Automatic Essay Grading System for Japanese Language Examination Using Winnowing Algorithm

The demand of scoring natural language responses has created a need for new computational tools that can be applied to automatically grade student essays. Systems for automatic essay assessment have been commercially available since 1990's. However, the progress in the field was obstructed by a lack of qualitative information regarding the effectiveness of such systems. Most of the research in automatic essay grading has been associated with English writing due to its widespread use and the availability of more learner collection and language processing software for the language. In addition, there is large number of commercial software for grading programming assignments automatically. In this work, we investigate document semantic similarity based on Latent Semantic Analysis (LSA) and on Latent Dirichlet Allocation (LDA). We use an open-source Python software, Gensim, to develop and implement an essay grading system able to compare an essay to an answer-key and assign to it a grade based on semantic similarity between the two. We test our tool on variable-size essays and conduct experimental tests to compare the results obtained from human grader (professor) and those obtained from the automatic grading system. Results show high correlation between the professor grades and the grades assigned by both modeling techniques. However, LSA-based modeling showed more promising results than the LDA-based method.

This paper aims to investigate the effectiveness of automatic grading systems, with a focus on their uses within Computer Science. Automatic grading systems have seen a rise in popularity in recent years with publications concerning automatic grading systems usually linked to a specific system. This paper will discuss the factors that need to be considered when using automatic grading, regardless of which system is being used, and will make recommendations for each factor. This discussion is based on the authors' experience of using an automatic grading system in a CS1 environment. From the research conducted, many elements should be considered when using these systems. These include how the code will be tested, the need for plagiarism checks and how marks are awarded. The findings of this study suggest there is a lack of defined standards when using these systems. This analysis of the considerations provides valuable insight into how these systems should be used and what the standards should be built on.

Multiple-choice (MC) forms have become the dominant, fast, reliable and the easiest tools of assessment in learning environments such as schools. Present automatic grading systems for such forms, though fast and accurate are either rated as expensive due to materials and equipments they require for their operation or can only operate on a particular set of forms or papers. Part of the main reasons this is so is due to the inability of such systems to generate a template from an unknown form or another form produced by a similar system. As a novel step to creating a more dynamic and efficient automatic MC test grading system, we present in this paper a template generation algorithm capable of generating a template from any type of bubble MC test form. The empty sheet is first scanned into digital form and then normalized to a given size. To consistently remove non-relevant objects such as lines from the image, a unique combination of Hough transform (HT) and Region of interest (ROI) is proposed. Finally, the true answer regions are segmented using region processing by calculating their sizes and positions, this result in the final template. The preliminary results showed that the algorithm is dynamic with overall high precision accuracy even at low resolutions.

The quality of malt barley is a key requirement in the brewery industry. Manual inspection of barley quality is time-consuming, expensive, and prone to human bias, which can ultimately affect the final product. Therefore, an automatic and accurate grading system for barley grain is essential to ensure the consistent quality of beer. This paper presents an automatic barley grain grading system using a deep learning approach based on the texture features of the grains. Specifically, we employed a convolutional neural network (CNN) to classify barley grains into predefined qualitycategories. Unlike the conventional approach of using raw images of barley grains, our method uses texture features as inputs to the CNN model. To train and test the system, we collected images of barley grains in Ethiopia and categorized them into four quality classes (Grade 1, Grade 2, Grade 3, and Under-grade) according to the Ethiopian Standards Authority specifications. Experimental results demonstrate that using texture features instead of raw images significantly improves accuracy in determining the quality of barley grains, achieving an average testing accuracy of 98.61% and a training accuracy of 99.81%. Additionally, we compared the performance of the proposed system with three other methods—ResNet-50, DenseNet-121, and VGGNet-19. The proposed system outperformed these methods, demonstrating superior accuracy. • A barley grading dataset was created using Ethiopian Standard Agency criteria. • A model was built by integrating segmentation, feature extraction, and CNN methods. • CNN training was modified by adding segmentation and texture feature extraction. • Performance was improved over ResNet-50, DenseNet-121, and VGGNet-19.

Reliable and pedagogically sound automated feedback and grading systems are highly coveted by educators. Automatic grading systems are useful for ensuring equity of grading of student submissions to assignments and providing timely feedback on the work. Many of these systems test submissions to assignments based on test cases and the outputs that they achieve, while others use unit tests to check the submissions. The approach presented in this paper checks submissions based on test cases but also analyses what the students actually wrote in their code. Assignment questions are constructed based around the concepts that the student are currently learning in lectures, and the patterns searched for in their submissions are based on these concepts. In this paper we show how to implement this approach effectively. We analyse the use of an automatic grading system within first year Computer Science programming modules and show that the system is straightforward to use and suited for novice programmers, while providing automatic grading and feedback. Evaluation received from students, demonstrators and lecturers show the system is extremely beneficial. The evaluation shows that such systems allow demonstrators more time to assist students during labs. Lecturers can also provide instant feedback to students while keeping track of their progress and identifying where the gaps in students’ knowledge are.

Optical mark recognition machines are used for performing automatic grading of the exam papers that have multiple choice answers. They use some mathematical operations to achieve recognizing the answers marked by the ones who take the exam. In this study, an automatic grading system developed by the use of Hough transform and a filtering system is proposed. The system introduced brings a new perspective for grading the multiple choice exam papers. It focuses on adapting the energy based circular Hough transform for identifying the marked answer bubbles. The procedure is also combined with a data filtering method known as casual median filter. The filtering system, which targets for detecting the outliers and removing them, is commonly used by the robotics and mechatronics researchers for cleaning the unwanted data. The whole system is verified by testing more than 2500 exam answer sheets of the Technical English course offered to the second year Mechanical Engineering students of the Bulent Ecevit University located in Zonguldak, Turkey. The system performance is also tested by observing the results obtained in three different case studies designed and conducted for different goals.

BackgroundEpidural spinal cord compression (ESCC) assessment is a critical step in surgical intervention for lung cancer spinal metastasis. However, clinicians can only assess compression grades through manual observation, which is subjective and inefficient. The metastatic cancer features on magnetic resonance imaging (MRI) slices are complex and difficult to identify, and there are no publicly labeled datasets currently. The objective of this study was to develop an automatic and efficient system for assessing spinal cord compression grading based on multi-tissue and feature segmentation.MethodsWe constructed a dataset containing 577 lung cancer spinal metastasis MRI slices with multi-tissue semantic segmentation masks and ESCC grading. We then proposed a progressive approach for grading ESCC via segmentation and transfer learning (TL) classification to improve clinical assessment accuracy and efficiency. First, we used an improved encoder-decoder segmentation network with a pyramid pooling module (PPM) on the bottleneck and cross-attention modules on the skip connections to segment the compression-related tissues, such as the spine, spinal cord, and bone tumors. Secondly, we proposed a dual attention network (DAN) with position and channel attention modules, and pre-trained it on ImageNet1000 to improve network performance. Finally, we used the multi-tissue features of the previous segmentation as training data and fine-tuned the DAN via TL to achieve accurate compression grading.ResultsThe segmentation results achieved an average pixel accuracy of 0.819 and an average intersection over union of 0.711. The classification accuracy reached 0.95, and the classification precision was 0.89. These results demonstrate the superiority of the proposed few-shot feature recognition and classification method in multi-target recognition. Compared to the traditional segmentation and classification models, our model exhibits enhanced performance and effectiveness in grading spinal compression.ConclusionsThe spinal cord compression grading in lung cancer spinal metastasis is an important surgical indication. The system developed herein has been verified with an accuracy of 0.95 in compression grading, which will be helpful for orthopedic surgeons to determine reasonable treatment plans for patients.

Handing in a paper or exercise and merely receiving “bad” or “incorrect” as feedback is not very helpful when the goal is to improve. Unfortunately, this is currently the kind of feedback given by Automatic Short Answer Grading (ASAG) systems. One of the reasons for this is a lack of content-focused elaborated feedback datasets. To encourage research on explainable and understandable feedback systems, we present the Short Answer Feedback dataset (SAF). Similar to other ASAG datasets, SAF contains learner responses and reference answers to German and English questions. However, instead of only assigning a label or score to the learners’ answers, SAF also contains elaborated feedback explaining the given score. Thus, SAF enables supervised training of models that grade answers and explain where and why mistakes were made. This paper discusses the need for enhanced feedback models in real-world pedagogical scenarios, describes the dataset annotation process, gives a comprehensive analysis of SAF, and provides T5-based baselines for future comparison.


 Classifiers belong to open class noun. All languages are naturally occupied with classifiers, yet the usage is various depending on how the language treats them. Japanese language, as an obligatory-classifier language, makes the classifiers compulsory, in contrast with an optional-classifier language, Indonesian language. Despite this, Japanese and Indonesian languages treat the classifiers syntactically and semantically similar. This article aims at revealing, both the semantic and syntactic differences and similarities between Japanese and Indonesian classifiers. The analysis shows that in terms of structure, Japanese and Indonesian classifiers are similar; however, each language demonstrates difference in classifier use. Japanese language has nineteen frequently used classifiers including two types which can replace others. They can change their form by fusing with its numeral adjective as in hitori (ichi-nin), ippiki (ichi-hiki), without affecting the whole meaning of the noun phrase. Indonesian classifiers cover of twenty-two optional types. Classifiers are often omitted in spoken form to avoid formality and some of them can also change their form by adding se- prior to the numeral adjective to indicate “one object” as in sebatang rokok.

Cataract is the leading cause of blindness worldwide. Two automatic grading systems are presented in this paper for nuclear cataract and cortical cataract diagnosis respectively. Model-based approach was applied to detect anatomical structure in slit-lamp images. Features were extracted based on the lens structure and severity of nuclear cataract was predicted using Support Vector Machines (SVM) regression. For cortical cataract, the opacity was detected using region growing. The seeds were selected by local thresholding and edge detection in radial direction. Cortical cataract was graded based on the area of cortical opacity. Both of the systems were tested by clinical data and results show that the automatic systems can provide objective grading of cataracts.

Image segmentation is a process that significantly important for machine vision system such as automatic fruit grading system. This process separates an image into several areas to extract the interest object from its background. However, the segmentation task is difficult for isolating the images that captured in outdoor environment. This is due to the existence of non-uniform illumination on the object surface. Technically, different illuminations lead to different intensity on the object surface colour. This condition leads to low quality segmented images and therefore reduces the accuracy of object classification. Image segmentation can be accomplished using several methods such as Otsu, K-means and Fuzzy C-means. However, these three traditional methods have limitations in producing accurate segmented areas due to the existence of illumination on the object surface. Therefore, this paper developed a rule-based segmentation method that is able to segment natural images correctly and accurately. This method uses IF-THEN algorithm to segment the images of interest object. All four segmentation methods are implemented on fruit images and their performance are compared based on visual and quantitative evaluations. The analysis results showed that the new method is capable to produce segmented images with high accuracy rate.

Under the influence of the COVID-19 pandemic, traditional in-person teaching has undergone significant changes. Online courses become an essential education method. However, online teaching lacks adequate evaluation approaches. That's why exams are still indispensable. However, grading short answer exam questions can be an onerous task. In this work, we propose a novel Automatic Short Answer Grading (ASAG) model based on the Sentence BERT model. On the Short Answer Scoring V2.0 dataset, our proposed model shows improvements on accuracy, Marco F1 score, and Weighted F1 score comparing to the results obtained from the BERT model. In addition, we also compare different task functions and different lengths of answers to further evaluate our model’s performance. A better result is achieved when using the regression task function. At the same time, we find that shorter answers’ result is better than the result obtained from longer answers.

Anatomic stage of cancer is a key factor in cancer prognosis and provides vital information in almost all types of cancer. The anatomic stage used to represent stage of cancer which shows attributes of tumor - T, number of affected lymph node -N and distant metastasis - M. Experts retrieve the anatomic stage information from clinical and pathological reports and most of these are unstructured. The main aim of the research study is to propose generalized decision-making tool for most accurate anatomic stage extraction from medical reports of different health organizations. Most previous work considered single health organization for research; extracted only TNM value, did not identify anatomic stage from TNM value; used pathology reports only. This research addresses the extraction of anatomic stage from Breast Cancer medical reports using Natural Language Processing because most of these reports are unstructured and require text processing to extract stage information from them. To detect most accurate anatomic stage, TNM information will be extracted from medical reports using NLP, machine learning technique and rule-based methodology. Average T, N and M accuracy for the rural region is 98%, 94% and 99% respectively whereas for urban region, average accuracy is 95%, 88% and 98% respectively. The average stage accuracy for rural and urban region is 93% and 85% respectively. 87% average stage accuracy achieved in both regions. Anatomic stage extraction with high accuracy in different health organizations of different regions shows that the proposed study is promising and generalized for most accurate anatomic stage extraction.

One thing that needs to be developed by members of POKDARWIS Jehem Village, Bangli Regency, Bali Province is the ability to serve tourists from Japan. Based on this background, Japanese Tourism language learning was carried out. The activity has been carried out twice, namely on July 26, 2024 and August 13, 2024. The name of the activity is Sakura (One day of gathering together). Each session was attended by around 50 participants. Based on the enthusiasm for the activity, it was grouped quite high. The first evaluation was carried out by means of a test. By giving questions, 93 people were able to correctly answer questions related to greetings and Japanese letters, 75 people were able to write self-introductions in Japanese, the rest still needed improvement in answering. Evaluation of Japanese language training was carried out by distributing questionnaires. The substance of the questions asked included satisfaction after learning Japanese with the results of 2 people saying they were quite satisfied, 77 people saying they were satisfied, and 21 people saying they were very satisfied. Some suggestions were given in the questionnaire, including improving the quality of the videos shown, more in-depth training on pronunciation and memorizing Japanese terms, and further training on the correct use of chopsticks. Practical evaluations include non-test evaluations. Non-test evaluations were conducted on the material on introducing oneself in Japanese and the use of chopsticks by participants. In the practice of introducing oneself, most were able to perform well, but they still seemed embarrassed because they were not yet confident in pronouncing Japanese terms. This was also seen when using chopsticks. Many were still not used to using chopsticks because it was not a habit in everyday life.

Diabetic maculopathy is the major cause of irreversible vision loss due to retinopathy and is found in 10% of the world diabetic population. Compulsory mass screening will help to identify the maculopathy at early stage and reduce the risk of severe vision loss. In this paper, we present a computer based system for automatic detection and grading of diabetic maculopathy severity level without manual intervention. The optic disc is detected automatically and its location and diameter is used to detect fovea and to mark the macular region respectively. Next, hard exudates are detected using clustering and mathematical morphological techniques. Based on the location of exudates in marked macular region the severity level of maculopathy is classified into mild, moderate and severe. The method achieves a sensitivity of 95.6% and specificity of 96.15% with 148 retinal images for detecting maculopathy stages in fundus images as comparable to that of human expert.