Indonesian J our nal of Electrical Engineering and Computer Science V ol. 25, No. 1, January 2022, pp. 460 ∼ 473 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v25.i1.pp460-473 ❒ 460 De v elopment of computer -based lear ning system f or lear ning beha vior analytics Kanyalag Phodong 1 , Thepchai Supnithi 2 , Rachada K ongkachandra 3 1 Department of Computer Science, F aculty of Science and T echnology , Thammasat Uni v ersity , P athumthani, Thailand 2 Language and Semantic T echnology Laboratory , Intelligent Informatics Research Unit, National Electronic and Computer T echnology , P athumthani, Thailand 3 Data Science and Inno v ation Program, Colle ge of Interdisciplinary Studies, Thammasat Uni v ersity , P athumthani, Thailand Article Inf o Article history: Recei v ed Apr 5, 2021 Re vised No v 23, 2021 Accepted No v 28, 2021 K eyw ords: Computer -based learning Learning analytics Natural language processing Self-re gulated learning ABSTRA CT This paper aims to analyze the learning beha vior of Thai learners by using a computer - based le arning system for English writing. Three main objecti v es were set: the de v el- opment of a computer -based learning system, automat ic beha vior data collection, and learning beha vior analytics. Firstly , the system is de v eloped under a multidisciplinary idea that is designed to inte grate tw o concepts between the self-re gulated learning model and components of natural language processing. The int e gration design en- courages self-learning in the digital learning en vironment and supports appropriate English writing by the pro vided component selection. Second, the system automati- cally coll ects the writing beha vior of a group of Thai learners. The data collected are necessary input for the process of learning analytics. Third, the writing beha viors data were analyzed to nd the learning beha vioral patterns of the learners. F or learning analytics, beha vior sequential ana lysis w as used to analyze the learning logs from the system. The 31 under graduate students are participated to record writing beha viors via the system. The learning patterns in relation to grammatical skills were compared between three groups: basic, intermediate, and adv anced le v els. The learning beha vior patterns of the three groups are dif ferent that use for reecting learners and impro ving the learning materials or curriculum. This is an open access article under the CC BY -SA license . Corresponding A uthor: Rachada K ongkachandra Data Science and Inno v ation Program, Colle ge of Interdisciplinary Studies, Thammasat Uni v ersity Rangsit Center , Phahon Y othin, Klong Luang, P athumthani 12121, Thailand Email: krachada@staf f.tu.ac.th 1. INTR ODUCTION The English language is considered as essential for Thai people and is therefore a fundamental part of the education system. Thai learners often e xperience dif culties in studying English as a foreign language (EFL), in reading, speaking and especially writing [1]. Most language teaching in Thailand is a one-size-ts-all that is unable to clearly identify the weaknesses of each learner . Personalized learning for the English language is one possible solution. This aims to analyze indi vidual learning beha vior in order to identify each learner’ s strengths and weaknesses. Computer technology increases learning beha vior analytics for personalized learning in terms of the storage and speed of analytics processing. Learning beha vior analytics using computer -based technology is quick er and cheaper than human analysis. Although computer technology supports data storage and f aster pro- cessing, language learning requires an underpinning pedagogy to foster self-learning for personalized learning J ournal homepage: http://ijeecs.iaescor e .com Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 ❒ 461 analytics. The self-re gulated learning model is an essential model to get positi v e outcomes in learning, such as encouraging learners’ skills to shape their o wn learning [2] and supporting lifelong autonomous learning [3]. There is v arious researches present model to encourage for learning of foreign language. The sel f- re gulated learning model is an ef cient f actor to impro v e the learning per formance [4]. The model is applied to foreign language learning. Incorporating the concepts of the self-re gulated learning model to the foreign lan- guage le arning that supports the de v elopment of autonomous learners [3]. The self-re gulated learning model f acilitates learners lead to higher ef cienc y in language skills such as comprehension of writing [5]. The model consists of three main phases: forethought, performance and self-reection [6], [7]. The self-re gulated learn- ing model encourages interaction between person, beha vior , and en vironme ntal f actors to increase ef fecti v e learning [8]. Computer technology is being e xtensi v ely used in the education eld [4], [9]. A computer -based learning syst em is a tool of computer t echnology that can be used for encouraging interacti v e beha vior between personal and learning en vironments. A computer -based learning system could support a better learning e xperi- ence that learners could eng age the interactions with learning tasks [10]. In addition, the computer -based learn- ing system supports automatic data collection for recording learning beha vior while using the digital system. The system can also automatically collect learning beha vior data to analyze the pattern of learning beha viors. Natural language processing (NLP) aims to mak e the computer able to understand the language through computer processing. There are six le v els of language processing: morphological, le xic al analysis, syntactic analysis, semantic analysis, pragmati cs, and discourse [11]. These processes are applied to de v elop man y NLP tools such as w ord se gmentation, le xical analysis and parsing [12]. Moreo v er , applying natural lan- guage processing is an ef fecti v e tool for enhancing the education eld. NLP can impro v e the learning ability of the student in case of student f ails to understand the conte xt due to the barrier of language. NLP and digital technology are combined to impro v e a computer -assisted teaching system [13]. Mathe w et al. [14] pro vide the application of NLP techniques for an assistant tool to support teachers get insights about each student’ s learning progress. Therefore, a computer -based learning system could inte grate the methods of NLP to assist Thai EFL learners in their understanding of language structure and encourage learners’ impro v ement in English writing, in particular . Learning analytics in digital learning en vironments is an inte gration of tw o research elds, which are those of education and computer technology . Learning analytics is an important issue in education. Learn- ing analytics is the analysis of ‘learning logs’ and education data for impro ving learning outcomes, learning designs, and learning en vironments [15]. On the other hand, computer technologies ha v e become popular for communications and learning. T echnologies are con v enience to access through portable de vices such as smartphones, tablets, and laptops. Therefore, the inte gration of these tw o elds can help impro ving education. This paper aims to acquire the learning beha vior by using the pro vided computer -based l earning system for composing the English sentences. The system is designed by incorporating concepts of the self- re gulated learning model and components of NLP . Self-re gulated learning encourages learners to set goals, as well as monitoring their beha viors and reect writing performance to learners. The NLP learning en vironment encourages action between learner and system to compose the tar get sentences. Furthermore, all beha viors are automatically recorded for use in the learning analytics proces s. The results of learning analytics are use- ful to demonstrate learning performance and to support the impro v ement of learning materials. This paper is structured as follo ws: Section 2 e xplains background information and related w orks about the model of self-re gulated learning, components of NLP , and learning analytics in foreign language learning. Section 3 de- scribes the computer -based learning system. Section 4 describes the e xperimental design. Section 5 describes the e xperimental results. Section 6 pro vides a discussion. Finally , section 7 gi v es conclusions. 2. B A CKGR OUND 2.1. Self-r egulated lear ning model The self-re gulated learning model is a conceptual frame w ork of i nteraction between person, beha vior and en vironment in a learning conte xt and comprises three main phases: forethought, performance and self- reection [16], as illustrated in Figure 1. a. F orethought phase: Learners set goals and learning plans. The learners plan ho w to reach them in the learning strate gies acti v ation process. b . Performance phase: Learners control themselv es while e x ecuting the task and the y monitor their progress in completing the task. De velopment of computer -based learning system for learning behavior analytics (Kanyala g Phodong) Evaluation Warning : The document was created with Spire.PDF for Python.
462 ❒ ISSN: 2502-4752 c. Self-reection phase: Learners e v aluate their satisf action in performing the task, making attrib utions for their achie v ement or f ailure. These attrib utions generate self-reactions that positi v ely or ne g ati v ely inuence learners. Figure 1. The three phases of self-re gulated learning model There are v arious w orks that present the benet of using the self-re gulated learning model in for - eign language learning. A range of research papers ha v e presented the benets of using the self-re gulated learning model in foreign language learning. In the English language learning conte xt, incorporating the self- re gulated learning model into the curriculum and training programs encourages autonomous, life-long learning. Abadikhah et al. [17] in v estig ated EFL uni v ersity learners’ attitudes to w ar ds the strate gies of self-re gulated learning in writing academic paper s. The study compared the attitudes of tw o groups in the application of the self-re gulated learning model. It set out to establish whether academic education assists learners in becoming self-re gulated writers. Assessing learners’ attitudes in applying self-re gulated strate gies in their writing may be benet the design of academic writing courses. The learners’ attitudes assessment can pro vide detailed and highly rele v ant information to help instructors enhance their learners’ performance. Instructors ha v e an impor - tant role in assisting learners to become self-re gulated writers. Moreo v er , Karami et al. [4] tried to answer the questions re g arding the ef fect of digital technology on the writing procienc y of learner and the self-re gulated strate gies usage in the conte xt of English learning as a foreign language. The ability of the self-re gulated strate gies is correlated to a higher le v el of writing achie v ement in an en vironment of digital technology . 2.2. Natural language pr ocessing (NLP) r esour ces and ser vices NLP aims to use the technique to mak e the computer system understand the natural language te xt or speech [18]. There are six le v els of NLP tasks [11]: morphological, le xical analysis, syntactic analysis, semantic analysis, pragmatics, and discourse. In this paper , le xical and syntactic NLP techniques were set as a learning en vironment to help the learners compose tar get sentences in English, as sho wn in T able 1. Moreo v er , pre vious w orks [19]–[21] relate to impro ving the NLP process with linguistic kno wledge for impro ving w ord alignment of SMT . Those proposed techniques are also applied to set as learning en vironments such as the dictionary , P art of Speech (POS) tagging, and tenses detection. T able 1. List of components in NLP and their grammatical aspects Le v el of NLP Processes Grammatical Aspects Components Le xical Le v el V ocab ulary Dictionary Plurality Syntactic Le v el Sentence Structure and T enses POS V erb P attern W ord Alignment 2.3. Backgr ound of lear ning analytics Interpreting and e v aluating the qualities o f acti vities, strate gies, goals and re gulation in v olv ed in self- re gulated learning model is some what complicated. Learning beha viors data g athered in a digital learning en vironment are instrumental to address these challenges [22]. Ho we v er , the ra w data alone are insuf cient to guide practice or shape theory . Therefore, learning analytics has a role to play in impro ving the ef fecti v eness of learning. Ther e are v arious w orks for applying learning analytics to the education eld. Learning analytics reports data analysis that describes features or f actors that inuence the self-re gulated model [23]. Analysis of Indonesian J Elec Eng & Comp Sci, V ol. 25, No. 1, January 2022: 460–473 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 ❒ 463 e-learning in f actors of culture, technology or infrastructure, and content satisf action that the analyzed results can be used to de v elop the proper e-learning in a remote city of Indonesia [24]. Since learning analytics are a supporting tool in the digital en vironment, this paper uses l earning analytics to analyze learning beha viors of writing. All learner beha viors in the log le are analyzed in the learning analytics process. This paper aims to in v estig ate learning beha vior that ho w the pro vided components in the learning en vironment reect the performance of English writing. In addition, learning analytics are used to nd learning beha vior patterns which cate gorize a group of the learner . 3. THE PR OPOSED SYSTEM The de v elopment of the computer -based learning system for English writing in Thai EFL learners ai ms for three tasks. First, the system inte grated tw o discipli n e s between the pedagogical model and components of NLP . The s elf-re gulated learning model is a pedagogical model that encourages self-learning, f acilitated by the use of a computer -based learning system. data analysis The components of NLP are helping to learn and compose English sentences. Second, the system aims to collect the learning beha vior in case: English writing for Thai EFL learners. The system is designed to automatically record learning beha viors while composing English sentences. Third, writing beha viors are analyzed for nding the English writing beha vioral patterns of Thai EFL learners. There are three main tasks that support designing and de v eloping processes of the system. 3.1. System pr ocess f or lear ning analytics This paper proposes three main processes : learning prole acquisition, learning beha vior and lear ning analytics, as sho wn in Figure 2. All three main processes w ork coherently a s starting with the process of learning prole acquisition. First, the learning prole acquisition process aims to get information on e xisting English writing skills. Ne xt, the learning beha vior collection is the process for recording the learning beha vior into the log les store in the data log storage. Finally , the process of learning analytics analyzes data of writing beha vior from the log le. The analysis result wil l conduct to dene the beha vior pattern of Thai EFL learners in the case of English writing. The patterns of learning beha vior use to reect learners or impro v e the learning materials or curriculum. Figure 2. System o v ervie w of the computer -based learning system for learning analytics 3.1.1. Lear ning pr ole acquisition The learning prole acquisition is an initial process that uses tw o steps, re gistration and acquisition of e xisting English skills, to get information from the learner . Firstly , learners pro vide personal and educational information on a re gistration form. Ne xt, the English grammatical skill acquisition step uses to get the e xisting writing skills. Learners test to compose the pro vided sentences without assisting tool for getting a learning prole that reects the learners e xisting grammatical skills in three aspects: v ocab ulary usage, sentence type understanding and tense usage. Then, all answers are scored [25] and cate gorized into one of three le v els (basic, intermediate or adv anced) in relation to their English grammatic al skill before learners access to the process of collection the learning beha vior . 3.1.2. Lear ning beha vior collection Learning beha vior collection is needed for the learning analytics process. This process connects to the data log store for col lecting learners’ beha vior that is important data to analyze by the process of learning analytics. Furthermore, this process is an inte grated process for encouraging learning skills by applying the concepts of self-re gulated learning model and components of NLP into four subprocesses: source sentence De velopment of computer -based learning system for learning behavior analytics (Kanyala g Phodong) Evaluation Warning : The document was created with Spire.PDF for Python.
464 ❒ ISSN: 2502-4752 assignment, component selection, writing beha vior monitoring and answering for self-reection, as sho wn in Figure 3. The strate gies of the self-re gulated learning are applied to the w orko w to support self-learning in the computer -based learning en vironment. The components of NLP are deplo yed to the component for writing guidelines into the system. When the collected beha viors are analyzed, t he data of component selection can reect the grammatical skills of learners. The model of self-re gulated learning encourages the interaction between person, beha vior , and en- vironmental f actors for ef fecti v e learning [8]. According to the denition of three main phases [6], [7], the forethought phase is a goal-setting about the learner’ s need to learn. The performance phase is collecting the learning beha vior . Learners’ actions with the pro vided learning en vironment and i n f orm their progress. The self-reection phase is self-assessment and beha vior adaption for increasing the ef fecti v e method of learning. Therefore, this process is designed according to the three main concepts of self-re gulated phases for process ef cienc y [26], as illustrated in Figure 3. (a) (b) Figure 3. A relation between phases of (a) self-re gulated learning model and (b) subprocesses of the system a. Source sentence assignment: After learners nish the learning prole acquisition process, the y access learning beha vior collection for recording writing beha vior . The process of learning beha vior collection starts wi th a subprocess of source sentence assignment to practice writing English sentences. The learner selects the source sentence by themselv es for trying to c o m pose the tar get sentence completely . Since learners’ decision to select source sentences by themselv es. This action relates to set the goal of the fore- thought phase in the self-re gulated learning model as sho wn in Figure 3. The source sentence selection indicates the learner set the goal for composing the complete tar get sentence in English. b . Component selection: This subprocess is designed to include the components of NLP that is an inte- gration pr o c ess between the method of NLP and the educational model. The details of the components of NLP are described in section 3.2. These components of NLP are designed to help learners compose English sentences and to moti v ate them in their writing. The selected component by learners will demon- strate their grammatical needs through the dened components selection. When the source sentence is assigned by the learner , the pro vided components are used to assist for tar get sentence composition. All selected components and time usage are recorded in the log le. Moreo v er , component selection is also related to the forethought phase of the self-re gulated learning model, as sho wn in Figure 3. Component selection by the users themselv es indicates the y ha v e a plan to write the English sentence properly . c. Writing beha vior monitoring: The subprocess is designed to allo w monitoring learners to monitor their progress in sentence composition. The system re cords all acti vities that since the learners select source sentences, selects all NLP components for writing guidelines, until the y submit the tar get sentences. When learners nish composing all tar get sentences, this subprocess will proces s the acti vity parameters in the log le and sho w results for learners’ observ ation, namely: amount of sentences, the selected component, and time usage. Since the learners monitor or observ e their writing performance results by themselv es that relates to the concept of self-observ ation in the performance phase (sho wn in Figure 3). d. Answering for self-reection: In the last subprocess of l earning beha vior collection, the learners answer a self-reection questionnaire containing questions to do with their writing [17]. This subprocess helps learners to reect on their writing beha vior , some of which learners may be able to use in adapting their subsequent writing. The learners’ reection and beha vior adaption that related to self-reaction of the self-reection phase in the model of self-re gulated learning are sho wn in Figure 3. Indonesian J Elec Eng & Comp Sci, V ol. 25, No. 1, January 2022: 460–473 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 ❒ 465 3.1.3. Lear ning analytics Learning analytics are designed to analyze learning beha vior from beha vioral data in the log les. The analysis process aims to ana lyze writing beha vior using the computer -based learning system. The process also analyzes the pro vided components in order to reect the writing performance. This process uses a statistical analysis method [27] to nd out the learning beha vior pattern. The statistical met hod determines the writing beha vior in the beha vior transition form. The result of beha vior patterns can support for consideration of writing procienc y . Moreo v er , the process supports nding the best practice of learning patterns that use the suggestions of other learners. 3.2. Lear ning en vir onment of the computer -based lear ning system The learning en vironment for writing gu i delines consists of tw o main materials: learning materi als and NLP materials as sho wn in Figure 4. The learning materials are composed of English writing tasks to practice for English sentence composition and instruction for introducing system usage. The NLP materials include components to help the English s entence composition. There are tw o reasons for setting NLP materials to support English writing tasks. Firstly , the NLP materials in v olv e linguistic understanding through NLP processes such as le xical, syntactic and semantic le v els. Second, since man y Thai EFL learners think in Thai before translating their ideas into English sentences, a better understanding of the components of linguistics guidelines can help in the writing of appropriate English sentences. The com po ne n t s of NLP are di vided into tw o le v els: le xical and syntactic, as sho wn in T able 1. The pro vided assisting components of the le xical le v el assist learners to write proper v ocab ulary i.e. dictionary and plurality . The components of the syntactic le v el guide lear ners to use appropriate grammar in sentence structure and tenses, including aspects such as part of speech (POS), v erb pattern and w ord alignment. The screen e xample for assisting components of NLP is sho wn in Figure 5. Figure 4. Learning en vironment of the computer -based learning system for English writing The background of NLP used for applying to create the pro vided components that assist learners to compose the complete tar get sentence as details belo w: a. POS component: This component uses w ord se gmentation and POS tagging by SW A TH [24]. The POS tag set is using based on the ORCHID corpus [28]. b . Dictionary component: This component uses w ord se gmentation by Le xT o+ [29]. Then, the w ord- se g- ment of Thai is matched with the English w ord by using the API of Thai-English LEXiTR ON dictionary [30]. c. V erb pattern component: This component denes the POS tag by SW A TH [31]. Then, the v erb or auxiliary v erb is identied in the tense of their w ord by grammatical attrib utes e xtraction [20]. d. Plurality component: This c o m ponent uses lemmatization to e xtract English plural w ords and transform the w ord using rules of plurality . Then, machine translation is used to match the plural w ord in English with its Thai equi v alent. e. W ord alignment component: This component uses w ord se gmentation and POS tagging by SW A TH [31]. Then, the w ord alignment uses the IBM model of GIZA w ord alignment [32] to align w ords of both languages. De velopment of computer -based learning system for learning behavior analytics (Kanyala g Phodong) Evaluation Warning : The document was created with Spire.PDF for Python.
466 ❒ ISSN: 2502-4752 Figure 5. The sample of the display for assisting components of NLP 4. EXPERIMENT AL DESIGN The e xperiment w as designed using beha vioral data to analyze the learning beha vioral patterns of Thai EFL learners. The beha vioral data were collected by automatic dat a collection (ADC) method [33] that automatically recorded into log les while learners write the English sentence via the computer -based system. The beha vioral sequential analysis method w as used to e xplore the learning beha vior pattern of Thai EFL learners in the case of English writing. 4.1. P articipants The system collects learning beha vior into a log le when learners were writing the English sentence via the computer -based learning system. The learners write English tasks for a duration of about 1 hour . A total of 31 under graduate students parti cipated in this study . Their personal information w as remo v ed during the research processing. All writing acti vities were recorded in the log le for analysis by the beha vioral sequential analysis method. 4.2. Coding scheme The coding schema is required for sequential analysis method [33], [34]. Ho we v er , this study uses the computer -based learning system for English writing that automatically records the learning beha vior log. The learning system is implemented for getting learning beha viors. The coding process is based on learner beha viors that operating with the system. When learners use the pro vided learning system to practice English composition, writing beha vior such as “composition”, “selection”, “insertion”, “modication” and “deletion” are recorded in the log les. Then, all data of writing beha vior are used to generate the patterns of learning beha vior . a. Composition (CP): When learners type to compose the tar get sentence in English, the y can type in the pro vided te xtbox. While learners type each w ord in the sentence, all typing will be recorded in the log le. b . Selection (SL): When learners are interested in the components of NLP for assisting sentence composi- tion, the y can select a particular component (or components). Then, all actions of component selection will be collected in the log le. The component of NLP consists of v e components: dictionary , POS, v erb pattern, plurality and w ord alignment. − Dictionary selection (SL-dict): When learners click the dictionary b utton, this indicates their inter - est in the appropriate w ords for composing each sentence. Indonesian J Elec Eng & Comp Sci, V ol. 25, No. 1, January 2022: 460–473 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 ❒ 467 − POS selection (SL-POS): Sometimes, learners are confused about which part of speech a w ord belongs to, such as mistaking noun forms and v erb forms in a sentence. When learners click the POS b utton, this indicates their desire to increase condence in the part of speech of w ords. − V erb pattern (SL-v erb): When learners click the v erb pattern b utton this indicates their interest in the structure of tenses in each sentence. − Plurality (SL-plural): Due to dif ferences in relation to s ingular and plural nouns between Thai and English, there are man y dif ferent rul es re g arding pluralization. When learners click the plurality b utton, this indicates their interest in using the singular or plural nouns in each sentence. − W ord alignment (SL-align): When learners click the w ord alignment b utton, this i nd i cates their interest in the order of w ords and pairs of w ords that are aligned in Thai and English sentences. c. Insertion (IS): When learners demand to add some w ords or phrases into the tar get sentence, the y can mo v e the cursor to the desired position and type additional w ords or phrases into the sentence. Then, these actions will be recorded in the log le. d. Modication (MD): When learners w ant to delete some w ords or partial in the tar get sentence, the y can mo v e the cursor to the desired position and click the backspace b utton to delete some w ords or parts of the sentence. Then, these actions will be recorded in the log le. e. Deletion (DL): Learners can click the “deletion” b utton when the y w ant to compose the ne w tar get sentence and delete a whole sentence. Then, the te xt box will be cleared. Ne xt, learners compose the ne w tar get sentence into the same te xt box. These actions are recorded in the log le. 5. BEHA VIORAL LEARNING AN AL YTICS In this paper , learning analytics aims to analyze writing beha viors by using the met hod of beha vioral sequential analysis [27] to determine beha vior transitions. The analysis process used to analyze learning beha v- ior with the assisting component in the pro vided learning en vironment reects the beha vior of Engl ish writing. The analysis of learning beha vior is used to in v estig ate all beha vior for nding the learning beha vior pattern. The beha vioral sequential analysis is a statistical anal ysis method that uses the sequential analysis matrix to calculate the beha vioral transition [34]. The method uses calculation of the frequenc y of the beha viors sequence and the z-v alue to determine the beha vior transition. Results greater than 1.96 indicated beha vior sequences that reached st atistical signicance [27], [34]. The sample of the matrix of a sequential beha vior series is calculated to z- v alue, as sho wn in Figure 6. Then, the z-v alues were greater than 1.96 were selected to generate the learning beha vior transition. Figure 6. The sample of calculation for sequential beha vior frequenc y to z-v alue 5.1. Analysis of indi vidual lear ning beha vior patter n based on existing english skills The 31 participants were separated into three groups based on e xisting English skills (basic, int erme- diate or adv anced). The writing beha viors of indi vidual learners were analyzed using the beha vioral sequential De velopment of computer -based learning system for learning behavior analytics (Kanyala g Phodong) Evaluation Warning : The document was created with Spire.PDF for Python.
468 ❒ ISSN: 2502-4752 analysis method. This method starts by dening the coding schemes from the writing beha viors which collect beha viors while learners use the pro vided computer -based system. The coding schemes represent the writing beha vior of learners. The frequencies of sequential beha vior were calculated into the matrix of a series of sequential beha vior . Then, all frequencies are calculated to z-v alue for conducting to e xplore the writing be- ha vior patterns. A z-v alue greater than 1.96 indicat es the beha vior sequences reach signicance. The beha vior transition of each learner used to represent the signicant beha vior sequences as illustrated in Figures 7 to 9. 5.1.1. The indi vidual lear ning beha vior patter n in the basic le v el The indi vidual beha vior pattern of 14 learners in the basic le v el as sho wn in Figure 7. All indi vidual beha vior patterns of basic le v el were separated into v e groups: a. Learning Beha vior P attern 1: “modication” has sequential correlations with “composition” b . Learning Beha vior P attern 2: “dictionary selection” has sequential correlations with “composition” c. Learning Beha vior P attern 3: “w ord alignment selection” has sequential correlations with “composition” d. Learning Beha vior P attern 4: “v erb pattern selection” has sequential correlations with “composition” e. Learning Beha vior P attern 5: “insertion” has sequential correlations with “composition” Analysis of the v e groups of learning beha vior patterns indicated that the ‘basic’ group l earners used the NLP components of dictionary , w ord alignment and v erb pattern to assist them in composing English sentences. Figure 7. The indi vidual learning beha vior transition of learner in the basic le v el Indonesian J Elec Eng & Comp Sci, V ol. 25, No. 1, January 2022: 460–473 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 ❒ 469 Figure 8. The indi vidual learning beha vior transition of learner in the intermediate le v el Figure 9. The indi vidual learning beha vior transition of learner in the adv anced le v el 5.1.2. The indi vidual lear ning beha vior patter n in the intermediate le v el The indi vidual beha vior pattern of 8 learners in the intermediate le v el as sho wn in Figure 8. All indi vidual beha vior patterns of intermediate le v el are separated into four groups: a. Learning Beha vior P attern 1: “modication” has sequential correlations with “composition” b . Learning Beha vior P attern 2: “dictionary selection” has sequential correlations with “composition” c. Learning Beha vior P attern 3: “w ord alignment selection” has sequential correlations with “composition” d. Learning Beha vior P attern 4: “insertion” has sequential correlations with “composition” Analysis of the four groups of learning beha vior patterns indicated that the ‘intermediate’ group l earn- ers used the NLP components of dictionary and w ord alignment for assisting to compose the English sentences. De velopment of computer -based learning system for learning behavior analytics (Kanyala g Phodong) Evaluation Warning : The document was created with Spire.PDF for Python.