A new open source dataset links human motion and language

February 10, 2017

Researchers have created a large, open source database to support the development of robot activities based on natural language input. The new KIT Motion-Language Dataset will help to unify and standardize research linking human motion and natural language, as presented in an article in Big Data, a peer-reviewed journal from Mary Ann Liebert, Inc., publishers. The article is available free on the Big Data website until March 9, 2017.

In the article "The KIT Motion-Language Dataset," Matthias Plappert, Christian Mandery, and Tamim Asfour, Institute for Anthropomatics and Robotics, Karlsruhe Institute of Technology (KIT), Germany, describe a novel crowd-sourcing approach and purpose-built web-based tool they used to develop their publicly available dataset that annotates motions. Their approach relies on a unified representation that is independent of the capture system or marker set to be able to merge data from different existing motion capture databases into the KIT Motion-Language Dataset. It currently includes about 4,000 motions and more than 6,200 annotations in natural language that contain nearly 53,000 words.

The article is part of a special issue of Big Data on "Big Data in Robotics" led by Guest Editors Jeannette Bohg, PhD, Matei Ciocarlie, PhD, Jaview Civera, PhD, and Lydia Kavraki, PhD.

"Human motion is complex and nuanced in terms of how it can be described, and it is surprisingly difficult to even retrieve motions from databases corresponding to natural language descriptions. There is a great need to describe robotic systems in natural language that captures the richness associated with motion, but doing this accurately is an extremely challenging problem," says Big Data Editor-in-Chief Vasant Dhar, Professor at the Stern School of Business and the Center for Data Science at New York University. "Plappert and his colleagues do a wonderful job using a novel crowd-sourcing approach and a tool to document the annotation process itself along with methods for obtaining high quality inputs and selecting motions that require further annotation automatically. They have constructed an impressive database of motions and annotations that can serve as a test-bed for research in this area. It is a great service to the research community."
-end-
About the Journal

Big Data, published quarterly online with open access options and in print, facilitates and supports the efforts of researchers, analysts, statisticians, business leaders, and policymakers to improve operations, profitability, and communications within their organizations. Spanning a broad array of disciplines focusing on novel big data technologies, policies, and innovations, the Journal brings together the community to address the challenges and discover new breakthroughs and trends living within this information. Complete tables of content and a sample issue may be viewed on the Big Data website.

About the Publisher

Mary Ann Liebert, Inc., publishers is a privately held, fully integrated media company known for establishing authoritative medical and biomedical peer-reviewed journals, including OMICS: A Journal of Integrative Biology, Journal of Computational Biology, New Space, and 3D Printing and Additive Manufacturing. Its biotechnology trade magazine, GEN (Genetic Engineering & Biotechnology News), was the first in its field and is today the industry's most widely read publication worldwide. A complete list of the firm's more than 80 journals, newsmagazines, and books is available on the Mary Ann Liebert, Inc., publishers website.

Mary Ann Liebert, Inc./Genetic Engineering News

Related Robotics Articles from Brightsurf:

Borrowing from robotics, scientists automate mapping of quantum systems
Riddhi Gupta has taken an algorithm used for autonomous vehicles and adapted it to help characterise and stabilise quantum technology.

COVID-19 should be wake-up call for robotics research
Robots could perform some of the 'dull, dirty and dangerous' jobs associated with combating the COVID-19 pandemic, but that would require many new capabilities not currently being funded or developed, an editorial in the journal Science Robotics argues.

How robots can help combat COVID-19: Science Robotics editorial
Can robots be effective tools in combating the COVID-19 pandemic?

Novel use of robotics for neuroendovascular procedures
The advanced technology has the potential to change acute stroke treatment.

Robotics: Teaming for future soldier combat
The US Army's investment for the 10 year, Army-led foundational research program has resulted in advanced science in four critical areas of ground combat robotics that affect the way US Warfighters see, think, move and team.

New haptic arm places robotics within easy reach
Imagine being able to build and use a robotic device without the need for expensive, specialist kit or skills.

AI-guided robotics enable automation of complex synthetic biological molecules
This article describes a platform that combines artificial intelligence-driven synthesis planning, flow chemistry and a robotically controlled experimental platform to minimize the need for human intervention in the synthesis of small organic molecules.

A step forward in wearable robotics: Exosuit assists with both walking and running
A soft robotic exosuit -- worn like a pair of shorts -- can make both walking and running easier for the wearer, a new study reports.

A first in medical robotics: Autonomous navigation inside the body
Bioengineers at Boston Children's Hospital report the first demonstration of a robot able to navigate autonomously inside the body.

Engineers build a soft robotics perception system inspired by humans
An international team of researchers has developed a perception system for soft robots inspired by the way humans process information about their own bodies in space and in relation to other objects and people.

Read More: Robotics News and Robotics Current Events
Brightsurf.com is a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for sites to earn advertising fees by advertising and linking to Amazon.com.