You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

You agree to our Responsible Use Guidelines.

Log in or Sign Up to review the conditions and access this dataset content.

Overview of MathNet57789

MathNet57789 is a dataset of K-12 students’ handwritten images given as a response to answering open-ended math questions. The dataset is collected from the ASSISTments online learning platform over the 2019-2023 school years, from grade 2 through high school students. The dataset consists of images that are redacted of Personally Identifiable Information(PII) by cropping the background and masking off any remaining PII with black rectangular boxes.

Data Source

This dataset includes two files: 1. MathNet57789 Problem Logs and 2. Problem Bodies.

MathNet57789 Problem Logs points to students' answers given as a response to math problems and this dataset maps to the Problem Bodies on the problem_id. Problem Bodies includes the math problems that students are answering.

The images in the MathNet57789 dataset were created during a joint project between Dr. Heffernan’s lab at WPI and ASSISTments, where students upload their handwritten math work and receive feedback from teachers. Dr. Heffernan’s lab is releasing this data set as part of their responsibility to release de-identified data under the tenets of Open Science as part of receiving NSF grants.

Sharing Agreement

This dataset is licensed under CC-BY-NC-4.0. We require that this dataset is used for research and educational purposes following this Responsible Use Guidelines.

We have taken measures to de-identify these images and are confident that they are free of PII. In the event that you encounter any PII, like their names. We ask you to report the file name to us and delete the image. Please report the details including the file name of such images by emailing etrials@assistments.org.

Describing the Dataset

The dataset was reviewed by a group of students at WPI who manually looked into the individual images in this dataset and cropped these images to remove any backgrounds, then masked off the remaining PIIs with black rectangular boxes. Blank or empty images were removed from the dataset.

The Mathnet57789 dataset contains 57,789 handwritten images from 21,212 different students answering 192 different math problems in the ASSISTments learning platform. Below is the summary statistics of the dataset:

Summary Statistics of the MathNet57789 Dataset

Metric Value
Total Students' Handwritten Images 57,789
Total Math Problems 192
Total Unique Students 21,212
Total Teachers 1,602

Statistics of Images per Problem

Statistic Value
Mean number of images per problem 301
Median number of images per problem 251
Maximum number of images per problem 800
Minimum number of images per problem 12

Data Dictionary

Row Each row is a student answer in this Database
Column The columns are shared below

The Columns in the MathNet57789 Problem Logs Dataset

Column Name Description
problem_log_id Identifier for logs per problem per student
assignment_log_id The identifier for logs per assignment
problem_id Unique identifier of problem
assistment_id Identifier of the Problem. It can have one problem_id or multiple problems depending on the type of problem. If the problem is multipart then one assistment_id can have multiple problem_ids in it
teacher_id Unique identifier of teacher
student_id Unique identifier of student
ImageURL URL pointing to the student's handwritten image
Image_sha256 Unique 256-bit hash generated using SHA-256 algorithm to represent the content of the student's handwritten image
path The file path for the student's handwritten image
image_name Filename for the student's handwritten image
answer_text Students' complete answer text, including the image that is an HTML blob. Students can upload images with additional text typed in the ASSISTments platform
score The score the teacher has assigned to the student’s answer. This score is an integer-valued scale ranging from 0 to 4. Each teacher uses their own judgment to create the score
continuous_score The score from the teacher converted to a decimal between 0 to 1

The Columns in the Problem Bodies Dataset for MathNet57789

Column Name Description
problem_id Unique Identifier for a problem
problem_code Alpha Numeric id for a problem, the problem code visible in reports and on the web app in ASSISTments
problem_image_url URL pointing to the image of the math problem. For multipart problems, the main problem is highlighted in a rectangular box with a red border
image_sha256 Unique 256-bit hash generated using the SHA-256 algorithm to represent the content of the math problem image
skill_codes Common Core State Standards (CCSS) associated with the problem
skill_names Name of the CCSS
curriculum The curriculum the problem belongs to
grade_or_subject Grade or the subject of the problem
unit Unit the problem belongs to
topic Math topic the problem belongs to
assistments_link The link of the problem in ASSISTments
problem_set_id Identifier/code for the problem set of the problem
problem_part If the problem is part of a multipart problem, problem_part indicate the order of this problem in the multipart problem
problem_body The HTML blob for the math problem that the student is solving
Downloads last month
53