Chatbots for Cultural Heritage: A Real Added Value

Fouad Nafis

, Ali Yahyaouy

and Badraddine Aghoutane

LISAC Laboratory, Department of Informatics, FSDM, Sidi Mohamed Ben Abdellah University, Fez, Morocco

IA Laboratory, Science Faculty, My Ismail University, Meknes, Morocco

Keywords: Chatbot, Cultural Heritage, Conversational Agent, Semantic Web, Ontologies, AIML, NLP.

Abstract: During the last few years, a lot of research work has been done on developing of chatbots that can support

humans in different functionalities. The role of a chatbot cannot be limited to being an important virtual

assistant to its users but can be exploited by organizations and governments as an adaptive application for the

promotion of certain sectors of the economy, such as tourism and cultural heritage (CH). In this paper, the

authors present a theoretical and historical background, then discuss the use of chatbots in the CH domain,

and finally describe the basic steps and challenges of programming a chatbot taking advantage of the advances

in machine learning (ML), deep learning (DL) and semantic web (SW) technologies.

1 INTRODUCTION

A chatbot is a computer program (web, mobile...)

designed to interact with users using text or voice so

that the user thinks they are interacting with a human.

To achieve this result, the use of machine learning

(ML) algorithms is essential. Older techniques

involved create an illusion of intelligence by

implementing much simpler techniques for matching

and processing strings for interaction with users,

using rule-based and generative models. These

techniques suffered from many problems and found it

difficult to respond reliably to user queries. However,

with the emergence of new machine learning

technologies, much more autonomous and especially

more intelligent systems have emerged. A text or

voice conversation is usually initiated by the main

actor, who is the user who formulates a question in

natural language, and the chatbot provides an answer

in natural language. It was one of the initial issues that

gave rise to Artificial Intelligence (AI) technology,

and its advancement has had a significant impact on

the development of Chatbots that can interpret and

process a human's enquiry. Artificial Intelligence

Markup Language (AIML), which is built from

Extensible Markup Language (XML) and is used to

https://orcid.org/0000-0001-6499-7151

https://orcid.org/0000-0003-1954-2734

https://orcid.org/0000-0002-9555-6786

artificially build a chatbot, is valuable in this regard

(Satu et al, 2015). The authors present a brief

overview of some applications that have used AIML

for their conversational service. These applications

are related to cultural heritage, e-learning, e-gov, and

many other fields. In the field of cultural heritage,

chatbots are a new area to explore, given the low

number of initiatives aiming at developing this field

despite the stakes it presents, especially in the

development of the economy and tourism. Indeed, we

might see a chatbot that automates the work of

guiding users and tourists interested in a region's

cultural heritage assets. This will provide a first idea

on the richness of this heritage, then assist the user in

repetitive tasks that previously consumed time and

energy and that can now be delegated to software to

save time and respond to a maximum number of

people.

The number of research papers in the subject of

developing generic or domain-specific virtual

assistants continues to rise year after year. This is

demonstrated in Figure 1, where a simple search of

the SCOPUS database reveals a considerable increase

in the number of research papers on chatbots.

Unfortunately, few of these works are tied to cultural

heritage, which is why the research presented in this

paper is so interesting.

502

Naﬁs, F., Yahyaouy, A. and Aghoutane, B.

Chatbots for Cultural Heritage: A Real Added Value.

DOI: 10.5220/0010737700003101

In Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning (BML 2021), pages 502-506

ISBN: 978-989-758-559-3

Figure 1: Number of documents about Chatbots and CH according to the SCOPUS DB

Several questions have motivated us in this work to

move towards a chatbot for cultural heritage:

Q1: What goals will the chatbot address?

Q2: Is it heavily used in the CH domain?

Q3: Which sectors of the economy will be affected?

Q4: What type of bot should I choose?

Q5: What problems should the chatbot solve?

Q6: Who are the potential users?

Q7: How to structure the conversational flow?

Q8: How to build a sufficiently rich knowledge base?

Chatbots in the field of CH raise challenges that are

comparable to those in other fields. Other challenges

related to this area may emerge, indicating how

conversational agents can give significant added

value, depending on the sections of the heritage

covered. Companies have established indices in the

commercial domain that allow them to evaluate the

influence of a chatbot on the company's everyday

operations in terms of turnover. Based on these

indices, managers can determine whether or not to

install a chatbot to support a portion of their job and

activities. Because prototypes are already in place and

can be customized to individual scenarios, developing

a chatbot is no longer a problem. This saves time and

allows decision-makers to reap the advantages much

faster.

In the sector of cultural heritage, we are also

beginning to gain from these services, but there are

some challenges that must arise owing to the

complexity of heritage data. An overview of the

design and implementation of a pilot audience

development project involving four museums in

Milan, Italy is provided in this case study. A

promising initiative that visualizes the narrative

utilizing a convergence of chatbot and gamification

platforms combining the latest artificial intelligence

(AI) technology in order to find new and fascinating

methods to engage adolescents in visiting museums

and art galleries. (Gaia et al, 2019)

2 HISTORY OF CHATBOTS

The first chatbot in history was launched in 1966. It's

a program called ELIZA, which was created by MIT

scientist Joseph Weizenbaum. ELIZA uses a keyword

recognition system to rephrase patients' statements

into interrogative form, simulating a psychotherapist.

The Jabberwacky program, created in 1988, simulates

a discussion in genuine human language in an

interesting and enjoyable way. ALICE is

unquestionably the precursor of chatbots (Artificial

Linguistic Internet Computer Entity). ALICE is a

computer program that was created in 1995 that can

simulate a beneficial conversation with a human. It

has an identification system that is adapted to the

personality of its interlocutor, as well as a larger

information base.

Watson was invented by IBM in 2005. Watson is

able to answer questions expressed in natural

Chatbots for Cultural Heritage: A Real Added Value

503

language because to the newest artificial intelligence

technologies. Apple (Siri), Google (Google Now),

Amazon (Alexa), Microsoft (Cortana), and Facebook

(Messenger) all joined the dance in the early 2010s,

launching their own natural language user interfaces.

We can construct conversational bots for Facebook

users to connect with via Facebook Messenger, for

example. This new ability to create a chatbot will

result in their massive democratization.

3 RELATED WORKS

Chatbots have been developed to overcome

communication barriers in a variety of fields. One of

the most important is cultural heritage. Chatbots are a

solution to employ as virtual assistants that answer

exactly to user queries at a lesser cost. In the cultural

heritage area, the usage of virtual assistants can have

a considerable positive impact on the preservation

and enhancement of a region's heritage. Despite the

significant stakes that it may offer, especially in areas

such as tourism, research and development work on

chatbots applied to heritage monuments remains

insufficient. In (Pilato et al, 2005), a chatbot system

for the cultural heritage domain is presented. This

system employs knowledge bases based on a

semantic approach, allowing them to assess their own

expertise in relation to the user's questions, which are

all mapped in the same semantic space.

A conversational agent would frequently employ

a vast database of questions and answers to train the

end system to respond to user inquiries automatically.

These databases can be done manually or using

publicly available data. This may bring us to another

issue that may have an impact on the system's quality:

the redundancy of the questions/answers. There are a

number of methods available to help solve this

challenge and arrive at a full data source with no

redundancy and enough questions and answers for the

learning stage (Nafis et al, 2015)

The authors in (Duguleană et al, 2020) offer an

intelligent conversational agent for increasing

museum information accessibility. Using NLP

approaches, the generated intelligent virtual agent

communicates with users. In (Lombardi et al, 2019)

the authors aims to take use of Italy's enormous

number of archaeological sites by creating a Chatbot

that can provide tourists with the relevant information

at the right moment. Through semantic analysis of

archaeological data, this Chatbot should be able to

provide users with contextual information. The goal

of (Casillo et al, 2020) is to propose a

recommendation system that may create a

personalized tourist itinerary for some of Campania's

most important cultural attractions. According on the

tourist's profile and contextual factors, this system

recommends sites of interest and related services. The

user interacts with the system using a chatbot, which

allows for a real conversation. Because the final aim

of a user in both cases is to have relevant information

that answers a defined query, chatbots and

recommendation systems can often support each

other (Atzori et al, 2017). Chatbots act as human

guides for users, and recommendation systems are

helpful tools for leading them to suitable services and

products. The purpose of both systems is to

understand and meet the user's demands (Nafis et al,

2020).

In a similar spirit, the authors of (Lombardi et al,

2019) present a Chatbot system capable of providing

contextual information to tourists visiting

archaeological sites in Italy, based on a semantic

analysis and the implementation of a

recommendation system that can provide users with

automatic assistance.

It should be noted that the social and affective

aspects of the human and chatbots interaction have

progressed significantly in recent years, proving to be

enjoyable for users and having a favorable impact on

the participants' perceived well-being. This is most

likely due to the features established in a chatbot that

support relationship building, such as acceptance,

understanding, and non-judgment (Skjuve et al, 2021)

4 BASIC FUNCTIONALITIES

Conversational agents can be classified according to

several criteria (Hussain et al, 2019). The most

important criteria are (Figure 2):

 Technology employed: ML and DL algorithms,

artificial intelligence, semantic Web

technologies such as linked data, ontologies

and RDF graphs, AIML, etc.

 Domain of knowledge: general or specialized

field

 Requirements: functional requirements are or

are not necessary.

 Final objective: oriented conversation or not

 Approach used: Rule-based, Retrieval-based,

or Generative based.

 Interaction method: vocal or text

conversational agent

 GUI: Web, mobile, etc.

BML 2021 - INTERNATIONAL CONFERENCE ON BIG DATA, MODELLING AND MACHINE LEARNING (BML’21)

504

Figure 2: Classification of conversational agents

Other criteria could be used depending on the context

of implementation and the scope, but the majority of

the criteria used in the literature can be summarized

in Figure 2.

Figure 3 presents the general architecture of a

conversational agent operating in the cultural heritage

domain. Other functionalities can be added depending

on the context and on the expected purpose and

results. A user provides a request through a mobile

device or through a web interface. An API receives

the request and processes it through ML and DL

algorithms that will use already stored data sources to

answer the request. For each user response, feedback

can be received to measure the degree of satisfaction

with the given answer in order to make updates or

confirm the answers for subsequent users. The

question and answer are automatically added to the

data source to enrich it. The API parameters are

continuously modified to improve its performance

and reliability. From time to time the data source is

enriched with a set of SCH data with well-chosen

questions/answers. This can be used with the help of

semantic web and Linked Data technologies to assign

a semantic layer on the manipulated data (Nafis &

Chiadmi, 2016). This addition is the responsibility of

one or more domain expert administrators who play

the role of user and admin, testing the different

features of the chatbot and then writing reports on

each feature developed. Then feeding the data source

with objects and questions/answers on the SCH to run

the machine learning algorithms that will be used

later. (Dimitris et al, 2019)

Figure 3: General architecture of a conversational agent

Chatbots for Cultural Heritage: A Real Added Value

505

5 CONCLUSION

This paper conducted study on the employment of

conversational agents to conserve a region's cultural

heritage. The architecture of a chatbot functioning in

the CH domain has been given as a generic

architecture. Implementing and testing a chatbot to

communicate the study region's rich cultural heritage

will be a promising project. The criteria for selecting

and developing appropriate technology will definitely

serve as a foundation for a successful first experience

with the region's cultural heritage. Several areas of the

local economy, such as tourism, will benefit as a

result of this. The next step will be to present and

develop an architecture for a conversational agent

specializing in the research region's scientific cultural

heritage. There will be a comparison of the results

acquired with those of other chatbots in the domain.

REFERENCES

Fararni, K. A, Nafis, F, Aghoutane, B., Yahyaouy, A., Riffi,

J. and Sabri, A. "Hybrid recommender system for

tourism based on big data and AI: A conceptual

framework," in Big Data Mining and Analytics, vol. 4,

no. 1, pp. 47-55, March 2021.

doi:10.26599/BDMA.2020.9020015

Atzori, M., Boratto, L., & Spano, L. D. (2017). Towards

Chatbots as Recommendation Interfaces. Proceedings

of the Second Workshop on Engineering Computer-

Human Interaction in Recommender Systems Co-

Located with the 9th ACM SIGCHI Symposium on

Engineering Interactive Computing Systems EICS

2017, 6.

S. Schaﬀer, O. Gustke, J. Oldemeier, et N. Reithinger,

« Towards Chatbots in the museum », Barcelona,

Spain, sept. 2018, vol. Volume 2176, p. 7.

Casillo, M., Clarizia, F., D’Aniello, G., De Santo, M.,

Lombardi, M., & Santaniello, D. (2020). Chatbot : A

cultural heritage aware teller-bot for supporting

touristic experiences. Pattern Recognition Letters, 131,

234‑243. https://doi.org/10.1016/j.patrec.2020.01.003

Dimitris C, G., Theodoridis, K. P., Tzavella, G.,

Vlachopoulou, G., Kondili, I., & Tzioli, M. (2019).

Chatbot tools evaluation. In Digital Phenotyping and

Mobile Sensing (p. 249‑260). Springer International

Publishing. https://doi.org/10.1007/978-3-030-31620-

4_16

Duguleană, M., Briciu, V.-A., Duduman, I.-A., &

Machidon, O. M. (2020). A Virtual Assistant for

Natural Interactions in Museums. Sustainability,

12(17), 6958. https://doi.org/10.3390/su12176958

Gaia, G., Boiano, S., & Borda, A. (2019). Engaging

Museum Visitors with AI : The Case of Chatbots. In T.

Giannini & J. P. Bowen (Éds.), Museums and Digital

Culture (p. 309‑329). Springer International Publishing.

https://doi.org/10.1007/978-3-319-97457-6_15

Hussain, S., Ameri Sianaki, O., & Ababneh, N. (2019). A

Survey on Conversational Agents/Chatbots

Classification and Design Techniques. In L. Barolli, M.

Takizawa, F. Xhafa, & T. Enokido (Éds.), Web,

Artificial Intelligence and Network Applications (Vol.

927, p. 946‑956). Springer International Publishing.

https://doi.org/10.1007/978-3-030-15035-8_93

Lombardi, M., Pascale, F., & Santaniello, D. (2019). An

application for Cultural Heritage using a Chatbot. 2019

2nd International Conference on Computer

Applications & Information Security (ICCAIS), 1‑5.

https://doi.org/10.1109/CAIS.2019.8769525

Nafis, F., & Chiadmi, D. (2016). Methods and Systems for

the Linked Data. In Proceedings of the Mediterranean

Conference on Information & Communication

Technologies 2015 (Vol. 381, p. 587‑592). Springer

International Publishing. https://doi.org/10.1007/978-

3-319-30298-0_62

Nafis, F., Fararni, K. A., Yahyaouy, A. and Aghoutane, B.

(2020). Towards a semantic recommender system for

cultural objects : Case study Draa-Tafilalet region.

2020 International Conference on Intelligent Systems

and Computer Vision (ISCV), 1‑4.

https://doi.org/10.1109/ISCV49265.2020.9204187

Nafis, F., Fararni, K. A., Yahyaouy, A. and Aghoutane, B.

(2021). An Approach based on Machine Learning

Algorithms for the Recommendation of Scientific

Cultural Heritage Objects, International Journal of

Advanced Computer Science and Applications

(IJACSA), 12(5). http://dx.doi.org/10.14569/

IJACSA.2021.0120529

Pilato, G., Vassallo, G., Augello, A., Vasile, M., & Gaglio,

S. (2005). Expert Chatbots for Cultural Heritage.

Intelligenza Artificiale, 2(2):25-31.

Satu, Md. S., Parvez, Md. H., & Shamim-Al-Mamun.

(2015). Review of integrated applications with AIML

based chatbot. 2015 International Conference on

Computer and Information Engineering (ICCIE),

87‑90. https://doi.org/10.1109/CCIE.2015.7399324

Skjuve, M., Følstad, A., Fostervold, K. I., & Brandtzaeg, P.

B. (2021). My Chatbot Companion—A Study of

Human-Chatbot Relationships. International Journal of

Human-Computer Studies, 149, 102601.

https://doi.org/10.1016/j.ijhcs.2021.102601

BML 2021 - INTERNATIONAL CONFERENCE ON BIG DATA, MODELLING AND MACHINE LEARNING (BML’21)

506