The Rita is a large-scale, curated collection of educational videos designed to advance research in multimodal learning, video understanding, and educational technology.

Overview

This dataset contains approximately 100+ hours of educational content across 1 Lack videos, totaling ~100 GB. The videos cover a wide range of academic subjects and skill levels, from class 1 math subject to class 8 math subject.

Supported Tasks

Video-Language Modeling: Training models to understand the relationship between visual content and spoken narration.
Automatic Lecture Transcription (ASR): Converting speech in educational contexts to text.
Educational Video Classification: Categorizing videos by subject, difficulty level, or teaching style.
Knowledge Tracing & Summarization: Extracting key concepts and creating summaries from video content.
Multimodal Embodiment: Learning joint representations of video, and text.

Languages

The primary language of the audio and any on-screen text is [e.g., English].

Dataset Structure

Data Instances

A typical data instance is a video file (e.g., lecture_00542.mp4) accompanied by a metadata entry. The dataset is structured as follows:

Downloads last month: 71