Chatbots: An Analysis of the State of Art of Literature

Abstract

Considering the context of Computer Science, Chatbots are computer programs that, through techniques of Artificial Intelligence, aim to simulate human behavior when at the moment of a dialogue through text messages. Despite being a relatively new area of study, the application of this concept has increased considerably, whether in academic settings or in commercial applications, such as situations involving assistants to clarify doubts on websites of manufacturers of electronic products or even in services such as telephony. Thus, this article proposes an analysis of the state of the art of research through a quantitative bibliometric analysis, supported by the Scopus database.

Full text

Chatbots: An Analysis of the State of Art of Literature
Andréia Ana Bernardini1, Arildo A. Sônego2, Eliane Pozzebon2
1 Escola Superior de Criciúma (ESUCRI) – Criciúma, SC - Brazil
2 Universidade Federal de Santa Catarina (UFSC) – Araranguá, SC – Brazil
andreia@esucri.com.br, asonego@gmail.com, eliane.pozzebon@ufsc.br
Abstract. Considering the context of Computer Science, Chatbots are
computer programs that, through techniques of Artificial Intelligence, aim to
simulate human behavior when at the moment of a dialogue through text
messages. Despite being a relatively new area of study, the application of this
concept has increased considerably, whether in academic settings or in
commercial applications, such as situations involving assistants to clarify
doubts on websites of manufacturers of electronic products or even in services
such as telephony. Thus, this article proposes an analysis of the state of the art
of research through a quantitative bibliometric analysis, supported by the
Scopus database.
1. Introduction
Artificial Intelligence (AI) denotes as a science that is dedicated to the study of systems
that, from the perspective of any observer, act intelligently [Coppin 2015]. This is a new
field of research, when compared to other sciences, since it began in 1956 [Russel and
Norvig 2013]. The most accepted concept is of an intelligent agent, in which the
symbolic and connectionist approaches can act in a collaborative way aiming at solving
problems through a computational system [Franco 2014].
The study of AI is interdisciplinary, since it makes use of a large amount of
knowledge from other areas, such as mathematical calculation and logic [Artero 2009].
Artificial Intelligence brings together a considerable variety of areas, such as learning
and perception to more specific tasks such as games, demonstration of mathematical
theorems, and diagnosis of diseases among others, being considered relevant to any
intellectual task [Russel and Norvig 2013].
A Chatbot is a computer program that works with programming languages
oriented to the manipulation of natural human language [Lima 2014]. The possibility of
training a computer to establish dialogues with humans dates back to the 1950s, with
Alan Turing. In his essay titled "Computing Machinery and Intelligence," Turing
proposed a test consisting of a program that developed a text message conversation with
an interrogator for five minutes [Coppin 2015]. The person should then guess if he was
talking to a human being or a computer and if the machine tricked the interrogator for
30% of the time, the program would pass the test.
The basic principle employed in a Chatbot consists of an environment that
receives questions in natural human language, associates these questions with a
knowledge base, and then issues an answer [Fryer and Carpenter 2006]. Usually, a
Chatbot works from the insertion by the user of a question or comment, and from this
DOI: 10.5753/wave.2018.1 1
I Workshop on Advanced Virtual Environments and Education (WAVE2)

moment the program answers the question, makes a comment or starts a new topic
[Huang, Zhou and Yang 2007].
Examples of Chatbots [Russel and Norvig 2013] are the ELIZA, MGONZ,
NATACHATA and CYBERLOVER projects. In addition to these, [Huang, Zhou and
Yang 2007] indicates the PARRY environment. This list [Lima 2014] is complemented
with the ALICE (Artificial Internet Computer Entity) system, activated in the year of
1995.
In this way, this article aims to demonstrate, based on the information reported
by the Scopus platform, the state of the art of research involving the theme Chatbots.
2. Methods, Results and Discussions
In order to strengthen the credibility of the research regarding its results, we opted for
the adoption of a systematic literature review and quantitative analysis. This process
consists of a data survey in which rigorous reviews of academic publications are
demanded for signs that may lead to the identification of evidences regarding a research
topic or even a topic in a desired area [Freire 2014].
In this way, for the elaboration of this research, the Scopus platform
(www.scopus.com) was chosen, international base responsible for scientific
publications of interdisciplinary nature.
Initially, the term "chatbot" was inserted in the search window of the platform.
This procedure reported the existence of 314 documents, distributed as shown in Figure
1. The originality of the theme is confirmed by the results presented, since the first
publication registered dates from the year of 2002. In the same way, the total amount of
documents about the subject can be considered small when compared to other topics,
such as Internet of Things (IoT) with about 32.949 publications at the time of this
research. It is also observed the linearity of these quantities over the years, and in the
year 2016 a slight increase occurred and in 2017 this number presented the highest
indices. The year 2018, being the current year, still presents little expressive numbers
when faced with the others.
Figure 1. Evolution of polls related to Chatbots
2
I Workshop on Advanced Virtual Environments and Education (WAVE2)

Regarding the publications according to the country of origin, Figure 2 presents
the results obtained. The ten countries that most published during the period chosen
were considered. We can see the predominance of the United States, followed by Italy.
In this context, the two countries hold 41.93% of all material produced. An interesting
point to consider is the position of China, ranking only sixth in this classification given
its traditional technological nature.
Figure 2. Publications by country
The next item, shown in Figure 3 and reported by the Scopus database, relates to
documents classified according to the subject area.
Figure 3. Publications by subject area
The absolute majority of publications are related to the area of Computer
Science, as would be expected, given the affinity with the main theme. Accordingly, the
universe of contexts to which the term chatbots is associated is broad and diversified,
confirming its interdisciplinarity.
3
I Workshop on Advanced Virtual Environments and Education (WAVE2)

For documents of the type, the search results are shown in Figure 4. Conference
papers, categorized as papers published in seminars and international conferences,
aggregate the largest amount with 63.38% of publications, followed by articles,
classified as papers published in journals and specialized websites, with 22.29%. The
conference revisions occupy third place with 7.64%, book chapters in fourth with 1.91%
and publications of other types 4.78%, in fifth place.
Figure 4. Publications by type
In terms of dates, the first published document is entitled "The catacomb project:
building a user-centered portal the conversational way", under the responsibility of
[Ginsburg 2002]. In the article, which received a quote, the author comments on the
need to unify business applications such as database and documents in general, to
promote centralized access to this information. As a solution, a conversation portal is
proposed, the Catacumb project, directed to the end user, which extends the
functionality of Chatbot ALICE. From the point of view of the author, this solution
promotes scalability, extensibility and coordination between end users and developers.
In relation to the most cited articles, [Turney and Littman 2003], entitled
"Measuring Praise and Criticism: Inference of Semantic Orientation from Association",
appears first with 748 citations. In the document, the authors discuss an automated
system with the objective of measuring the semantic orientation (positive or negative) of
words of a text. According to the authors, this orientation is associated with the
expressions used in the sentences, indicating a character of criticism or praise. This
system would be applied to the classification of texts, the tracking of opinions in online
dialogues, analysis of responses to surveys and in chatbots. In the case of chatbots, the
suggested system could help in choosing the most appropriate response type for the
moment.
3. Conclusion
This document demonstrates, based on the information contained in the Scopus
database, a portion of the research involving the theme Chatbots.
4
I Workshop on Advanced Virtual Environments and Education (WAVE2)

An initial feature concerns the little published material (to the total were 314
documents) when compared to other subjects. This condition is credited to the fact that
the first publication occurred in the year 2002, being considered in this way a recent
theme. However, a significant increase in the number of publications has been detected
since 2016, possibly driven by the increase in investments in the area of Artificial
Intelligence by organizations, by the competitive and increasingly globalized market, as
well as by the increase in performance of the hardware of the computers in general, thus
allowing the execution of more robust applications. In this context, it is suggested to
extend the research in other international databases, such as the IEEE or even the Capes
Portal, afterwards making a comparison with the data collected in this research.
In relation to the content of the analyzed articles, one observes its adherence to
academic environments and researches, usually oriented to educational contexts, in
order to help students in their daily activities. However, recent products such as
Amazon's Echo, Microsoft's Cortana, Apple's Siri or even Google Now, strengthen the
advancement of research in the area aimed at the general public. These products, as
indicated by their manufacturers, make intensive use of neural networks, considered an
emerging concept in the field of Artificial Intelligence. Here is a potential subject to be
explored in future works.
Considering the quantity and diversity of titles, we can note the significant
presence of interdisciplinarity, ratifying that Artificial Intelligence contemplates several
distinct areas, such as health, education, computing, linguistics, psychology, among
others. Thus, the relevance of the theme is emphasized, with a vast horizon to be
explored.
References
Artero, A. (2009). Inteligência Artificial: teoria e prática. 1st ed. São Paulo: Livraria da
Física.
Coppin, B. (2015). Inteligência Artificial. 1st ed. Rio de Janeiro: Ltc.
Franco, C. (2014). Inteligência Artificial. 1st ed. Londrina: Educacional.
Freire, P. (2013). Aumente a qualidade e a quantidade de suas publicações científicas:
manual para elaboração de projetos e artigos científicos. 1st ed. Curitiba: Crv.
Fryer, L. and Carpenter, R. (2006). Bots as language learning tools. Language Learning
and Technology. 10 (3) 8-14.
Ginsburg, M. (2002). The catacomb project: building a user-centered portal the
conversational way. Fourth International Workshop on Web Information and Data
Management. 84-87, ACM Press.
Huang, J., Zhou, M. and Yang, D. (2007). Extracting Chatbot Knowledge from Online
Discussion Forums. In: 20TH International Joint Conference on Artificial Intelligence,
Iijcai 2007, 423 - 428.
Lima, L. A. (2014). Estudo de implementação de um robô de conversação em curso de
língua estrangeira em ambiente virtual: um caso de estabilização do Sistema
Adaptativo Complexo: Doctoral Thesis. Belo Horizonte: Universidade Federal de Minas
Gerais.
5
I Workshop on Advanced Virtual Environments and Education (WAVE2)

Russell, S. and Norvig, P. (2013). Inteligência Artificial. 3rd ed. Rio de Janeiro:
Elsevier.
Turney, P. and Littman, M. (2003). Measuring praise and criticism: Inference of
Semantic Orientation from Association. Acm Transactions On Information Systems. 21
(4) 315-346, Association for Computing Machinery (ACM).
6
I Workshop on Advanced Virtual Environments and Education (WAVE2)