Wordinary: A Software Tool for Teaching Greek Word Families to Elementary School Students

Learning to spell is a rather challenging task. In order to do so, teaching approaches focus on phoneme-grapheme correspondences in the early school years. Orthographic and morphological knowledge is usually targeted later on along with accumulating reading experience [18].

Recent research, however, has questioned the aforementioned teaching tradition. Bowers and Bowers [4] posit that phonology, morphology as well as etymology should be taught jointly from the beginning of literacy instruction. Their argument holds particularly true in morphologically rich languages, such as Greek.

Most of the roots of the Greek words (etymology) are traced back to Ancient Greek or Latin languages. Although a highly transparent orthography in the feedforward direction (from orthography to phonology with 95.1% overall consistency at the grapheme-phoneme level), consistency level drops at 80.3% in the feedback direction (from phonology to orthography) [17]. One significant source of opacity stems from multiple representations for certain vowels. For example, the vowel phoneme /i/ has six different spellings (η, υ, ι, ɛι, οι, υι). Hence, children may spell phonologically acceptable words but struggle with successful choice of vowel graphemes [7].

One way to enhance children's spelling ability is a shift of attention towards word roots through learning of word families. Words from the same family share common vowel graphemes hence reducing the mnemonic load of discrete orthographic representations. For instance, the words "οικογένɛια" /ikojenia/ (family), "κάτοικος" /katikos/ (resident), "πολυκατοικία" /polikaticia/ (apartment block) derive from the root word "οίκος" /ikos/ (house) where the vowel phoneme /i/ is spelled with the vowel grapheme "οι" (not η, υ, etc.). As perceived, children's realization of common roots within multiple words might make spelling attempts less confusing and more accurate.

In accordance with this notion, a study [8] found that an approach which focuses on the interconnections between phonology, morphology and etymology yielded promising results regarding the reading and spelling skills of persistently poor Grade 3 readers. This approach called the Structured Word Inquiry (SWI) [4] trains children to attend to grapho-morphemic patterns, generate and test hypotheses about the spelling-meaning connection of words when there is a conflict of grapheme choice, such as "Why is there a ‘οι’ /i/ and not a ‘ι’ or ‘η’ in the word ‘κάτοικος’ /katikos/ (resident)?" in the aforementioned example.

Although ongoing research informs educational practice on the effectiveness of various approaches, practical implications cannot be disregarded. Educators are confronted with multiple challenges throughout the school curriculum. Thus, it is of outmost importance to be equipped with the appropriate technological tools that would motivate the implementation of alternative methodologies to spelling instruction. That is the purpose of the current research.

This paper presents Wordinary, an interactive desktop application that supports teachers in their efforts of teaching Greek word families to elementary school students. The application presented in this paper is the first of its kind that supports the Greek language. Given the connection between learning word families and developing spelling skills, the importance of such tools to support the teacher becomes apparent.

This section presents software tools that have a similar purpose to that of the application presented in this paper. These tools provide functionality to search for the family, derivatives or even compound words of a specific word entered by the user. No tool has been found that supports searching for the word family of a user-provided Greek word in general, and a word that is included in the books of the Greek elementary school in particular.

2.1 Word family framework

The British Council is an organization of the United Kingdom that, among other things, aims to aid people around the world in learning English by providing them helpful tools. One of these tools is the Word Family Framework¹ (WFF), an online application where people can find the families of words for any level of the English language, classified according to the scale of the Common European Framework of Reference for Languages (CEFR) [6].

Users can either select a specific word to see its related words in the various CEFR levels, or a specific level to see the words, and their families, included in that level (Figure 1). WWF consists of 22000 words and for each of them information is included about their meaning and what part of speech they are. This tool only supports the English language.

Screenshot of the Word Family Framework website displaying a search for words in the B2 level of the CEFR scale.

2.2 OWID elexiko

OWID² is a portal for scientific, corpus-based lexicography of German developed by the Leibniz Institute for the German Language (IDS). Its website contains multiple tools that support the learning and understanding of the German language.

One of these tools is the OWID elexiko³ (Figure 2). This tool provides functionality for a user to search for compound words or derivatives of a particular word. A user can also, among other things, define what parts of speech will be returned in the results of the search as well as whether these results will contain the selected word at the beginning or at the end of them, a handy feature when searching for compound words.

Screenshot from the OWID website displaying a usage case of the elexiko tool. Specifically, a user has searched for compound words beginning with the German word "Arbeit" and the search results are showcased.

Wordinary's main functionality is searching for any word contained in the schoolbooks of the Greek elementary school grades and returning its family of words. This could even be considered the sole requirement for the proposed application. However, our goal when developing Wordinary was to create an easy-to-use application that is highly configurable so that users can adapt it to their own needs. Thus, extra requirements were put into place to achieve this. Table 1 presents the main high-level requirements for Wordinary. Development followed the user-centered software design.

3.1 Architecture and development technologies

The Wordinary internal architecture comprises two conceptual layers: the vocabulary generation layer (named WordinaryKernel), and the user interface layer (named WordinaryUI). Figure 3 showcases the information flow presented in the following.

Wordinary uses the Python library tika to extract all words contained in the schoolbooks' PDF files. Then, the ILSP Lemmatizer is used to keep only the uninflected form of words. Then, Wiktionary is used to determine the family of each uninflected word. This process results in an SQLite3 database that stores the words and their families. The Wordinary user can then use its GUI to search for words and/or add user data.

The WordinaryKernel is responsible for extracting the words contained in the elementary school books and transforming them in a way that the WordinaryUI can use in order to support the functionality of searching in them to find the family of a user-provided word. First, the WordinaryKernel parses the PDF files of the schoolbooks to extract all words contained in them. The parsing is implemented using the Python programming language and specifically its module, tika. Subsequently, the extracted words are lemmatized using the ILSP Lemmatizer⁴ tool of the CLARIN:EL⁵ platform. This results in keeping only the uninflected form of words. Finally, the Greek version of the Wiktionary⁶ is used to determine the family of each uninflected word. Wiktionary also contains words in their uninflected form, which is typical for any dictionary, and thus lemmatization was a necessary step in the process. The end result of the WordinaryKernel is an SQLite3 database containing the schoolbook words and their families.

The WordinaryUI is responsible for the interaction with the tools’ user. The application's graphical user interface (Figure 4) was developed using the PyQt6 Python library. The PyInstaller Python module was used to create an executable file for the Windows operating system.

Screenshot displaying the main graphical user interface of the Wordinary application.

3.2 User flows

Wordinary has a single main flow that branches into two separate ones as well as multiple secondary flows that have to do with managing the application data and settings.

The main flow serves the application's main purpose, that is searching for the family of specific user-provided words. Users can achieve this by either typing in the search bar located in the top of the main UI (Figure 4) or by selecting a word from the recent searches or starred words (i.e. favorites) sections which are located in the left part of the main UI (Figure 4). For both ways, the filters in the right part of the main UI (Figure 4) can be used to adjust the set of words a user can search from (i.e., select a specific grade and/or a specific subject), while the search results (i.e. the family of the user-provided word) are displayed in the bottom part of the main UI. As seen in Figure 4, some results are displayed with a black border while others with a blue one. The former corresponds to words contained in the books of the selected grade while the latter to words that are not contained in these books but are returned by Wiktionary as part of the searched word's family.

The secondary user flows have to do with modifying the Wordinary vocabulary (Figure 5, top-left) and user data (Figure 5, top-right), editing the application settings (Figure 5, bottom-left) and reading the application's tutorial (Figure 5, bottom-right). These actions are all accessible through popup windows opened by corresponding buttons in the main window of Wordinary.

Screenshots displaying the following dialogue windows of the Wordinary app: Manage vocabulary, Manage students and profiles, Settings, and Tutorial.

Regarding the evaluation of Wordinary, the key question that we wanted to answer was whether this first version of the tool, which was developed in the context of the first author's undergraduate thesis, could satisfy the needs of its end users. To this end, two methods were employed: a) heuristic evaluation, b) user testing.

This paper presented Wordinary, a free desktop application for teaching Greek word families to elementary school children. Development followed the user-centered software design. Teachers, speech and language therapists and students in educational sciences and computer science participated in the design and evaluation process. A user testing study involving 12 end users found that Wordinary has good-to-excellent usability. Participants characterized Wordinary as a simple to use and useful application, but they have also identified areas where it could be improved and expanded in future work.

Wordinary is not without limitations. One such limitation is related to the parsing algorithm of the elementary level schoolbooks. On one hand, the pdf files of the books retrieved by the website of the Greek Ministry of Education, contain text in a format that is error-prone when parsed by a computer. On the other hand, the Python library used for parsing these files did not always work correctly with some Greek characters. An additional limitation is related to the process of lemmatization. Wordinary uses the ILSP Lemmatizer of the CLARIN:EL platform. However, there isn't any lemmatizer that produces 100% correct results, particularly for the Greek language that is of interest in our work. As a result, some words in the schoolbooks might be not captured correctly or at all.

Lastly, to find the family of a word, Wordinary uses an algorithm that relies on the Wiktionary website. Wiktionary may be a rather large and popular dictionary supporting multiple languages thanks to a large worldwide community that constantly enhances it, but its data still contain some deficiencies, especially in languages like Greek which are not widely used. As a result, the families of the words that it contains might not be complete. In addition, the Wiktionary website does not have an API. Therefore, to extract any information, the corresponding HTML pages of the website had to be parsed. Although some patterns had been identified regarding the arrangement of information on the Wiktionary webpages, these patterns were not consistently followed on every webpage. Even though special handling for the various cases that were found was implemented, some information is probably still being lost in the process.

Limitations related to the user testing study, is the relatively small sample size (N=12) and the fact that the sample was severely gender-skewed. In specific, almost all participants (92%) were women, and thus the results may not be representative of male users of Wordinary.

Future work includes implementing some of the suggestions reported by the user testing participants in order to further improve the usefulness of Wordinary. First, some participants mentioned the enrichment of the information displayed about words contained in the family of a specific word as a significant enhancement that could be made to the current implementation. According to these participants, this information could include the meaning of these words, their synonyms and antonyms as well as example sentences of the schoolbooks in which they are contained. An additional idea for future work was to add a quiz to test the students' knowledge of word families. This could be achieved by various types of exercises (e.g., fill in the blanks with the appropriate word for a word family, match words with other words that are included in their families). Future work also includes the addition of the ability to create an account and log into it while launching the application. This would be quite helpful in cases where the application is installed on a communal computer (e.g. at the school) and possibly used by multiple teachers of different elementary school grades. Finally, the introduction of technology to support a learning context does not guarantee by itself the learning effectiveness of the approach. Based on our previous experience [2,9,10,12,13,19], we plan to evaluate the learning effectiveness of a Wordinary-supported approach for teaching Greek word families.