Introduction 1.Motivation

Nowadays, the rapid evolution of computer technologies changes scientific approaches to modern issues and provides ways for creating novel and enhancing existing research methods. It is especially considerable for applied science wherein computational technique implementation accelerates the complex applied problem solution requiring large volumes of calculations. Social sciences, which have historical relations with philosophy, have a peculiar wide range of research methods. However, social sciences are also in a transformation state and increasingly using computing technology for research problem solving, which has led to the emergence of computational social sciences. Initially, computational social sciences were associated with agent-based modeling for the simulation of the behavior of an individual or social group under certain conditions. Nevertheless, the Internet spreading, social networks and online platforms popularity increasing within the growth of numbers of Internet users provoked a new large stream of digital data which has become a valuable source of information for social sciences researchers and has led to the expansion of the concept of "computational social science" [1]. Although earlier researchers have argued that digital data analysis-based computational social science has developed slowly [2], more recent studies show that in recent times increased the interest of social sciences scholars in using computational techniques for research [1].

Related Work

The use of information technologies for political science research is not as new as it may seem, and began in the second half of the 20th century. The first experiments using computers were aimed at trying to predict election results. Typically, all studies were based on the use of agent-based modeling in different form, using classical theories of political interactions [3]. The further evolution of the Internet, increasing the power and availability of computer technology provided new research fields for political scientists and tools for expanding methodology. Political science methodology expansion led to the fact that in addition to agent-based modeling, political scientists more often were beginning to use methods related to big data analysis [4].

Scholars in social and political science address "big data" as a broad concept that includes any digital elements left by users or organizations on the Internet that can be read by information technologies [5]. In the computer science field, big data is usually associated with so-called "5Vs" used to describe its characteristics (value, variety, velocity, veracity, and volume) [6].

Today, for social sciences fields and particularly for political science, when working with big data, it is important to use web scraping tools, which is the automatic extraction of data from websites for further analytical processing [7]. For political science, this approach can be valuable for defining features of how political parties or government agencies use their websites [5] and for researching local politics through the mining and analysis of unstructured data from the websites of local government institutions [8]. Some researchers conclude and we agree that the use of automated data mining in social sciences opens a new way for cooperation between social and computer science researchers [6].

It should be stressed that political science must be in the continuous dynamic movement condition and must permanently react to political changes in the modern world. Today, the policies of many countries aim at the formation of an inclusive information society, which includes widespread digital transformation. Governments create and support open e-government web portals, which aim to facilitate citizens' access to government information and improve the process of providing government services to citizens. However, it should be noted that the politics of different states differ from each other and may have various accents. Under such circumstances, it may be interesting how the policies of different states influence their e-government web portals.

Thus, we propose an information technology that can help researchers to explore services provided on government web portals, which can be useful for further analysis of different countries' policies. It aims to improve e-government web portal evaluation processes by using web harvesting techniques.

Therefore, this study is expected to answer the following research questions:  What reference model can be used to evaluate the e-government web portal?  What algorithms can be used to process and harvest the desired e-government web portal data?  How can the extracted data be quantitatively evaluated to compare the policies of different countries and define the prevailing citizen services?

Materials and Methods

E-Government Web Portal Services Model

Let us formally describe the set of services that the e-government web portal is expected to provide:

𝑒𝐺𝑆 = {𝑒𝐺𝑆 1 , 𝑒𝐺𝑆 2 , … , 𝑒𝐺𝑆 𝑛 }. (1)

Here 𝑛 is the number of services 𝑒𝐺𝑆 1 , 𝑒𝐺𝑆 2 , … , 𝑒𝐺𝑆 𝑛 the e-government web portal is expected to provide, 𝑖 = 1, 𝑛.

Moreover, for each of the e-government web portal services 𝑒𝐺𝑆 𝑖 , 𝑖 = 1, 𝑛 we propose to define the set of keywords 𝑊 𝑖 , 𝑖 = 1, 𝑛 which completely describes the mentioned service:

𝛿: 𝑒𝐺𝑆 𝑖 → 𝑊 𝑖 = {𝑤 𝑖1 , 𝑤 𝑖2 , … , 𝑤 𝑖𝑚 𝑖 }. (2)

Here 𝑚 𝑖 is the number of synonymic keywords 𝑤 𝑖1 , 𝑤 𝑖2 , … , 𝑤 𝑖𝑚 𝑖 in 𝑊 𝑖 , 𝑖 = 1, 𝑛 defined for 𝑖-th egovernment web portal service 𝑒𝐺𝑆 𝑖 , 𝑗 = 1, 𝑚 𝑖 .

Hence, the formal definition of E-Government Web Portal Services (EGWPS) model can be formulated as given below: 𝑒𝐺𝑊𝑃𝑆 = 〈𝑒𝐺𝑆, 𝛿, 𝑊〉.

Here 𝑊 is the set of keyword sets mapped to each of the e-government web portal services, 𝑊 = {𝑊 1 , 𝑊 2 , … , 𝑊 𝑛 }. Let us graphically illustrate in Fig. 1 the proposed e-government web portal services model. Fig. 1 demonstrates the set of expected services and their keywords used to detect such services on the e-government web portals under assessment. Using the proposed EGWPS model (Fig. 1), we propose to find the "distance" between the e-government web portal under assessment and the so-called "perfect" e-government web portal (in terms of its contents) described by this model. The e-government services data extracted from HTML (Hyper Text Markup Language) pages of a corresponding web portal is represented as the set of HTML hyperlink tags [10]:

Web Data Extraction Algorithm𝐻 = {ℎ 1 , ℎ 2 , … , ℎ 𝑝 }. (4)

Here 𝑝 is the number of hyperlinks ℎ 1 , ℎ 2 , … , ℎ 𝑝 located on e-government web portal HTML pages, 𝑘 = 1, 𝑝.

Each hyperlink tag includes the text and URL (Unified Resource Locator) as it is demonstrated in the meta-model diagram in Fig. 2.

The set of e-government web portal services 𝑒𝐺𝑆 extracted from the respective HTML pages is represented as follows: 𝑆 = {𝑆 1 , 𝑆 2 , … , 𝑆 𝑛 }.

(5) Each service 𝑆 𝑖 , 𝑖 = 1, 𝑛 is expected to be digitally implemented by one or multiple hyperlinks located on the e-government web portal pages:

𝑆 𝑖 = 𝐻 𝑆 𝑖 ⊆ 𝐻. (6) Here 𝐻 𝑆 𝑖 is the sub-set of hyperlinks extracted from the e-government web portal that implement 𝑖th service detected in the e-government web portal 𝑆 𝑖 , 𝑖 = 1, 𝑛. Therefore, to detect services provided by the e-government web portal using the proposed EGWPS model (Fig. 1) and the meta-model (Fig. 2), the following algorithm should be used:

Given: set of extracted e-government web portal hyperlinks 𝐻 EGWPS model 〈𝑒𝐺𝑆, 𝛿, 𝑊〉 empty set of detected e-government web portal services 𝑆 for each 𝑒𝐺𝑆 𝑖 in 𝑒𝐺𝑆:

for each ℎ 𝑘 in 𝐻: 𝑊 𝑖 = 𝛿(𝑒𝐺𝑆 𝑖 ) for each 𝑤 𝑖𝑗 in 𝑊 𝑖 : if 𝑤 𝑖𝑗 is a substring of ℎ 𝑘 text:

𝑆 𝑖 ← ℎ 𝑘 end end end

The input set 𝐻 of the e-government web portal hyperlinks can be extracted using web scraping tools in Python or other programming languages.

The output set 𝑆 basically represents the instances of Service class defined in the proposed metamodel (Fig. 2). Furthermore, each service 𝑆 𝑖 , 𝑖 = 1, 𝑛 has multiple hyperlinks that belong to 𝐻.

Fig. 3 graphically illustrates the proposed algorithm using the UML activity diagram [9].

E-Government Portal Assessment Metrics

Finally, we propose the following metrics to assess the e-government web portal in terms of detected services. The following metric allows to find the number of services detected in the e-government web portal under assessment: 𝑆𝐷 = |{𝑆 𝑖 ∈ 𝑆, 𝑆 𝑖 ≠ ∅}|.

(7) The following metric allows to find the "service richness" calculated as the relative number of services of the e-government web portal under assessment in comparison to the reference EGWPS model (Fig. 1) [11]:

𝑆𝑅 = 1 𝑛 𝑆𝐷.(8)

The following metric allows to find the "relative cardinality" of the particular service calculated as the relative number of hyperlinks used to implement the 𝑖-th service detected in the e-government web portal under assessment in comparison to the maximum possible number of hyperlinks used in the same web portal for a certain service [12]:

𝑆𝐶 𝑖 = { 1 max 𝑖=1,𝑛 |𝑆 𝑖 | |𝑆 𝑖 |, max 𝑖=1,𝑛 |𝑆 𝑖 | > 0 0, max 𝑖=1,𝑛 |𝑆 𝑖 | = 0(9)

The following metric allows to find the total "service balance" to assess the balance of hyperlinks related to services detected in the e-government web portal under assessment:

𝑆𝐵 = 1 𝑛 ∑ 𝑆𝐶 𝑖 𝑛 𝑖=1 . (10)

Using the following algorithm, it is possible to evaluate an e-government web portal against the EGWPS model. Thus, as a reference model, we can use the experience and best practices of the most advanced e-government web portals, define the set of services 𝑒𝐺𝑆 a portal is expected to provide, and the keywords 𝑊 to detect such services in corresponding HTML web pages.

Fig. 4 graphically illustrates the proposed algorithm using the UML activity diagram [9].

Information Technology for E-Government Portal Assessment

Finally, the information technology for e-government web portal assessment can be formally described using the following tuple: 𝑒𝐺𝑊𝑃𝐴𝐼𝑇 = 〈𝑒𝐺𝑊𝑃𝑆, 𝑆, 𝐴𝑀〉.

(11) Here 𝐴𝑀 is the algorithmic model, which includes the proposed algorithms (Fig. 3 and Fig. 4) [13]:

𝐴𝑀 = (𝐴 = {𝐴 1 , 𝐴 2 }, 𝑅 ⊂ 𝐴 × 𝐴). (12)

Here 𝐴 is the set of algorithms, where 𝐴 1 is the data extraction algorithm and 𝐴 2 is the evaluation algorithm; 𝑅 describes the interconnections between the proposed algorithms when used to assess an e-government web portal. Selected e-government web portals will be analyzed using the proposed information technology implemented using Python, in-build packages, and third-party libraries:

 "urllib" -used the "request" module to open and work with URLs [14];  "re"for regular expressions operations to parse web pages [15];  "json"to save results as JSON (JavaScript Object Notation) [16];  "bs4" -used the "Beautiful Soup" library to scrape information from web pages of citizen portals [17].

Fig. 5 below demonstrates the Data Flow Diagram (DFD) [18] of the data processing workflow implemented by the proposed information technology.

Figure 5: Data processing workflow for e-government portal assessment

According to Fig. 5, obtained results are displayed using Power BIa high-performance Business Intelligence (BI) tool for advanced data visualization and data-driven decision making [19].

Fig. 5 illustrates the hands-on usage of the proposed information technology.

Results and Discussion

E-Government Portal Data Extraction: Citizens Viewpoint

Let us form the reference EGWPS model considering the Integrated Architecture Framework for E-Government (IAFEG) shown in Fig. 6 [20]. In this study, we focus on the "Social Sub-system" layer of this framework, in particular -on its "Citizens" perspective [20]. The citizens' viewpoint according to IAFEG [20] includes the following services (or topics) expected from an e-government web-portal:

 taxation;  education;  health;  immigration;  employment. The IAFEG-based set of services the e-government web portals are expected to provide 𝑒𝐺𝑆 (from the citizens' viewpoint of the IAFEG "Social Sub-system" layer) and the keywords 𝑊 used to describe each of the services on HTML web pages are given in Table 1.

According to the "UN E-Government Knowledgebase" and its UN (United Nations) E-Government Survey 2022, top five countries with the highest E-Government Development Index are Denmark (0.9717), Finland (0.9533), Republic of Korea (0.9529), New Zealand (0.9432), and Iceland (0.9410).

Denmark citizen portal "Life in Denmark.dk" is shown in Fig.

7 [21]. The "Life in Denmark.dk" portal offers topics related to immigration, housing, working, family and children, money and taxation, education, healthcare, travel and transportation, pension, rights, leisure and networking, as well as stand-alone digital services (Fig. 7) [21]. Table 2 shows hyperlinks detected on the "Life in Denmark.dk" citizen portal according to IAFEG taxation, education, health, immigration, and employment services [20]. The "Suomi.fi" portal offers similar topics to "Life in Denmark.dk". These topics are connected to family, social security, healthcare, education, working, housing, rights and obligations, finances and taxation, moving and travelling (Fig. 8) [22]. Table 3 shows hyperlinks detected on the "Suomi.fi" citizen portal according to IAFEG [20]. Finland citizen portal "Suomi.fi" is shown in Fig. 8 [22]. In this study we focus on the top two countries (Denmark and Finland) and their citizen portals. First of all, their impact according to the E-Government Development Index is greater than 0.95. Another country, which aspirations were highly estimated is Republic of Korea, however, we failed to access its English web portal version. Therefore, we obtained the following e-government web portals evaluation results (Table 4). Here 𝑆𝐶 𝑖 , 𝑖 = 1,5 describe taxation, education, health, immigration, and employment services.

Here ∑ 𝐹𝑖𝑡𝑛𝑒𝑠𝑠 = 𝑇𝑟𝑢𝑒 is the number hyperlinks estimated as correctly categorized against IAFEG services [20], and ∑ 𝐹𝑖𝑡𝑛𝑒𝑠𝑠 = 𝐹𝑎𝑙𝑠𝑒 is vice versa (see Table 2 -3). Hence, the accuracy of proposed information technology for e-government web portal assessment is 0.80 for Denmark and 0.88 for Finland. However, the total accuracy for both estimated citizen web portals "Life in Denmark.dk" and "Suomi.fi" is 0.83.

Therefore, the proposed information technology allows to obtain accurate (of 83%) e-government web portal assessment results and can be suggested scholars in social political science fields.

Conclusion and Future Work

In this paper we proposed the information technology for e-government portal assessment based on web data extraction techniques. The study aims to improve the processes of e-government web portal assessment by using web harvesting and data analysis approaches. Therefore, we developed algorithms to extract and assess e-government web portals using the proposed E-Government Web Portal Services reference model and evaluation metrics. The software implementation of the proposed technology is based on Python programming language and Power BI data visualization tool. Such a tool allows nontechnical users, i.e. social or political science scholars, to configure the desired references models and automatically assess e-government web-portals as part of their studies with the accuracy of 83%.

The following conclusions can be made after the obtained results analysis:  this approach has a room to identify the differences between e-government web portals and the services they provide;

 consequently, for researchers who study and compare the state of information society formation in different countries, in terms of digital services provision, conducting such experiments can be a useful complement to other data-driven methods;

 the need for interdisciplinary cooperation between social and computer science is increasing and such interdisciplinary studies can benefit both domains with new methods and solutions.

In the future we plan to elaborate metrics proposed to evaluate e-government web portals, as well as conduct a large-scale study, with more government portals and more careful sampling, to identify and study their differences. From the information technology viewpoint, such experiments require advanced techniques to be applied, such as data warehousing, data mining, and data visualization.