Bart Pursel - Academia.edu (original) (raw)

Papers by Bart Pursel

Proceedings of the AAAI Conference on Artificial Intelligence

Concept prerequisite learning focuses on machine learning methods for measuring the prerequisite ... more Concept prerequisite learning focuses on machine learning methods for measuring the prerequisite relation among concepts. With the importance of prerequisites for education, it has recently become a promising research direction. A major obstacle to extracting prerequisites at scale is the lack of large-scale labels which will enable effective data-driven solutions. We investigate the applicability of active learning to concept prerequisite learning.We propose a novel set of features tailored for prerequisite classification and compare the effectiveness of four widely used query strategies. Experimental results for domains including data mining, geometry, physics, and precalculus show that active learning can be used to reduce the amount of training data required. Given the proposed features, the query-by-committee strategy outperforms other compared query strategies.

Proceedings of the AAAI Conference on Artificial Intelligence

Prerequisite relations among concepts play an important role in many educational applications suc... more Prerequisite relations among concepts play an important role in many educational applications such as intelligent tutoring system and curriculum planning. With the increasing amount of educational data available, automatic discovery of concept prerequisite relations has become both an emerging research opportunity and an open challenge. Here, we investigate how to recover concept prerequisite relations from course dependencies and propose an optimization based framework to address the problem. We create the first real dataset for empirically studying this problem, which consists of the listings of computer science courses from 11 U.S. universities and their concept pairs with prerequisite labels. Experiment results on a synthetic dataset and the real course dataset both show that our method outperforms existing baselines.

Proceedings of the AAAI Conference on Artificial Intelligence

We demonstrate BBookX, a novel system that auto-matically builds in collaboration with a user onl... more We demonstrate BBookX, a novel system that auto-matically builds in collaboration with a user online openbooks by searching open educational resources (OER).This system explores the use of retrieval technologies todynamically generate zero-cost materials such as text-books for personalized learning.

Volume 3: 19th International Conference on Advanced Vehicle Technologies; 14th International Conference on Design Education; 10th Frontiers in Biomedical Devices, 2017

Due to the internet’s increasing global availability, online learning has become a new paradigm f... more Due to the internet’s increasing global availability, online learning has become a new paradigm for distance learning in higher education. While student interactions and reactions are readily observable in a physical classroom environment, monitoring student interactions and quantifying divergence between lecture topics and the topics that interest students are challenging in online learning platforms. Understanding the effects of this divergence is important for monitoring student engagement and aiding instructors, who are focused on improving the quality of their online courses. The authors of this paper propose a topic modeling method, based on latent Dirichlet allocation (LDA), that quantifies the effects of divergence between course topics (mined from textual transcriptions) and student-discussed topics (mined from discussion forums). Correlations between the measured dissimilarities and (a) the number of posts and comments in discussion forums, (b) the number of submitted assi...

2008 11th International Conference on Information Fusion, 2008

Rapid advances in visualization technology and virtual world tools provide opportunities for impr... more Rapid advances in visualization technology and virtual world tools provide opportunities for improvements in multisensor data fusion. These technologies can re-engage the human user in the fusion process, improving multi-analyst collaboration, enhancing data understanding by engaging the analystpsilas visual pattern recognition capabilities, and providing new mechanisms for hypothesis generation and understanding. The virtual world environments can leverage gaming concepts to provide rich story-telling capabilities. Much like the traditional use of cases or logical templates for target identification or event/activity detection, gaming concepts involving characterization of characters and world views can assist the formulation and evaluation of hypotheses for non-traditional targets. As new requirements emerge for fusion systems to support asymmetric warfare and non-traditional operations, these technologies become increasingly important. This paper provides a perspective on these c...

Although learning spaces research is not new, research approaches that target the specific teachi... more Although learning spaces research is not new, research approaches that target the specific teaching and learning experiences of faculty and students who occupy active learning classrooms (ALCs) is nascent. We report on two novels data collection approaches: Flashbacks and Re-captures. Both leverage faculty reflective practice and provide windows into the rich and varied teaching and learning activities that active learning spaces afford. Findings suggest that in ALCs, faculty are easily able to design “activity strings,” multiple active learning activities knitted together within the same instructional period. Further, over time, activity strings become regular occurrences, manifesting as “instructional routines.”

ArXiv, 2019

Strict partial order is a mathematical structure commonly seen in relational data. One obstacle t... more Strict partial order is a mathematical structure commonly seen in relational data. One obstacle to extracting such type of relations at scale is the lack of large-scale labels for building effective data-driven solutions. We develop an active learning framework for mining such relations subject to a strict order. Our approach incorporates relational reasoning not only in finding new unlabeled pairs whose labels can be deduced from an existing label set, but also in devising new query strategies that consider the relational structure of labels. Our experiments on concept prerequisite relations show our proposed framework can substantially improve the classification performance with the same query budget compared to other baseline approaches.

With the emergence of information technology tools, organizational teams often work virtually, re... more With the emergence of information technology tools, organizational teams often work virtually, relying on IT tools to successfully collaborate. Early reports indicated that many of these partially distributed teams (PDTs) experience difficulty, particularly in the areas of geographic distance, temporal distance and cultural distance. To date, the common tools used to facilitate PDT communication and coordination are email, instant messenger, conference calls, and collaborative Internet environments such as Basecamp, Drupal, and others that have features such as wikis, message boards and shared file space. With the emergence of 3D virtual worlds, the technology is present to begin a new era of experiments in PDT collaboration. This research study examined PDTs collaborating in primarily 2D, text-based environments to PDTs collaborating in both 2D environments and a 3D virtual world, ProtoSphere. Data were collected around nine different dependent variables pulled from virtual teaming...

Online Learning Analytics, 2021

Proceedings of the Knowledge Capture Conference, 2017

Distractor generation is a crucial step for fill-in-the-blank question generation. We propose a g... more Distractor generation is a crucial step for fill-in-the-blank question generation. We propose a generative model learned from training generative adversarial nets (GANs) to create useful distractors. Our method utilizes only context information and does not use the correct answer, which is completely different from previous Ontology-based or similarity-based approaches. Trained on the Wikipedia corpus, the proposed model is able to predict Wiki entities as distractors. Our method is evaluated on two biology question datasets collected from Wikipedia and actual college-level exams. Experimental results show that our context-based method achieves comparable performance to a frequently used word2vec-based method for the Wiki dataset. In addition, we propose a second-stage learner to combine the strengths of the two methods, which further improves the performance on both datasets, with 51.7% and 48.4% of generated distractors being acceptable.

2014 ASEE Annual Conference & Exposition Proceedings

focusing on the intersection of technology and pedagogy. Barton works collaboratively with facult... more focusing on the intersection of technology and pedagogy. Barton works collaboratively with faculty across disciplines to explore how emerging technologies and trends, such as MOOCs, digital badges, and learning analytics, impacts both students and instructors.

Contemporary Educational Technology, 2014

As social and collaborative technologies emerge, educators and scholars continue to explore and e... more As social and collaborative technologies emerge, educators and scholars continue to explore and experiment with how these tools might impact pedagogy. For over a decade, educators experimented with the use of blogs in academic settings, a tool that allows for students and instructors to enter into a rich dialogue. With most technology tools, users often leave 'digital footprints' throughout the environment. These footprints, in combination with other sources of data, allow researchers to explore relationships between the tool itself and the different types of end users. This study examines two years of institutional blog data, combined with demographic data to help describe the users of a blog platform. Different clusters of users are uncovered, and various use cases are explored, illustrating how different instructors choose to leverage blogs in the flow of a course. Using analysis of variance (ANOVA) to compare different blogging groups, results show a strong correlation between entry-dominant bloggers and growth in Grade Point Average (GPA) over time. With the rise in popularity of learning analytics, the results of this study might influence future learning analytics tools and systems.

Proceedings of the Thirteenth Workshop on Innovative Use of NLP for Building Educational Applications, 2018

We investigate how machine learning models, specifically ranking models, can be used to select us... more We investigate how machine learning models, specifically ranking models, can be used to select useful distractors for multiple choice questions. Our proposed models can learn to select distractors that resemble those in actual exam questions, which is different from most existing unsupervised ontology-based and similarity-based methods. We empirically study feature-based and neural net (NN) based ranking models with experiments on the recently released SciQ dataset and our MCQL dataset. Experimental results show that feature-based ensemble learning methods (random forest and LambdaMART) outperform both the NN-based method and unsupervised baselines. These two datasets can also be used as benchmarks for distractor generation.

Computer Applications in Engineering Education, 2018

Due to the increasing global availability of the internet, online learning platforms such as Mass... more Due to the increasing global availability of the internet, online learning platforms such as Massive Open Online Courses (MOOCs), have become a new paradigm for distance learning in engineering education. While interactions between instructors and students are readily observable in a physical classroom environment, monitoring student engagement is challenging in MOOCs. Monitoring student engagement and measuring its impact on student performance are important for MOOC instructors, who are focused on improving the quality of their courses. The authors of this work present a semantic network model for measuring the different word associations between instructors and students in order to measure student engagement in MOOCs. Correlation analysis is then performed for identifying how student engagement in MOOCs affect student performance. Real-world MOOC transcripts and MOOC discussion forum data are used to evaluate the effectiveness of this research.

TechTrends, 2017

Interest towards implementing educational gaming into courses within higher education continues t... more Interest towards implementing educational gaming into courses within higher education continues to increase, but it requires extensive amounts of resources to create individual games for each course. This paper is a description of a university's effort to create a custom educational game engine to streamline the game development process within the university. This paper includes a discussion of the institutional effort, the game engine itself, as well as a case study of a game created with the custom FLAG (an HTML5 game engine built to run 2D games on any HTML5 compatible device) game engine to illustrate how the FLAG engine can simplify the game development process and promote game-based learning in higher education.

Journal of Computer Assisted Learning, 2016

Massive open online courses (MOOCs) continue to appear across the higher education landscape, ori... more Massive open online courses (MOOCs) continue to appear across the higher education landscape, originating from many institutions in the USA and around the world. MOOCs typically have low completion rates, at least when compared with traditional courses, as this course delivery model is very different from traditional, fee-based models, such as college courses. This research examined MOOC student demographic data, intended behaviours and course interactions to better understand variables that are indicative of MOOC completion. The results lead to ideas regarding how these variables can be used to support MOOC students through the application of learning analytics tools and systems.

Proceedings of the Association for Information Science and Technology, 2015

While discussion forums in online courses have been studied in the past, no one has proposed a mo... more While discussion forums in online courses have been studied in the past, no one has proposed a model linking messages in discussion forums to a learning taxonomy, even though forums are widely used as educational tools in online courses. In this research, we view forums as information seeking events and use a keyword taxonomy approach to analyze a large amount of MOOC forum data to identify the types of learning interactions taking place in forum conversations. Using 51,761 forum messages from 8,169 forum threads from a MOOC with a 50,000+ enrollment, messages are analyzed based on levels of Bloom's Taxonomy to categorize the scholarly discourse. The results of this research show that interactions within MOOC discussion forums are a learning process with unique characteristics specific to particular cognitive learning levels. Results also imply that different types of forum interactions have characteristics relevant to particular learning levels, and the volume of higher levels of cognitive learning incidents increase as the course progresses.

Proceedings of the 25th International Conference Companion on World Wide Web - WWW '16 Companion, 2016

We describe BBookX, a web-based tool that uses a human-computing approach to facilitate the creat... more We describe BBookX, a web-based tool that uses a human-computing approach to facilitate the creation of open source textbooks. The goal of BBookX is to create a system that can search various Open Educational Resource (OER) repositories such as Wikipedia, based on a set of user-generated criteria, and return various resources that can be combined, remixed, and re-used to support specific learning goals. As BBookX is a work-in-progress, we are in the midst of a design-based research study, where user testing guided multiple rounds of iteration in the design of the user interface (UI) as well as the query engine. From an interface perspective, the challenges we present are the matching of the UI to users' mental models from similar systems, as well as educating users how to best work with the algorithms in an iterative manner to find and refine content for inclusion into open textbooks.

Proceedings of the 25th ACM International on Conference on Information and Knowledge Management, 2016

We present a framework for constructing a specific type of knowledge graph, a concept map from te... more We present a framework for constructing a specific type of knowledge graph, a concept map from textbooks. Using Wikipedia, we derive prerequisite relations among these concepts. A traditional approach for concept map extraction consists of two sub-problems: key concept extraction and concept relationship identification. Previous work for the most part had considered these two sub-problems independently. We propose a framework that jointly optimizes these sub-problems and investigates methods that identify concept relationships. Experiments on concept maps that are manually extracted in six educational areas (computer networks, macroeconomics, precalculus, databases, physics, and geometry) show that our model outperforms supervised learning baselines that solve the two sub-problems separately. Moreover, we observe that incorporating textbook information helps with concept map extraction.

Proceedings of the AAAI Conference on Artificial Intelligence

Concept prerequisite learning focuses on machine learning methods for measuring the prerequisite ... more Concept prerequisite learning focuses on machine learning methods for measuring the prerequisite relation among concepts. With the importance of prerequisites for education, it has recently become a promising research direction. A major obstacle to extracting prerequisites at scale is the lack of large-scale labels which will enable effective data-driven solutions. We investigate the applicability of active learning to concept prerequisite learning.We propose a novel set of features tailored for prerequisite classification and compare the effectiveness of four widely used query strategies. Experimental results for domains including data mining, geometry, physics, and precalculus show that active learning can be used to reduce the amount of training data required. Given the proposed features, the query-by-committee strategy outperforms other compared query strategies.

Proceedings of the AAAI Conference on Artificial Intelligence

Prerequisite relations among concepts play an important role in many educational applications suc... more Prerequisite relations among concepts play an important role in many educational applications such as intelligent tutoring system and curriculum planning. With the increasing amount of educational data available, automatic discovery of concept prerequisite relations has become both an emerging research opportunity and an open challenge. Here, we investigate how to recover concept prerequisite relations from course dependencies and propose an optimization based framework to address the problem. We create the first real dataset for empirically studying this problem, which consists of the listings of computer science courses from 11 U.S. universities and their concept pairs with prerequisite labels. Experiment results on a synthetic dataset and the real course dataset both show that our method outperforms existing baselines.

Proceedings of the AAAI Conference on Artificial Intelligence

We demonstrate BBookX, a novel system that auto-matically builds in collaboration with a user onl... more We demonstrate BBookX, a novel system that auto-matically builds in collaboration with a user online openbooks by searching open educational resources (OER).This system explores the use of retrieval technologies todynamically generate zero-cost materials such as text-books for personalized learning.

Volume 3: 19th International Conference on Advanced Vehicle Technologies; 14th International Conference on Design Education; 10th Frontiers in Biomedical Devices, 2017

Due to the internet’s increasing global availability, online learning has become a new paradigm f... more Due to the internet’s increasing global availability, online learning has become a new paradigm for distance learning in higher education. While student interactions and reactions are readily observable in a physical classroom environment, monitoring student interactions and quantifying divergence between lecture topics and the topics that interest students are challenging in online learning platforms. Understanding the effects of this divergence is important for monitoring student engagement and aiding instructors, who are focused on improving the quality of their online courses. The authors of this paper propose a topic modeling method, based on latent Dirichlet allocation (LDA), that quantifies the effects of divergence between course topics (mined from textual transcriptions) and student-discussed topics (mined from discussion forums). Correlations between the measured dissimilarities and (a) the number of posts and comments in discussion forums, (b) the number of submitted assi...

2008 11th International Conference on Information Fusion, 2008

Rapid advances in visualization technology and virtual world tools provide opportunities for impr... more Rapid advances in visualization technology and virtual world tools provide opportunities for improvements in multisensor data fusion. These technologies can re-engage the human user in the fusion process, improving multi-analyst collaboration, enhancing data understanding by engaging the analystpsilas visual pattern recognition capabilities, and providing new mechanisms for hypothesis generation and understanding. The virtual world environments can leverage gaming concepts to provide rich story-telling capabilities. Much like the traditional use of cases or logical templates for target identification or event/activity detection, gaming concepts involving characterization of characters and world views can assist the formulation and evaluation of hypotheses for non-traditional targets. As new requirements emerge for fusion systems to support asymmetric warfare and non-traditional operations, these technologies become increasingly important. This paper provides a perspective on these c...

Although learning spaces research is not new, research approaches that target the specific teachi... more Although learning spaces research is not new, research approaches that target the specific teaching and learning experiences of faculty and students who occupy active learning classrooms (ALCs) is nascent. We report on two novels data collection approaches: Flashbacks and Re-captures. Both leverage faculty reflective practice and provide windows into the rich and varied teaching and learning activities that active learning spaces afford. Findings suggest that in ALCs, faculty are easily able to design “activity strings,” multiple active learning activities knitted together within the same instructional period. Further, over time, activity strings become regular occurrences, manifesting as “instructional routines.”

ArXiv, 2019

Strict partial order is a mathematical structure commonly seen in relational data. One obstacle t... more Strict partial order is a mathematical structure commonly seen in relational data. One obstacle to extracting such type of relations at scale is the lack of large-scale labels for building effective data-driven solutions. We develop an active learning framework for mining such relations subject to a strict order. Our approach incorporates relational reasoning not only in finding new unlabeled pairs whose labels can be deduced from an existing label set, but also in devising new query strategies that consider the relational structure of labels. Our experiments on concept prerequisite relations show our proposed framework can substantially improve the classification performance with the same query budget compared to other baseline approaches.

With the emergence of information technology tools, organizational teams often work virtually, re... more With the emergence of information technology tools, organizational teams often work virtually, relying on IT tools to successfully collaborate. Early reports indicated that many of these partially distributed teams (PDTs) experience difficulty, particularly in the areas of geographic distance, temporal distance and cultural distance. To date, the common tools used to facilitate PDT communication and coordination are email, instant messenger, conference calls, and collaborative Internet environments such as Basecamp, Drupal, and others that have features such as wikis, message boards and shared file space. With the emergence of 3D virtual worlds, the technology is present to begin a new era of experiments in PDT collaboration. This research study examined PDTs collaborating in primarily 2D, text-based environments to PDTs collaborating in both 2D environments and a 3D virtual world, ProtoSphere. Data were collected around nine different dependent variables pulled from virtual teaming...

Online Learning Analytics, 2021

Proceedings of the Knowledge Capture Conference, 2017

Distractor generation is a crucial step for fill-in-the-blank question generation. We propose a g... more Distractor generation is a crucial step for fill-in-the-blank question generation. We propose a generative model learned from training generative adversarial nets (GANs) to create useful distractors. Our method utilizes only context information and does not use the correct answer, which is completely different from previous Ontology-based or similarity-based approaches. Trained on the Wikipedia corpus, the proposed model is able to predict Wiki entities as distractors. Our method is evaluated on two biology question datasets collected from Wikipedia and actual college-level exams. Experimental results show that our context-based method achieves comparable performance to a frequently used word2vec-based method for the Wiki dataset. In addition, we propose a second-stage learner to combine the strengths of the two methods, which further improves the performance on both datasets, with 51.7% and 48.4% of generated distractors being acceptable.

2014 ASEE Annual Conference & Exposition Proceedings

focusing on the intersection of technology and pedagogy. Barton works collaboratively with facult... more focusing on the intersection of technology and pedagogy. Barton works collaboratively with faculty across disciplines to explore how emerging technologies and trends, such as MOOCs, digital badges, and learning analytics, impacts both students and instructors.

Contemporary Educational Technology, 2014

As social and collaborative technologies emerge, educators and scholars continue to explore and e... more As social and collaborative technologies emerge, educators and scholars continue to explore and experiment with how these tools might impact pedagogy. For over a decade, educators experimented with the use of blogs in academic settings, a tool that allows for students and instructors to enter into a rich dialogue. With most technology tools, users often leave 'digital footprints' throughout the environment. These footprints, in combination with other sources of data, allow researchers to explore relationships between the tool itself and the different types of end users. This study examines two years of institutional blog data, combined with demographic data to help describe the users of a blog platform. Different clusters of users are uncovered, and various use cases are explored, illustrating how different instructors choose to leverage blogs in the flow of a course. Using analysis of variance (ANOVA) to compare different blogging groups, results show a strong correlation between entry-dominant bloggers and growth in Grade Point Average (GPA) over time. With the rise in popularity of learning analytics, the results of this study might influence future learning analytics tools and systems.

Proceedings of the Thirteenth Workshop on Innovative Use of NLP for Building Educational Applications, 2018

We investigate how machine learning models, specifically ranking models, can be used to select us... more We investigate how machine learning models, specifically ranking models, can be used to select useful distractors for multiple choice questions. Our proposed models can learn to select distractors that resemble those in actual exam questions, which is different from most existing unsupervised ontology-based and similarity-based methods. We empirically study feature-based and neural net (NN) based ranking models with experiments on the recently released SciQ dataset and our MCQL dataset. Experimental results show that feature-based ensemble learning methods (random forest and LambdaMART) outperform both the NN-based method and unsupervised baselines. These two datasets can also be used as benchmarks for distractor generation.

Computer Applications in Engineering Education, 2018

Due to the increasing global availability of the internet, online learning platforms such as Mass... more Due to the increasing global availability of the internet, online learning platforms such as Massive Open Online Courses (MOOCs), have become a new paradigm for distance learning in engineering education. While interactions between instructors and students are readily observable in a physical classroom environment, monitoring student engagement is challenging in MOOCs. Monitoring student engagement and measuring its impact on student performance are important for MOOC instructors, who are focused on improving the quality of their courses. The authors of this work present a semantic network model for measuring the different word associations between instructors and students in order to measure student engagement in MOOCs. Correlation analysis is then performed for identifying how student engagement in MOOCs affect student performance. Real-world MOOC transcripts and MOOC discussion forum data are used to evaluate the effectiveness of this research.

TechTrends, 2017

Interest towards implementing educational gaming into courses within higher education continues t... more Interest towards implementing educational gaming into courses within higher education continues to increase, but it requires extensive amounts of resources to create individual games for each course. This paper is a description of a university's effort to create a custom educational game engine to streamline the game development process within the university. This paper includes a discussion of the institutional effort, the game engine itself, as well as a case study of a game created with the custom FLAG (an HTML5 game engine built to run 2D games on any HTML5 compatible device) game engine to illustrate how the FLAG engine can simplify the game development process and promote game-based learning in higher education.

Journal of Computer Assisted Learning, 2016

Massive open online courses (MOOCs) continue to appear across the higher education landscape, ori... more Massive open online courses (MOOCs) continue to appear across the higher education landscape, originating from many institutions in the USA and around the world. MOOCs typically have low completion rates, at least when compared with traditional courses, as this course delivery model is very different from traditional, fee-based models, such as college courses. This research examined MOOC student demographic data, intended behaviours and course interactions to better understand variables that are indicative of MOOC completion. The results lead to ideas regarding how these variables can be used to support MOOC students through the application of learning analytics tools and systems.

Proceedings of the Association for Information Science and Technology, 2015

While discussion forums in online courses have been studied in the past, no one has proposed a mo... more While discussion forums in online courses have been studied in the past, no one has proposed a model linking messages in discussion forums to a learning taxonomy, even though forums are widely used as educational tools in online courses. In this research, we view forums as information seeking events and use a keyword taxonomy approach to analyze a large amount of MOOC forum data to identify the types of learning interactions taking place in forum conversations. Using 51,761 forum messages from 8,169 forum threads from a MOOC with a 50,000+ enrollment, messages are analyzed based on levels of Bloom's Taxonomy to categorize the scholarly discourse. The results of this research show that interactions within MOOC discussion forums are a learning process with unique characteristics specific to particular cognitive learning levels. Results also imply that different types of forum interactions have characteristics relevant to particular learning levels, and the volume of higher levels of cognitive learning incidents increase as the course progresses.

Proceedings of the 25th International Conference Companion on World Wide Web - WWW '16 Companion, 2016

We describe BBookX, a web-based tool that uses a human-computing approach to facilitate the creat... more We describe BBookX, a web-based tool that uses a human-computing approach to facilitate the creation of open source textbooks. The goal of BBookX is to create a system that can search various Open Educational Resource (OER) repositories such as Wikipedia, based on a set of user-generated criteria, and return various resources that can be combined, remixed, and re-used to support specific learning goals. As BBookX is a work-in-progress, we are in the midst of a design-based research study, where user testing guided multiple rounds of iteration in the design of the user interface (UI) as well as the query engine. From an interface perspective, the challenges we present are the matching of the UI to users' mental models from similar systems, as well as educating users how to best work with the algorithms in an iterative manner to find and refine content for inclusion into open textbooks.

Proceedings of the 25th ACM International on Conference on Information and Knowledge Management, 2016

We present a framework for constructing a specific type of knowledge graph, a concept map from te... more We present a framework for constructing a specific type of knowledge graph, a concept map from textbooks. Using Wikipedia, we derive prerequisite relations among these concepts. A traditional approach for concept map extraction consists of two sub-problems: key concept extraction and concept relationship identification. Previous work for the most part had considered these two sub-problems independently. We propose a framework that jointly optimizes these sub-problems and investigates methods that identify concept relationships. Experiments on concept maps that are manually extracted in six educational areas (computer networks, macroeconomics, precalculus, databases, physics, and geometry) show that our model outperforms supervised learning baselines that solve the two sub-problems separately. Moreover, we observe that incorporating textbook information helps with concept map extraction.