) ben (†eb) be (†e)
kor + ben= kor + o = kor + ish = kor + ben = kor + be =
(Ki+‡eb= Ki‡eb) (Ki+A=K‡iv) (Ki+Bm= (Ki+eb= (Ki+†e= Ki‡e)
korben koro Kwim) Ki‡eb) korbe
korish korben
Table 4: Agreement of Person + class with the verb auxiliary and subject for Habitual Past:
Person and Class
FP SPH SPNH SPP TPH TPNH
Subject ami/amara apani/apanara tumi/tomara tui/tora tini/tara se/tara
(Avwg/Avgiv) (Avcwb/Avcbvi (Zzwg/‡Zvgiv) (ZzB/‡Zviv) (wZwb/Zviv) (‡m/Zviv)
v)
tam=to+am ten = to + en te = to + e ti = to + i ten = to + en to (Z)
Indefinite (Zvg = Z+Avg) (†Zb = Z+Gb) (†Z= Z+G) (wZ=Z+B) (‡Zb=Z+Gb) kor + to =
kor + tam = kor + ten = kor + te = kor + ti = kor + ten= (Ki+Z=KiZ)
(Ki+Zvg=KiZvg) (Ki+†Zb = Ki‡Zb) (Ki+†Z=Ki‡Z) (Ki+wZ=KiwZ) korten korto
kortam korten korte korti (Ki‡Zb)
Aspect

te thaktam te thakten te thakte te thakti te thakten te tha kto

(†Z _vKZvg) (†Z _vK‡Zb) (†Z _vK‡Z) (†Z _vKwZ) (†Z _vK‡Zb) (†Z _vK‡Zv)
Continuous
e thaktam e thakten e thakte e thakti e thakten e thakto
(G _vKZvg) (G _vK‡Zb) (G _vK‡Z) (G _vKwZ) (G _vK‡Zb) (G _vKZ)
Perfect
Imperative Not applicable
 Table 5: Agreement of Person + class with the verb auxiliary and subject for Present Tense:
Person and Class
FP SPH SPNH SPP TPH TPNH
Subject ami/amara apani/apanara tumi/tomara tui/tora tini/tara se/tara
(Avwg/Avgiv) (Avcwb/Avcbvi (Zzwg/‡Zvgiv) (ZzB/‡Zviv) (wZwb/Zviv) (‡m/Zviv)
v)
i (B) en (Gb) o (A) ish (Bm) en (Gb) e (G)
Indefinite kor + i =kori kor + en = kor + o = kor + ish = kor + en = kor + e = kore
(Ki + B =Kwi) koren (Ki+Gb) koro (Ki+A) korish (Kwim) koren (K‡ib) (K‡i)
chhi/chchhi = chhen/ (†Qb) cho/ (†Qv) chhish/ (wQm) chhen/ (†Qb) chhe/ (†Q)
wQ/w”Q chehhen (†”Qb) chchho (†”Q chchhish chchhen chchhe (†”Q
Continuous kor + chhi (Ki + wQ kor + chhen kor + chho kor + chhish = kor + chhen = kor + chhe
korchhi (KiwQ) = korchhen = korchho korchhish korchhen = korchhe
(Ki‡Qb) (Ki‡Qv) (KiwQm) (Ki‡Qb) (Ki‡Q)
Aspect

echhi/ echhen/(G‡Qb) echho/ (G‡Qv) echhish/ echhen/ echhe/ (G‡Q

(BwQ/Bw”Q) echchhen echchho (GwQm) (G‡Qb) echchhe
Perfect echchhi kor+echhen kor+echho echchhish echchhen kor+echhe
kor + echhi = korechhen = korecho kor+echhish = kor+echhen = korechhe
= korechhi (K‡i‡Qb) (K‡i‡Qv) korechish = korechhen (K‡i‡Q)
(Ki+BwQ= (K‡iwQm) (K‡i‡Qb)
K‡iwQ)
Not applicable un (Db) o (A) Null un (Db) uk (DK)
Imperative kor + un = kor + o = kor + Null kor + un = kor + uk =
korun (Ki“b) koro (K‡iv) = kor (Ki) korun (Ki“b) koruk (Ki“K)
5. TYPES OF SENTENCES:
Structurally, there are three (3) types of sentence in trees or an SR. By using the context-sensitive PS rule
Bangla [3]. So, a PS rule for different kinds of sentences of table 1, 6, and 7, we can parse the simple sentence
of Bangla can be expressed as table 7: shown in figure 1.
Table 7: PS rules for Bangla Sentence. Table 6: Context – Sensitive PS rules for Bangla with
Rule No. PS Rule for Simple Sentence Subject -Verb and Person + Class agreement.
01. S SS / CS / COMS Rule CSG’s Rule
02. SS Simple Sentences No.
03. CS Complex Sentences 01. VP (NP) (AP) VF
04. COMS Compound Sentences 02. VF VR AUX
03. VR kor (Ki) /por (co) /other verb roots
5.1. Structure of the Simple Sentence: 04. AUX ASP T CON
A simple sentence is formed by an independent clause 05. T Pr/Pt/Fu
or principle clause. A principle clause in Bangla can be 06. Pr Null
rewritten as (or decomposes into, or consist of) a NP 07. Pt lo (j)/ il (Bj)
08. Fu b (e)/bo (†ev)
followed by VP. For example - “se kall ashbe” (®p
09. ASP I/Con/P/IMP/Hab
L¡m Bp-h) or “Rahim eshkole jae” (l¢qj ú¤-m k¡u). 10. I Null
Therefore, a simple sentence can be represented by the 11. Con chh (Q)/ chchh (”Q)
following PS rule- 12. P echh (GQ)
SS NP VP 13. IMP Null
S 14. Hab to (Z)
16. NICON NICONFP/
NICONSPH/NICONSPNH
NP VP 17. NCON NICONSPP/NICONTPH/
NICONTPNH/
18. ICON ICONSPH/ICONSPNH/
SPR N NP VF ICONSPP/ICONTPH/ICONTPNH
19. NICONFP i (B)/am (Avg)/Null/o (A)
Per N 20. NICONSPH en (Gb)
QFR PP VR
21. NICONSPNH Null/e (G)
TPNH Per ASP T CON 22. NICONSPP ish (Bm)/I (B)
23. NICONTPH en (Gb)
TPNH Con Pt NICON 24. NICONTPNH e (G)/Null
25. ICONSPH un (Db)/en (Gb)/ben (†eb)
NICONTPNH 26. ICONSPNH o (A)
ek ti Chele boi por chh Null e 27. ICONSPP ish (Bm)/Null/i (B)
28. ICONTPH en (Gb)/un (Db)
The structure of sentences that are allowed by grammar 29. ICONTPNH e (G)/uk (DK)/be (†e)
Fig. 1: SR for the simple sentences “ Ekti Chele boi
Porche” (HL¢V ®R-m hC fs-R).
5.2. Structure of the Complex Sentence: analyzed the independent clauses separately. If
independent clause is same as simple sentence then
A complex sentence is formed by one principle clause
analyze that clause as simple sentence or if the
and one or more subordinate clause. For example -
independent clause is same as complex sentence then
“ami jani je tumi ashbe” (B¢j S¡¢e ®k a¤¢j Bp-h). In
analyze that clause as complex sentence. For example -
this sentence, Principle Clause: “ami jani” (B¢j S¡¢e),
“Tomora manush noo, ar jader chalaoo taraoo manush
and Subordinate Clause: “je tumi ashbe” (®k a¤¢j
noo”(®a¡jl¡ j¡e¤o eJ, Bl k¡-cl Q¡m¡J a¡l¡J j¡e¤o eu). It
Bp-h). This two-sentence “ami jani” (B¢j S¡¢e) and “je
is a compound sentence and it consists of two
tumi ashbe” (®k a¤¢j Bp-h) are simple form but
independent clause- (i) “Tomara manush noo” (®a¡jl¡
second part is the subordinate of the first part. Clauses
j¡e¤o eJ) and (ii) “Jader chalaoo taraoo manush noo”
are combined with subordinator and/ or corresponding
(k¡-cl Q¡m¡J a¡l¡J j¡e¤o eu). These two independent
subordinator complement.
clauses are combined a connective ar (Bl). First
Some subordinator and their respective subordinator
independent clause is a simple sentence and second
complement used in Bangla are given in table 8.
independent clause is a complex sentence. In the
Table 8: Some Subordinators and their respective
second independent clause is the combination of two
Subordinator Complements.
clauses: (i) Principle clause: “Tomora manush noo”
Subordinator (SUBORD) Subordinator Complement
(SUBCOM) (®a¡jl¡ j¡e¤o eJ) and (ii) Subordinate Clause: “jader
jadi (hw` ) tahole (Zvn‡j) chalaoo” (k¡-cl Q¡m¡J). These two clauses are simple
jadi (hw` ) tobe (Z‡e) sentence. Therefore, first two independent clause are
jadi-na (hw`-bv) tahole (Zvn‡j) separated and first one is analyze as simple sentence
jekhane (‡hLv‡b) sehkane (‡m‡nZz) and second one is analyze as complex sentence and
je (‡h) se (‡m) then the two clause under the complex sentence are
jemon (‡hgb) temon (‡Zgb) analyze as simple sentence separately. Compound
jehetu (‡h‡nZz) sehetu (‡m‡nZz) sentence can be constructed in several ways [8]. Such
jara (hviv) tara (Zviv)
as-
S
From the above analysis, proposed a set of PS rules for
parsing the complex Sentences of Bangla. Apply the CS
rules of table 1, 6, 7, and 8 for parse the complex
sentence is shown in fig. 2. IC DC
Table 9: A set of Rules for parsing Complex Sentence.
SS
SUBORD SS
Rule No. Rule for Complex Sentence NP VP
01. C DC IC N VF
02. CS IC DC NP VP
DC NP (SUBORD) VP Per
03. VR AUX N
04. DC (SUBORD) SS FP NP VF
05. IC NP (SUBCOM) VP Per
ASP T CON N
06. IC (SUBCOM) SS SPNH VR AUX
07. SS All Simple Sentenc I Pr NICON Per
08. SUBORD Jadio, Jadi, Jekhane, Jehetu TPNH ASP T CON
09. SUBCOM Tahole, Tobe, Sekhane, Sehetu Null
I Fu NICON
Null kal ash NICONSPNH
5.3. Structure of the Compound Sentence:
A compound sentence is formed by two or more ami jan i je tumi Null b e
principle clause joined by an indeclinable or Fig. 2: SR for the Complex Sentence “Ami Jani Je
connectives. Tumi kal ashbe” (B¢j S¡¢e ®k a¤¢j Bp-h).
5.3.1 Connectives or Indeclinable: (i) Simple + Simple:
Indeclinable means no changes i.e. word in the Consider an example- "se soth kingtu tar bhai osoth"
sentence is unchanged. The part of speech, which is (®p pv ¢L¿º a¡l i¡C Apv). In this sentence, "se soth"
always, unchanged form and which makes the sentence (®p pv) and "tar bhai osoth" (a¡l i¡C Apv) are principal
more meaningful and also used, as a connective for clause or simple sentence and kingtu (¢L¿º) is a
more than one part of speech, clauses or sentences is connective.
called indeclinable. As a PS rule Connectives can be (ii) Simple + Complex:
represented as follows: Consider an example- "se soth kingtu je bonduti tar
Conj O (J), Ebong (Hhw), Fale (g-m), Sutorang sange ashechilo se osoth" (®p pv ¢L¿º ®k h夢V a¡l
(Bl p¤al¡w), Kingtu (¢L¿º). p-‰ H-p¢Rm ®p Apv). In this sentence, "se soth" (®p
pv) is a simple sentence and "je bonduti tar sange
5.3.2 Decomposition Technique of Compound ashechilo se osoth" (®k h夢V a¡l p-‰ H-p¢Rm ®p
Sentence: Apv) is a complex sentence.
To analyze the compound sentence, breaks the (iii) Complex + Complex:
sentence into several independent clauses then
Consider a long sentence as an example -"jadi jante • CSG’s rules to handle various types of Bangla
chaoo se keno asheni ta hole bolbo ami jani Na, ar jadi punctuation symbols.
jante chaoo ami keno jaini ta hole bolbo amar echhe • The concepts of Voice, narration, composition, of
holona Tai" (k¢c S¡e-a Q¡J ®p ®Le B-p¢e a¡ q-m words and inflection of noun or pronoun should be
hm-h¡ B¢j S¡¢e e¡, Bl k¢c S¡e-a Q¡J B¢j ®Le S¡¢e further analyzed.
a¡q-m hm-h¡ Bj¡l C-µR q-m¡ e¡ a¡C). In this example, • Pragmatic analysis on complex and compound
two complex sentences are connected by a connective sentences may be carried on.
ar (Bl). In this similar manner, compound sentence can
be constructed as simple + simple, simple + complex, 7. CONCLUSION:
complex + simple, Complex + Complex etc. Now, we Parser is one of the most important parts of an NLP
represent a set of rules to express the Compound and an MT systems. Parsing a sentence then involves
sentence in table 10. finding a possible legal structure for sentence and
Table 10: A set of rules for Compound Sentence. finally gets an SR. Our proposed context-sensitive PS
rules can be assigned all types of Bangla sentences into
RULE NO. RULE for COMPOUND SENTENCE
an SR. To represents a complex and a compound
01. COMS SS Conj SS sentence into an SR, we uses the decomposition
02. COMS SS Conj CS technique. An SR generally graphic object and this
03. COMS CS Conj CS
04.
representation cannot be deals with computer. The
COMS CS Conj SS
05. SS standard representation of an SR is a list that are one of
All Simple Sentence
06. CS All Complex Sentence the data structure that can be implemented and
07. Conj O, Ebong, Noile, Kintu, Fale, manipulated very easily within a computer. Further
Notuba, Sutorang,... modify and extending the CSG rules (with the addition
of sentences consisting of idioms and phrases, double
Now, we apply the rule of table 1,6, 7, 8 and 10 to word, change of voice, and narration), we can able to
parse the compound sentence of “Karim Danio lok represents the all kinds of Bangla sentences which are
sutorang Tini Jakat Diben” in the following fig. 3. used as an input of an MT engine to produced other
S equivalent sentences.
COMS
8. REFERENCES:
[1] A. Trujillo, Translation Engines: Techniques for
SS SS Machine Translation, Springer-verlag, London, (1992).
NP [2] D. Arnold, L. Balkan, S. Meijer, R. L. Humphreys,
VP
NP VP and L. Sadler, Machine Translation An Introductory
N Guide, Ncc, Blackwell Ltd., London, (1994).
NP VF [3] A. Humayan, Bakyattava, 2nd ed., The University of
Per N NP VF
Dhaka, 1994.
TPH imp- N [4] M. M. Murshed, Parsing of Bengali Natural
licit VR AUX
AP N Per Per Language Sentences, in Proc., International
Conference on Computer and Information Technology,
SPNH ASP T CON ICCIT’ 98, (Dhaka, Bangladesh), PP. 185-189, (1998).
Null I Fu NICON [5] M. R. Selim and M. Z. Ikbal, Syntax analysis of
TPH
doni Phrases and Different types of Sentences in Bangla, in
sutorang NICONTPH
Proc., International Conference on Computer and
Karim lok tini jakat di Null b en Information Technology, ICCIT’ 99”, (Sylhet,
Bangladesh), pp. 175-186, (1999).
Fig. 3: SR for the Compound sentence “Karim Doni [6] M. I. A. Khan, A. K. M. A. Hossian and R. C.
lok sutorang Tini Jakat Diben” (L¢lj de£ Debnath, “A Bangla Conversion Processor using
®m¡L p¤al¡w, ¢a¢e k¡L¡a ¢c-he). Natural Language Processing”, In Proc. International
Conference on Computer and Information Technology,
6. FUTURE IMPROVMENTS: ICCIT’02, (Dhaka, Bangladesh), PP. 262-266, (2002).
Our proposed CSG’s PS rules can be parsed the all [7] L. Mehedy, S. M. Arefin, and M. Kaykobad,
kinds of Bangla sentences in syntactic patterns. There Bangla Syntax Analysis: A Comprehensive Approach,
are lots of researches scopes on the filed of Bangla In Proc. International Conference on Computer and
NLP. The possible future development may mention in Information Technology, ICCIT’03, (Dhaka,
the following: Bangladesh), PP. 287-293, (2003).
• CSG’s for semantic analysis of Bangla sentences. [8] M. M. Hoque and M. M. Ali, A Parsing
• CSG’s for morphological analysis of Bangla Methodology for Bangla Natural Language Sentences,
sentences. In Proc. International Conference on Computer and
• Efficient algorithm for Bangla NL parser to solve Information Technology, ICCIT’03, (Dhaka,
the problem of ambiguity. Bangladesh), PP. 277-282, (2003).