{
  "id": 188911,
  "title": "Papers on Knowledge Tracing and Education",
  "url": "/competitions/riiid-test-answer-prediction/discussion/188911",
  "author_name": "Charlie Craine",
  "post_date": "2020-10-05T20:04:29.759000",
  "votes": 158,
  "comment_count": 33,
  "views": 0,
  "content": "<p>Hey everyone!</p>\n<p>I wanted to start a papers thread and build on it, and hope others share as well, papers to gain domain knowledge. I have no domain knowledge in this area so I have downloaded papers that appeared to be relevant after reading their abstracts. I'll be reading these over the coming days and commenting more as I go through all of them. Hope this helps!</p>\n<p>Research Papers:</p>\n<ul>\n<li><p><a href=\"https://arxiv.org/abs/1907.06837\" target=\"_blank\">A Self-Attentive model for Knowledge Tracing</a> - we develop an approach that identifies the KCs from the student's past activities that are \\textit{relevant} to the given KC and predicts his/her mastery based on the relatively few KCs that it picked. </p></li>\n<li><p><a href=\"https://arxiv.org/abs/1910.13197\" target=\"_blank\">Knowledge Tracing with Sequential Key-Value Memory Networks</a> - Can machines trace human knowledge like humans? Knowledge tracing (KT) is a fundamental task in a wide range of applications in education, such as massive open online courses (MOOCs), intelligent tutoring systems, educational games, and learning management systems.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/1912.03072\" target=\"_blank\">EdNet: A Large-Scale Hierarchical Dataset in Education</a> - With advances in Artificial Intelligence in Education (AIEd) and the ever-growing scale of Interactive Educational Systems (IESs), data-driven approach has become a common recipe for various tasks such as knowledge tracing and learning path recommendation.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2001.04841\" target=\"_blank\">Domain Adaption for Knowledge Tracing</a> - With the rapid development of online education system, knowledge tracing which aims at predicting students' knowledge state is becoming a critical and fundamental task in personalized education.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2001.10617\" target=\"_blank\">Systematic Review of Approaches to Improve Peer Assessment at Scale</a> - Peer Assessment is a task of analysis and commenting on student's writing by peers, is core of all educational components both in campus and in MOOC's.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2001.09830\" target=\"_blank\">What's happened in MOOC Posts Analysis, Knowledge Tracing and Peer Feedbacks? A Review</a> - Learning Management Systems (LMS) and Educational Data Mining (EDM) are two important parts of online educational environment with the former being a centralised web-based information systems where the learning content is managed and learning activities are organised (Stone and Zheng,2014) and latter focusing on using data mining techniques for the analysis of data so generated.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2002.07033\" target=\"_blank\">Towards an Appropriate Query, Key, and Value Computation for Knowledge Tracing</a> - Knowledge tracing, the act of modeling a student's knowledge through learning activities, is an extensively studied problem in the field of computer-aided education. Although models with attention mechanism have outperformed traditional approaches such as Bayesian knowledge tracing and collaborative filtering, they share two limitations.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2005.00869\" target=\"_blank\">Generalized Knowledge Tracing: A Constrained Framework for Learner Modeling</a> - Adaptive learning technology solutions often use a learner model to trace learning and make pedagogical decisions.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2005.06139\" target=\"_blank\">Towards Interpretable Deep Learning Models for Knowledge Tracing</a> - As an important technique for modeling the knowledge states of learners, the traditional knowledge tracing (KT) models have been widely used to support intelligent tutoring systems and MOOC platforms.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2005.09109\" target=\"_blank\">Dynamic Knowledge embedding and tracing</a> - In this paper we propose a novel approach to knowledge tracing that combines techniques from matrix factorization with recent progress in recurrent neural networks (RNNs) to effectively track the state of a student's knowledge.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2005.12442\" target=\"_blank\">qDKT: Question-centric Deep Knowledge Tracing</a> - We introduce qDKT, a variant of DKT that models every learner's success probability on individual questions over time. </p></li>\n<li><p><a href=\"https://arxiv.org/abs/2006.16915\" target=\"_blank\">HGKT : Introducing Problem Schema with Hierarchical Exercise Graph for Knowledge Tracing</a> - In this paper, we propose a hierarchical graph knowledge tracing model framework (HGKT) which could leverage the advantages of hierarchical exercise graph and sequence model to enhance the ability of knowledge tracing. Besides, we introduce the concept of problem schema to better represent a group of similar exercises and propose a hierarchical graph neural network to learn representations of problem schemas. </p></li>\n<li><p><a href=\"https://arxiv.org/abs/2007.12324\" target=\"_blank\">Context-Aware Attentive Knowledge Tracing</a> - We propose attentive knowledge tracing (AKT), which couples flexible attention-based neural network models with a series of novel, interpretable model components inspired by cognitive and psychometric models.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2009.05991\" target=\"_blank\">GIKT: A Graph-based Interaction Model for Knowledge Tracing</a> - From the model perspective, previous models can hardly capture the long-term dependency of student exercise history, and cannot model the interactions between student-questions, and student-skills in a consistent way. In this paper, we propose a Graph-based Interaction model for Knowledge Tracing (GIKT) to tackle the above problems.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2008.01169\" target=\"_blank\">Deep Knowledge Tracing with Convolutions</a> - We propose a Convolutional Knowledge Tracing (CKT) model in this paper.</p></li>\n</ul>",
  "messages": [
    {
      "id": 1038477,
      "postDate": "2020-10-05T20:04:29.760Z",
      "content": "<p>Hey everyone!</p>\n<p>I wanted to start a papers thread and build on it, and hope others share as well, papers to gain domain knowledge. I have no domain knowledge in this area so I have downloaded papers that appeared to be relevant after reading their abstracts. I'll be reading these over the coming days and commenting more as I go through all of them. Hope this helps!</p>\n<p>Research Papers:</p>\n<ul>\n<li><p><a href=\"https://arxiv.org/abs/1907.06837\" target=\"_blank\">A Self-Attentive model for Knowledge Tracing</a> - we develop an approach that identifies the KCs from the student's past activities that are \\textit{relevant} to the given KC and predicts his/her mastery based on the relatively few KCs that it picked. </p></li>\n<li><p><a href=\"https://arxiv.org/abs/1910.13197\" target=\"_blank\">Knowledge Tracing with Sequential Key-Value Memory Networks</a> - Can machines trace human knowledge like humans? Knowledge tracing (KT) is a fundamental task in a wide range of applications in education, such as massive open online courses (MOOCs), intelligent tutoring systems, educational games, and learning management systems.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/1912.03072\" target=\"_blank\">EdNet: A Large-Scale Hierarchical Dataset in Education</a> - With advances in Artificial Intelligence in Education (AIEd) and the ever-growing scale of Interactive Educational Systems (IESs), data-driven approach has become a common recipe for various tasks such as knowledge tracing and learning path recommendation.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2001.04841\" target=\"_blank\">Domain Adaption for Knowledge Tracing</a> - With the rapid development of online education system, knowledge tracing which aims at predicting students' knowledge state is becoming a critical and fundamental task in personalized education.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2001.10617\" target=\"_blank\">Systematic Review of Approaches to Improve Peer Assessment at Scale</a> - Peer Assessment is a task of analysis and commenting on student's writing by peers, is core of all educational components both in campus and in MOOC's.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2001.09830\" target=\"_blank\">What's happened in MOOC Posts Analysis, Knowledge Tracing and Peer Feedbacks? A Review</a> - Learning Management Systems (LMS) and Educational Data Mining (EDM) are two important parts of online educational environment with the former being a centralised web-based information systems where the learning content is managed and learning activities are organised (Stone and Zheng,2014) and latter focusing on using data mining techniques for the analysis of data so generated.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2002.07033\" target=\"_blank\">Towards an Appropriate Query, Key, and Value Computation for Knowledge Tracing</a> - Knowledge tracing, the act of modeling a student's knowledge through learning activities, is an extensively studied problem in the field of computer-aided education. Although models with attention mechanism have outperformed traditional approaches such as Bayesian knowledge tracing and collaborative filtering, they share two limitations.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2005.00869\" target=\"_blank\">Generalized Knowledge Tracing: A Constrained Framework for Learner Modeling</a> - Adaptive learning technology solutions often use a learner model to trace learning and make pedagogical decisions.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2005.06139\" target=\"_blank\">Towards Interpretable Deep Learning Models for Knowledge Tracing</a> - As an important technique for modeling the knowledge states of learners, the traditional knowledge tracing (KT) models have been widely used to support intelligent tutoring systems and MOOC platforms.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2005.09109\" target=\"_blank\">Dynamic Knowledge embedding and tracing</a> - In this paper we propose a novel approach to knowledge tracing that combines techniques from matrix factorization with recent progress in recurrent neural networks (RNNs) to effectively track the state of a student's knowledge.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2005.12442\" target=\"_blank\">qDKT: Question-centric Deep Knowledge Tracing</a> - We introduce qDKT, a variant of DKT that models every learner's success probability on individual questions over time. </p></li>\n<li><p><a href=\"https://arxiv.org/abs/2006.16915\" target=\"_blank\">HGKT : Introducing Problem Schema with Hierarchical Exercise Graph for Knowledge Tracing</a> - In this paper, we propose a hierarchical graph knowledge tracing model framework (HGKT) which could leverage the advantages of hierarchical exercise graph and sequence model to enhance the ability of knowledge tracing. Besides, we introduce the concept of problem schema to better represent a group of similar exercises and propose a hierarchical graph neural network to learn representations of problem schemas. </p></li>\n<li><p><a href=\"https://arxiv.org/abs/2007.12324\" target=\"_blank\">Context-Aware Attentive Knowledge Tracing</a> - We propose attentive knowledge tracing (AKT), which couples flexible attention-based neural network models with a series of novel, interpretable model components inspired by cognitive and psychometric models.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2009.05991\" target=\"_blank\">GIKT: A Graph-based Interaction Model for Knowledge Tracing</a> - From the model perspective, previous models can hardly capture the long-term dependency of student exercise history, and cannot model the interactions between student-questions, and student-skills in a consistent way. In this paper, we propose a Graph-based Interaction model for Knowledge Tracing (GIKT) to tackle the above problems.</p></li>\n<li><p><a href=\"https://arxiv.org/abs/2008.01169\" target=\"_blank\">Deep Knowledge Tracing with Convolutions</a> - We propose a Convolutional Knowledge Tracing (CKT) model in this paper.</p></li>\n</ul>",
      "rawMarkdown": "Hey everyone!\n\nI wanted to start a papers thread and build on it, and hope others share as well, papers to gain domain knowledge. I have no domain knowledge in this area so I have downloaded papers that appeared to be relevant after reading their abstracts. I'll be reading these over the coming days and commenting more as I go through all of them. Hope this helps!\n\nResearch Papers:\n* [A Self-Attentive model for Knowledge Tracing](https://arxiv.org/abs/1907.06837) - we develop an approach that identifies the KCs from the student's past activities that are \\textit{relevant} to the given KC and predicts his/her mastery based on the relatively few KCs that it picked. \n\n* [Knowledge Tracing with Sequential Key-Value Memory Networks](https://arxiv.org/abs/1910.13197) - Can machines trace human knowledge like humans? Knowledge tracing (KT) is a fundamental task in a wide range of applications in education, such as massive open online courses (MOOCs), intelligent tutoring systems, educational games, and learning management systems.\n\n* [EdNet: A Large-Scale Hierarchical Dataset in Education](https://arxiv.org/abs/1912.03072) - With advances in Artificial Intelligence in Education (AIEd) and the ever-growing scale of Interactive Educational Systems (IESs), data-driven approach has become a common recipe for various tasks such as knowledge tracing and learning path recommendation.\n\n* [Domain Adaption for Knowledge Tracing](https://arxiv.org/abs/2001.04841) - With the rapid development of online education system, knowledge tracing which aims at predicting students' knowledge state is becoming a critical and fundamental task in personalized education.\n\n* [Systematic Review of Approaches to Improve Peer Assessment at Scale](https://arxiv.org/abs/2001.10617) - Peer Assessment is a task of analysis and commenting on student's writing by peers, is core of all educational components both in campus and in MOOC's.\n\n* [What's happened in MOOC Posts Analysis, Knowledge Tracing and Peer Feedbacks? A Review](https://arxiv.org/abs/2001.09830) - Learning Management Systems (LMS) and Educational Data Mining (EDM) are two important parts of online educational environment with the former being a centralised web-based information systems where the learning content is managed and learning activities are organised (Stone and Zheng,2014) and latter focusing on using data mining techniques for the analysis of data so generated.\n\n* [Towards an Appropriate Query, Key, and Value Computation for Knowledge Tracing](https://arxiv.org/abs/2002.07033) - Knowledge tracing, the act of modeling a student's knowledge through learning activities, is an extensively studied problem in the field of computer-aided education. Although models with attention mechanism have outperformed traditional approaches such as Bayesian knowledge tracing and collaborative filtering, they share two limitations.\n\n* [Generalized Knowledge Tracing: A Constrained Framework for Learner Modeling](https://arxiv.org/abs/2005.00869) - Adaptive learning technology solutions often use a learner model to trace learning and make pedagogical decisions.\n\n* [Towards Interpretable Deep Learning Models for Knowledge Tracing](https://arxiv.org/abs/2005.06139) - As an important technique for modeling the knowledge states of learners, the traditional knowledge tracing (KT) models have been widely used to support intelligent tutoring systems and MOOC platforms.\n\n* [Dynamic Knowledge embedding and tracing](https://arxiv.org/abs/2005.09109) - In this paper we propose a novel approach to knowledge tracing that combines techniques from matrix factorization with recent progress in recurrent neural networks (RNNs) to effectively track the state of a student's knowledge.\n\n* [qDKT: Question-centric Deep Knowledge Tracing](https://arxiv.org/abs/2005.12442) - We introduce qDKT, a variant of DKT that models every learner's success probability on individual questions over time. \n\n* [HGKT : Introducing Problem Schema with Hierarchical Exercise Graph for Knowledge Tracing](https://arxiv.org/abs/2006.16915) - In this paper, we propose a hierarchical graph knowledge tracing model framework (HGKT) which could leverage the advantages of hierarchical exercise graph and sequence model to enhance the ability of knowledge tracing. Besides, we introduce the concept of problem schema to better represent a group of similar exercises and propose a hierarchical graph neural network to learn representations of problem schemas. \n\n* [Context-Aware Attentive Knowledge Tracing](https://arxiv.org/abs/2007.12324) - We propose attentive knowledge tracing (AKT), which couples flexible attention-based neural network models with a series of novel, interpretable model components inspired by cognitive and psychometric models.\n\n* [GIKT: A Graph-based Interaction Model for Knowledge Tracing](https://arxiv.org/abs/2009.05991) - From the model perspective, previous models can hardly capture the long-term dependency of student exercise history, and cannot model the interactions between student-questions, and student-skills in a consistent way. In this paper, we propose a Graph-based Interaction model for Knowledge Tracing (GIKT) to tackle the above problems.\n\n* [Deep Knowledge Tracing with Convolutions](https://arxiv.org/abs/2008.01169) - We propose a Convolutional Knowledge Tracing (CKT) model in this paper.\n\n",
      "votes": 158
    },
    {
      "id": 1039155,
      "postDate": "2020-10-06T11:34:07.097Z",
      "content": "<p>Perhaps these papers with their code implementations will be useful to you:</p>\n<p><a href=\"https://paperswithcode.com/task/knowledge-tracing\" target=\"_blank\">https://paperswithcode.com/task/knowledge-tracing</a></p>",
      "rawMarkdown": "Perhaps these papers with their code implementations will be useful to you:\n\nhttps://paperswithcode.com/task/knowledge-tracing",
      "votes": 3,
      "replies": [
        {
          "id": 1039270,
          "postDate": "2020-10-06T13:10:26.720Z",
          "content": "<p>I saw some of these. It seemed like they were often very old.. which is why I didn't include that link. But they could be useful. I was really surprised that there weren't more. </p>",
          "rawMarkdown": "I saw some of these. It seemed like they were often very old.. which is why I didn't include that link. But they could be useful. I was really surprised that there weren't more. "
        }
      ]
    },
    {
      "id": 1038729,
      "postDate": "2020-10-06T02:30:59.693Z",
      "content": "<p><a href=\"https://www.mendeley.com/community/riiid-kaggle-research-papers/\" target=\"_blank\">https://www.mendeley.com/community/riiid-kaggle-research-papers/</a></p>",
      "rawMarkdown": "https://www.mendeley.com/community/riiid-kaggle-research-papers/",
      "votes": 4
    },
    {
      "id": 1043631,
      "postDate": "2020-10-09T05:45:39.893Z",
      "content": "<p>Does this competition require time series knowledge?</p>",
      "rawMarkdown": "Does this competition require time series knowledge?",
      "votes": 1,
      "replies": [
        {
          "id": 1045529,
          "postDate": "2020-10-10T18:13:34.740Z",
          "content": "<blockquote>\n  <p>Does this competition require time series knowledge?</p>\n</blockquote>\n<p>It will certainly help. So far, no public kernels have taken a sequential approach, but this will soon change. We are at the early stages of the competition and it is not so easy to apply sequential methods, mainly because this is an autoregressive problem as well, so we don't have access to the full test sequences.</p>\n<p>The current methods do not take into consideration the order in which students answer questions (at least, the kernels I have seen so far don't), so the models don't really capture the process of learning. They are looking at all the answers at once and finding patterns among them to apply to the test set. This works well if 1) you have the same users in the test set and 2) if you have the same questions in the test set. A smart enough model is able to find patterns among users and questions that generalize well, but only to the users and questions it has already seen. </p>\n<p> Update 10/14/2020: as per <a href=\"https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/191106\" target=\"_blank\">this discussion post</a>, the hidden test set contains new users but <em>not</em> new questions.</p>\n<p>and for 1),  we can have new users, The sequential approach bypasses this because it transforms the data into sequences where each user is assigned a single sequence and this sequence is treated like a sentence and the questions in the sequence like a corpus of words. In this context, it doesn't really matter which sequence is which. This is my rudimentary understanding of how one would treat this problem sequentially, but please correct me if I am confused. </p>",
          "rawMarkdown": "> Does this competition require time series knowledge?\n\nIt will certainly help. So far, no public kernels have taken a sequential approach, but this will soon change. We are at the early stages of the competition and it is not so easy to apply sequential methods, mainly because this is an autoregressive problem as well, so we don't have access to the full test sequences.\n\nThe current methods do not take into consideration the order in which students answer questions (at least, the kernels I have seen so far don't), so the models don't really capture the process of learning. They are looking at all the answers at once and finding patterns among them to apply to the test set. This works well if 1) you have the same users in the test set and 2) if you have the same questions in the test set. A smart enough model is able to find patterns among users and questions that generalize well, but only to the users and questions it has already seen. \n\n~~But this is not the case. 2) is clearly not true as per the [data description tab](https://www.kaggle.com/c/riiid-test-answer-prediction/data).~~ Update 10/14/2020: as per [this discussion post](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/191106), the hidden test set contains new users but *not* new questions.\n\nand for 1), ~~I'm pretty sure~~ we can have new users,~~ but I'm not entirely sure.~~ The sequential approach bypasses this because it transforms the data into sequences where each user is assigned a single sequence and this sequence is treated like a sentence and the questions in the sequence like a corpus of words. In this context, it doesn't really matter which sequence is which. This is my rudimentary understanding of how one would treat this problem sequentially, but please correct me if I am confused. ",
          "votes": 7
        },
        {
          "id": 1046074,
          "postDate": "2020-10-11T10:22:42.500Z",
          "content": "<p>Thank You for the detailed Explanation. This is very informative!</p>",
          "rawMarkdown": "Thank You for the detailed Explanation. This is very informative!",
          "votes": 1
        }
      ]
    },
    {
      "id": 1043976,
      "postDate": "2020-10-09T11:38:03.950Z",
      "content": "<p>A very interesting update. Has anyone gone to Arvix and scrolled down and noticed there is now a code tab? Very cool!!</p>\n<p>See attached screenshot. </p>",
      "rawMarkdown": "A very interesting update. Has anyone gone to Arvix and scrolled down and noticed there is now a code tab? Very cool!!\n\nSee attached screenshot. ",
      "votes": 2
    },
    {
      "id": 1222683,
      "postDate": "2021-03-02T03:31:27.487Z",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/crained\" target=\"_blank\">@crained</a> . can you suggest the paper Recommendation after the output of Knowedge Tracing?</p>\n<p>Thanks you!</p>",
      "rawMarkdown": "Hi @crained . can you suggest the paper Recommendation after the output of Knowedge Tracing?\n\nThanks you!",
      "replies": [
        {
          "id": 1226035,
          "postDate": "2021-03-04T07:06:27.620Z",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/crained\" target=\"_blank\">@crained</a> , pls help me</p>",
          "rawMarkdown": "Hi @crained , pls help me"
        },
        {
          "id": 1244376,
          "postDate": "2021-03-19T00:28:55.820Z",
          "content": "<p><a href=\"https://www.kaggle.com/phucpx\" target=\"_blank\">@phucpx</a> glad I was able to help!</p>",
          "rawMarkdown": "@phucpx glad I was able to help!"
        }
      ]
    },
    {
      "id": 1050806,
      "postDate": "2020-10-15T18:35:41.087Z",
      "content": "<p>Beautiful, so much energy saved!</p>",
      "rawMarkdown": "Beautiful, so much energy saved!",
      "replies": [
        {
          "id": 1244377,
          "postDate": "2021-03-19T00:29:13.203Z",
          "content": "<p>Glad it was helpful!</p>",
          "rawMarkdown": "Glad it was helpful!"
        }
      ]
    },
    {
      "id": 1048836,
      "postDate": "2020-10-13T20:38:15.023Z",
      "content": "<p>Deep Knowledge Tracing (NIPS 2015): <a href=\"https://geometry.stanford.edu/paper.php?id=pbhgsgs-dkt-15\" target=\"_blank\">https://geometry.stanford.edu/paper.php?id=pbhgsgs-dkt-15</a></p>",
      "rawMarkdown": "Deep Knowledge Tracing (NIPS 2015): https://geometry.stanford.edu/paper.php?id=pbhgsgs-dkt-15"
    },
    {
      "id": 1044072,
      "postDate": "2020-10-09T13:49:45.413Z",
      "content": "<p><a href=\"https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189962\" target=\"_blank\">Github-code of few above papers and various deep learning models for  Knowledge tracing in PyTorch  and tensorflow</a></p>",
      "rawMarkdown": "[Github-code of few above papers and various deep learning models for  Knowledge tracing in PyTorch  and tensorflow](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189962)"
    },
    {
      "id": 1043749,
      "postDate": "2020-10-09T08:06:32.517Z",
      "content": "<p>Thanks <a href=\"https://www.kaggle.com/crained\" target=\"_blank\">@crained</a>. This is really helpful! Looking forward to sharing ideas with everyone!</p>",
      "rawMarkdown": "Thanks @crained. This is really helpful! Looking forward to sharing ideas with everyone!"
    },
    {
      "id": 1043503,
      "postDate": "2020-10-09T03:28:58.800Z",
      "content": "<p>thanks for sharing, here's more <a href=\"https://paperswithcode.com/task/knowledge-tracing/codeless\" target=\"_blank\">https://paperswithcode.com/task/knowledge-tracing/codeless</a> </p>",
      "rawMarkdown": "thanks for sharing, here's more https://paperswithcode.com/task/knowledge-tracing/codeless "
    },
    {
      "id": 1041916,
      "postDate": "2020-10-08T01:09:16.330Z",
      "content": "<p>Google QUEST問答標籤</p>",
      "rawMarkdown": "Google QUEST問答標籤"
    },
    {
      "id": 1041571,
      "postDate": "2020-10-07T20:20:30.187Z",
      "content": "<p>These will be really helpful. Thank you. </p>",
      "rawMarkdown": "These will be really helpful. Thank you. "
    },
    {
      "id": 1040446,
      "postDate": "2020-10-07T06:29:24.530Z",
      "content": "<p>Thanks! You saved me a bunch of time spent on Google 🖖😊😊</p>",
      "rawMarkdown": "Thanks! You saved me a bunch of time spent on Google 🖖😊😊"
    },
    {
      "id": 1039161,
      "postDate": "2020-10-06T11:37:43.557Z",
      "content": "<p>Thank you <a href=\"https://www.kaggle.com/crained\" target=\"_blank\">@crained</a>. This is a really great collection of papers. </p>",
      "rawMarkdown": "Thank you @crained. This is a really great collection of papers. ",
      "replies": [
        {
          "id": 1039269,
          "postDate": "2020-10-06T13:09:46.887Z",
          "content": "<p>No problem. I was really surprised at how many there were!</p>",
          "rawMarkdown": "No problem. I was really surprised at how many there were!"
        }
      ]
    },
    {
      "id": 1038663,
      "postDate": "2020-10-06T00:32:41.893Z",
      "content": "<p><a href=\"https://www.kaggle.com/crained\" target=\"_blank\">@crained</a> Thank you for creating this thread. This definitely is the first step for me, too.</p>",
      "rawMarkdown": "@crained Thank you for creating this thread. This definitely is the first step for me, too."
    },
    {
      "id": 1038584,
      "postDate": "2020-10-05T22:26:54.160Z",
      "content": "<p>I think this will be useful, I have read a couple of these and will share my ideas as the completion goes on</p>",
      "rawMarkdown": "I think this will be useful, I have read a couple of these and will share my ideas as the completion goes on\n",
      "replies": [
        {
          "id": 1039271,
          "postDate": "2020-10-06T13:11:13.537Z",
          "content": "<p>That'd be really great. I probably won't get to coding for the competition until the weekend so I'd be super interested in any ideas that these helped spark!</p>",
          "rawMarkdown": "That'd be really great. I probably won't get to coding for the competition until the weekend so I'd be super interested in any ideas that these helped spark!"
        },
        {
          "id": 1104609,
          "postDate": "2020-12-07T05:47:30.113Z",
          "content": "<p>Thanks for initiating this thread…<br>\nMy summary of a few of these papers: <a href=\"https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/201481\" target=\"_blank\">https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/201481</a></p>",
          "rawMarkdown": "Thanks for initiating this thread...\nMy summary of a few of these papers: https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/201481\n"
        }
      ]
    },
    {
      "id": 1042858,
      "postDate": "2020-10-08T14:14:48.110Z",
      "rawMarkdown": "",
      "votes": 2,
      "isDeleted": true
    },
    {
      "id": 1040514,
      "postDate": "2020-10-07T07:23:51.970Z",
      "content": "<p>A lot of effort saved here. Thanks</p>",
      "rawMarkdown": "A lot of effort saved here. Thanks",
      "votes": 1
    },
    {
      "id": 1045967,
      "postDate": "2020-10-11T07:39:28.207Z",
      "content": "<p>Thanks for sharing!!</p>",
      "rawMarkdown": "Thanks for sharing!!"
    },
    {
      "id": 1045011,
      "postDate": "2020-10-10T09:28:52.160Z",
      "content": "<p>Thanks for sharing!!</p>",
      "rawMarkdown": "Thanks for sharing!!"
    },
    {
      "id": 1043973,
      "postDate": "2020-10-09T11:34:58.817Z",
      "content": "<p>Thanks for sharing!</p>",
      "rawMarkdown": "Thanks for sharing!"
    },
    {
      "id": 1041966,
      "postDate": "2020-10-08T01:58:57.200Z",
      "content": "<p>Thank you!</p>",
      "rawMarkdown": "Thank you!"
    },
    {
      "id": 1040853,
      "postDate": "2020-10-07T12:00:46.867Z",
      "content": "<p>Thanks for creating this. very helpful.</p>",
      "rawMarkdown": "Thanks for creating this. very helpful."
    },
    {
      "id": 1040674,
      "postDate": "2020-10-07T09:31:20.277Z",
      "content": "<p>Thank you. These will be very useful</p>",
      "rawMarkdown": "Thank you. These will be very useful"
    }
  ],
  "comments": [
    {
      "id": 1039155,
      "author_name": "Pedro Novas",
      "author_url": "",
      "post_date": "2020-10-06T11:34:07.097000",
      "content": "<p>Perhaps these papers with their code implementations will be useful to you:</p>\n<p><a href=\"https://paperswithcode.com/task/knowledge-tracing\" target=\"_blank\">https://paperswithcode.com/task/knowledge-tracing</a></p>",
      "votes": 3,
      "replies": [
        {
          "id": 1039270,
          "author_name": "Charlie Craine",
          "author_url": "",
          "post_date": "2020-10-06T13:10:26.720000",
          "content": "<p>I saw some of these. It seemed like they were often very old.. which is why I didn't include that link. But they could be useful. I was really surprised that there weren't more. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1038729,
      "author_name": "Carlos Souza",
      "author_url": "",
      "post_date": "2020-10-06T02:30:59.693000",
      "content": "<p><a href=\"https://www.mendeley.com/community/riiid-kaggle-research-papers/\" target=\"_blank\">https://www.mendeley.com/community/riiid-kaggle-research-papers/</a></p>",
      "votes": 4,
      "replies": []
    },
    {
      "id": 1043631,
      "author_name": "Aravind P",
      "author_url": "",
      "post_date": "2020-10-09T05:45:39.893000",
      "content": "<p>Does this competition require time series knowledge?</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1045529,
          "author_name": "Tucker Arrants",
          "author_url": "",
          "post_date": "2020-10-10T18:13:34.740000",
          "content": "<blockquote>\n  <p>Does this competition require time series knowledge?</p>\n</blockquote>\n<p>It will certainly help. So far, no public kernels have taken a sequential approach, but this will soon change. We are at the early stages of the competition and it is not so easy to apply sequential methods, mainly because this is an autoregressive problem as well, so we don't have access to the full test sequences.</p>\n<p>The current methods do not take into consideration the order in which students answer questions (at least, the kernels I have seen so far don't), so the models don't really capture the process of learning. They are looking at all the answers at once and finding patterns among them to apply to the test set. This works well if 1) you have the same users in the test set and 2) if you have the same questions in the test set. A smart enough model is able to find patterns among users and questions that generalize well, but only to the users and questions it has already seen. </p>\n<p> Update 10/14/2020: as per <a href=\"https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/191106\" target=\"_blank\">this discussion post</a>, the hidden test set contains new users but <em>not</em> new questions.</p>\n<p>and for 1),  we can have new users, The sequential approach bypasses this because it transforms the data into sequences where each user is assigned a single sequence and this sequence is treated like a sentence and the questions in the sequence like a corpus of words. In this context, it doesn't really matter which sequence is which. This is my rudimentary understanding of how one would treat this problem sequentially, but please correct me if I am confused. </p>",
          "votes": 7,
          "replies": []
        },
        {
          "id": 1046074,
          "author_name": "Aravind P",
          "author_url": "",
          "post_date": "2020-10-11T10:22:42.500000",
          "content": "<p>Thank You for the detailed Explanation. This is very informative!</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1043976,
      "author_name": "Charlie Craine",
      "author_url": "",
      "post_date": "2020-10-09T11:38:03.950000",
      "content": "<p>A very interesting update. Has anyone gone to Arvix and scrolled down and noticed there is now a code tab? Very cool!!</p>\n<p>See attached screenshot. </p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 1222683,
      "author_name": "Phuc Phan",
      "author_url": "",
      "post_date": "2021-03-02T03:31:27.487000",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/crained\" target=\"_blank\">@crained</a> . can you suggest the paper Recommendation after the output of Knowedge Tracing?</p>\n<p>Thanks you!</p>",
      "votes": 0,
      "replies": [
        {
          "id": 1226035,
          "author_name": "Phuc Phan",
          "author_url": "",
          "post_date": "2021-03-04T07:06:27.620000",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/crained\" target=\"_blank\">@crained</a> , pls help me</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1244376,
          "author_name": "Charlie Craine",
          "author_url": "",
          "post_date": "2021-03-19T00:28:55.820000",
          "content": "<p><a href=\"https://www.kaggle.com/phucpx\" target=\"_blank\">@phucpx</a> glad I was able to help!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1050806,
      "author_name": "miklos kralik",
      "author_url": "",
      "post_date": "2020-10-15T18:35:41.087000",
      "content": "<p>Beautiful, so much energy saved!</p>",
      "votes": 0,
      "replies": [
        {
          "id": 1244377,
          "author_name": "Charlie Craine",
          "author_url": "",
          "post_date": "2021-03-19T00:29:13.203000",
          "content": "<p>Glad it was helpful!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1048836,
      "author_name": "Nya 🚀",
      "author_url": "",
      "post_date": "2020-10-13T20:38:15.023000",
      "content": "<p>Deep Knowledge Tracing (NIPS 2015): <a href=\"https://geometry.stanford.edu/paper.php?id=pbhgsgs-dkt-15\" target=\"_blank\">https://geometry.stanford.edu/paper.php?id=pbhgsgs-dkt-15</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1044072,
      "author_name": "Shivanand",
      "author_url": "",
      "post_date": "2020-10-09T13:49:45.413000",
      "content": "<p><a href=\"https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189962\" target=\"_blank\">Github-code of few above papers and various deep learning models for  Knowledge tracing in PyTorch  and tensorflow</a></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1043749,
      "author_name": "Sujay Rokade",
      "author_url": "",
      "post_date": "2020-10-09T08:06:32.517000",
      "content": "<p>Thanks <a href=\"https://www.kaggle.com/crained\" target=\"_blank\">@crained</a>. This is really helpful! Looking forward to sharing ideas with everyone!</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1043503,
      "author_name": "frozhen",
      "author_url": "",
      "post_date": "2020-10-09T03:28:58.800000",
      "content": "<p>thanks for sharing, here's more <a href=\"https://paperswithcode.com/task/knowledge-tracing/codeless\" target=\"_blank\">https://paperswithcode.com/task/knowledge-tracing/codeless</a> </p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1041916,
      "author_name": "howard zhou ",
      "author_url": "",
      "post_date": "2020-10-08T01:09:16.330000",
      "content": "<p>Google QUEST問答標籤</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1041571,
      "author_name": "Aman Mittal",
      "author_url": "",
      "post_date": "2020-10-07T20:20:30.187000",
      "content": "<p>These will be really helpful. Thank you. </p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1040446,
      "author_name": "Matt Wagner",
      "author_url": "",
      "post_date": "2020-10-07T06:29:24.530000",
      "content": "<p>Thanks! You saved me a bunch of time spent on Google 🖖😊😊</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1039161,
      "author_name": "Shivanand",
      "author_url": "",
      "post_date": "2020-10-06T11:37:43.557000",
      "content": "<p>Thank you <a href=\"https://www.kaggle.com/crained\" target=\"_blank\">@crained</a>. This is a really great collection of papers. </p>",
      "votes": 0,
      "replies": [
        {
          "id": 1039269,
          "author_name": "Charlie Craine",
          "author_url": "",
          "post_date": "2020-10-06T13:09:46.887000",
          "content": "<p>No problem. I was really surprised at how many there were!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1038663,
      "author_name": "VidyaSagarPanati",
      "author_url": "",
      "post_date": "2020-10-06T00:32:41.893000",
      "content": "<p><a href=\"https://www.kaggle.com/crained\" target=\"_blank\">@crained</a> Thank you for creating this thread. This definitely is the first step for me, too.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1038584,
      "author_name": "Henry Agbaeze",
      "author_url": "",
      "post_date": "2020-10-05T22:26:54.160000",
      "content": "<p>I think this will be useful, I have read a couple of these and will share my ideas as the completion goes on</p>",
      "votes": 0,
      "replies": [
        {
          "id": 1039271,
          "author_name": "Charlie Craine",
          "author_url": "",
          "post_date": "2020-10-06T13:11:13.537000",
          "content": "<p>That'd be really great. I probably won't get to coding for the competition until the weekend so I'd be super interested in any ideas that these helped spark!</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1104609,
          "author_name": "Allohvk",
          "author_url": "",
          "post_date": "2020-12-07T05:47:30.113000",
          "content": "<p>Thanks for initiating this thread…<br>\nMy summary of a few of these papers: <a href=\"https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/201481\" target=\"_blank\">https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/201481</a></p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1042858,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-10-08T14:14:48.110000",
      "content": "",
      "votes": 2,
      "replies": []
    },
    {
      "id": 1040514,
      "author_name": "deHack",
      "author_url": "",
      "post_date": "2020-10-07T07:23:51.970000",
      "content": "<p>A lot of effort saved here. Thanks</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1045967,
      "author_name": "Kohama Hayato",
      "author_url": "",
      "post_date": "2020-10-11T07:39:28.207000",
      "content": "<p>Thanks for sharing!!</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1045011,
      "author_name": "Hitesh Tripathi",
      "author_url": "",
      "post_date": "2020-10-10T09:28:52.160000",
      "content": "<p>Thanks for sharing!!</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1043973,
      "author_name": "Michael Oswaldo",
      "author_url": "",
      "post_date": "2020-10-09T11:34:58.817000",
      "content": "<p>Thanks for sharing!</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1041966,
      "author_name": "Connor Shorten",
      "author_url": "",
      "post_date": "2020-10-08T01:58:57.200000",
      "content": "<p>Thank you!</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1040853,
      "author_name": "Hasanga Methmal",
      "author_url": "",
      "post_date": "2020-10-07T12:00:46.867000",
      "content": "<p>Thanks for creating this. very helpful.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 1040674,
      "author_name": "Keelisetti Lokesh",
      "author_url": "",
      "post_date": "2020-10-07T09:31:20.277000",
      "content": "<p>Thank you. These will be very useful</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1038477": "Hey everyone!\n\nI wanted to start a papers thread and build on it, and hope others share as well, papers to gain domain knowledge. I have no domain knowledge in this area so I have downloaded papers that appeared to be relevant after reading their abstracts. I'll be reading these over the coming days and commenting more as I go through all of them. Hope this helps!\n\nResearch Papers:\n* [A Self-Attentive model for Knowledge Tracing](https://arxiv.org/abs/1907.06837) - we develop an approach that identifies the KCs from the student's past activities that are \\textit{relevant} to the given KC and predicts his/her mastery based on the relatively few KCs that it picked. \n\n* [Knowledge Tracing with Sequential Key-Value Memory Networks](https://arxiv.org/abs/1910.13197) - Can machines trace human knowledge like humans? Knowledge tracing (KT) is a fundamental task in a wide range of applications in education, such as massive open online courses (MOOCs), intelligent tutoring systems, educational games, and learning management systems.\n\n* [EdNet: A Large-Scale Hierarchical Dataset in Education](https://arxiv.org/abs/1912.03072) - With advances in Artificial Intelligence in Education (AIEd) and the ever-growing scale of Interactive Educational Systems (IESs), data-driven approach has become a common recipe for various tasks such as knowledge tracing and learning path recommendation.\n\n* [Domain Adaption for Knowledge Tracing](https://arxiv.org/abs/2001.04841) - With the rapid development of online education system, knowledge tracing which aims at predicting students' knowledge state is becoming a critical and fundamental task in personalized education.\n\n* [Systematic Review of Approaches to Improve Peer Assessment at Scale](https://arxiv.org/abs/2001.10617) - Peer Assessment is a task of analysis and commenting on student's writing by peers, is core of all educational components both in campus and in MOOC's.\n\n* [What's happened in MOOC Posts Analysis, Knowledge Tracing and Peer Feedbacks? A Review](https://arxiv.org/abs/2001.09830) - Learning Management Systems (LMS) and Educational Data Mining (EDM) are two important parts of online educational environment with the former being a centralised web-based information systems where the learning content is managed and learning activities are organised (Stone and Zheng,2014) and latter focusing on using data mining techniques for the analysis of data so generated.\n\n* [Towards an Appropriate Query, Key, and Value Computation for Knowledge Tracing](https://arxiv.org/abs/2002.07033) - Knowledge tracing, the act of modeling a student's knowledge through learning activities, is an extensively studied problem in the field of computer-aided education. Although models with attention mechanism have outperformed traditional approaches such as Bayesian knowledge tracing and collaborative filtering, they share two limitations.\n\n* [Generalized Knowledge Tracing: A Constrained Framework for Learner Modeling](https://arxiv.org/abs/2005.00869) - Adaptive learning technology solutions often use a learner model to trace learning and make pedagogical decisions.\n\n* [Towards Interpretable Deep Learning Models for Knowledge Tracing](https://arxiv.org/abs/2005.06139) - As an important technique for modeling the knowledge states of learners, the traditional knowledge tracing (KT) models have been widely used to support intelligent tutoring systems and MOOC platforms.\n\n* [Dynamic Knowledge embedding and tracing](https://arxiv.org/abs/2005.09109) - In this paper we propose a novel approach to knowledge tracing that combines techniques from matrix factorization with recent progress in recurrent neural networks (RNNs) to effectively track the state of a student's knowledge.\n\n* [qDKT: Question-centric Deep Knowledge Tracing](https://arxiv.org/abs/2005.12442) - We introduce qDKT, a variant of DKT that models every learner's success probability on individual questions over time. \n\n* [HGKT : Introducing Problem Schema with Hierarchical Exercise Graph for Knowledge Tracing](https://arxiv.org/abs/2006.16915) - In this paper, we propose a hierarchical graph knowledge tracing model framework (HGKT) which could leverage the advantages of hierarchical exercise graph and sequence model to enhance the ability of knowledge tracing. Besides, we introduce the concept of problem schema to better represent a group of similar exercises and propose a hierarchical graph neural network to learn representations of problem schemas. \n\n* [Context-Aware Attentive Knowledge Tracing](https://arxiv.org/abs/2007.12324) - We propose attentive knowledge tracing (AKT), which couples flexible attention-based neural network models with a series of novel, interpretable model components inspired by cognitive and psychometric models.\n\n* [GIKT: A Graph-based Interaction Model for Knowledge Tracing](https://arxiv.org/abs/2009.05991) - From the model perspective, previous models can hardly capture the long-term dependency of student exercise history, and cannot model the interactions between student-questions, and student-skills in a consistent way. In this paper, we propose a Graph-based Interaction model for Knowledge Tracing (GIKT) to tackle the above problems.\n\n* [Deep Knowledge Tracing with Convolutions](https://arxiv.org/abs/2008.01169) - We propose a Convolutional Knowledge Tracing (CKT) model in this paper.\n\n",
    "1039155": "Perhaps these papers with their code implementations will be useful to you:\n\nhttps://paperswithcode.com/task/knowledge-tracing",
    "1038729": "https://www.mendeley.com/community/riiid-kaggle-research-papers/",
    "1043631": "Does this competition require time series knowledge?",
    "1043976": "A very interesting update. Has anyone gone to Arvix and scrolled down and noticed there is now a code tab? Very cool!!\n\nSee attached screenshot. ",
    "1222683": "Hi @crained . can you suggest the paper Recommendation after the output of Knowedge Tracing?\n\nThanks you!",
    "1050806": "Beautiful, so much energy saved!",
    "1048836": "Deep Knowledge Tracing (NIPS 2015): https://geometry.stanford.edu/paper.php?id=pbhgsgs-dkt-15",
    "1044072": "[Github-code of few above papers and various deep learning models for  Knowledge tracing in PyTorch  and tensorflow](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189962)",
    "1043749": "Thanks @crained. This is really helpful! Looking forward to sharing ideas with everyone!",
    "1043503": "thanks for sharing, here's more https://paperswithcode.com/task/knowledge-tracing/codeless ",
    "1041916": "Google QUEST問答標籤",
    "1041571": "These will be really helpful. Thank you. ",
    "1040446": "Thanks! You saved me a bunch of time spent on Google 🖖😊😊",
    "1039161": "Thank you @crained. This is a really great collection of papers. ",
    "1038663": "@crained Thank you for creating this thread. This definitely is the first step for me, too.",
    "1038584": "I think this will be useful, I have read a couple of these and will share my ideas as the completion goes on\n",
    "1042858": "",
    "1040514": "A lot of effort saved here. Thanks",
    "1045967": "Thanks for sharing!!",
    "1045011": "Thanks for sharing!!",
    "1043973": "Thanks for sharing!",
    "1041966": "Thank you!",
    "1040853": "Thanks for creating this. very helpful.",
    "1040674": "Thank you. These will be very useful"
  }
}