This review pursues a twofold goal, the first is to preserve and enhance the chronicles of recent educational data mining (EDM) advances development; the second is to organize, analyze, and discuss the content of the review based on the outcomes produced by a data mining (DM) approach. Thus, as result of the selection and analysis of 240 EDM works, an EDM work profile was compiled to describe 222 EDM approaches and 18 tools. A profile of the EDM works was organized as a raw data base, which was transformed into an ad-hoc data base suitable to be mined. As result of the execution of statistical and clustering processes, a set of educational functionalities was found, a realistic pattern of EDM approaches was discovered, and two patterns of value-instances to depict EDM approaches based on descriptive and predictive models were identified. One key finding is: most of the EDM approaches are ground on a basic set composed by three kinds of educational systems, disciplines, tasks, methods, and algorithms each. The review concludes with a snapshot of the surveyed EDM works, and provides an analysis of the EDM strengths, weakness, opportunities, and threats, whose factors represent, in a sense, future work to be fulfilled.

https://romisatriawahono.net/lecture/rm/survey/machine%20learning/Ayala%20-%20Educational%20Data%20Mining%20-%202014.pdf

Review: Educational data mining: A survey and a data mining-based analysis of recent works

While exam-style questions are a fundamental educational tool serving a variety of purposes, manual construction of questions is a complex process that requires training, experience, and resources. This, in turn, hinders and slows down the use of educational activities (e.g. providing practice questions) and new advances (e.g. adaptive testing) that require a large pool of questions. To reduce the expenses associated with manual construction of questions and to satisfy the need for a continuous supply of new questions, automatic question generation (AQG) techniques were introduced. This review extends a previous review on AQG literature that has been published up to late 2014. It includes 93 papers that were between 2015 and early 2019 and tackle the automatic generation of questions for educational purposes. The aims of this review are to: provide an overview of the AQG community and its activities, summarise the current trends and advances in AQG, highlight the changes that the area has undergone in the recent years, and suggest areas for improvement and future opportunities for AQG. Similar to what was found previously, there is little focus in the current literature on generating questions of controlled difficulty, enriching question forms and structures, automating template construction, improving presentation, and generating feedback. Our findings also suggest the need to further improve experimental reporting, harmonise evaluation metrics, and investigate other evaluation methods that are more feasible.

/pdf/a-systematic-review-of-automatic-question-generation-for-4or9ww94ie.pdf

A Systematic Review of Automatic Question Generation for Educational Purposes

The field of adaptive e-learning is continuously developing. More research is being conducted in this area as adaptive e-learning aims to provide learners with adaptive learning paths and content, according to their individual characteristics and needs, which makes e-learning more efficient and effective. The learner model, which is a representation of different learner’s characteristics, plays a key role in this adaptation. This paper presents a systematic literature review about learner modelling during the last 5 years, describing the different modelled characteristics and the adopted modelling techniques and modeling types: automatic modeling and collaborative modeling. 107 publications were selected and analyzed, and six categories of the modelled characteristics were identified. This literature review contributes to the identification of the learners’ individual traits and presents the most used modelling techniques for each of them. It also identifies the latest research trends of Learner Modeling and generates future research directions in this field.

Learner modelling: systematic review of the literature from the last 5 years

Designing good multiple choice questions (MCQs) for education and assessment is time consuming and error-prone. An abundance of structured and semi-structured data has led to the development of automatic MCQ generation methods. Recently, ontologies have emerged as powerful tools to enable the automatic generation of MCQs. However, current question generation approaches focus on knowledge recall questions. In addition, questions that have so far been generated are, compared to manually created ones, simple and cover only a small subset of the required question complexity space in the education and assessment domain. In this paper, we focus on addressing the limitations of previous approaches by generating questions with complex stems that are suitable for scenarios beyond mere knowledge recall. We present a novel ontology-based approach that exploits classes and existential restrictions to generate case-based questions. Our contribution lies in: (1) the specification of procedure for generating case-based questions which involve (a) assembling complex stems, (b) selecting suitable options, and (c) providing explanations for option correctness/incorrectness, (2) an implementation of the procedure using a medical ontology and (3) and evaluation of our generation technique to test question quality and their suitability in practise. We implement our approach as an application for a medical education scenario on top of a large knowledge base in the medical domain. We generate more than 3 million questions for four physician specialities and evaluate our approach in a user study with 15 medical experts. We find that using a stratified random sample of 435 questions out of which 316 were rated by two experts, 129 (30%) are considered appropriate to be used in exams by both experts and a further 216 (50%) by at least one expert.

/pdf/ontology-based-generation-of-medical-multi-term-mcqs-2ed0fnamf4.pdf

Ontology-Based Generation of Medical, Multi-term MCQs

When information from traditional recommender systems is augmented with information about user relationships that social networks store, more successful recommendations can be produced. However, this information regarding user relationships may not always be available, since some users may not consent to the use of their social network information for recommendations or may not have social network accounts at all. Moreover, the rating data (categories and characteristics of products) may be unavailable for a recommender system. In this paper, we present an algorithm that can be applied in any social network-aware recommender system that utilizes the users' ratings on items and users' social relations. The proposed algorithm addresses the issues of limited social network information or limited collaborative filtering information for some users by adapting its behavior, taking into account the density and utility of each user's social network and collaborative filtering neighborhoods. Through this adaptation, the proposed algorithm achieves considerable improvement in rating prediction accuracy. Furthermore, the proposed algorithm can be easily implemented in recommender systems.

An Adaptive Social Network-Aware Collaborative Filtering Algorithm for Improved Rating Prediction Accuracy

Constraint-based student modelling (CBM) is an important technique employed in intelligent tutoring systems to model student knowledge to provide relevant assistance. This paper introduces the Math Story Problem Tutor (MAST), a Web-based intelligent tutoring system for probability story problems, which is able to generate problems of different contexts, types and difficulty levels for self-paced learning. Constraints in MAST are specified at a low-level of granularity to allow fine-grained diagnosis of the student error. Furthermore, MAST extends CBM to address errors due to misunderstanding of the narrative story. It can locate and highlight keywords that may have been overlooked or misunderstood leading to an error. This is achieved by utilizing the role of sentences and keywords that are defined through the Natural Language Generation (NLG) methods deployed in the story problem generation. MAST also integrates CBM with scaffolding questions and feedback to provide various forms of help and guidance to the student. This allows the student to discover and correct any errors in his/her solution. MAST has been preliminary evaluated empirically and the results show the potential effectiveness in tutoring students with a decrease in the percentage of violated constraints along the learning curve. Additionally, there is a significant improvement in the results of the post–test exam in comparison to the pre-test exam of the students using MAST in comparison to those relying on the textbook

/pdf/constraint-based-student-modelling-in-probability-story-23th1963fi.pdf

Constraint-based Student Modelling in Probability Story Problems with Scaffolding Techniques

To address the increase in volume of data streams online users interact with, there are a growing number of tools and models to summarize and extract information. These tools use prediction models to personalize and extract useful information. However, data streams are highly prone to the phenomena of concept drift, in which the data distribution changes over time. To maintain the performance level of these models, models should adapt to handle the existence of adrift. In this work, we present the Incremental Knowledge Concept Drift (IKCD) algorithm, an adaptive unsupervised learning algorithm for recommendation systems in news data stream. Data modelling in IKCD uses k-means clustering to determine the occurrence of a drift while avoiding the dependency on the availability of data labels. Once a drift is detected, new retraining data is composed from the old and new concept. IKCD is tested using synthetic and real benchmark datasets from various domains, which demonstrate the different drift types and with different rate of change. Experimental results illustrate an enhanced performance with respect to (a) reducing model sensitivity to noise, (b) reducing model rebuilding frequency up to 50% in case of re-occurring drift and (c) increasing accuracy of the model by about 10% with respect the accuracy of confidence distribution batch detection algorithm.

/pdf/detection-and-handling-of-different-types-of-concept-drift-2bfswjnvnb.pdf

Detection and Handling of Different Types of Concept Drift in News Recommendation Systems

Bayesian based adaptive question generation technique

Purpose




The purpose of this paper is to present an algorithm to generate story problems via controlled parameters in the domain of mathematics. The generation process is performed in the problem generation module in the context of an intelligent tutoring system suggested in this paper. Controlling the question parameters allows for adapting the generated questions according to the specific student needs. Story problems are selected since they are one of the most important types of problems in mathematics, as they help train students to apply their knowledge to real-world problems. Such problems target improving different student’s skills including literacy skills through reading the problem, recognizing the embedded mathematical information, and applying the required arithmetic operators.




Design/methodology/approach




Natural language generation (NLG) techniques are used to control the difficulty level of the generated story problem header in addition to effecting variations from the natural language point of view. The proposed NLG technique is based on different separated knowledge categories to provide flexibility in the generation process and allow porting the module to other contexts, domains, and to other natural languages without a complete redesign.




Findings




The approach has been empirically evaluated, and the results show that the generated problems are sound, clear, and naturally readable. This is in addition to the usability of the tutoring system itself.




Research limitations/implications




The generation technique is confined to the problem described using rhetorical schemas. Nevertheless, it can generate any problem provided that the rhetorical schema is available.




Originality/value




Most story problems generation systems limit the variation of the story problems to formulating the sentences that describe the story problem and the associated mathematical operations. In contrast, this paper presents a story problems generation technique that allows variations in the structure of the narrative story as well as the context, sentences, wordings, and mathematical operations. This variability allows assessing different student skills along different dimensions with gradually increasing difficulty levels.

Generating story problems via controlled parameters in a web-based intelligent tutoring system

In this paper we aim to estimate the differential student knowledge model in a probabilistic domain within an intelligent tutoring system. The student answers to questions requiring diagnosing skills are used to estimate the actual student model. Updating and verification of the model are conducted based on the matching between the student's and model answers. Two different approaches to updating are suggested, i) coarse and ii) refined model updating. Moreover, the effect of the order of which questions are presented to the student is investigated. Results suggest that the refined model, although takes more computational resources, provides a slightly better approximation of the student model. In addition, the accuracy of the algorithm is highly insensitive to the order of which the questions are presented, more so when using the refined model updating approach..

/pdf/utilizing-diagnosing-problems-in-a-probabilistic-domain-to-3ri5v6pprf.pdf

Nabila Khodeir

Papers

Constraint-based Student Modelling in Probability Story Problems with Scaffolding Techniques

Detection and Handling of Different Types of Concept Drift in News Recommendation Systems

Bayesian based adaptive question generation technique

Generating story problems via controlled parameters in a web-based intelligent tutoring system

Utilizing Diagnosing Problems in a Probabilistic Domain to Build Student Models