PTADisc

This release contains 939 extracted data files (125.50 GB), with 128 configured subsets. Original-project overview statistics below are retained for context and are not recomputed counts for this distribution.

This dataset card reuses the original PTADisc README, with download links, paths, and release-specific notes adapted for this Hugging Face distribution. The original authors’ contact information is retained below. Licensed under CC BY-NC-SA 4.0.

PTADisc is sourced from PTA (Programming Teaching Assistant), a learner-based online system for universities and society developed by Hangzhou PAT Education Technology Co., Ltd.

PTADisc is a diverse, immense, student-centered dataset that emphasizes its sufficient cross-course information for personalized learning. It includes $74$ courses, $1,530,100$ students, $4,504$ concepts, $225,615$ problems, and over $680$ million student response logs.

We illustrate the characteristics of PTADisc from the following four aspects:

• Diverse: PTADisc contains rich concept-related information and fine-grained records of a large number of student behaviors.
• Immense: PTADisc is the largest dataset in personalized learning. It also includes different courses of different data scales which offer options for various studies.
• Student-centered: In PTADisc, problems and knowledge concepts are validated based on student response logs, leading to better consistency and well-maintainability in terms of diagnostic tasks.
• Cross-course: PTADisc covers a significant amount of students taking multiple courses as students are likely to take a series of courses according to their training program.

Loading the data

Select a course or experiment subset in the dataset viewer, or stream a subset with datasets:

from datasets import load_dataset

ds = load_dataset("whning0513/PTADisc", "cd_Probability_and_statistic")
logs = load_dataset(
    "whning0513/PTADisc", "non_programming_Python_programming",
    split="train", streaming=True,
)
row = next(iter(logs))

Configured CSV columns are exposed as strings to preserve source identifiers, timestamps, code, and empty fields without numeric inference. Cast columns as needed for analysis. Schemas follow each source CSV header. Unsplit source tables use the train split as a storage convention; it does not denote a curated training partition. Existing experiment train/validation/test splits are preserved. Supplementary JSON dictionaries, course information, matrices, and experimental variants are available as individual files.

To download all extracted files:

from huggingface_hub import snapshot_download
snapshot_download("whning0513/PTADisc", repo_type="dataset", local_dir="PTADisc")

Data repository structure

Data files are published directly in this dataset repository as extracted CSV, JSON, text, and model files. Browse the Files and versions tab or download individual files; no archive extraction or access request is required. The original project is available on GitHub.

The directory is organized as:

PTADisc/
├── scripts/
├── structured data/
│   ├── course_info/
│   ├── global data/
│   ├── non programming data/
│   └── programming data/
└── task specific dataset/
    ├── non_programming_dataset
    │   ├── baseline_dataset_for_CD
    │   ├── baseline_dataset_for_KT
    │   ├── cross_course_datasets
    │   └── datasets_for_CCLMF
    └── programming_datasets

Leaf directory	Content
baseline_dataset_for_CD	$4$ selected courses' datasets for cognitive diagnosis.
baseline_dataset_for_KT	$4$ selected courses' datasets for knowledge tracing.
cross_course_datasets	datasets of $5$ courses which are simultaneously taken by $29,454$ students.
datasets_for_CCLMF	Java Programming datasets and pretrained NCD model on Python Programming, which is used to conduct CCLMF experiments.
programming_datasets	$4$ selected courses' datasets for knowledge tracing.

Release note: The table above retains the original project’s dataset descriptions. The received source has empty baseline_dataset_for_KT and programming_datasets directories. Data files are published in extracted form, preserving their contents. Course archives lacking an enclosing directory have been placed in a directory named after the course. Embedded TAR.GZ backups are extracted under a separate _nested directory. See manifest.json for source mappings.

Detailed information of structured data

For futher information, please contact: liyahu@zju.edu.cn.

Global Data

Description

Global data includes:

Filename	Description
knowledge_point_all.csv	The tree structure of all the knowledge concepts in PTADisc, which is in forms of knowledge_point_id, name, parent_id.
problem_difficulty.json	Global dictionary of problem_id to difficulty.
problem_knowledge.json	Global dictionary of problem_id to knowledge_id.
problem_reference_count.json	Global dictionary of problem_id to reference_count.
problem_type	Global dictionary of problem_id to problem_type.
psp_full_score.json	Global dictionary of problem_set_problem_id to full_score.
psp_problem.json	Global dictionary of problem_set_problem_id to problem_id.
psp_problem_set_id.json	Global dictionary of problem_set_problem_id to problem_set_id.

The filenames above follow the received source data. The problem_type dictionary described in the original README is not provided as a separate global file in this release.

Knowledge Concept Example

parent_id=-1 indicates that it's the course's name, which is the root of its knowledge concept tree.

knowledge_point_id	name	parent_id
383	概率论与数理统计	-1
384	随机事件和概率	383
393	随机变量及概率分布	383
389	全概率公式	384
390	贝叶斯公式	384

Problem Type

enum ProblemType {
    NO_PROBLEM_TYPE = 0;
    TRUE_OR_FALSE = 1;  
    MULTIPLE_CHOICE = 2;  
    MULTIPLE_CHOICE_MORE_THAN_ONE_ANSWER = 3; 
    FILL_IN_THE_BLANK = 4; 
    FILL_IN_THE_BLANK_FOR_PROGRAMMING = 5; 
    CODE_COMPLETION = 6;  
    PROGRAMMING = 7; 
    SUBJECTIVE = 8;  
    MULTIPLE_FILE = 9;  
    }

Non-programming data

Description

Each course contains the following files:

Filename	Description
info/info.json	Record the number of the course's concepts, students, problems and logs.
info/concept_list.txt	Record the course's concept IDs.
info/student_list.txt	Record the course's student IDs.
info/problem_list.txt	Record the course's problem IDs.
problem_info.csv	Record problems' problem_set_problem_id, problem_id, knowledge_id, full_score in the course.
response_log.csv	Record logs' submission_id, user_id, create_at, problem_type, score, problem_set_id, problem_set_problem_id, status in the course.

response_log.csv fields description:

• submission_id: the ID of one commit record of a problem set.
• problem_type: 1 represents true or false, 2 represents single-choice, 3 represents multiple-choice, and 4 represents fill-in-the-blank.
• score: the scoring ratio the student get on this problem.
• status: auxiliary field to calculate student's score on this problem, contains ACCEPTED, WRONG_ANSWER, PARTIAL_ACCEPTED, etc.

Logs Example

submission_id	user_id	create_at	problem_type	score	problem_set_id	problem_set_problem_id	status
178060	0fa9733a132497dd515d426df206b67106364354b579231879e3c8f70630e431	2015/10/13 16:47	2	2	139	1746	ACCEPTED
178060	0fa9733a132497dd515d426df206b67106364354b579231879e3c8f70630e431	2015/10/13 16:47	2	0	139	1747	WRONG_ANSWER
178060	0fa9733a132497dd515d426df206b67106364354b579231879e3c8f70630e431	2015/10/13 16:47	2	0	139	1748	WRONG_ANSWER
178060	0fa9733a132497dd515d426df206b67106364354b579231879e3c8f70630e431	2015/10/13 16:47	2	2	139	1749	ACCEPTED
178060	0fa9733a132497dd515d426df206b67106364354b579231879e3c8f70630e431	2015/10/13 16:47	2	2	139	1750	ACCEPTED

Programming data

Description

Each course contains a response_log.csv with fields: submission_id, user_id, create_at, language , score, problem_set_problem_id, problem_id, skill_id, code, response, time_consume, memory_consume. Some of the fields description:

• submission_id: the same as in non_programming data.
• language: the programming code
• score: the same as in non_programming data.
• code: the code submitted by the student.
• response: the same as statusin non_programming data.
• time_consume: time consumed by the submitted code.
• memory_consume: memory consumed by the submitted code.

Logs Example

submission_id	user_id	create_at	language	score	problem_set_problem_id	problem_id	skill_id	code	response	time_consume	memory_consume
1523963863660281856	82e2f8cab241392f2ffae8614297ed854b6113256c5c8d31cf2d14740a68706b	2022-05-10 17:51:13	Java	0.0	1523908485467762710	1013962033606774784	201	import java.util.*;\npublic class Main {\n public static void main(String[] args) {\n Scanner sc = new Scanner(System.in);\n int n = sc.nextInt();\n List list = new ArrayList();\n for (int i = 0; i < n; i++) {\n list.add(sc.nextInt());\n }\n list.sort(Comparator.naturalOrder());\n System.out.println(list.get(list.size()-2));\n }\n}	WRONG_ANSWER	125	18404
1523963973706231808	82e2f8cab241392f2ffae8614297ed854b6113256c5c8d31cf2d14740a68706b	2022-05-10 17:51:39	Java	0.0	1523908485467762710	1013962033606774784	201	import java.util.*;\npublic class Main {\n public static void main(String[] args) {\n Scanner sc = new Scanner(System.in);\n int n = sc.nextInt();\n List list = new ArrayList();\n for (int i = 0; i < n; i++) {\n list.add(sc.nextInt());\n }\n list.sort(Comparator.naturalOrder());\n System.out.print(list.get(list.size()-2));\n }\n}	WRONG_ANSWER	113	15232

Course Info

Description

The files of course info provide additional information of user group and problem set. Each course contains the following files:

Filename	Description
user_group_member.csv	Record the user group members.
user_group_problem_set.csv	Record the opened problem sets for user groups.
problem_set.csv	Record the problem set's start time and end time.

User Group Member Example

user_group_member_id is the id of a (user_id, user_group_id) pair.

user_group_member_id	user_id	user_group_id
1041674784427282432	e7de9a3c1163736606a53e738cdb82a29f860903e84e14c61f13b31f82706d28	1041661419302744064
1041680824921296896	95c2f1226984fee4c1748e567279808b02e5128dbd1060ebd4692dbe164000a6	1041661579701325824
1041680824535420928	8e3f445a0fae30aef3daf067d1bebf648ca84181c21a924560f9a9a1f1d1504c	1041661579701325824

User Group - Problem Set Example

user_group_problem_set_id	user_group_id	problem_set_id
1053296011098509312	1041661579701325824	1053295797428080640
1053296069663563776	1041661419302744064	1053295797428080640
1057397021588357120	1041661419302744064	1057375475356430336

Problem Set Example

problem_set_id	start_at	end_at
1057375475356430336	2018-11-02 12:00:00	2018-11-21 12:00:00
1118521024520945664	2019-04-17 22:24:00	2019-04-18 22:24:00
1119973878406443008	2019-04-21 22:38:00	2019-04-28 22:38:00

Statistics

Detailed statistics of each course's students, problems, concepts, non_programming_logs and programming_logs can be found in statistics.xlsx.

CCLMF

Details can be found in README.md.

Citation

@inproceedings{hu2023ptadisc,
  title={PTADisc: A Cross-Course Dataset Supporting Personalized Learning in Cold-Start Scenarios},
  author={Hu, Liya and Dong, Zhiang and Chen, Jingyuan and Wang, Guifeng and Wang, Zhihua and Zhao, Zhou and Wu, Fei},
  booktitle={Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
  year={2023}
}