=Paper=
{{Paper
|id=Vol-3797/paper13
|storemode=property
|title=
Can Natural Language Processing Technologies Help the Digital Transformation of Local Public Administrations?
|pdfUrl=https://ceur-ws.org/Vol-3797/paper13.pdf
|volume=Vol-3797
|authors=Ángel Lloret
|dblpUrl=https://dblp.org/rec/conf/sepln/Lloret24
}}
==
Can Natural Language Processing Technologies Help the Digital Transformation of Local Public Administrations?==
https://ceur-ws.org/Vol-3797/paper13.pdf
Can Natural Language Processing Technologies Help
the Digital Transformation of Local Public
Administrations?
Ángel Lloret1,∗
1
Department of Software and Computing Systems, University of Alicante, 03690 Alicante, Spain
Abstract
In this article, we present an investigation with its conclusions and results on the pivotal role of Natural
Language Processing (NLP) technologies in the ongoing digital transformation of our society, specifically
within local public administrations. Our research is focused on the utilization of chatbots as tools for
interaction, information retrieval, and provision, which we justify as the central objective to our study.
This advanced tool facilitates the acquisition of crucial information for public managers regarding
usage patterns, behaviors, statistics, and other data collected from citizens, tourists, or other groups
interacting with their local public administration. Monitoring will be conducted via indicators and data
compiled in a control panel or dashboard, aiding data-driven decision-making, enhancing user experience,
and increasing satisfaction. The justification for employing chatbots as a communication channel lies
in their simplicity, ease of understanding, and user-friendly nature. Key contributions of this NLP tool
include the selection, compilation, and classification of datasets, as well as the provision of various
functionalities (such as search, reading comprehension, chat, recommendation, and classification) that
ensure users receive the requested information with the highest possible accuracy. The research based
on the hypothesis outlined below will ultimately confirm the results presented in this study affirmatively
answer the question posed in the title of the article.
Keywords
Natural Language Processing, Local Public Administrations, chatbot, Smart City, Citizen,
1. Introduction: Justification of the research
Firstly, it is crucial to note that this research is part of a Doctoral Thesis titled ”Modeling the
Degree of Digital Transformation of Local Public Administrations: NLP and IoT Technologies
as Catalysts for Change.” The thesis aims to develop a methodology to measure digital transfor-
mation in local public administrations. This work will build on the ASIS study by Lloret et al.
(2021) [1], which involved creating a technological questionnaire, collecting technical data, and
developing a dashboard. The thesis proposes the hypothesis that natural language processing
technologies have significantly contributed to this digital and cultural shift, and the proposed
article will explore this hypothesis in greater depth.
In recent years, digital technologies have profoundly transformed the economy and society,
affecting all sectors and daily life. The European Union has promoted digitalization strategies
Doctoral Symposium on Natural Language Processing, 26 September 2024, Valladolid, Spain.
Envelope-Open lloret@ua.es (Á. Lloret)
GLOBE https://cvnet.cpd.ua.es/curriculum-breve/es/lloret-rivera-angel-rafael/164903 (Á. Lloret)
Orcid 0009-0008-2665-309X (Á. Lloret)
© 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
to support this shift [1]. Specifically, the cities are pivotal in the digital transformation process,
making research into digital components and their implementation levels highly valuable.
This study, which is part of our doctoral investigation, emphasizes the role of local public
administrations as they are closest to citizens and crucial in driving this ongoing change.
One of the earliest discussions on digital transformation in academic literature appeared in
2005 by Larsen and Milakovich (2005) [2], who emphasized the need for strategies to manage
relationships between public administrations and citizens through electronic government. They
highlighted the use of Customer Relationship Management (CRM) strategies, originally from
the private sector and primarily based on telephone communication, as essential for addressing
the challenges faced by public administrations. The article also identified digital divides and
the necessity to overcome them to improve citizen engagement, governance, democracy, and
public administration.
Within the different Natural Language Processing technologies we are going to focus in
this paper on chatbots, but it is important to concisely list the natural language processing
techniques along with their most common uses [3]:
• Machine Translation. Automatic translation converts texts from one language to another
using complex algorithms. It is based on machine learning of semantics, syntax and
real-world context.
• Discourse Analysis. Discourse analysis identifies the structure of the text used. It is based
on the design of algorithms that classify and categorize speech structures.
• Morphological Analysis, Analysis, morphological analysis separates words into mor-
phemes and categories based on the complexity of the structures and words of the
language.
• Natural Language Generation and Understanding. Natural language generation trans-
lates data into human-readable language, while understanding converts text into formal
notations, involving semantic recognition and the creation of ontologies.
• Named Entity Recognition. This technique classifies phrases in the text as named entities,
such as organizations or people.
• Text Analysis. Text analysis creates trees from initial study and analysis in order to study
the grammar of sentences, address ambiguity in natural language grammars, and attempt
to resolve the multiple interpretations that can occur in natural language.
• Speech Recognition. Speech recognition transcribes audio into written text. This is
a complex task that requires continuously segmenting speech into units that can be
processed independently and the result makes sense and is coherent.
• Sentiment Analysis. Sentiment analysis extracts subjective information from texts, such as
reviews on social networks, to determine the polarity of certain elements. It is commonly
used in advertising or in the detection and identification of fake news.
• Word Segmentation and Detection. This technique divides continuous text into individual
words. It is straightforward in languages with spaces, like Spanish or English, but
challenging in languages like Japanese, Chinese, and Thai where spaces are not present
in the grammar.
• Word Sense Disambiguation. Word sense disambiguation determines the correct meaning
of a word based on the context in which it is used.
In the case of chatbots, which use natural language processing (NLP) to simulate conversations
with users, they have been widely adopted by both local public administrations and private
companies, occupying a prominent place on their main websites. The main features that
drive its use are constant availability, 24/7 access, multilingual support and anonymity in web
interactions. Notable examples are virtual assistants such as Siri and Alexa, widely used for
efficient information retrieval.
Specifically, Van Noordt and Misuraca (2019) [4] highlight in their article that advances in
artificial intelligence have generated growing interest in chatbots in both the public and private
sectors. In public administration, chatbots can improve service delivery by managing frequent
queries and facilitating transactions, thus alleviating routine staff tasks. Note that the proposed
article makes the following contributions:
• Justify the importance of natural language processing technologies in the digital transfor-
mation of Local APPP.
• Specify the Chatbot tool as a driver of this digital transformation.
• Transform the social change that these technologies have brought about in the way in
which the local public administration communicates with its citizens.
• Improve, local public administrations can use human language technologies to modernize
the quality and management of their data in order to make decisions that benefit their
cities, providing a better service to their citizens.
• Propose, a chatbot architecture that meets the needs of local public administration
• Create an information retrieval system, validated by experts in its content and legality
(to allow the citizen to search for the requested information), a question generator (to
help the citizen) and an administration information classifier system ( to quickly and
successfully identify the information requested by the citizen).
2. Research background
In this section we will mention the various research studies carried out in this regard. We will
focus on human language processing technologies, specifically chatbots, and influence within
public administration. The theoretical basis of modern virtual assistants and chatbots originates
from Alan Turing’s 1951 work on machine intelligence, particularly the ”Turing Test”, which
assesses a machine’s ability to exhibit intelligence similar to human through conversational
indistinguishability [5]. Since then, the development of chatbots has evolved significantly, start-
ing with Joseph Weizenbaum’s ELIZA in 1966, which simulated a psychotherapist’s responses
based on the recognition of keywords [6]. This was followed by Kenneth Colby’s 1971 chatbot,
PARRY, which mimicked a paranoid patient with a more advanced set of predefined responses
[7]. The evolution continued with the creation of human-like avatars, which improve user
interaction by adding gestures and expressions, as commented by Klopfenstein et al. (2017) [8].
Today, conversational agents such as Google Assistant, Siri, Cortana, and Watson have become
an integral part of modern technology, performing a variety of tasks through text and voice
interfaces. Recently, there has been a notable increase in the integration of chatbots on local
public administration websites, a trend that accelerated significantly starting in 2021 [9]. This
rapid increase is attributed to several factors:
• The COVID pandemic and its confinement forced the accelerated implementation of these
communication systems with citizens.
• The advancement of Natural Language processing technologies at the level of Research
and transfer to society.
• And as a consequence of the above, the sudden appearance of artificial intelligence
3. Hypothesis and Methodology
The hypothesis presented in this article, which extends from the Doctoral Thesis as explained
in Section 1, is to verify that the implementation of natural language processing tools in public
administrations, specifically chatbots, aids in the digital transformation of these organizations.
To prepare this article, we have undertaken the following steps:
• Use bibliometric methods to investigate the most relevant scientific literature related to
the topic.
• Propose a chatbot architecture, based on scientific studies, that is most suitable for the
public administration environment, as outlined in Section 4, including a specific use case
for citizen interaction.
• Propose three successful use cases derived from scientific publications that help to confirm
the proposed hypothesis, as discussed in Section 5.
We adopted the bibliometric review method. Bibliometrics is defined as a scientific field that
uses mathematical and statistical methods to study and analyze scientific literature (Vlachý,
1985) [10]. In the mid-20th century, E. Garfield [11] introduced bibliometrics to scientific studies,
making it widely used in scientific research to facilitate the review of knowledge (Karaskus
et al., 2019) [12]. Therefore, bibliometric indicators measure information about the results of
scientific activity in various manifestations. Bibliometric analysis collects citations from articles
and publications to determine the impact of the topics, authors, institutions, countries, journals,
and keywords mentioned in this article (Zupic & Cater, 2015) [13].
In the case of our article, an analysis was carried out using bibliometric techniques to search for
scientific articles related to the topic presented. This approach aims to provide a comprehensive
vision of the development and evolution of the proposed research, evaluating the collaboration
between authors and institutions, as well as key terms related to natural language processing
techniques and digital transformation in local public administration.
The research was carried out using the Web of Science as the main scientific source, due to
its extensive and useful background for this article. Our research began by identifying search
terms in the Web of Science database. The initial search used a combination of terms related
to digital transformation in local public administration, chatbots, citizen processes and the
Boolean operators ”AND” and ”OR” to refine the results. (“local public administration” AND
“digital transformation”) OR (“Natural language processing” AND “chatbot”) OR “local public
administration”) AND (“digital transformation”)
4. Architecture of a chatbot for local public administration
The architecture of chatbots is based on a set of applications whose objective is to help, in our
use case, interactions between citizens and local public administrations.
The essential natural language processing components or applications that make up a chatbot
are part-of-speech taggers, parsers, and semantic resources from FreeLing [14], as well as
the Natural Language Toolkit (NLTK). Additionally, information retrieval tools and machine
learning classifiers based on language-specific tags.
Task flow: Initially, users access the chatbot via computers or mobile devices to ask their
questions. If the interaction involves procedures with the public administration, users must
register to receive updates on the status of their requests. The chatbot architecture is based on
the interaction of several key modules:
• (1) Speech manager
• (2) Information retrieval engine
• (3) Question generator
• (4) Classifier
Below we will explain these modules and their operation.
Following the scheme in figure 1, the operation would be as follows:
• Request or initial interaction: a citizen initiates a request to the chatbot through different
devices (for example, mobile phones, tablets, computers...), starting a conversation with
the chatbot of the local public administration.
• NLP module management: Citizen interactions are managed using various natural lan-
guage processing (NLP) modules. The Discourse Manager (DM) determines the user’s
intent and directs them to the appropriate module, such as Information Retrieval or
Question Classification, while simultaneously providing a help icon to assist the user.
• Interruption Handling: The chatbot allows citizens to interrupt the conversation at any
time using natural language or by clicking on a help icon. The Discourse Manager module
handles these interruptions to adjust the flow of interaction and resume the conversation
if necessary.
• Topic Classification: During interactions, the chatbot starts by asking a question and
after each user response the Topic Classification engine identifies and confirms the most
relevant topic based on the user input.
• Data storage and tracking: User conversations and interactions are recorded in files, such
as XML, with timestamps, to allow tracking of user progress and collection of usage
statistics.
5. How chatbots have helped the digital transformation of local
public administrations
In this section, we will examine three chatbots implemented in Lithuania, Vienna and Bonn,
which were documented in the referenced article.[4]. The first significant implementation of a
Figure 1: Chatbot architecture (own source)
chatbot for citizen services occurred in Latvia in 2018. The Latvian Business Registry launched
a chatbot called UNA to address frequently asked questions about business registration. UNA,
”Future support for entrepreneurs” in Latvian [4], is available both on the Commercial Registry’s
website and on its Facebook page, taking advantage of the Facebook messaging application, the
most popular social network in 2018, figure 2 . UNA was designed to address common questions
about business registration, including tax and fee queries, and to allow citizens to check the
status of their ongoing applications.
The UNA chatbot was implemented to solve the problem of managing the large volume of
phone calls that were produced and emails that were received, requesting the same information
from users. These repetitive tasks consumed a lot of officials’ time and frustrated citizens due
to the long wait times to receive the response. The introduction of this chatbot, UNA, led to
a significant reduction in the time that officials dedicated to these low-value tasks, favoring
dedication to higher-value tasks and improving the quality of the services provided to the user.
Overall, the use of artificial intelligence and natural language processing techniques simplified
administrative operations.
A second notable example of chatbot implementation in local public administration occurred
in Vienna in 2017 with the launch of WienBot [4]. This conversational agent was designed to
Figure 2: Digital in 2018: Social networks add 11 new users every second. (source https://www.expan-
sion.com/blogs/think-social/2018/02/02/digital-in-2018-las-redes-sociales-suman.html)
answer frequently asked questions from residents and improve access to information from the
city’s online services, which previously required browsing the municipal website, searching and
finding the requested information. WienBot was developed to simplify access to various city
services, such as public parking availability, opening hours... and won the World Summit Award
in 2017 for the best Government and Citizen Participation application. Although it provides
extensive information, WienBot does not facilitate service transactions, but instead directs users
to relevant pages for more details, this functionality was developed later.
Our last case is the implementation of GovBot in Bonn, Germany, in 2018 [4]. GovBot was
developed to help citizens with administrative tasks, such as requesting forms, checking opening
hours, and scheduling appointments with local public administration. The success of GovBot
was such that other German public administrations adopted this technological solution for their
processes. It is based on the use of machine learning and a comprehensive knowledge base,
which aims to reduce the workload of public personnel by handling repetitive queries and, in
turn, guiding users in completing forms. The implementation took into account the growing
trend of using smartphones to access municipal services, figure 3
6. Conclusions: How chatbots have helped the digital
transformation of local public administrations
In conclusion, this preliminary analysis demonstrates that chatbots using natural language
processing (NLP) tools provide substantial benefits to citizens and serve as critical drivers for the
digital transformation of local public administrations,[15], [9] and [4] highlighting the following
key points:
• Provide a new digital communication channel to the citizen.
Figure 3: Devices used to access the internet (source Meeker, M., & Wu, L. (2018). Internet trends 2018)
• Implementing, a chatbot cuts across the entire organization of a local public administration,
which requires training in digital capabilities in all municipal areas.
• Modernizing, requires the implementation of ICT infrastructure, data processing cen-
ters, software tools, hardware systems... which requires having more budget for the IT
departments of local public administrations.
• Change, change management, serves as a lever to promote change management towards
a digital local public administration, it will be done in a much more participatory way
and with more involvement on the part of municipal workers.
• Innovate, the application of natural language processing technologies will be accepted by
IT departments that will require further developments in NLP technologies to apply in
their city council processes, both internal and external.
• Data-Driven Analysis, The analysis of data derived from chatbot interactions can inform
policy decisions, enhancing satisfaction among citizens and municipal workers alike. This,
in turn, encourages further investment in digital transformation projects and progress
towards Smart City initiatives.
References
[1] A. Rafael, L. Rivera, M. Jose, R. Jaume, J. H. Meseguer, Asis study: Analysis and observation
of the digital status and situation of the municipalities in the province of alicante, in:
CENID2021, 2021.
[2] B. Larsen, M. Milakovich, Citizen relationship management and e-government, in:
International Conference on Electronic Government, Springer, 2005, pp. 57–68.
[3] J. Bharadiya, A comprehensive survey of deep learning techniques natural language
processing, European Journal of Technology 7 (2023) 58–66.
[4] C. Van Noordt, G. Misuraca, New wine in old bottles: Chatbots in government: Exploring
the transformative impact of chatbots in public service delivery, in: Electronic Participation:
11th IFIP WG 8.5 International Conference, ePart 2019, San Benedetto Del Tronto, Italy,
September 2–4, 2019, Proceedings 11, Springer, 2019, pp. 49–59.
[5] B. J. Copeland, The turing test, Minds and Machines 10 (2000) 519–539.
[6] J. Weizenbaum, Eliza—a computer program for the study of natural language communica-
tion between man and machine, Communications of the ACM 9 (1966) 36–45.
[7] K. M. Colby, S. Weber, F. D. Hilf, Artificial paranoia, Artificial intelligence 2 (1971) 1–25.
[8] L. C. Klopfenstein, S. Delpriori, S. Malatini, A. Bogliolo, The rise of bots: A survey of
conversational interfaces, patterns, and paradigms, in: Proceedings of the 2017 conference
on designing interactive systems, 2017, pp. 555–565.
[9] A. Ferrández, R. Lavigne-Cerván, J. Peral, I. Navarro-Soria, Á. Lloret, D. Gil, C. Rocamora,
Cuentosie: can a chatbot about “tales with a message” help to teach emotional intelligence?,
PeerJ Computer Science 10 (2024) e1866.
[10] J. Vlachỳ, Citation histories of scientific publications. the data sources, Scientometrics 7
(1985) 505–528.
[11] H. D. White, Pennants for garfield: bibliometrics and document retrieval, Scientometrics
114 (2018) 757–778.
[12] C. Karakus, Y. Sun, S. Diggavi, W. Yin, Redundancy techniques for straggler mitigation in
distributed optimization and learning, Journal of Machine Learning Research 20 (2019)
1–47.
[13] I. Zupic, T. Čater, Bibliometric methods in management and organization, Organizational
research methods 18 (2015) 429–472.
[14] L. Padró, E. Stanilovsky, Freeling 3.0: Towards wider multilinguality, in: LREC2012, 2012.
[15] A. Miklosik, N. Evans, A. M. A. Qureshi, The use of chatbots in digital business
transformation: A systematic literature review, IEEE Access 9 (2021) 106530–106539.
doi:10.1109/ACCESS.2021.3100885 .