Blog

Syllabus Explorer

We are happy to introduce the Harvard Syllabus Explorer project. Currently, only members of the Faculty of Arts and Sciences have access. If you feel like you should have access but do not, please email syllabusexplorer@harvard.edu

This project collects syllabi from Canvas course pages and the corresponding course descriptions from the Registrar’s Office. Using natural language processing tools, the Syllabus Explorer analyzes the text from these sources and...

Read more about Syllabus Explorer

DART in-video search: Connecting search queries to point of occurrence in videos

Daniel Seaton, David House, Elliott Yates - a collaboration between VPAL and HUIT

Have you ever wanted to navigate to a specific topic within an hour long video? Video recordings of classroom lectures, invited seminars, and speeches are generally rich with information, but difficult to explore due to their length and information density. DART now provides functionality that connects search terms directly to their point of occurrence inside videos. Below we describe the general functionality of this newly released feature and point toward future use cases....

Read more about DART in-video search: Connecting search queries to point of occurrence in videos

Working with edX content data: Five years of courses in the edX consortium

Daniel Seaton, Adam Nahari, Glenn Lopez

It is well understood that the MOOC movement has led to an unprecedented number of learners and clicks, but we often forget to discuss the catalog of digital-learning assets that facilitates those clicks. Since 2012, HarvardX alone has created over 100 unique courses that account for over 26,000 resources. Taking into account all members of the edX consortium, we estimate over 400,000 individual videos, problems, and html pages have been created since the beginning of edX. Pause and let that sink in… That is a...

Read more about Working with edX content data: Five years of courses in the edX consortium

SearchBetter: helping DART find even more search results

Neel Mehta, Harvard CS 2018
 

There are numerous open source packages that enable text-search of custom text corpuses. A major challenge for these implementations is addressing zero result searches, where a user may misunderstand that a small corpus should not return millions of results like Google or other Internet search engines. Harvard has recently implemented search across a large body of education content from HarvardX and Harvard YouTube channels. The new project called Digital Assets for...

Read more about SearchBetter: helping DART find even more search results