See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327397470

Arabic Chatbots: A Survey

Article  in  International Journal of Advanced Computer Science and Applications · January 2018

DOI: 10.14569/IJACSA.2018.090867

CITATIONS READS
4 1,395

3 authors:

Sarah O Al-Humoud Al Asma

Imam Muhammad bin Saud Islamic University Tanjungpura University
25 PUBLICATIONS   160 CITATIONS    8 PUBLICATIONS   4 CITATIONS   

SEE PROFILE SEE PROFILE

Wafa Aldamegh
Imam Muhammad bin Saud Islamic University
1 PUBLICATION   4 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

phytochemical qualitative test View project

Arabic Sentiment Analysis tool View project

All content following this page was uploaded by Wafa Aldamegh on 22 September 2018.

The user has requested enhancement of the downloaded file.

 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 9, No. 8, 2018

Arabic Chatbots: A Survey

Sarah AlHumoud, Asma Al Wazrah, Wafa Aldamegh
College of Computer and Information Science
Al-Imam Mohammad Ibn Saud Islamic University (IMSIU)
Riyadh, Saudi Arabia

Abstract—A Chatbot is a programmed entity that handles formal written and spoken forms in mediums such as news,
human-like conversations between an artificial agent and education, and literature [4]. The third category is Dialectal
humans. This conversation has attracted the attention of Arabic (DA), which is used daily in spoken and written
researchers who are interested in the interaction between personal communication and in informal settings, where each
humans and machines to make the conversation more rational country and region has its own dialect [2].
and hence pass the Turing test. The available research done in
the field of Arabic chatbots is comparably scarce. This paper The rest of the paper is organized as follows: Section II III
presents a review of the published Arabic chatbots studies to presents the background. Section III presents the survey
identify the gap of knowledge and to highlight the areas that methodology. Section IV discusses the Arabic chatbots
needs more study and research. This study concluded the rarity researches. Section V presents the conclusion.
of available research on Arabic chatbots and that all available
works are retrieval based. II. BACKGROUND
There are generally three components of Chabot [5]: the
Keywords—Artificial intelligence; Arabic chatbot;
interface, which interacts with user's input and output, the
conversational agent; ArabChat; human-machine interaction;
Knowledge Base or brain, which include the content of the
utterance
conversation and keep truck of the domain, and the
I. INTRODUCTION Conversation Engine, which manages the semantic context of
the conversation.
Artificial Intelligence (AI) is focused on the learning
processes. The idea of using a human language to There are two types of dataset models, which represent the
communicate with computers is holding merit to AI. A chatbot knowledge source type in chatbots: the retrieval-based model
or a chat-agent is an intelligent conversation agent which and the generative-based model. In the retrieval-based model, a
interacts with human users via natural language and emulates chatbot uses a pool of predefined responses and employs a type
human conversation. This area has attracted more interest from of heuristics to select the proper response to the input, but it
both research and industry fields in the past few years [1]. The may not be applicable when there is no existing predefined
first chatbot was developed in Massachusetts Institute of response. In the generative-based model, a chatbot uses a set of
Technology (MIT), where Weizenbaum implemented the the techniques for generating new responses and could utilize
ELIZA chatbot to emulate a psycho-therapist in 1966 [2]. predefined responses as well using deep learning and neural
network (NN). In the following, we will discuss the literature
Nowadays, a variety of chatbots are available online to based on the different techniques used, the length of the
serve in different domains ranging from customer service and conversation, the domain of conversation, and the dataset
information acquisition to entertainment where primarily users model.
react with these applications to make a small conversation.
They extend from unpretentious systems which extract answers In the retrieval-based model, there are common techniques
from datasets when they match specific keywords to more to build the conversational agent, using pattern matching,
advanced ones which utilize Natural Language Processing Artificial Intelligence Markup Language (AIML), Ontologies,
(NLP) techniques. A chatbot could be programmed to serve Parsing, Markov Chain Model, and ChatScript. While in the
almost any human language. Although the research on English generative-based model, the different techniques are neural
chatbots is diffuses heavily, there is a Scarcity in the Arabic network and deep learning techniques. Seq2seq based on
chatbots due to difficulties in the Arabic language. In this neural network will be introduced.
paper, we present a survey on Arabic chatbots. The number of
Pattern matching is used mainly in the question/answer
online Arabic users have increased which motivated to build
chatbots, where the system matches the input with a predefined
Arabic chatbots.
structure to create a response. AIML is widely used in chatbot
Processing Arabic language texts have a lot of challenges design, it is a language derived of XML. It represents the
[3] such as rich morphology, the high degree of ambiguity, knowledge as objects, consisting of topics and categories. The
orthographic variations, and the existence of multiple dialects. AIML pattern consists of words having letters and numerals
Moreover, written Arabic text can be classified into three but no special characters or spaces [6]. Many chatbot
categories. First, Classical Arabic (CAL) or Quranic Arabic in applications have emerged in English. A.L.I.C.E. [7] is a
the Holy Quran. Second, the Modern Standard Arabic (MSA) retrieval-based model and it is using advanced pattern
that is the official language in the Arab world and used in a matching and AIML approach. It is the first AIML-based

535 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 9, No. 8, 2018

chatbot and won the Loebner Prize in 2000, 2001 and 2004. aspects in this survey and will be discussed. The conversation
A.L.I.C.E is a supervised learning and it is based on categories length is classified into short and long conversation. A short
containing a pattern, and a template for the response. Category conversation is a single response produced for a single input
patterns are matched to find the most appropriate response to a such as a question/ answer conversation. Where a long
user input. AIML tags provided for consideration of context, conversation indicates that a large amount of information is
conditional branching to produce new responses. Some Arabic exchanged during the conversation lifetime, and this
studies are applied to A.L.I.C.E. information is tracked and may be present in the output. On the
conversation domain, chatbots are classified into two types,
Ontology [8] is a set of interconnected hierarchy classes. closed and open. The closed domain is designed to serve a
The knowledge base can be described as a graph that contains specific purpose, where the knowledge that is required to
classes, each class describes the concepts and the properties. generate a suitable response to an input is limited. While the
The classes that have a logical relationship are also connected, open domain is like human‟s conversation, the domain may
and use these relationships to imply new statements change with the time, supporting more than one conversation
(reasoning). Examples of ontologies are OpenCyc and domain.
Wordnet.
Textual Parsing [9] is a method which converts the text into III. METHODOLOGY
a set of words (lexical parsing) to determine its grammatical The research survey methodology consists of scanning
structure. After the tree is built from these words, the lexical different literature databases. A similar methodology is used in
structure can be then checked if it forms the rule of the the literature review here [11]. The literature collection was
language (syntactical parsing). The latter parsers are getting done in highly cited computer science libraries like: IEEE,
more complex using natural language processing. ACM, Springer, Science Direct and Google Scholar. The
search was done using ten keywords coupled with the
The Markov Chain Model [8] depend on the probability of
keyword „Arabic‟. Those keywords are „chatbot‟, „chatterbot‟,
occurrence of a word or letter in the input text, this method
„ArabChat‟, „chat agent‟, „interactive agent‟, „conversational
helps in building responses that are probabilistically more
agent‟, „conversational robot‟, „artificial conversational‟,
suitable and hence more correct. For example, if an input text
„dialogue‟, and „utterance‟. The result consisted of 184 papers
is "xyyyzxyzyyyzy", then the Markov model of order 0
as shown in Fig. 1. Those papers collected from 2004 to 2017
predicts that letter "x" occurs with a probability 2/13. The
and evaluated according to the title and the abstract of the
Markov model of order 1 predicts the fixed probability for
paper, eliminating the papers that do not present an
every letter depends on the previous letter.
implementation of an Arabic chatbot. After evaluation and
ChatScript [8], [9] aims to be easier to maintain than AIML elimination, we found that there are fourteen papers present
by focusing on better syntax, it fixes the zero-word matching Arabic chatbot application, which was from the IEEE and
problems. The Chatscript first finds the best topic that matches Springer libraries and Google Scholar.
the user query string and executes a rule in that topic. Rather
than using separated categories for each word as in AIML, IV. ARABIC CHATBOT RELATED WORK
Chatscript uses „concepts‟ to merge similar words with Although there are available developed Arabic chatbots
meanings or parts of speech. Suzette (written in ChatScript) applications, the research available on Arabic chatbots is
won the 2010 Loebner Prize. limited. Some examples on the former are services as Al-Haj
On the other hand, there is a common technique to build a Bot, Rammas, Msa3ed, Theyabi, El-Kahwagy, and others
generative-based model chatbot using a Recurrent Neural provide an Arabic chatbot application developed for a
Network (RNN). Seq2seq model [10] is an encoder-decoder commercial purpose. Also, there are platforms that provide
model that uses RNN and it is primarily used for translating developers with coding facility and aid such as Watson by
from one language to some other language, but in the context IBM, Messenger Bot by Facebook, Telegram Bot, PandoraBot.
of chatbots, the input is translated to a response. The seq2seq These platforms and applications are excluded from this review
model is composed of two main RNNs, an encoder RNN for the lack of published research on them, making it difficult
which takes the input sequence and encapsulates the to analyze and compare fairly. The purpose of this survey is to
information into a fixed representation one cell at time, and a highlight state of the art Arabic chatbot research.
decoder RNN which take that representation, and generates a The fourteen collected papers present twelve different
variable length text that best responds to it also one cell at time. Arabic chatbot applications, that classified and evaluated in a
Seq2seq encodes only the important information in the manner similar to what is applied here [12], [8], [9]. Based on
sequence and convert a sequence of symbols into a fixed size the data type a chatbot processes, it is classified into two
feature vector. The cell used in RNN is long short-term categories, text and speech conversation chatbot. The
memory (LSTM), It allows the cells to remember what classification relies on the chatbot input and output interaction.
information needs to be remembered or updated from the In each category, chatbots are classified based on the
previous cells [10]. On scanning the literature, no studies were implementation technique into two subcategories namely,
found to apply a neural network in Arabic chatbot design. pattern matching and AIML approach. In addition to that, we
In addition to the technique used and the type of data being will discuss chatbot aspects such as the length of the
processed, the length of the conversation and the domain of conversation in terms of interaction duration, the domain of
conversation, as well as the dataset model are considered conversation as topics domain that chatbot can interact with,
and the dataset model of the chatbot.

536 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 9, No. 8, 2018

Fig. 1. Total Results from Literature Databases For Search Keywords.

A. Text Conversational Chatbot English and Arabic. The Java program reads the Quran text
from a corpus and converts it to the AIML format to be used by
In this category, the interaction with the chatbot ALICE chatbot. The domain of the conversation is closed
conversation is through textual input and output. The chatbot in based on the content of the Quran in both the English and
this category is classified based on the implementation Arabic languages. Also, this is a retrieval-based model as the
technique into two subcategories, those are AIML and pattern dataset is limited by the content of the Quran. The challenge in
match approach. this work is to show how ALICE chatbot adapted to learn from
1) AIML approach: Among the earliest research studies non-conversational text.
on Arabic chatbot application in the collected related work is The Arabic Web Question Answering (QA) chatbot [2],
Quran chatbot [13] by Shawar and Atwell. It is a chatbot on [15] is a web interface chatbot based on an Arabic QA corpus,
the Quran Islamic holy book. The Quran contains 6236 verses that was built from five different web pages with 412 Arabic
or „Ayahs‟ and 114 „Surahs‟ which is a set of verses. The question and answer. Those web pages‟ cover topics such as
format of the user inputs are Arabic words with „Tashkil‟ or motherhood and pregnancy, dental care, fasting and health,
diacritics, that is used as phonetic guides. The chatbot replies blood disease such as cholesterol and diabetes, and blood
by finding the „Ayahs‟ from the Quran that contain the user‟s charity. The chatbot supports more than one closed domain,
and thus regarded as a closed conversation domain. The
input. The nature of Quran text is non-conversational, a Java
chatbot conversation is short, where the user inputs a textual
program is developed to adopt a learning process. The question in MSA about one of the supported domains and, the
learning process was based on the most significant word of the chatbot responses with the answer without using sophisticated
„Ayha‟ that represents the category in the AIML file and NLP. Also, a Java program was developed to convert the text
template is the Ayah. Arabic AIML file is generated by the corpus to create two AIML files atomic and default. The
Java program. The conversation length is short since the atomic file contains the questions and answers that appear in
chatbot responses with a single response to a single user input. the corpus. The default file is used to guarantee that the user
The domain of the conversation is limited by the content of question is mapped to the appropriate question stored in the
the Quran. From that, the sources of the chatbot dataset are knowledge base. Moreover, the file is built using the first word
retrieval-based. The interaction and response of the chatbot and the most significant word approach. The first word acts as
a classifier to the question and the most significant word is the
are limited by the pool of the most significant word of the
least frequent in the question.
„Ayhas‟ that are extracted by the java program.
The latter is done by building questions‟ frequency list after
The Quran14-114 chatbot by Shawar and Atwell [6] is a applying a tokenization process to the question. The generated
version of the ALICE chatbot [14] added to the Java program list contains the question‟s words along with their frequencies.
to interact as the Quran chatbot. The conversation in the Then, the approach extracts the two most significant words in
chatbot is short. The user inputs a question or a statement in the list, those are the two least frequent words, used as
English, and the chatbot responses with one or more keywords to map the question to an answer. The purpose of
appropriate „Surahs‟ and „Ayahs‟ from the Quran in both

537 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 9, No. 8, 2018

using the most significant word approach was to increase the 2) Pattern matching approach: Mohammad Hijjawi,
rate of the expected output. The chatbot was tested by entering Zuhair Bandar, Keeley Crockett and David Mclean [17]
fifteen questions and the result was 93% correct answers. The implemented the ArabChat, which is a conversational agent
main drawback in this model appears when the structure of the web interface. The chatbot conversation domain is closed,
question is changed or altered from the stored in the knowledge
designed to serve the students of the Applied Science
base, then the chatbot responses with wrong answers. That
happens because the chatbot does not use a heuristic to select a University in Jordon. The interaction between the user and the
proper response and it is based on a direct retrieval model. chatbot is through textual Arabic MSA language. The
Also, a success rate of to be 93% is not justifiable having a conversation remains ongoing until one of the conversation‟s
dataset of fifteen questions only. parties terminates it. The ArabChat reuses the previously
exchanged information during the conversation as a response
BOTTA chatbot by Ali and Habash [3] is a female chatbot,
to the user input, creating long conversations. The core
that simulates friendly conversations with users. The chatbot
supports Egyptian Arabic dialect for both input and output. components of the ArabChat chatbot are the scripting engine
BOTTA is available to the public. It simulates the English and a scripting language. The scripting engine is divided into
chatbot Rosie [16]. The knowledge base is made up of Rosie‟s subcomponents, that allows handling topics of the
AIML files set. Some of Rosie‟s AIML files are translated conversations. ArabChat knowledge base contains 1218
directly to Arabic, and the others are modified according to the utterances, that are classified into contexts, each context
use of the Arabic dialects. Also, for each conversation, contains rules. The rules consist of patterns and associated
BOTTA chatbot temporally stores the basic information about textual responses. ArabChat was tested over 174 users, the
the user such as age, gender, and nationality by asking average input for each user was 7 inputs per user. The result
questions yielding a conversation that is open since the chatbot shows that 73.56% of the inputs matched the expected output.
can response to different topics domain. The length of the
conversation is long where chatbot can response to the user Enhanced ArabChat [18] is an updated version of ArabChat
based on previous information in conversation. However, it [17] by Hijjawi, Bandar and Crockett. This version uses extra
does not update the knowledge base and add new responses, so features including Utterance Classification and Hybrid Rule.
it is based on retrieval-based model. It depends on a pool of These improvements were at the engine level while some
predefined responses using heuristics to response with an additional improvements need to be added to the scripting
appropriate output. Also, the chatbot does not perform the text language and knowledge base to meet the changes needs.
normalization on the user input to get the suitable response. It Utterance classification feature aims to distinguish between a
performs orthographic transformations, that includes correcting question and non-question utterances. It works by adding extra
common spelling mistakes of the user input. With this method, keywords to the pattern of the question-based rule, to deal with
BOTTA was able to resolve 85.1% of the common spelling keyword matching. Hybrid Rule is the second feature and it
mistakes in Arabic typing. focuses on how to reply and deal with an utterance that request
many topics. Although ArabChat gave a better result of Ratio
Table I shows a summary on the discussed textual of Matched Utterances to the Total (RMUT) than enhanced one
conversation chatbots that uses AIML approach. due to unserious users, the manual checking gives more
accurate results and showed improvement in performance. By
TABLE I. A SUMMARY OF TEXTUAL ARABIC CHATBOTS USING AIML
APPROACH analyzing logs manually, Enhanced ArabChat deals
successfully with 82% of utterances with two topics and this
Interaction ratio is decreasing when the number of topics is increased in
Type the utterance. Using manual checking, classifying utterance
Chatbo Langua Conversati Domai Dataset shows a high percentage of question-based utterances due to
t ge on Length n Model three factors: the selected domain, the users' needs, that implies
Inpu Outp
t ut that they are more likely to ask rather than discuss, and
difficulties to script a large number of rules.
Closed ArabChat with classification methodology [19] is another
Quran Classical Retrieva
Text Text Short (Quran ArabChat [17] update by Hijjawi, Bandar and Crockett. Using
[13] Arabic l- Based
book) a new classification methodology for Arabic utterances. This
Closed
new approach classifies the sentences into questions and non-
Quran1 English/
4-114 classical Text Text Short
Retrieva questions including assertions and instructions. The benefit of
(Quran l- Based
[6] Arabic book)
applying this approach is that the number of patterns required
per rule will decrease and hence increase the performance by
Web Closed firing the suitable rule, depending on the utterance type being a
Arabic Retrieva
QA [2], Text Text Short (Medic question or non-question. Different topics and list of function
(MSA) l- Based
[15] al care) words have been used from domains such as politics, religion,
Egyptian
sports, education, business and adding some synthetic non-
Botta Retrieva question sentences and indirect questions. This classification is
Arabic Text Text Long Open
[3] l- Based done by pre-processing the Arabic sentence into equivalent
dialect
numeric tokens and then importing the tokens into a machine

538 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 9, No. 8, 2018

learning toolkit in WEKA. In WEKA, a Decision Tree, which TABLE II. A SUMMARY OF TEXTUAL ARABIC CHATBOTS USING PATTERN
achieve the highest accurate classifier to be applied on the MATCHING APPROACH
tokenized numeric dataset, is generated and then is converted
into a standard IF-THEN classification rule to classify Interaction
Type Conversat Datase
utterances. Chatbot
Langua
ion Domain t
ge Inp Outp Length Model
Mobile ArabChat [20] is based on the original ArabChat ut ut
[17] and it is a mobile-based conversational agent and it is also
used to work as an advisor for students in Applied Science Closed
(for
University in Amman. It is a light version of ArabChat students
implemented in Android. Although there are some challenges Retriev
ArabChat Arabic of
Text Text Long al-
facing users in the Arab Countries such as slow and unstable [17] (MSA) Applied
Based
internet connection and limited bandwidth, this application Science
works even with limited Internet bandwidth. Mobile ArabChat Universi
ty)
implemented pattern matching approach based on the text. This
framework consists of the same component as in ArabChat: Close
scripting engine, scripting language and a knowledge base. (for
students
Based on a subjective approach, 96% of users agree that using Enhanced
Arabic of
Retriev
Mobile ArabChat via mobile is better than using the same ArabChat Text Text Long al-
MSA Applied
[18] Based
system via desktop. However, Mobile ArabChat needs an Science
internet connection to work. Universi
ty)
Abdullah [4] is an Arabic Conversational Intelligent
ArabChat
Tutoring System (CITS) that teaches children aged 10 to 12 with Arabic
Retriev
years old essential topics about Islam. This online system can Text Text Long Open al-
classificat MSA
Based
engage with students using MSA. That asking a series ion [19]
questions to the students, and discuss with them their answers, Close
using Classical Arabic to give evidences from the Quran and (for
Hadith, which is the sayings and traditions of the Prophet of students
Mobile Retriev
Islam Muhammed. The system is using images and sound Arabic of
ArabChat Text Text Long al-
MSA Applied
effects to interact with students and can determine the student's [20]
Science
Based
knowledge level and hence direct the conversation. Abdullah Universi
CITS can distinguish between the user's questions and answers. ty)
The framework is based on a Pattern Matching approach, it Close
consists of knowledge base having subject topics, the Classica (teach
Retriev
Conversational Agent scripting language to deliver the tutorial Abdullah l Islam
Text Text Long al-
CITS [4] Arabic/ for
conversation to the learners, and The Tutorial Knowledge Base MSA children
Based
to determine the level of individual student knowledge and the )
subject. Close
(for
LANA [21] is another CITS and it was developed for
children
children with Autism Spectrum Disorder (ASD) that are 10 to LANA English/ with Retriev
16 years old who have reached a basic competency with the CITS classica Text Text Long Autism al-
mechanics of Arabic writing to teach them topics on science [21] l Arabic Spectru Based
using MSA. Children with ASD have difficulties in traditional m
learning because the teacher can't meet the need of every Disorder
)
individual student. LANA engages children with a science
tutorial delivered in MSA. It is similar to Abdullah CITS, but it B. Speech Conversation Chatbot
offers different learning style models such as visual, auditory
and kinesthetic, enabling children to practice learning skills The interaction in the speech conversation chatbot is based
independently based on their needs using pattern matching and on the voice as an input, or output, or both. Also, the textual
short text similarity algorithm. This system also interacts with interaction for input or output in this type of chatbot is
children using materials such as picture, audio, or instructions supported as well with the voice interaction. However, the
according to the user‟s learning style. Table II shows a brief research on speech conversation chatbots is limited in Arabic.
review of the discussed related pattern matching text This section presents two related works, that are classified
conversational chatbots. based on the used approach into AIML and pattern matching.

539 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 9, No. 8, 2018

1) AIML Approach: Hala [22] is a female robot However, it does not generate new responses, as it depends
receptionist located at Carnegie Mellon University in Qatar on the information predefined in the dataset to respond with an
(CMU-Qatar). Hala accepts and speaks English and Arabic. appropriate output. The IbnSina conversation system is
There are three possible input modes the English, MSA or designed based on object-oriented classes as Wikipedia class
and Quran class, that allows the robot to reply with the
„Arabizi‟ which is Arabic written in English letters. The users
expected response such as chatterbot class, that enables making
interact with the chatbot through the keyboard. Hala responses a simple conversation and reply to user inquiries. Also, there is
by producing a voice reply and a text appears next to her face a developed chatterbot module, that replies to user inputs.
on the screen. The response language depends on the user
input language. Hala provides information about campus There are two modules supported in the conversation. First,
directions, weather, local events and answer queries regarding text-to-speech, where the input is text, and the output is audio.
Second, speech-to-text, the input is an audio that is converted
her personal life, in an open domain and long conversation
to text then processed as the first module to get the speech
style. The conversation between Hala and the user takes an output. In addition, IbnSina robot supports other features such
equal number of turns. When the user leaves, Hala will detect as read-aloud-text by reading a text image through the camera
the conversation was ended after a defined timeout. The located in the robot eyes area. The robot interacts with the user
purpose of implementing Hala project was to explore culture by body interactions such as real-time lip syncing, eye
of the human-robot interaction in the CMU-Qatar by studying blinking, face movement, facial expressions, and shaking
the dialogue patterns such as robot's attributes, covered hands. Table III shows a summary of the discussed related
knowledge bases, and cultural variation in the community of work for both AIML and pattern matching approach of the
users. speech Arabic chatbots.
2) Pattern Matching Approach: IbnSina [23], [24] is a From the reviewed studies, we notice that all presented
multilingual conversational robot, that supports Arabic MSA work on Arabic chatbot applications is employing retrieval-
and English. The user interacts with it through text or voice based model. That is, the chatbot responses are based on the
inputs. IbnSina robot responses with audio output, where the data pool from AIML files, database, or web pages. Which can
language of the response matches the user input language. limit the capability and usability of the chatbot. Also, we notice
IbnSina robot generates human interaction dialogue by that all related work relay on AIML or pattern matching
accessing the online Wikipedia and the stored Quran database. approaches. That may lead to 1) a small size of chatbot's
dataset and a restriction to closed domain 2) limits chatbot's
This makes the Chabot of IbnSina robot covering a wide area response to the user where it requires that the user input
of topics. Because of that, it replies to general questions, matches the chatbot dataset to get the correct response.
translating words, or answering the question by giving online Moreover, the complexity of Arabic grammar and the user's
information, or from the stored books in its database. Also, it spelling and grammar mistakes could be one of the reasons
gives the user feedback when there are missing information or for the shortage in capabilities of Arabic chatbots in the
incorrect spelling. That makes the conversation style of literature. Which can explain the limited number of the text and
IbnSina open and long. speech Arabic chatbot applications.

TABLE III. A SUMMARY OF SPEECH ARABIC CHATBOTS V. CONCLUSIONS

This paper presents a survey on Arabic chatbots covering
Interaction twelve different Arabic chatbot studies. They are classified
Type based on the chatbot conversation interaction type into two
Conversa Datas
Chat Langu Appro
tion
Dom
et
groups, text and speech conversational chatbots. The studies
bot age ach ain were presented and evaluated based on the implementation
Length Model
Inp Outp technique, the conversation length and domain, and the model
ut ut
used for the chatbot dataset. The evaluation shows that all the
reviewed chatbots were incorporating retrieval-based dataset
model. This rises a flag to focus on studying and formalizing
Englis Arabic NLP for the conversational agent research. Linguistic
h/ Text complexities hindering Arabic NLP such as morphological
Retrie
Hala Arabic Tex /
[22] (Arabi t Voic
AIML Long Open val- ambiguities which means that the word has many meanings,
Based and syntactic ambiguities which means that the sentence has
zi / e
MSA) more than one structure, along with the diversity of Arabic
dialects are open research challenges that needs cooperation of
linguists and computer scientists. Hence, until now, pattern
matching and AIML are the ways used to build Arabic
IbnSi Englis Tex conversational agents. Additionally, generative-based and deep
Pattern Retrie
na h/ t/ Voic
[23], Arabic Voi e
Matchi Long Open val- learning models are challenging to achieve in Arabic, for the
ng [25] Based lack of available resources to train the learning model
[24] (MSA) ce
compared to the pool of resources available in English for
example. Moreover, there is a shortage in the published

540 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 9, No. 8, 2018

research on Arabic chatbot compared to the available [13] B. Shawar and E. Atwell, “An Arabic chatbot giving answers from he
commercial applications that are more sophisticated and Qur‟an,” in Proceedings of TALN04: XI Conference sur le Traitement
Automatique des Langues Naturelles, 2004, vol. 2, pp. 197–202.
developed than the chatbots available in the literature.
[14] R. Wallace, The Elements of AIML Style. Alice AI Foundation, 2003.
REFERENCES [15] W. Brini, M. Ellouze, S. Mesfar, and L. Belguith, “An Arabic question-
[1] Z. X. Bingquan Liu, B. W. Chengjie Sun, D. F. W. Xiaolong Wang, and answering system for factoid questions,” in 2009 International
Min Zhang, “Content-Oriented User Modeling for Personalized Conference on Natural Language Processing and Knowledge
Response Ranking in Chatbots,” IEEEACM Trans. Audio Speech Lang. Engineering, 2009, pp. 1–7.
Process., vol. PP, no. 99. [16] Pandorabots, Rosie: Base content for AIML 2.0 chatbot. pandorabots,
[2] B. A. Shawar, “A Chatbot as a Natural Web Interface to Arabic Web 2018.
QA,” Int. J. Emerg. Technol. Learn. IJET, vol. 6, no. 1, pp. 37–43, Mar. [17] M. Hijjawi, Z. Bandar, K. Crockett, and D. Mclean, “ArabChat: An
2011. Arabic Conversational Agent,” in 2014 6th International Conference on
[3] D. A. Ali and N. Habash, “Botta: An Arabic Dialect Chatbot.,” in Computer Science and Information Technology (CSIT), 2014, pp. 227–
COLING (Demos), 2016, pp. 208–212. 237.
[4] O. Alobaidi, K. Crockett, J. O‟Shea, and T. Jarad, “Abdullah: An [18] M. Hijjawi, Z. Bandar, and K. Crockett, “The Enhanced Arabchat: An
intelligent arabic conversational tutoring system for modern islamic Arabic Conversational Agent,” Int. J. Adv. Comput. Sci. Appl., vol. 7,
education,” in Proceedings of the World Congress on Engineering, 2013, no. 2, 2016.
vol. 2. [19] M. Hijjawi, Z. Bandar, and K. Crockett, “User‟s utterance classification
[5] G. F. L. Chayan Chakrabarti, “A Semantic Architecture For Artificial using machine learning for Arabic Conversational Agents,” in 2013 5th
Conversations,” presented at the The 6th International Conference on International Conference on Computer Science and Information
Soft Computing and Intelligent Systems, and The 13th International Technology, 2013, pp. 223–232.
Symposium on Advanced Intelligence Systems. [20] M. Hijjawi, H. Qattous, and O. Alsheiksalem, “Mobile Arabchat: An
[6] B. Shawar and E. Atwell, “Accessing an information system by Arabic Mobile-Based Conversational Agent,” Int. J. Adv. Comput. Sci.
chatting,” Nat. Lang. Process. Inf. Syst., pp. 570–584, 2004. Appl. IJACSA, vol. 6, no. 10, 2015.
[7] R. Wallace, “The Anatomy of A.L.I.C.E.,” in Parsing the Turing Test: [21] S. Aljameel, J. O‟Shea, K. Crockett, A. Latham, and M. Kaleem,
Philosophical and Methodological Issues in the Quest for the Thinking “Development of an Arabic Conversational Intelligent Tutoring System
Computer, R. Epstein, G. Roberts, and G. Beber, Eds. Dordrecht: for Education of children with ASD,” in 2017 IEEE International
Springer Netherlands, 2009, pp. 181–210. Conference on Computational Intelligence and Virtual Environments for
[8] K. Ramesh, S. Ravishankaran, A. Joshi, and K. Chandrasekaran, “A Measurement Systems and Applications (CIVEMSA), 2017, pp. 24–29.
Survey of Design Techniques for Conversational Agents,” in [22] M. Makatchev et al., “Dialogue patterns of an Arabic robot
Information, Communication and Computing Technology, 2017, pp. receptionist,” in 2010 5th ACM/IEEE International Conference on
336–350. Human-Robot Interaction (HRI), 2010, pp. 167–168.
[9] L. Bradeško and D. Mladenić, “A survey of chatbot systems through a [23] N. Mavridis, A. AlDhaheri, L. AlDhaheri, M. Khanii, and N.
loebner prize competition,” in Proceedings of Slovenian Language AlDarmaki, “Transforming IbnSina into an advanced multilingual
Technologies Society Eighth Conference of Language Technologies, interactive android robot,” in 2011 IEEE GCC Conference and
2012, pp. 34–37. Exhibition (GCC), 2011, pp. 120–123.
[10] K. Cho et al., “Learning Phrase Representations using RNN Encoder– [24] L. Riek et al., “Ibn sina steps out: Exploring arabic attitudes toward
Decoder for Statistical Machine Translation,” 2014, pp. 1724–1734. humanoid robots,” in Proceedings of the 2nd international symposium
[11] H. N. Io and C. B. Lee, “Chatbots and conversational agents: A on new frontiers in human–robot interaction, AISB, Leicester, 2010, vol.
bibliometric analysis,” in 2017 IEEE International Conference on 1.
Industrial Engineering and Engineering Management (IEEM), 2017, pp. [25] N. Mavridis, “About the Chatterbot Used in IbnSina Conversational
215–219. Robot. Personal E-mail,” 23-Feb-2018.
[12] S. AbdulKader and J. Woods, “Survey on Chatbot Design Techniques in
Speech Conversation Systems,” Int. J. Adv. Comput. Sci. Appl.
IJACSA, vol. 6, no. 7, 2015.

541 | P a g e
www.ijacsa.thesai.org
View publication stats