Google Tech Talk - Data Commons

Speaker: Ramanathan V. Guha , Google

Date: Thursday, November 30, 2023

Time: 2:00 PM to 3:00 PM Note: all times are in the Eastern Time Zone

Public: Yes

Location: Remote at https://mit.zoom.us/j/96487703855

Event Type: Seminar

Room Description:

Host: Jim Glass, CSAIL MIT

Contact: Hongyin Luo, hyluo@mit.edu

Relevant URL: https://mit.zoom.us/j/96487703855

Speaker URL: http://www.guha.com/cv.html

Speaker Photo:
None

Reminders to: csail-local@lists.csail.mit.edu

Reminder Subject: TALK: Google Tech Talk - Data Commons

Abstract of the talk:
Statistical data is vital to understanding the world around us, especially for tackling the impact of crises ranging from the pandemic to climate change. Thus requires a deep understanding of not just the virus or the climate, but the various stresses that will impact access to food, shelter, healthcare and other aspects of life. Unfortunately, this data required for this understanding is fragmented across thousands of organizations, in different schemas and a multitude of databases, making it very expensive, if not impossible to use.

Google has ‘organized and made easily accessible’ many kinds of information — web pages, images, maps, etc. One of our contributions to tackling these crises is to organize this statistical data and make it easily accessible to consumers, journalists, policy makers and researchers through our Data Commons effort. The Data Commons approach is to do the data wrangling once and make the processed data widely available. It is a single unified knowledge graph, with over 250 billion data points from hundreds of sources, created by normalizing/aligning the schemas and entity references across data from these sources. The data is available via standard schemas and Cloud APIs and more recently, using developments in language models, via natural language.

Research Areas:
AI & Machine Learning

Impact Areas:
Big Data, Education, Health Care

This event is not part of a series.

Created by Hongyin Luo Email at Monday, November 27, 2023 at 1:13 PM.