The actuality of the problem how to design and to develop advanced solutions in the domain of electronic dictionary systems (EDS) is motivated, in particular, with respect to the such linguistic phenomena as lexical parallels in Ukrainian and German languages. The briefly overview of some existing domestic and international EDS solutions is done, and some their specific features and shortcoming are analyzed. The own framework for EDS development is proposed, which supposes a data-centered system design approach, where the specific noSQL data model for the dictionary entry's representation with its existing lexical parallels is presented. The microservices system software architecture is elaborated in order to increase system's availability and scalability. The special attention is paid to the problem how to design and to implement the advanced UI solutions for the target Webbased EDS. For its quality assessment the enhanced version of the Goals, Objects, Methods and Selection rules (GOMS) model was presented and applied successfully (titled as the GOMS+), especially the final quality estimation shows the increasing about 12,35%. The obtained test results shown the acceptable functionality assessment of the elaborated Web-based EDS, and it allows to make the positive conclusions about the proposed approach, and to define the further steps to be done in this research.
functionality of these software solutions, in particular, taking into account the phenomena of lexical
parallels in different languages, and these problems are discussed intensively in such works as, e.g., [
! АБІТУРІЄНТ, -а, ч: той, хто вступає до вищого або середнього спеціального навчального закладу: заява абітурієнта - Bewerber(in) um die Zulassung zu einem Studium. АБСОЛЮТН/ИЙ, -а, -е: 1. безвідносний, узятий поза зв'язком, без порівняння з чим-н.; безумовний: абсолютна маячня; 2. цілковитий, повний; необмежений; досконалий, довершений: абсолютна монархія; 3. матем. про величину без знаків плюс або мінус: абсолютна величина; 4*. цілковито чистий, без домішок: абсолютний спирт - rein.
⁕ ⁕ ⁕ ⁕ ⁕
! Abiturient, m (-en, -en): випускник середньої школи: die Abiturienten feiern den Schulabschluss. absolut, adj: 1.= абсолютний 1: absolute Höhe; 2.= абсолютний 2: absolute Monarchie, absolutes Halteverbot; 3.= абсолютний 3: absoluter Wert (einer Zahl).
More detailed the linguistic issues of this approach are discussed in [
The rest of our paper is structured in the following way: Section 2 provides the short overview of some existing EDS; Section 3 presents the main project artifacts developed: the data model to store dictionary entries with lexical parallels, the functional requirements specifications and microservice system architecture, and the intelligent user interface layout and the quantitative metrics to evaluate its quality; Section 4 shous the method to system testing, and discusses the obtained results. The paper concludes with Section 5 giving the brief summary and outlook on future work, Section 6 includes our acknowledgements, and Section 7 provides the list of the information sources used in this research.
As it is mentioned above, Web-based EDS already are used intensively, and inn order to implement own solution with an advanced functionality and nigher quality, the existing ones were investigated to define common functionality, some advantages and shortcomings, to be taken into account in their further design and implementation.
In Figure 2 the main interface of “Ukrainian academic dictionary” website is shown [5], and it is to mention that it was the digital implementation of the first explanatory dictionary of the Ukrainian language.
The user interface (UI) of this dictionary contains the following elements: • • • • •
Menu with buttons for sign in and sign up.
Input field for a word, hints appear during entering a word.
Checkbox to search entered text in word definitions.
Information about dictionary.
List of last viewed words.
The next observed dictionary “СЛОВНИК.ua” [6] is based on the same paper dictionary, but it has additionally 21000 words meanings that were added by the dictionary users, and it contains the official “passport” (CMU 2010) for online transliteration (see in Figure 3 in Ukrainian language).
This UI consists of the next elements: • Menu with links to other project services and link to sign in page. • Input field for a word. • Alphabet to search dictionary articles by word’s first letter. • Other service components.
In Figure 4 the home page of “УКРЛІТ.ORG” website is shown. It includes 3 separate dictionaries: Ukrainian academic dictionary, Grinchenko’s dictionary and Zhaivoronka’s dictionary [7]. All their UI include the similar elements: menu with links to another project services, input field for a word, dropdown with dictionaries, alphabet to search word articles by first two letters, dictionary switching.
Additionally, three recognized international EDS were analyzed briefly. The first one is the “DICTIONARY.COM” [8], and its UI consists of the next elements (see in Figure 5): • Menu with links to other project services and link for authentication. • Dropdown with dictionaries. • Input field for a word, hints appear during entering a word. • Trended words. • Word of the day.
• Other services components.
In Figure 6 the home page of “Cambridge Dictionary” website is shown [9]. It has meaning dictionary for English language and translation dictionary that allows translation to multiple languages. The UI consists of the following elements: menu with the links to other project services, the links to the selected social networks, the links for authentication option, the dropdown menu with languages, the input field for a word, hints appear during entering a word, dictionaries select button, and the list of dictionaries.
The “Macmillan Dictionary” website [10] is shown in Figure 7. Its UI consist of the next elements: menu with links to another project services, input field for a word, hints appear during entering a word, links to other project services.
Based on performed investigate we can make conclusion about the set of typical UI - options that present the main functionality of the observed Web-based EDS, namely: • menu bar that contains some controls for user’s authentication and navigation, • input fields to customizing search for dictionary articles, • another services components (if there are any), e.g.controls for dictionary selection, alphabet to search articles by first letter(s) of word.
At the same time as some common shortcomings of the EDS / UI presented above the following points can be mentioned: 1. almost all reviewed solutions provide ability to search articles by entering full word or its part into a search input fields, but only some of them provide ability to list articles by word’s first letter. This functionality should allow users to explore dictionary. 2. all reviewed solutions provide simple search by exact match of substring. It means that if word has alternation of letters, like in Ukrainian language, or different form of word formatted by changing it parts (prefixes, suffixes, endings, etc.) then it would be harder to find a target article.
The proposed EDS solution with the lexical parallels (see the example in Figure 1) is supposed to provide the search functionality that takes into account the different forms of an appropriate word.
It is necessary to define the appropriate data model, that will be processed by software solution, before
start to design graphical UI and implement its functionality. The dictionary article’s structure with linguistic
parallel discussed in [
The equivalent words and definitions are represented by class EquivalentWordDefinitions that contains information about comment to word or definition (field precedingComment), and word or definition itself (field definition).
Basing on the UI analysis in Section 2, the functional requirements for a target EDS was elaborated, and they are represented as the UML use case diagram [11] in Figure 9.
There were requirements to consider system for future extension and decouple different pieces of functionality from each other. So, it was decided to use microservice architecture to meet these requirements. The proposed system architecture is shown as UML component diagram [11] in Figure 10.
This diagram consists of the following software components: • Client, in this case client is web application created using React.js library [12], however since software system exposes REST API [13] this means that other web application or mobile application, or other software system could be a client. • API Gateway, implementation of design pattern of the same name, provides one entry point to the system, communication with Client and other system components implemented using HTTP. • Service Registry, implementation of design pattern of the same name as well, registers address of active system components (IP address and port) and provides this data to other components for further communication. • Word Article service, manages dictionary articles, provides functionality for unauthenticated users and system administrators. Request authorization is performed by using self-encoded JSON Web Token (JWT) token, that should be passed in Authorization HTTP header, when request is received, and token extracted then it’s sent to User service for validation. • User service, stores information about administrator, root administrators and issued JWT tokens, provides REST API for registering, removing administrators, password updating, registers and removes root administrators, also provides REST API for issuing, updating and validating JWT tokens, after new administrator registration send data about email, that should be sent to the Queue component using AMQP [14] protocol. • Queue, message queue, that is used to store information about emails that should be sent. • Email sender service, service for sending emails, retrieves data about emails from Queue using AMQP protocol.
As the data storage model for dictionary articles, it was decided to use document-oriented model with MongoDB, since all operations related to articles management operates with whole article object. There is no functionality that updates only some part of article in one place, and it should be updated in other places. Also, there are no complicated queries that would be effective to implement with relation data model.
Considering the listed above shortcomings 1 – 2 of the reviewed existing EDS (see in Section 2), it was decided to add alphabet to the UI of developed solution to implement search by first letter of word, and the text index was created in the system database.
The UI defines how interface will look like. It defines what colors would be used, font styles etc. Graphical interface is a main way to interact with modern information systems. If graphical interface not changed, you could change interaction object without changing interaction principle with it. If user know how to use one system, he could easily learn how to use another system with similar interface. Combination of convenient interface with well-structured navigation increase users’ impressions. The main task during any UI design is to simplify its functions, so user can get more results with less effort [15]. The Web-UI was implemented using JavaScript library “React.js” and component library “Material-UI”. Dictionary article creation form is shown in Figure 11.
There are different ways to evaluate designed interface, one of them is to invite stakeholders, ask them to use interface to complete list of tasks. Users’ interaction with interface is recorded and then analyzed. If you don’t have ability to perform such tests, you could use quantitative methods. Such methods allow to replace contentious questions about interface evaluation to simple calculations.
There are some quantitative methods to evaluate interface: the Hick’s Law, the Fitts’s law, and the model of Goals, Objects, Methods and Selection rules (GOMS) [16].
The Fitts’s law evaluates efficiency based on the size of the element and the distance from the cursor to its center and shown by the formula (1)
D T = a + b log2 + 1 (1)
S where is a mean time spent to perform an action, is a mean time of starting/ending movement, – is a value that depends on the typical speed, is a distance from the starting point to the center of the target, is a width of the target measured along the axis of motion.
The Fitts’s law is usually accompanied by Hick’s law, which states that the reaction time when choosing one of several alternatives depends on the number of options. Hick’s law is shown by the formula (2) T = a + b log2 (n +1) (2) where – total reaction time, – delay before the task, – individual coefficient of decision-making speed. – the number of alternative elements from which to choose one.
The GOMS model allows to evaluate UI using time intervals for specific simple actions that have been determined experimentally. GOMS has the next list of actions with its time intervals [16]: • Key pressing, denoted as “K” = 0.2 s. • Cursor moving to a specific position, denoted as “P” = 1.1 s. • Moving hand from mouse to keyboard or vice versa, denoted as “H” = 0.4 s. • Mental decision of next action, denoted as “M” = 1.35 s. • Waiting response from system, denoted as “R” (this action depends on a concrete computer system and it should not be used in the calculation).
However, such actions could be extended with respect to the specifics features of a target UI. In our case we are dealing with search results that could contain quite big amount of dictionary articles ordered alphabetically. To navigate throw it is common practice to use a scrolling function. That’s why the standard GOMS model presented above could be extended by the two additional actions • Scrolling time that is required for navigation to the target element of the UI, denoted as “S” and in average = 1,5 s • Object focusing time that is required to catch a necessary information from the UI, denoted as “F” and in average = 0.3s
Thus, the new enhanced GOMS+ model allows to provide a more accurate assessment of the EDS / UI quality.
It was decided to use the GOMS model to evaluate the implemented UI, since it provides all necessary values and doesn’t require to define some values in experimental way like Fitts’s and Hick’s laws (see the formulas (1)-(2) above). We will evaluate UI by performing search of word articles that are represents lexical parallel in our eUGDLP facility. Using GOMS rules we will receive next navigation sequence: “HPPKKKKKKKPPHPKPMSF”. This means next: first “H” letter means taking computer mouse in the hand; first “P” letter means moving cursor to specific position; first “K” letter means click with mouse; second “H” letter means moving hand from mouse to keyboard; eight “K” letters mean input of eight characters of the word; third “H” letter means moving hand from keyboard to mouse; second “P” letter means moving cursor; last “K” letter means mouse click.
So, we have done the next actions: selected input field using mouse, entered eight characters and pressed search button using mouse. Next, we need to apply the GOMS rules to evaluate navigation sequence. After applying all GOMS+ model rules we receive next sequence “8K6P2H1M1S1F”. Next, we need to map each letter to its time interval and sum all numbers and we will get 12.15 s. This is because in other interfaces you can submit your request by pressing “Enter” key, and in the developed solution you need to press the “Search” button with the mouse. The best result was achieved not only by submitting requests by “Enter” key, but also by setting focus on input field after page is upload, which allows to omit mouse movements to select the input field. Taking into account all the mentioned above, it was decided to implement an autofocus feature on an input field, after a web page is already loaded. Having the such improvements with the new GOMS+ rules, the generated activity sequence would be the next: “6K5P2H1M1S1F”, which is equal to 10.65 s. In the same way we have estimated the UI in the existing EDS described in Section 2. The obtained results are shown in Table 1.
As it can be seen from the Table 1, the implemented eUGDLP solution has the best evaluated value since it has specific data structure for lexical parallels that allows users to search more efficiently comparing with other solutions. The eUGDLP provides functionality that is not available fast in all considered analogs, like ability to search by input string, or search by alphabet letters that is absent for several analogs. Besides that, analyzing the data presented in the Table 1, it is to mention, that the improvements done in the target UI design version has the better estimation GOMS+ value in compare with the initial one, which can be calculated as follows (approx.): (1 – 10,65/12.15) * 100% = 12,35%.
Finally, based on the results shown on Table 1, we can state, that the developed eUGDLP solution provides the best usability for lexical parallels than the considered analogs. That’s for the front-end functionality, and for the back-end solutions we developed the several microservices that communicate using REST API [13] and AMQP [14], in order to develop software services independently, and to increase system’s availability and scalability.
In this paper we have motivated an actuality of the problem how to design and to implement the advanced solutions in the domain of electronic dictionary systems (EDS) creation, in particular, with respect to the such linguistic phenomena as lexical parallels in Ukrainian and German languages. The performed overview of some existing domestic and international EDS solutions has shown their specific features and shortcoming as well, that allows us to propose the own framework for EDS development. Within this one the data-centered system design approach was chosen, where the specific noSQL data model was elaborated for the dictionary entry’s representation with its lexical parallels. The microservices system software architecture was proposed in order to increase system’s availability and scalability. Furthermore, the special attention was paid to the problem to design and to implement the advanced UI solutions for the target EDS, and for its quality assessment the enhanced version of the so-called Goals, Objects, Methods and Selection rules (GOMS) model was proposed and applied successfully (titled as the GOMS+). The obtained test results shown the acceptable quality of the elaborated Web-based EDS, and it allows to make the positive conclusions about the proposed approach.
Further work in this research is supposed to compare the quality assessment results with usage of another UI models, to provide more representative amount and complexity of testing data, and to develop a comprehensive methodology for a performance evaluating of a prospective EDS with respect to the phenomena of lexical parallels in other languages (for instance, English and Polish) for the specific professional groups of their end users.
The authors would like to thank prof. Volodymyr Dubichynskyi (University of Warsaw, Poland) for his useful advises during the system requirements analysis for the target eUGDLP solution, and bachelor student Bach Viet Yen (V.N. Karazin Kharkiv National University, Ukraine) for her active participation on the software development in this project. [5] Ukrainian academic dictionary, 2018. URL: http://sum.in.ua/. [6] Slovnyk.ua Portal of Ukrainian language and culture, 2022. URL: https://slovnyk.ua/. [7] Ukrlit.org Public electronic dictionary of the Ukrainian language, 2021. URL: http://ukrlit.org/slovnyk. [8] Dictionary.com Meanings and Definitions of Words at Dictionary.com, 2022. URL: https://www.dictionary.com/. [9] Cambridge Dictionary Make your words meaningful, 2022. URL: https://dictionary.cambridge.org/ [10] Macmillan Dictionary Free English Dictionary and Thesaurus, 2022. URL: https://www.macmillandictionary.com/ [11] I. Sommerville, Software Engineering, 10th Edition, Addison Wesley (2015) [12] React – A JavaScript library for building user interfaces, 2022. URL: https://reactjs.org/ [13] REST API Tutorial, 2022. URL: https://restfulapi.net/ [14] AMQP Tutorial, 2022. URL: https://www.wallarm.com/what/what-is-amqp [15] G. Krishna, The Best Interface Is No Interface: The Simple Path to Brilliant Technology (Voices That
Matter), 1st. ed., New Riders Pub (2015). [16] J. Marsh, UX for Beginners: a Crash Course in 100 Short Lessons, 1st. Edition, O'Reilly Media (2016).