Building an Annotated L1 Arabic/L2 English Bilingual Writer Corpus: The Qatari Corpus of Argumentative Writing (QCAW)

1.1 Research Aims

The aims of the current research are threefold. The first aim of this study is to develop a bilingual corpus, referred to as the Qatari Corpus of Argumentative Writing (QCAW), comprising 200,000 tokens of argumentative writing produced by Qatari university students in both L1 Arabic and L2 English. This aim focuses on the three critical stages of corpus construction: corpus design, development, and annotation. The second aim of the present study is to highlight how the QCAW enhances our understanding of texts written by the same bilingual writers in both L1 Arabic and L2 English, specifically metadiscourse markers. The last aim of the current study is to shed light on some specific applications derived from the Qatari Corpus of Argumentative Writing (QCAW) to enhance our understanding of bilingual writing in both L1 Arabic and L2 English contexts, with specific reference to form variability, learner errors, and writing assessment.

1.2 Research Significance

The current study is of considerable scientific significance for multiple reasons. First, building the Qatari Corpus of Argumentative Writing (QCAW) addresses a discernible gap in existing research by creating the first bilingual corpus in L1 Arabic and L2 English that encapsulates the nuanced dynamics of argumentative writing in both languages. This innovative approach provides a distinctive opportunity to explore the intricate interplay of linguistic features – ranging from vocabulary use to metadiscourse markers, syntactic structures, and part-of-speech distribution, writing errors – within the same set of writers across two languages. Secondly, the study contributes to advancing corpus linguistics, a pivotal field for understanding language patterns and structures. The meticulous annotation of the corpus, incorporating Part-Of-Speech (POS) annotations, enhances the scientific rigour of linguistic analyses, deepening our comprehension of specific linguistic features in argumentative writing and offering a valuable resource for developing and refining theories applicable to bilingual contexts.

Moreover, the demographic analysis, including considerations of gender and major in relation to writing proficiency, aligns with the broader goals of educational research, contributing to a more comprehensive understanding of factors influencing writing proficiency in bilingual university students. Additionally, the detailed procedural aspects of corpus collection and analysis outlined in the study contribute methodologically to the field, providing valuable insights into best practices for corpus creation and analysis and enhancing the robustness and replicability of future research endeavours. Lastly, the study’s exploration of the implications of the QCAW as a bilingual data source is of paramount importance. By emphasising its potential contributions to language acquisition, argumentative writing, and bilingual studies in both L1 Arabic and L2 English contexts, the research extends the practical applications of its findings, fostering advancements in pedagogy and linguistic research. In summary, the scientific justification for this study resides in its distinctive contribution to the fields of corpus linguistics, bilingual education, and educational research, presenting a valuable resource for scholars, educators, and researchers interested in the intricacies of argumentative writing proficiency in a bilingual setting.

2.1 Types of Learner Corpora

There are six different types of learner corpora, each with unique characteristics and uses. Firstly, the written learner corpora consist of written texts produced by language learners, such as essays, journals, and emails (Coxhead 2000; Gilquin and Granger 2015). These corpora are useful for studying language learners’ errors, error patterns, and language development over time. Secondly, the spoken learner corpora consist of spoken language produced by language learners, such as oral interviews, dialogues, and conversations (Caines, McCarthy, and O’Keeffe 2016; Yoon 2020). These corpora study language learners’ pronunciation, fluency, and spoken discourse strategies. Thirdly, the learner-compared corpora consist of written or spoken texts produced by language learners and native speakers, allowing for a direct comparison of language use between the two groups. These corpora are useful for identifying the specific areas in which language learners struggle and for identifying patterns of language use unique to language learners (Gilquin, Granger, and Paquot 2007).

In addition, learner corpora differ in multiple dimensions, including the time of collection, the scope of the collection, the targeted language (L2), the learner’s mother tongue (L1), the medium, and the text type (Granger 2011). In reference to the time of collection, there are two types of learner corpora: cross-sectional learner corpora and longitudinal learner corpora. The former consists of instances of learner writing or speech collected from various categories of learners at a particular moment. In contrast, the latter monitors the progress of identical learners over a specific time frame. In relation to the scope of the collection, two types of learner corpora are identified: global and local. Global learner corpora are large data collections from diverse learners that inform SLA theory and teaching tools. On the other hand, local learner corpora are smaller collections gathered by teachers in their routine teaching practices, used as the foundation for classroom materials.

Another way to categorise learner corpora is based on the language they focus on, such as L2 English learner corpora and L1 learner corpora. In terms of the medium, there is a written learner corpus, which refers to corpora of learner writing. In contrast, a spoken learner corpus may refer to transcriptions of oral production data. Finally, based on the text types, the two most commonly represented text types in learner corpora are argumentative texts for writing and informal interviews for speaking.

2.2 Differences Between Arabic and English

Salloum et al. (2023) outlined eight notable distinctions between Arabic and English. The first contrast lies in character connections: English utilises diagonal strokes to link characters, while Arabic connects the baseline with horizontal strokes. Second, English character versions exhibit limited shape variations. In contrast, Arabic characters display significant variability, showcasing up to four distinct shapes based on word position. Capitalisation is a feature exclusive to English. The direction of writing diverges, with English following a left-to-right pattern and Arabic adopting a right-to-left orientation. Additional differences include gender differentiation in Arabic verb and sentence structure, plural forms (singular and plural in English versus singular, dual, and plural in Arabic), adjective placement (before the noun in English and after in Arabic), and segmentation methods in handwriting. Moreover, the alphabet size differs (26 letters in English and 28 in Arabic), and sentence types vary (verbal in English and nominal and verbal in Arabic). Lastly, the total number of speakers is substantial, with English at 1.348 billion and Arabic at 274 million.

2.3 Arabic–English Bilingual Corpora

The Zayed Arabic–English Bilingual Undergraduate Corpus (ZAEBUC) corpus is the only Arabic–English bilingual corpus available online. The ZAEBUC corpus was developed by Habash and Palfreyman (2023). It comprises bilingual writing samples from the same writers on different occasions, matching comparable texts in different languages. Specifically, it currently contains short essays from several hundred Freshman students, predominantly Emirati. The corpus includes 388 English essays (88,000 words) and 214 Arabic essays (33,000 words).

The Qatari Corpus of Argumentative Writing (QCAW), under investigation, was published in the Linguistic Data Consortium by Ahmed et al. (2022). It comprises writing samples in L1 Arabic and L2 English written by the same Qatari students on two different Argumentative topics. It shows the same Qatari university students’ argumentative writing in L1 Arabic and L2 English. It includes 195 essays in L1 Arabic (97,248 tokens) and 195 in L2 English (98,379 tokens). The next section sheds light on the features of L1 Arabic writing and L2 English writing.

2.4 Features of L1 Arabic Writing

Arabic written language is characterised by distinctive features that set it apart from other languages. Kaye (2017) identified Arabic as a Semitic language spoken by over 200 million people as a mother tongue. Arabic speakers primarily live in Southwest Iran, Iraq, Syria, the Arabian Peninsula, the Maghreb region of North Africa, Egypt, and Mauritania (Al-Khatib 2000). The Arab world is considered a diglossic speech community, where the language has two forms: colloquial Arabic, which exists as the vernacular varieties of the major Arabic-speaking nations, and classical Arabic, the language of the Quran, which provides a common, standard written form for all the vernacular variants and a shared medium for state affairs, religion, and education across the Arab world (Al-Khatib 1988, 1995).

The Arabic script comprises a set of 28 letters, each of which can take different forms depending on its position in the word (Khorsheed 2002). Additionally, Arabic script includes diacritical marks that indicate vowel sounds not represented in the script (Hamed and Zesch 2017). In addition, Arabic script uses ligatures, which are combinations of two or more letters written as a single unit (Naz et al. 2016). Arabic grammar includes two genders (feminine and masculine), three numbers (singular, dual, and plural), and three grammatical cases (nominative, genitive, and accusative) (Chen and Gey 2002).

Arabic written language is marked by its use of the definite article, represented by the prefix “al-” (Al-Jarf 2022). This noun prefix indicates that it is definite and changes the form of the noun depending on its grammatical case (Chen and Gey 2002). Besides, Arabic has a complex grammatical structure, with a system of nouns, verbs, and other parts of speech that are inflected to indicate tense, mood, and other grammatical features (Sawalha and Atwell 2013).

Arabic written language has an inflexion system, known as declensions, which indicate the grammatical function of nouns and adjectives (Saiegh-Haddad and Henkin-Roitfarb 2014). This system depends on the use of patterns of consonants and vowels, which change depending on the grammatical case and the number of words (Abu-Rabia and Awwad 2004). The inflexion system is clear in Arabic nouns, which have three different cases (nominative, genitive, and accusative) and three different numbers (singular, dual, and plural) (Chen and Gey 2002). Another characteristic of the inflexion system is that Arabic verbs have a complex conjugation system based on the person, gender, and number of subjects (Kusters 2003).

In addition, the written Arabic language also includes a set of grammatical particles known as particles of negation, which are used to indicate negation and other grammatical functions (Al-Momani 2011). Furthermore, the Arabic written language has a rich system of idiomatic expressions, proverbs, and colloquial expressions, which convey meaning and emphasise certain ideas (Alqahtni 2014).

To summarise, Arabic has a distinctive script (the Arabic alphabet). It uses a complex system of declensions and particles of negation. It also has a rich tradition of idiomatic expressions, proverbs, and colloquial expressions, making it a unique and complex language.

2.5 Features of L2 English Writing

L2 English writing is characterised by some features different from native speakers. Grammatical, lexical, syntactical and orthographic errors are prevalent in L2 English learners’ writing (Olsen 1999). For example, Arab students often struggle with L2 English grammar, vocabulary, organisation and coherence in their English writing (Khuwaileh and Shoumali 2000). These errors are often caused by the influence of the learners’ first language (L1) on their second language (L2) (Crompton 2011). These errors may also be attributed to students’ problems with the cultural and linguistic differences between their native language and English (Al-Jarf 2013).

Overgeneralization is another feature of L2 English learners in writing. It occurs when learners apply the rules of their L1 to the L2 (Mourssi 2013). Overgeneralization is particularly common in L2 English learners using irregular verb forms and verb tenses (Kirmizi and Karci 2017). Additionally, learners may overgeneralise grammatical structures from their L1, such as articles or word order (Hertel 2003).

Many English learners have a limited vocabulary, sometimes resulting in repetitive words and phrases (Ahmed 2010a). Learners’ limited vocabulary repertoire may lead to problems with word order and collocation, showing an insufficient command of more complex vocabulary that enables them to express their ideas precisely (Phoocharoensil 2013). Additionally, L2 English learners have problems with coherence, cohesion, lexis, grammar and mechanics (Ahmed 2010b), making it difficult for readers to understand the intended meaning. These problems are attributed to socio-cultural issues (Ahmed and Myhill 2016). Moreover, English learners may also have difficulties using cohesive devices such as referencing, substitution, and ellipsis, which are crucial for text coherence (Ahmed 2010b).

2.6 Research Questions

1. How was the Qatari Corpus of Argumentative Writing (QCAW) built in terms of corpus design, development and annotation?

2. How does the Qatari Corpus of Argumentative Writing (QCAW) enhance our understanding of texts written by the same bilingual writers in both L1 Arabic and L2 English, specifically metadiscourse markers?

3. What specific applications can be derived from the Qatari Corpus of Argumentative Writing (QCAW) in enhancing our understanding of bilingual writing in both L1 Arabic and L2 English contexts, with specific reference to form variability, learner errors, and writing assessment?

3.1 Corpus Design

3.2 Corpus Development

3.3 Corpus Annotation

3.4 Rating Essays for Writing Quality and Voice

In the corpus annotation stage, both English and Arabic texts underwent separate rating processes to assess writing quality and dimensions of voice. The essays were initially graded for writing quality using a five-category analytical rubric, emphasising students’ stances over structural elements. A benchmarking procedure involving four raters ensured reliability, with alignment checks and discussions to maintain consistency. For voice assessment, a holistic voice rubric was employed, scored by four native Arabic-speaking raters, including both Egyptian and Tunisian perspectives. Benchmarking procedures for voice included norming sessions, consensus-building discussions, and reliability measures to ensure consistent evaluations. Raters focused solely on assessing voice salience, and the general instructions emphasised reliability through breaks, rubric reviews, and double-rating each writing sample. These rigorous procedures contribute to the validity of the assessments and enhance the overall quality of the Qatari Corpus of Argumentative Writing (QCAW).

3.5 Standardising Argumentative Writing

We standardised argumentative writing within the QCAW to establish a benchmark or reference point for evaluating and understanding bilingual university students’ proficiency levels in argumentative writing in L1 Arabic and L2 English. It enables comparative analyses to highlight areas of strength and improvement in bilingual writing skills.

3.6 Research Limitations

While the current study aims to create a robust bilingual corpus for analysing argumentative writing, some limitations should be acknowledged. Firstly, the students participating may exhibit a wide range of proficiencies in both L1 Arabic and L2 English. This variability in language capabilities could impact the quality and sophistication of argumentative writing samples collected, introducing potential variation. Secondly, some students might be more accustomed to the argumentative writing task assigned based on prior coursework or experience, while it may be new for others. This could influence the structure and development of their written arguments. Thirdly, L1 Arabic language structures might interfere with or influence students’ L2 English writing and vice versa. Such cross-linguistic influences could contribute to variations in writing style, rhetorical preferences, or organisational approaches. Finally, the in-class timed writing situation, while well-controlled, limits student access to online resources, references, or assistance compared to a take-home writing assignment. The absence of such support conceivably has some impact on the writing process.

While the Qatari Corpus of Argumentative Writing incorporates rigorous design to ensure a robust dataset, these limitations acknowledge potential sources of variability stemming from the student writers, which could impact generalizability. Further studies controlling for some of these factors would prove valuable.

3.7 Ethical Considerations

Several measures were implemented to ensure this study was conducted ethically and safeguard participants’ rights. Approval was obtained from the university Institutional Review Board (IRB) overseeing studies involving human subjects before any participant recruitment or data collection.

Once approved, student participants were given information sheets outlining the study’s purposes and procedures in accessible language. Informed written consent was collected from all participants, specifying that their involvement was voluntary with no impact for declining. Steps were taken to allow anonymous participation, with all identifying information removed from essays during corpus compilation, and participants were assigned ID numbers. Students could withdraw their consent at any time without repercussion.

Collected written data was securely stored in encrypted files, accessible only to the research team members. Future users of the corpus for research purposes are appropriately bound by end-user agreements prohibiting attempts to identify participants. Any excerpts selected for publication are subject to stringent anonymisation practices.

This comprehensive approach ensured that participant anonymity, autonomy and confidentiality were preserved and maintained throughout the project in alignment with prevailing ethical standards for educational research. The procedures received full board approval, signifying adherence to established ethical guidelines for working with student subjects.

4.1 QCAW Building

The answer to the first research question was answered in depth in the methodology section of this research paper, where the Qatari Corpus of Argumentative Writing (QCAW) went through the stages of corpus design, development and annotation.

4.2 QCAW and Metadiscourse Markers

In an attempt to answer the second research question, following Hyland’s model of metadiscourse (2010), we have created Tables 2 and 3 to highlight how QCAW can be used to enhance our understanding of interactive and interactional metadiscourse categories with their specific functions, sub-categories and examples in both L1 Arabic corpus and L2 English corpus of Qatari university students.

Table 2:

Analysis of interactive metadiscourse markers in QCAW.

1. N.B. All provided L1 Arabic and L2 English examples are verbatim from students’ texts in the QCAW without corrections.

Table 3:

Analysis of interactional metadiscourse markers in QCAW.

1. N.B. All provided L1 Arabic and L2 English examples are verbatim from students’ texts in the QCAW without corrections.

How does the Qatari Corpus of Argumentative Writing (QCAW) enhance our understanding of texts written by the same bilingual writers in both L1 Arabic and L2 English, specifically metadiscourse markers?

Table 2 outlines the interactional category of metadiscourse, showing their functions, and provides examples in L1 Arabic and L2 English corpora extracted from the Qatari Corpus of Argumentative Writing (QCAW). The examples extracted from QCAW illustrate how these metadiscourse markers operate in the respective languages. The interactive category includes the following metadiscourse sub-categories: transitions, frame markers, endophoric markers, evidentials, and code glosses.

Table 3 outlines the interactional category of metadiscourse, showing their functions, and provides examples in L1 Arabic and L2 English corpora extracted from the Qatari Corpus of Argumentative Writing (QCAW). The examples extracted from QCAW illustrate how these metadiscourse markers operate in the respective languages. The interactional category includes the following metadiscourse sub-categories: hedges, boosters, attitude markers, engagement markers, and self-mentions.

4.3 QCAW Applications

This section attempts to answer the last research question stated below. What specific applications can be derived from the Qatari Corpus of Argumentative Writing (QCAW) in enhancing our understanding of bilingual writing in both L1 Arabic and L2 English contexts, with specific reference to form variability, learner errors, and writing assessment?

4.3.1 Applications of the Qatari Corpus of Argumentative Writing (QCAW)

Interpretation of learner corpus data differs from native language in many respects, among which the most important are form variability, learner errors, and writing assessment. In the case of learner English studies, form variability is usually investigated in terms of variations of linguistic forms or language-specific variations. The term learner errors in this paper refers to using a word or an expression to denote a different meaning from native speakers, which could cause misinterpretation, ambiguity, or illogical statements. Writing assessment is inseparable from interpreting learner corpus data. It is especially applicable for teachers and researchers because written texts are a stable source for investigating the longitudinal progress of L1-specific learners and improving teaching strategies. To contribute to the aspects above, we will demonstrate how to use QCAW as a source of a learner corpus.

4.3.1.1 Form Variability in Learner English Corpus

Form variability is important in learner corpus research for two reasons. First, variations become unneglectable when the fact is “that the number of non-native speakers far outnumbers that of native speakers” (Granger, Gilquin, and Fanny 2015, p. 1). This results in an enormous number of language users, leading to higher variability in language forms. Second, to determine and describe the proficiency levels of learners, it is crucial to be aware of the differences or changes in language forms developed by the learners (Ädel 2015; Gilquin and Granger 2015; Gries and Wulff 2020; Hendriks 2005; Jarvis 2000; Mollin 2006; Paquot and Fairon 2006; Pendar and Chapelle 2008; Regan 2013; Vyatkina 2013; Wulff and Gries 2021).

One of the most common form variabilities in learner English is L1 specific variations, which are also our focus in this section. L1 specific variations, as the name suggests, refer to unique patterns found in L2 production of learners from specific L1 language backgrounds. They are not seen in the language use of native speakers. L1 specific variations vary from different language groups, i.e., variations used by one language group may rarely be seen in the language production of other language groups.

The QCAW is used here to exemplify L1 specific variations used by L1 Arabic speakers in their L2 English argumentative writing. Unique uses that occur more than twice are considered to be variations. Underlined parts in examples (1) to (3) illustrate the phenomenon of L1 specific variations.

1. On the first hand, many people agree that emails and telephones are the best ways of communication in our life. [48B]

2. On the first hand, the first team believes that nowadays technology is the only way to get any kind of information because it is easier, and not only that it is easier but also it is fast and quick with that students are more likely to use the technology whether it is on their mobile or tablets or even computers. [199B]

3. On the first hand, it is believed that students need technology to further expand their knowledge because of the fact that it is very easy to access. [52B]

As the above examples show, “on the first hand” seemed inappropriate. However, they were not incorrect because they were coherent at the discourse level and did not interfere with readers’ understanding. Examples of another transition marker, “on the one hand”, are shown in Figure 1 below to supplement this case.

Figure 1:

Examples of “on the one hand”.

We further compared instances found in QCAW with argumentative writing by American university students provided by The Louvain Corpus of Native English Essays (LOCNESS).^[1] Unsurprisingly, we found that “on the first hand” was not used by native speakers.

Another example of L1 specific variations found by our comparison between QCAW and LOCNESS-USARG is shown in examples (4)–(6).

1. Some people say that technology affected education positively. On the other side, some people disagree to this and state that technology affected education in a bad way. [17B]

2. On the other side, some people tell that the application of means of technology to education waste students time as some of them get to social media websites while being in classes instead of searching for information. [17B]

3. On the other side, the people who oppose preferring studying on the technology support their point of view by many supports. [152B]

We found that L2 learners and native speakers used “on the other hand” to introduce a contrasting point of view in their argumentative writing. But, its equivalent, “on the other side”, was used only by L2 learners. This was also an example of L1 specific variation similar to examples (1)–(3). Figure 2 shows examples of “on the other hand” used by L2 learners.

Figure 2:

Examples of “on the other hand”.

L1 specific variations not only contribute to the interpretation of learner corpus data, they also render visible the differences between natives and non-natives. Furthermore, L1 specific variations are text-internal measures; they can be necessary to analyse the lexical diversity of L2 texts with quantitative data.

4.3.1.2 Learner Errors and Learner English Corpus

When comparing learners’ performance across proficiency groups or native speakers, error analysis can help researchers with signalling developmental stages. Meanwhile, learner corpus can provide traceable contexts when researchers want to analyse errors based upon certain hypotheses. QCAW was not annotated for learner errors. This makes it open to hypothesis-specific studies in the future, and researchers can develop their error annotation scheme.

The notion of learner errors can be traced back to as early as Corder’s work in 1967. From then on, many studies have discussed learner errors from various perspectives, for example, parts of speech, syntax, morphology, register, appropriateness, cross-corpora … etc. Following a thorough examination of previous studies in learner errors and learner corpus research (e.g., Corder, 1967; Cowan, Choi, and Kim, 2003; Dobrić 2023; Dobrić and Sigott 2014; Ellis and Barkhuizen, 2005; Granger 1996; Lennon 1991; McEnery et al. 2019; Paquot 2008; Satake 2020; Valero Garcés 1997; Yoo and Shin 2019; Yoon and Jo 2014), the working definition of learner errors in this paper is: Errors are identified when learners use a word or an expression to denote a different meaning from native speakers, which could cause misinterpretation, ambiguity, or illogical statement. Learner errors are considered to be a necessary part of acquisition developed by learners in the process of acquiring a second language. They are erroneous outputs which learners’ native-speaker counterparts would not produce.

In addition, from the perspective of linguistic forms, errors can be divided into errors of form, grammar, lexico-grammar, lexis, word redundant, word missing, word order, punctuation, style, and infelicities. While based on the causes of errors, learner errors may be induced by L1 specific factors or other factors such as intralingual factors, developmental factors, teaching-induced factors or communication strategies (Cowan, Choi, and Kim, 2003; Cross and Papp, 2008; James 1980). When errors are induced by learners’ L1, they are defined as interlingual errors. Meanwhile, when induced by other factors regardless of learners’ L1s, they are defined as intralingual errors, including over-generalisation errors, ignorance of rule restrictions, incomplete application of rules, false concepts hypothesised, and other unique types identified by researchers.

The example below highlights how to use QCAW to illustrate what it can offer when interpreting learner corpus data from the angle of learner errors.

1. People that agree with the usage of technology in studying say that technology has no limits which means that it is really rare for someone to not be able to find results of what they are searching for. As well as no limitation of books and studying resources, hundreds of results appear when someone searches for a specific subject. Secondly, technology is considered an open education. [142B]

“as well as” is used to mention another item or point connected with the subject discussed in a sentence as an addition to the subject. It usually emphasises the expression preceding “as well as” than the one following it. In example (7), on the surface, “as well as” was used by the writer to add “no limitation of books and studying resources” to the context as supplementary information. However, according to the context, “as well as” did not make the argument more convincing or provide more detailed information or explanation by linking “no limitation of books and studying resources” with the sentence. This clause did not function as added supplementary information to the sentence. The meaning of “as well as” is used incorrectly in this case.

However, most students correctly used “as well as” as an addition marker, as shown in Figure 3. This indicated that cases like example (7) are more likely to be individual-specific.

Figure 3:

Example of “as well as”.

1. Recent studies and professors experiences’ prove that students who use mobiles in classroom are not attentive or in the same level as disciplined ones. Therefore, a professor decides to ban using phones in class indirectly; thus he banned all electronic devices usage at all, such as laptops, except with disability students. The main problem behind banning the smartphone is the attention; students who use their smartphones lose half or more of their attention and focusing in class. They do not listen carefully nor taking notes. [204B]

“therefore” is used to introduce a logical result or conclusion of a fact or reason mentioned previously in the context. Similarly, the information following “thus” is expressed as a result of something mentioned in the immediate context. The writer of (8) wanted to establish a cause-and-effect relationship by using “therefore” and “thus” in this example, a declarative sentence which described a fact or phenomenon to serve as an example of the argument. The clauses were arranged sequentially. “Therefore” and “Thus” should not be used.

Examples (7) and (8) were noticed when we investigated transition markers (Ahmed et al. 2023) in the process of annotation. They were not as easily identified as word missing or inflectional errors. It is noteworthy that, first, high-proficiency participants found the above two examples in writing. Learners at this proficiency level do not usually make errors, such as incorrect word forms. However, infelicities do occur in their writing, which could imply that errors like (7) and (8) are more likely to occur when learners’ writing performance achieves an advanced level. Thus, they can signal the stages of learners’ writing development. The second point is that these two examples may enlighten researchers to use a hypothesis-driven approach to investigate learner errors in argumentative writing.

Apart from errors like examples (7) and (8), we found some typical and prevalent errors in learning English easily, as shown in the underlined words in examples (9) and (10).

1. It is highly debates issue to determine whether or not email and telephone has made communication between people less personal. [220B]

2. Furthermore, technology has make the communication between the schools management and parents easier than before and more quickly than parents attending, successful positive communication is essential to the success of the educational process. [127B]

Errors may occur in writing by proficient and less proficient learners. Our examples provide additional support for analysing writings by L1 specific bilingual speakers. For example, the activation of the L1 influenced word forms and word sequences or the impact of errors on cross-language interaction.

4.3.1.3 Learner Corpus Data and Writing Assessment

Besides investigating variability and errors, learner corpus has been proven highly significant in writing assessment. QCAW collected texts from L1 Arabic and L2 English argumentative writing tasks, which reflected the nature of task-based first/second language learning and teaching. This design focused on language performance and acquisition and promoting in-class language learning. QCAW is expected to contribute to developing assessments in task-based L1 Arabic and L2 English writing contexts from two aspects.

First, language use represented by learner corpus is not the same as in the real world. The reason is that texts collected by a learner corpus, no matter whether a corpus of spoken or written language, is usually attained in an environment that induces learners to produce certain types of language use, and they are meaning-focused (Müller-Hartmann and Schocker-von Ditfurth 2011; Skehan 1998). Therefore, a corpus containing written texts by learners from a specific L1 background will prove that some unique features can be used in writing assessment. For instance, many instances of long sentences were found in QCAW. They are much longer than the average sentence length in the writing of native speakers, which could be a result of L1 Arabic influence. For example, we even found a sentence that contained 97 words. This feature could help writing assessors determine the complexity and coherence of L2 English writing by L1 Arabic speakers.

Second, a more detailed assessment of variables which drew less attention in this area should be considered in terms of methodology. For example, QCAW collected texts from male and female participants. Gender, as a potential variable, can be combined with other variables, such as lexical complexity or proficiency level, to benefit future research on writing assessment.

5.1 Implications of QCAW as a Learner Corpora

Building learner corpora has important implications for language teaching, learning, and linguistic research. One of the main implications of building learner corpora is that it allows for a more detailed and accurate understanding of the language learning process. For example, analysing the errors and patterns of usage in learner corpora can enable researchers to identify common problems that learners face (Blagoeva 2004). In addition, using learner corpora can help increase learners’ understanding of usage patterns, reduce collocation errors, and enhance students’ autonomy and self-correction skills (Smirnova 2017). This can apply to QCAW, where researchers can identify common writing problems that Qatari university students encounter in L1 Arabic and L2 English and develop tailored teaching materials accordingly.

Another implication of building learner corpora is that teachers and researchers can develop more authentic and relevant language materials and assessments (Chen 2011). For example, by analysing learners’ written language in QCAW, developers of L1 Arabic and L2 English materials can create texts and tasks more appropriate for learners’ abilities and interests.

Another important implication of building learner corpora is the ability to evaluate language teaching materials and methods (Man and Chau 2019). Researchers can evaluate the teaching materials and methods by comparing the writing produced by Qatari L1 Arabic and L2 English learners with the language presented in teaching materials.

Moreover, considering learners’ corpora’s strengths and weaknesses can ensure they are successfully integrated into the language classes (Kaltenböck and Mehlmauer-Larcher 2005). One form of this successful integration could be teachers’ provision of focused feedback on their students’ mistakes (Crosthwaite 2017). L1 Arabic writing instructors and L2 English writing instructors in the Arab world, in general, and Qatar, in particular, can incorporate QCAW in their teaching and feedback to enrich students’ learning experiences.

In addition to the implications mentioned above, building learner corpora allows for examining the lexical development of second language learners (Berger, Crossley, and Kyle 2019). It helps develop their collocational competence (Li 2017). This applies to our QCAW, where teachers and researchers can analyse the students’ lexical complexity and develop students’ collocational competence in their L1 Arabic and L2 English writing and can, therefore, provide insightful implications to the teaching and learning of vocabulary acquisition.

Furthermore, building learner corpora also allows for examining second-language learners’ discourse markers and organisation. Studies have used learner corpora to investigate discourse markers (Gilquin 2016) and discourse organisation (e.g., Swales 1990) in second language learners. Consequently, QCAW can be a great opportunity for L1 Arabic and L2 English writing instructors and researchers to analyse the patterns used in discourse markers and discourse organisation, giving insightful implications to the teaching and assessment of L1 Arabic and L2 English writing.

Moreover, building learner corpora can provide insights into the pragmatic development of second language learners in different genres or contexts in characterising their use (Vaughan and Clancy 2013). Pragmatic features such as hedging (e.g., Hyland 1998) in second language learners can be analysed in our QCAW in L2 English, providing an insight into whether they are appropriately or inappropriately used.

Furthermore, building learner corpora can provide insights into the writing development of second language learners. Studies have used learner corpora to investigate cohesive devices (e.g., Biber and Finegan 1989; Hyland 1998) and text organisation (e.g., Bhatia 1993; Swales 1990) in second language learners. Using QCAW can help students identify how different transition markers and paragraph development in L1 Arabic and L2 English were used by analysing patterns in the corpus.

Besides, building learner corpora can provide insights into the interlanguage development of second language learners. For example, in Hernández and Paredes’ study (2005), students, in their corpus-based research, overused highly technical and general vocabulary in their writing, which confirms the interlanguage factor. Similarly, researchers can analyse the QCAW and find L2 English writers’ patterns that could be attributed to L1 Arabic interlanguage or vice versa.

In conclusion, building learner corpora has numerous implications for language teaching, learning, and linguistic research. Learner corpora provide insights into the language learning process, allow for the development of more authentic and relevant language materials and assessments, enable the evaluation of language teaching materials and methods, and help develop learners’ linguistic competence. Furthermore, building learner corpora can provide insights into second language learners’ writing, pragmatic, and interlanguage development. Thus, creating and analysing learner corpora, such as the Qatar Corpus of Argumentative Writing (QCAW), can significantly improve language education and research in L1 Arabic and L2 English.

5.2 Suggestions for Further Research

The Qatari Corpus of Argumentative Writing (QCAW) can be a valuable resource for the following areas of future research.

1. Contrastive Rhetoric: The corpus can investigate the rhetorical and linguistic differences between Arabic and English argumentative writing, particularly in terms of organisation, coherence, and argumentation strategies. This can provide insights into the cultural and linguistic factors that shape argumentation in different contexts and inform the development of cross-cultural communication skills.

2. L1 Arabic/L2 English Writing Research: The corpus can be used to investigate the development of argumentative writing skills among Arabic-speaking students in L1 Arabic and L2 English. This can involve exploring the influence of L1 transfer, the role of feedback and instruction, or the impact of individual differences on writing proficiency.

3. Studying Language Transfer: The corpus can be used to study the influence of students’ first language (L1 Arabic) on their second language (L2 English) argumentative writing and vice versa. This can help identify common errors and challenges that Arabic-speaking students face when writing in English and inform the design of more effective writing instruction and feedback.

4. Automated Essay Scoring: The corpus can be used to develop and evaluate automated scoring systems for argumentative essays written by Arabic-speaking students in L1 Arabic and L2 English. This can involve exploring the performance of different machine learning algorithms, features, or scoring models or investigating the impact of factors such as prompt complexity or genre.

5. Writing Pedagogy: The corpus can inform the design and evaluation of writing pedagogy for Arabic-speaking students learning argumentative writing in L1 Arabic and L2 English. This can involve exploring the effectiveness of different instructional approaches, feedback strategies, and writing tasks or investigating the impact of individual differences on writing proficiency.

6. Cross-linguistic Research: The corpus could be used to conduct cross-linguistic research to analyse linguistic aspects (.e.g. metadiscourse and voice; syntactic complexity, use of collocations, grammatical complexity, style of writing, cohesion, coherence, mechanics of writing) in L1 Arabic and L2 English. The corpus can explore linguistic phenomena in L1 Arabic and L2 English argumentative writing, including discourse markers, rhetorical devices, or lexical and grammatical features. This can involve investigating the role of these phenomena in argumentative writing, their relationship to other linguistic and rhetorical factors, or their impact on text quality or persuasiveness.