Simplification Complex Sentences in Indonesia

Language using Rule-Based Reasoning

Novi Yusliani, S.Kom, M.T M. Fachrurrozi, M.T
Rifka Widyastuti
Informatics Engineering, Informatics Engineering,
Informatics Engineering, Faculty of Computer Science, Faculty of Computer Science,
Faculty of Computer Science Sriwijaya University, Sriwijaya University,
Sriwijaya University 30163, Palembang, Indonesia, 30163, Palembang, Indonesia,
30163, Palembang, Indonesia [email protected] [email protected]
[email protected]

Abstract-Complex sentence consist of two or more single

sentence and one or more conjunctions. Complex sentences have II. RELATED WORK
many meaning, so that is quite difficult to be simplified because
the results of simplification complex sentence does not match the Some research on text simplification done. In this study,
original sentence. To solve this problem, this research using Rule- will try simplifying the text in Indonesian to simplify complex
Based Reasoning method and Surface Expression Rules. Because
sentences. Previous research [2] studied the question and
Rule-Based Reasoning method can be adapted to the rules of
surface expression that can look for patterns of complex answer system development for Non-factiod question for
sentences, so thus simplifying precise and accurate results. The Indonesian. Non-factoid question is a question-answer
results of researchis the Rule-Based Reasoning methods used in questions that generally require a fairly lengthy explanation as
accordance with the accuracy of 93.8% and an assessment of a definition of a term method for doing something or the cause
questionnaires obtained values 76-100. of an incident. This indicates that non-factoid question is more
complex, because it requires the classification of questions to
Keyword–complex sentences, simplification, rule-based get the expected answer. Wear pattern classification question
reasoning, surface expression, pattern complex sentences. classification based on pattern Surface Expression. Types of
questions can be included to predict the type of response
I. INTRODUCTION generated.

Complex sentence consist two or more single sentence. Future studies related on the simplification of text in
Every complex sentences has different conjunctions and some French[3] entitled Acquisition of Syntactic Simplification
are not wearing but use a comma.Complex sentences can be Rules for French. His research describes the simplification of
determined by looking at the use of conjunctions or syntax (syntactic simplification) is a data-driven approach that
punctuation used. People often difficult to understand what implements two methods. The first method is manual corpus
you want delivered in complex sentences. Using conjunctive analysis that aims to identify the word you would be
or sign can make a complex sentence into a different meaning simplified, then the second method is a semi-automatic that
and content of information. automatically identifies the simplified function informs
sentence simplification rules. The results of his research to
The problem is when simplification the sentence which overcome obstacles no longer need the data as parallel
divides complex sentences into a single sentence in which a resources and increase flexibility. In particular, syntactic
single sentence that has been simplified to alter the meaning simplification can explore domains on user-generated content
and content of information. Simplifying text (text as pre-editing for statistical machine translation.
simplification) is one of the fields of natural language
processing (NLP) which rewrite a sentence to reduce syntactic Research simplification of text above, this study will
complexity (syntactic complexity) and lexical complexity examine the simplification of complex sentences in Indonesian
(lexical complexity) without changing or eliminating the using Rule-Based Reasoning method based on rules and their
meaning of the sentence and fill in the information surface expression tagger post on the introduction of a class of
sentence[1]. In particular, the simplification of the sentence words in a sentence. In a complex sentence simplification is
has been developed in various countries. Development is done expected to facilitate the delivery of simplifying complex
by a variety of methods and rules. However, in Indonesia the sentences and the meaning of the sentence.
development of sentence simplification is not much, especially
simplification complex sentences III. PERFORMANCE

RIFKA WIDYASTUTI | INFORMATIC ENGINEERING | DEPARTEMENT OF COMPUTER SCIENCE | SRIWIJAYA UNIVERSITY 2014 1
 Natural Language Processing is the area of research and
application that explores how computers can be used to
understand and manipulate the natural language text or speech c. Rule-Based Reasoning
to do something useful [4]. Rule-Based Reasoning is a decision support system which
also has a knowledge base. In this method, the settlement of
a. Preprocessing the problem based on an artificial intelligence approach using
Preprocessing is process of managing the data before the problem-solving techniques based on the rules contained in
processing data [5]. Preprocessing consist of case folding and the knowledge base [7].
tokenizing. Case folding is process of changing all the letters
in a document / sentence to lowercase. Only the letters 'a' [2] uses the rules of the component surface expression
through 'z' received [6] while the characters other than letters answer finder. Surface expression is the surface expression of
received are considered delimiter. Examples delimiter can be the sentence or the pattern used in the sentence. regulation of
seen in Table I. surface expression in the study can be found in appendix
Table I. Daftar Delimiter
Daftar Delimiter d. Complex Sentences
0 5 [ % ` . ? | ) ≥ Complex sentence is a merger of two or more single
1 6 ] ^ ~ , : ! - ∞ sentences using conjunctions. Examples of complex sentence
simplification can be seen in Figure II.
2 7 { & \\ / ; @ _ π
1. Tini berbelanja sayuran.
3 8 } * £ < ‘ # + ±
2. Tini memasak sayuran
4 9 \ ( € > ‘’ $ = ɸ 3. Tini berbelanja sayuran dan memasaknya
Tokenizing is process of identification the smallest units
(tokens) of a sentence structure (Triawati, 2009). Breaking
sentences into single words performed by scanning a sentence
using white space separators such as spaces, tabs, and newline.
Schematic of the process of folding and tokenizing case can be
seen in Table II.
Tabel II. Preprocessing Sentences Scheme
Preprocessing Sentences
Sentences : Ibu Pergi Ke Pasar
Case folding : ibu pergi ke pasar
Tokenizing : “ibu” “pergi” “ke” “pasar” Figure II. Chart of Complex Sentences (3)

b. Part of Speech
Part of Speech (POS) tagging is a process that is done to IV. EXPREMENTAL
determine the type of a word in the text. A simple form of this
process is the identification of words as adjectives, adverbial, Simplification complex sentence is not as easy as one
interjection, conjunction, noun, numerial, prepositions, might imagine, some people find it difficult to simplification
pronouns, verbs, etc. [5]. The process of determining the type complex sentences, especially during the learning process in
of words in a sentence can be seen in Figure I. schools. Therefore, need a applications to help the learning
process and make it more attractive. In this research,
simplification complex sentence process starts from
preprocessing which case folding and tokenizing. The results
of the research complex sentence preprocessing can be seen in
Figure III.

Figure I.Process of Identification Type of Word[5]

RIFKA WIDYASTUTI | INFORMATIC ENGINEERING | DEPARTEMENT OF COMPUTER SCIENCE | SRIWIJAYA UNIVERSITY 2014 2
 Preprocessing Kalimat Majemuk

Contoh Kalimat :
“Tini Berbelanja Sayuran dan Ibu Memasaknya”
Hasil Proses Case Folding:
“tini berbelanja sayuran dan ibu memasaknya”
Hasil Proses Tokenizing:

Figure III. Preprocessing of Simplification Complex

Sentences
After we get a results in the form of preprocessing tokens Figure V. Simplification Complex Sentences using Rule-
(word class), tokenizing on this journal wear NLP_ITB Based Reasoning
package where the package is Indonesian word dictionary.
Token can make easy to process of simplification of complex This research used 60 samples were taken from the
sentences. Further, simplification complex sentences of the complex sentence http://bse.kemdikbud.go.id/. Based on the
process that is using the Rule-Based Reasoning with rules experiment results of the software by entering the 60 samples
Surface Expression. Process simplification of complex of complex sentences, obtained 4 sample of complex
sentences can be seen in Figure IV and Figure V. sentences that can not be simplified accurately. Experiment
result on this research using 60 sample of complex sentences
can be seen on appendix B. This is due to several factors that
the sentences can not simplified accurately, there are:
1. The token tagging errors occurred in the compound
sentence "his face is thin and pale". The error occurs on
the token marking words that should generate "n, v, c, n,
v", but in POSTagging generated token is "v, n, c, v, n".
The error occurs from the package NLP_ITB.
2. Sentence Compound "It's fun playing ball so spaced out"
can not be reduced to a single sentence and a two
conjunctions appropriately. The fault lies in the pattern
of compound sentence has no subject. So that can not be
simplified complex sentences correctly.
3. Compound sentence "the birthday party would not be
more festive if you come to attend". The error occurs on
the token marking the word "if" is a word that should be
connecting, but the token POSTagging recognizable
words with a noun. The error occurs from the package
NLP_ITB.
4. Sentence compound "People panic because there was an
earthquake" in POSTagging identified by the token of the
Figure IV. Surface Expression Rules in Simplification word "n, n, c, v, v", but should have obtained a token
Complex Sentences word is "n, v, c, v, n". The error occurs from the package
NLP_ITB.

Based on the experimental results of 60 samples of

complex sentences obtained 4 sample of complex sentences
that can not be simplified appropriately. Therefore, the
percentage of success of software obtained for 93.3% of the
software is built. Word tokens not generated as expected.
However, the word is sometimes different tokens if put in a
different sentence. Therefore, there are some words that can
not be replaced tokens he said. For example the word "if" is
not a word said base so that the resulting token is different.

CONCLUSION

RIFKA WIDYASTUTI | INFORMATIC ENGINEERING | DEPARTEMENT OF COMPUTER SCIENCE | SRIWIJAYA UNIVERSITY 2014 3
 ICITA 2005. Third International Conference on,
The conclusion that can be take from this study are 2005, pp. 52-57 vol.1.
1. Methods of Rule-Based Reasoning can be used to [8] A. Chaer, Sintaksis Bahasa Indonesia: Pendekatan
simplification sentence and can be applied to the case of Proses: Rineka Cipta, 2009.
complex sentences which basically has a single and a [9] B. S. R. Chandrasekar "Automatic Induction of Rules
two-sentence conjunctions. for Text Simplification," Institute for Research in
2. Rules Surface Expression can be used to describe the Cognitive Science, 1996.
word before and after the conjunctive. So that the [10] P. M. Nugues, An Introduction to Language
compound sentence can be simplified by appropriate Processing with Perl and Prolog. Germany:
because it does not change the meaning and information Springer-Verlag Berlin Heidelberg, 2006.
after simplifying complex sentences. [11] W. Duch, "Rule-Based Methods," Department of
3. Sentence of 60 samples were available, the percentage of Informatics, Nicolaus Copernicus University, Poland,
complex sentences simplification results in Indonesian 2010.
using Rule-Based Reasoning on software as much as [12] S. D. Hasan Alwi, Hans Lapoliwa, Anton M.
93.3% of the 60 samples in which the existing manjemuk Moelino, "Tata Bahasa Baku Bahasa Indonesia," vol.
sentences, compound sentences there are four samples EdisiKetiga, ed. Jakarta: Pusat Bahasa
that can not be simplified appropriate. This is because an danBalaiPustaka, 2003, p. 475.
error occurred while defining the token word and sample [13] A. O. Hatem, N. Shaker, "Morphological Analysis
sentences compound does not have a compound sentence for Rule-Based Machine Translation," in Semantic
patterns that have been defined. Technology and Information Retrieval (STAIR), 2011
4. Results simplification of complex sentences are split into International Conference on, 2011, pp. 260-263.
two single sentences and the conjunctive word is [14] R. Ismoyo, Nasarius Sudaryono, Bahasa Indonesia
determined by the class defined. Just a word class of untuk Sekolah Dasar/MI Kelas 6. Jakarta: Pusat
each word in a sentence compound sentence is used to
simplify the process of using Rule-Based Reasoning.
Therefore, the software can simplify complex sentences
are not appropriate when an error in the definition of the
word class by NLP_ITB package.

REFERENCES

[1] A. Siddharthan, "An Architecture for A Text

Simplification System," in Language Engineering
Conference, 2002. Proceedings, 2002, pp. 64-71.
[2] N. Yusliani, "Sistem Tanya-Jawab Bahasa Indonesia
untuk 'Non-Factoid Question'," Master, Program
Studi Informatika, Institut Teknologi Bandung,
Bandung, 2010.
[3] V. Seretan, "Acquisition of Syntactic Simplification
Rules for French," 2012.
[4] G. G. Chowdhury, "Natural Language Processing,"
Annual Review of Information Science and
Technology (ARIST), vol. 37, 2003.
[5] R. A. Sukamto, "Penguraian Bahasa Indonesia
dengan Menggunakan Penguraian Collins," Magister,
Program Magister Informatika, Institut Teknologi
Bandung, Bandung, 2009.
[6] C. Triawati, "Metode Pembobotan Statistical Concept
Based untuk Klastering dan Kategorisasi,"
Informatika, ITTELKOM, Bandung, 2009.
[7] H. M. L. S. Jani, Peck, "Applying Machine Learning
Using Case-Based Reasoning (CBR) and Rule-Based
Reasoning (RBR) Approaches to Object-Oriented
Application Framework Documentation," in
Information Technology and Applications, 2005.

RIFKA WIDYASTUTI | INFORMATIC ENGINEERING | DEPARTEMENT OF COMPUTER SCIENCE | SRIWIJAYA UNIVERSITY 2014 4
 APPENDIX A
Surface Expression Rules in Simplification Complex Sentences

No. Conjunction Before Conjuntion Word After Conjunction Word Information

1. Dan Objek (noun) Predikat (verb) Conjunction on middle of complex sentences
2. Dan Subjek (noun) Subjek (noun) Conjunction on middle of complex sentences
3. Dan Objek (noun) Objek (noun) Conjunction on middle of complex sentences
4. Dan Predikat (verb) Predikat (verb) Conjunction on middle of complex sentences
5. Tetapi Predikat (kata kerja) Predikat (verb) Conjunction on middle of complex sentences
6. Tetapi Objek (noun) Pelengkap (noun) Conjunction on middle of complex sentences
7. Tetapi Keterangan (noun) Predikat (verb) Conjunction on middle of complex sentences
8. Tetapi Objek (noun) Predikat (verb) Conjunction on middle of complex sentences
9. Tetapi Predikat (verb) Subjek (noun) Conjunction on middle of complex sentences
10. Tetapi Keterangan (noun) Keterangan (noun) Conjunction on middle of complex sentences
11. Jika - Subjek (noun) Kata penghubung di awal kalimat
12. Melainkan Predikat (kata kerja) Predikat (verb) Conjunction on middle of complex sentences
13. Melainkan Keterangan (noun) Predikat (verb) Conjunction on middle of complex sentences
14. Melainkan Objek (noun) Objek (noun) Conjunction on middle of complex sentences
15. Bahkan Predikat (verb) Predikat (verb) Conjunction on middle of complex sentences
16. Atau Predikat (verb) Predikat (verb) Conjunction on middle of complex sentences
17. Atau Keterangan (noun) Predikat (verb) Conjunction on middle of complex sentences
18. Atau Objek (noun) Objek (noun) Conjunction on middle of complex sentences
19. Biarpun - Subjek (noun) Kata penghubung di awal kalimat
20. Jangankan - Predikat (verb) Kata penghubung di awal kalimat
21. Sedangkan Objek (noun) Subjek (noun) Conjunction on middle of complex sentences
22. Sedangkan Predikat (verb) Subjek (noun) Conjunction on middle of complex sentences
23. Karena Predikat (verb) Predikat (verb) Conjunction on middle of complex sentences
24. Karena Predikat (verb) Subjek (noun) Conjunction on middle of complex sentences
25. Karena Keterangan (noun) Predikat (verb) Conjunction on middle of complex sentences
26. Karena - Predikat (verb) Conjunction in Front of Sentences
27. Daripada - Subjek (noun) Conjunction in Front of Sentences
28. Maka Predikat (verb) Subjek (noun) Conjunction on middle of complex sentences
29. Sehingga Predikat (verb) Subjek (noun) Conjunction on middle of complex sentences
30. Sehingga Predikat (verb) Predikat (verb) Conjunction on middle of complex sentences
31. Sehingga Keterangan (noun) Predikat (verb) Conjunction on middle of complex sentences
32. Sehingga Objek (noun) Predikat (verb) Conjunction on middle of complex sentences
33. Saat - Subjek (noun) Conjunction in Front of Sentences

RIFKA WIDYASTUTI | INFORMATIC ENGINEERING | DEPARTEMENT OF COMPUTER SCIENCE | SRIWIJAYA UNIVERSITY 2014 5
34. Kemudian Keterangan (noun) Predikat (verb) Conjunction on middle of complex sentences
35. Kemudian Objek (noun) Predikat (verb) Conjunction on middle of complex sentences
36. Meskipun - Subjek (noun) Conjunction in Front of Sentences
37. Meskipun - Predikat (verb) Conjunction in Front of Sentences
38. Lalu Objek (noun) Predikat (verb) Conjunction on middle of complex sentences
39. Ketika - Subjek (noun) Conjunction in Front of Sentences
40. Walaupun - Subjek (noun) Conjunction in Front of Sentences
41. Walaupun - Objek (noun) Conjunction in Front of Sentences
42. Agar Objek (noun) Predikat (verb) Conjunction in Front of Sentences

RIFKA WIDYASTUTI | INFORMATIC ENGINEERING | DEPARTEMENT OF COMPUTER SCIENCE | SRIWIJAYA UNIVERSITY 2014 6
 APPENDIX B
EXAMPLE OF EXPERIMENT SIMPLIFICATION COMPLEX SENTENCES

No. Complex Sentences Output Result Conclusion

1. Ibu dan ayah pergi Kalimat 1 : ibu pergi ke pasar Kalimat 1 : ibu pergi ke pasar TRUE
ke pasar Kalimat 2 : ayah pergi ke pasar Kalimat 2 : ayah pergi ke pasar
Konjungsi : dan Konjungsi : dan
2. Kegiatan pramuka Kalimat 1 : kegiatan pramuka bukan Kalimat 1 : kegiatan pramuka bukan TRUE
bukan hanya hanya berkemah hanya berkemah
berkemah melainkan Kalimat 2 : kegiatan ada yang lain Kalimat 2 : kegiatan ada yang lain
ada yang lain Konjungsi : melainkan Konjungsi : melainkan
3. Ita dan kakaknya Kalimat 1 : ita sedang berada di Kalimat 1 : ita sedang berada di TRUE
sedang berada di jakarta jakarta
jakarta Kalimat 2 : kakaknya sedang berada Kalimat 2 : kakaknya sedang berada di
di jakarta jakarta
Konjungsi : dan Konjungsi : dan
4. Meskipun ia telah Kalimat 1 : ia telah lelah Kalimat 1 : ia telah lelah TRUE
lelah, ia tetap Kalimat 2 : ia tetap menyelesaikan Kalimat 2 : ia tetap menyelesaikan
menyelesaikan tugasnya tugasnya
tugasnya Konjungsi : meskipun Konjungsi : meskipun
5. Amir tidak pergi ke Kalimat 1 : amir tidak pergi ke Kalimat 1 : amir tidak pergi ke stasiun TRUE
stasiun melainkan stasiun Kalimat 2 : amir pergi ke terminal
pergi ke terminal Kalimat 2 : amir pergi ke terminal Konjungsi : melainkan
Konjungsi : melainkan
6. Ketika ayah datang, Kalimat 1 : ayah datang Kalimat 1 : ayah datang TRUE
aku sedang mandi Kalimat 2 : aku sedang mandi Kalimat 2 : aku sedang mandi
Konjungsi : ketika Konjungsi : ketika
7. Berbicara dan Kalimat 1 : berbicara merupakan Kalimat 1 : berbicara merupakan TRUE
menulis merupakan keterampilan kreatif keterampilan kreatif
keterampilan kreatif Kalimat 2 : menulis merupakan Kalimat 2 : menulis merupakan
keterampilan kreatif keterampilan kreatif
Konjungsi : dan Konjungsi : dan
8. Aziz bukan anak Kalimat 1 : aziz bukan anak pak Kalimat 1 : aziz bukan anak pak jamal TRUE
pak jamal melainkan jamal Kalimat 2 : aziz keponakannya yang
keponakannya yang Kalimat 2 : aziz keponakannya yang selama ini tinggal
selama ini tinggal selama ini tinggal Konjungsi : melainkan
Konjungsi : melainkan
9. Saya pergi ke Kalimat 1 : saya pergi ke sekolah Kalimat 1 : saya pergi ke sekolah TRUE
sekolah kemudian Kalimat 2 : saya pergi ke mesjid Kalimat 2 : saya pergi ke mesjid dekat
pergi ke mesjid dekat alun-alun alun-alun
dekat alun-alun Konjungsi : kemudian Konjungsi : kemudian
10 Walaupun veri Kalimat 1 : veri sedang sedih Kalimat 1 : veri sedang sedih TRUE
sedang sedih, ia bisa Kalimat 2 : ia bisa tersenyum Kalimat 2 : ia bisa tersenyum
tersenyum Konjungsi : walaupun Konjungsi : walaupun

RIFKA WIDYASTUTI | INFORMATIC ENGINEERING | DEPARTEMENT OF COMPUTER SCIENCE | SRIWIJAYA UNIVERSITY 2014 7