Design and Implementation of a Language Assistant for English – Arabic Texts

Design and Implementation of A Language Assistant
For English – Arabic Texts
A. A. Alabi
Computer Science Department,
College of Technology,
The Polytechnic,
Ile-Ife, Osun, State, Nigeria.
newalabikompurer@gmail.com
O. A. Ayilara
Computer Science Department,
College of Natural Sciences,
Oduduwa University,
Ipetumodu, Ile-Ife, Nigeria
Talk2tobi2004@gmail.com
K. O. Jimoh
Computer Engineering Department
Federal Polytechnic, Ede,
Osun State, Nigeria
oyewumikudirat@gmail.com
ABSTRACT
People across the globe have access to materials such as journals, articles, adverts etc. via the internet. However
many of these resources come in diverse nature of languages. Although, English language seems most suitable to
most people, some readers do believe that working on materials in one’s native language is more enjoyable than in
other languages. Researches have shown that Arabic language has not been prominent in terms of online materials
and the few existing are most times ignored due to the peculiar nature of its various characters and constructs.
Hence, a proper study of its relationship with English language with a view to bringing people closer to its
understanding becomes necessary. The system scenarios were modeled and implemented using Unified Modeling
Language and Microsoft C# respectively in a way that the expected set of characters of the language of interest was
automatically formed with respect to a given input. The procedural steps were properly followed in the development
and running of the code using Context-Free Rule Based Technique with the availability of hardware required as
clearly described in the design. The system’s workability was tested with different source texts as inputs and in each
case the resulting outputs were very effective with respect to the translation process. The design here is expected to
serve as a tool for assisting beginners in these two languages and so, showcases a one-to-one form of
correspondence, hence, more rules and functions for ensuring a more robust are expected in future works.
Keywords: English language, Arabic language, Language Assistant, Machine Translation
1. INTRODUCTION
People across the globe have access to multiple forms of relevant materials such as texts, journals, articles, adverts
and websites via the internet. However not all these resources (especially the text-based) are in the languages most
users understand. Many of these resources come in diverse nature of languages ranging from English language (US
and/or British) which is a popular language, French language, Arabic language and a host of others. Many people
today believe that reading and working on materials in one’s native language seems more enjoyable than other forms
of languages thereby calling for the need to translate texts into languages of interest.
Translation as a process, is an activity comprising of the interpretation of the meaning of a text in one language and
the production of a new, equivalent text in another; thus, ensuring that both the source and the target texts
communicate the same message while taking into account a number constraints such as context, the grammar of the
source language, its writing conventions, idioms and the likes. The real challenge in this isn’t only about translating
the text(s) of a language into equivalent text(s) of another language but the generation of meanings instantaneously
and accurately. In real sense, No automated system is meant to replace human translation system but could help in
some areas where necessary. For instance, a properly implemented system could assist those seeking to understand
certain text materials which aren’t written in the language(s) they understand for important decisions making and
action taking. Designers in Machine Translation (MT) mostly care about the provision of the meaning of a particular
input text in a general form to the user when designing systems. These systems (i.e. Translators) come in different
forms; for example, one type functions to provide assistance and help to human translators during translation
processes, such as the help in linguistic rules and language grammar. Another type is the automatic system that
attempts to directly translate sentences and texts without any human intervention while the third is designed to
operate according to different laid down algorithms and translation policies with respect to the contents and contexts
of the translation. The key item of interest in this translation mechanism is “languages” which across the globe are
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 16, No. 5, May 2018
211 https://sites.google.com/site/ijcsis/
ISSN 1947-5500

made up constructs derived or created out of syntaxes of different classes. This is simply because of the differences
in grammars of the world existing languages. Some languages seem relatively simple and easy to learn to some
people while others look difficult to some extent. Meanwhile, information in some languages considered difficult are
rarely accessible on net by people; a condition that tends to portray these languages as irrelevant.
Although, some readers and information users have access to many texts, papers, articles and websites over the
internet, not all of these texts are in languages of interest, thereby making some people either ignore or neglect some
of these materials. Many even find some of these language texts difficult to pronounce and so, making it necessary
to get or find the exact translator which could either be a person very familiar with the language or software to do a
form of translation into target language before they could understand anything. A good example is Arabic language
which most people usually ignore any time they come across while sourcing for information due to its peculiar
nature of its various characters and constructs. This, we know could amount to serious waste of time and resources
and also loss of credible information in some cases. People do find it extremely uneasy to either source or make
reference to Arabic texts due to some reasons and so, making the language less relevant on most of our online
platforms. The salient issue here is that, some of the ignored contents could also be of relevance and could probably
be the only means of writing and reading especially while dealing with people with only knowledge of Arabic
language. The fact still remains that people (i.e. Non-Arabs) have a need to understand Arabic, and Arabs need to
understand other languages for handling meaningful materials over the internet, all by an automated system and at
no cost. Many non-Arabs today are faced with the need to understand Arabic for the existence of interactive
communication in matters of interest. This Arabic language seems very special in terms of its lexis and syntax thus
posing serious problems in the area of correspondence for people who do not understand the language. On the other
hand, some Arabs (mostly the beginners) sometimes find it difficult to pronounce English text the exact way English
speaking people do. What happens in most cases is that the Arabic way (tone) does come into play such whenever
people in this regard tend to pronounce letters in English language. Affected people in this regards do improvise by
employing a third party for assistance which in most cases tends to turn private information into a public type.
The truth is that no translator can implicitly take care of the syntactic relationships between English and Arabic
language because of this peculiar nature of Arabic language but the burden here could drastically be minimized by
creating a form of facilitator that could serve as a bridge between the languages in question. The strategy here is to
find a form of computer-based language (Rule-Based) tool that could assist in generating any form of English texts
into their respective Arabic - based form in a very fast and easy manner of response. This shall increase people with
little or no knowledge of the Arabic language in developing interest in the language thus creating a way out for those
who want to start working on materials written in such a language. It could also serve as a language assistant for
Arabs and others Arabic speakers (people whose sole communication tool is Arabic Language) thus improving their
pronunciation of English texts.
2. RELATED WORK
Machine Translation is about the translation of natural languages (Albat & Fritz, 2012); an area which has recently
witnessed series of developments. However, lots of works have been pondered in this area of Natural Language
Processing; a phenomenon also referred to as Machine Translation (MT). For instance, Nadkarni et al (2011)
provided a brief description of common machine learning approaches that are been used for diverse Natural
Language Processing (NPL) sub-problems. They also discussed how modern NPL architectures are designed with a
summary of the Apache Foundation Unstructured Information Management Architecture.
Carbonell et al (1981), developed a tool for addressing the several translation problems with examples of English-to-
Spanish and English-to-Russian translations. In this, the source was first analyzed and mapped into a language-free
conceptual representation with an inference mechanism for the application of contextual world knowledge about
items that were only implicit in the input text. The final step then involves the mapping of appropriate sections of the
language-free representation into the target language by the natural language generator. Another design was carried
out as regards to this but only for the extraction of molecular pathways from journal articles (Friedman et al, 2001).
Corresponding results from this demonstrated the value of the underlying techniques for the purpose of acquiring
valuable knowledge from biological journals. In another development, the use and efficiency of Statistical Machine
Translation (SMT) has been lauded then in a model by Shwenk et al (2008) comprising of the Open-Source Moses
decoder, the integration of a bilingual dictionary and a continuous space target language model for a general purpose
French/English statistical machine translation system. Other forms involved the introduction of a Unified Neural
Network Architecture and Learning Algorithm by Collobert et al (2011). This technique was discovered useful in
Vol. 16, No. 5, May 2018
ISSN 1947-5500

various Natural Language Processing tasks including part-of-speech tagging, chunking, named entity recognition
and the likes. This system learnt internal representations on the basis of vast amounts of mostly unlabeled training
data instead of exploiting man-made input features and was said to be a basis for building a freely available tagging
system with good performance and minimal computational requirements. William et al (2011), also introduced the
Multiscale Geometric Multi-Resolution Analysis (GMRA) for handling the investigation of detection, measurement
and modeling techniques to exploit low-dimensional intrinsic structures with a view to improve procedure such as
machine learning. The results obtained showed that the approximation error of the GMRA is completely
independent of the ambient dimension; thus establishing GMRA as a provably fast algorithm for dictionary learning
with approximation and guarantees. Folajimi and Isaac (2012) came up a tool for understanding Yoruba Language
using Statistical Machine Translation (SMT). The software employs a machine translation paradigm where
translations are generated on the basis of statistical models whose parameters are derived from the analysis of
bilingual text corpora. It was discovered that SMT seems to be a veritable tool for translating between English
Language and Yoruba Language because of the non-existence of parallel corpus between the two languages. In
their work (Kaufman et al, 2016), a technique known as Generic Notions of Complexity for two dominant
frameworks was introduced as an improvement on the stochastic multi-armed bandit model which proved
significantly positive. Also, Maggioni (2016), extended work on Multiscale Geometric Multi-Resolution Analysis
(GMRA) application-wise using a Non-Asymptotic Bounds and Robustness procedure (Maggioni, 2016).
However, several other research works were carried out and the trend still evolves on a daily basis. This is as a result
of the different forms of scope embedded in most available publications. While certain people argue on the need for
totally new designs, some prefer improving on the techniques demonstrated in existing publications. An example of
this was the study to determine the relationship between grammar efficacy and grammar performance among Arabic
learners on aspects such as Correction of grammar errors, Vocalization of words, and Construction of sentences
through questionnaire and it was observed that a moderate exist correlation between grammar efficacy and grammar
performance with efficacy of sentence construction as the most noticeable result (Mustapha, 2017).
3. SYSTEM DESIGN
3.1 Methodology
The work here work is designed to replace the involvement of humans in the translation of English (source
language) to Arabic (target language). In most cases, the system needs to emulate the thinking strategy as humans
do during translation. The various characters sets making up the alphabets in both English and Arabic languages
(fig. 1) are to be identified and analyzed. A database of the English alphabets is to be created and made to represent
the set of characters representing the source with those of Arabic language taken in to consideration. Also, the
syntaxes of both languages are to be made reference to with the aid of in-built tools and some other forms of
program segments when needed in terms of their relationship character-wise. The various scenarios in the design of
the system were modeled using Unified Modeling Language (UML). For instance, the modeling section involving
the Use-Case diagram (figure 2) has to with the demonstrations of the relationship among the various entities
making up the system in terms of functions and possibly dependencies. A logical way of representing this user-data
relationship is also shown in the class diagram (figure 3). Another aspect of this modeling includes the procedural
flow among the various class objects involved in the translation mechanism. The said scenario is diagrammatically
illustrated using the Activity diagram (figure 4).
The program is designed in such a way that the expected set of characters (word, sentence etc.) of the
language of interest (target language) is automatically formed with respect to a call (an instruction code) by the user
with English language as the source language and Arabic language as the target language. This was made possible
with the aid of Microsoft C#. So many programming languages were considered in the cause of designing the
system. A lot of factors were put into consideration which includes database access, data transmission via networks,
database security, database retrieval, multi user network access, data capture, etc. The choice “Microsoft C# (C
sharp)” programming language was made to achieve the above set of objectives. Microsoft C# programming
language is a user friendly platform that gives room for the design of an interface that can be modified
programmatically. The language has the advantage of easy development, flexibility, and it has the ability of
providing the developer/programmer with possible hints and it produces a beautiful graphical interface.
Vol. 16, No. 5, May 2018
ISSN 1947-5500

Send in a text in a language
(i.e. English)
Generate the Arabic equivalent
of the sent text
View the Arabic form of the
English text
Figure 2: Use-Case Diagram
Figure 1: English - Arabic Alphabets
Vol. 16, No. 5, May 2018
ISSN 1947-5500

TextType:
char/string
TextLength: int
display():
Translation
TranslatorType:
string
TranslatorLength:
int
gen():
Text(s)(English)
TextType:
string/char
TextLength: int
real():
Figure 3: Class Diagram
Figure 4: Activity diagram
User Reporting tool
Select: To send
text(s) in English
format
Select: To view
translated form of
texts
Generate Arabic
form of texts
Display result
on screen
Vol. 16, No. 5, May 2018
ISSN 1947-5500

3.2 Input Design and Output Specification
It is necessary to denote that data inputted in the computer for processing determines what the output
usually is. Screen designs are generally or basically made for data entry or capture. With English text as input, the
output section of the system in this research work is designed to automatically display its result in Arabic language
(based on the input supplied); i.e. generate response immediately after the input is received by the system. However,
the nature of the output strictly depends on that of the input as well as the correctness of the code implementation
with respect to the rules governing the writing and transformation of both character and words in the concerned
languages.
3.3 Program Design and Specification
The system’s structure, as shown in the flowchart (fig. 5) has the general form as a task divided into several sub-
tasks, which come together to give the solution to the problem with the translation process as the core stage (fig. 6).
The program is designed with the specification of having two languages modules namely English Language and
Arabic Language.
Figure 5: System Flowchart
Start
Translate
Source Language
Target Language
Translate
More?
Stop
Yes
No
Vol. 16, No. 5, May 2018
ISSN 1947-5500

Figure 6: Translation Process
Input Text (English)
English
I- Rules
ENGLISH DICTIONARY
ARABIC DICTIONARY
Output Text (Arabic)
Arabic
I-Rules
Vol. 16, No. 5, May 2018
ISSN 1947-5500

4. SYSTEM IMPLEMENTATION AND EVALUATION
The procedural steps were properly followed in the development and running of the code with the availability of
hardware required as clearly described in the design section. The resulting outputs were very effective with respect
to the translation process. The system’s workability was tested with three different source texts as inputs and in each
case, the result came out accurately. For example, the first three sets of English characters (alphabets a, b and c)
where the first point of call for translation .This is captured in figure 7; with its first part (upper part) showing the
inputs (in English) while the second part (lower part) displays results (in Arabic). Sufficed to that was the translation
of some phrases as shown in figures 8and 9. The various displays in the translation exercise also came up with the
processing speed in each of the instances.
Figure 7: Character Translation
Vol. 16, No. 5, May 2018
ISSN 1947-5500

Figure 8: Translation: phrase 1
Figure 9: Translation Phrase 2
Vol. 16, No. 5, May 2018
ISSN 1947-5500

5. FURTHER ISSUES AND
CONCLUSION
The parameters of interest in this study as demonstrated in the various results obtained show that the system
was simply meant to handle the text-to-text conversion between English language and Arabic language. This
showcases the form of representation one expects as regards to the Arabic form of a given English text. However,
some improvements are clearly necessary to increase both the quality of the study scenario with a view to addressing
the possible meaning of the Arabic form of any given English text and possibly make it speech based. This is
simply because of the draconian nature of spellings and pronunciations involved in Arabic language. To achieve
this, a robust system that takes care of both texts as well as the voice part is required. This shall take care of the
targeted goal in future designs as regards to this area of Natural Language Processing.
The developed system proved very efficient with respect to its outputs and the translation was very accurate
within reasonable time frame. This represents a useful tool for assisting those seeking redress in a form of assistance
in the lexical understanding of texts between English language and Arabic language. Understanding how characters
in English are written in their Arabic and vice-versa shall promote a form of familiarity between the two languages
thus assisting stakeholders in not just writing of letters but knowing the exact of any character thus enhancing how
certain letters are to be pronounced in their respective languages. The different tasks accomplished in the process of
designing and implementing the system were properly displayed in the various figures shown and their needed
explanations included accordingly. The design here showcases a translation form considered as “One-to-One
Correspondence” and so, some future works are still required as stated earlier for the incorporation of more rules as
well as functions for ensuring a more robust and more useful form of language translation.
In theory, one is expected to get increased in comprehending a particular language, therefore, getting use to
the Arabic form of English letters is expected to create a form of familiarity that could assist Arabs who sometimes
find it uneasy to either write or pronounce such texts. This could also be a tool for beginners in Arabic language.
6. REFERENCES
[1] Albat A. and Fritz T., (2012) . Systems and Methods for Automatically Estimating a Translation Tie.
US Patient 0185235.
[2] Carbonell J. G., Cullingford R.E. and Gershman A. V., (1981). Steps Towards Knowledge-Based Machine
Translation. IEEE Transactions on Pattern Analysis and Machine Intelligence. Vol. PAM 1-3, No. 4.
[3] Collobert R., Weston J., Bottou L., Karlen M., Kavukcuoglu K. and Kulksa P., (2011). Natural Language
Processing (Almost) From Scratch. Journal of Machine Learning Research. 12, 2493-2537
[4] Folajimi Y. O. and Isaac O., (2012). Using Statistical Machine Translation (SMT) as a Language
Translation Tool for Understanding Yoruba Language. EIE Second International Conference; Computer,
Energy Net., Robotics and Telecom. eleCon2012.
[5] Friedman C, Kra P., Yu H., Krauthmmer M. and Rzhetsky A., (2001). GENIES: A Natural-Language
Processing System for the Extraction of Molecular Pathways from Journal Articles. BIOINFORMATICS.
Oxford University Press. Vol. 17, Suppl.1.
[6] Kaufman E., Cappe O. and Garivier A., (2016). On The Complexity of Best-Arm Identification in Multi-
Armed Bandit Models. Journal of Machine Learning Research. 17, 1-42.
[7] Maggioni M., Minsker S. and Strawn N., (2016). Multiscale Dictionary Learning: Non-Asymptotic
Bounds and Robustness. Journal of Machine Learning Research. 17.
[8] Mustapha N. F. (2017). Grammar Efficacy and Grammar Performance: An Exploratory Study on
Arabic Learners. Mediterranean Journal of Social Sciences, Vol 8 No 4.
[9] Nadkarni P., Ohno-Machado L and Chapman W. (2011). Natural Language Processing: An Introduction
Vol. 16, No. 5, May 2018
ISSN 1947-5500

Journal of The American Medical Informatics Association.; 18:544-551. doi: 101136.
[10] Shwenk H. Fouet J. and Senellart J., (2008). First Step towards a general-purpose French-English Statistical
Machine Translation System Proceedings of the third Workshop on Statistical Machine Translation
pages 119-122 Columbus.
[11] William K., Chen G. and Maggioni M., (2011). Multiscale Geometric Method for Data Sets II:
Geometric Analysis. Cornell University Library, New York, USA
.
Vol. 16, No. 5, May 2018
ISSN 1947-5500

Design and Implementation of a Language Assistant for English – Arabic Texts

More Related Content

What's hot (19)

Similar to Design and Implementation of a Language Assistant for English – Arabic Texts (20)

Recently uploaded (20)

Design and Implementation of a Language Assistant for English – Arabic Texts