Behind the scenes @ Labs
Showing results for 
Search instead for 
Do you mean 

HP Labs creates Echo, a powerful new tool for understanding opinions in context

‎01-25-2013 02:31 PM - edited ‎09-30-2015 06:45 AM

Contributed by Simon Firth, freelance technology journalist


Josh-Hailpern_Jan-2013_web.jpgIt’s a common challenge in the Information Age: thanks to modern storage and search technology, we can easily acquire huge sets of documents that will likely be useful to us—but we then have to read through them all.


“You could be a doctor reading up on a new research field, a lawyer looking for legal precedents or a marketer gauging what people have been saying about your product. They’re all tasks typical of our information-rich, time-poor world,” suggests HP Labs researcher Joshua Hailpern.  


In each case, Hailpern argues, it would be highly useful to know whether the opinions in a specific document are typical of the whole set, or outliers—and to have that information created automatically. It would help us research faster and more accurately, and fight back against information overload.


Last summer Hailpern and HP Senior Fellow Bernardo Huberman set out to create a tool that would do exactly that. The result is Echo – a software system that can take any set of documents, automatically register the opinions contained in each and then display the results via an interface that’s both powerful and intuitive to navigate.


Complex algorithm, simple interface


At the heart of Echo is a Java-based, five-stage algorithm that analyzes what each document is about. Going sentence by sentence, it extracts key words and picks out sentiments (agreement, for example, or concern, derision or praise) expressed in each article about specific nouns that represent concepts, people, places or things.


Also key is an interface, called the Echo Reader, that uses a magazine-style metaphor to direct your attention to the specific opinions expressed in each document. Changes in typeface and visual call-outs let you “read as normal,” while focusing your attention on representative or unusual opinions within each document.


For any article, it’s possible to highlight opinions that are typical (‘representative’) when compared to the others, or atypical (‘out-of-line’).  And for any highlighted opinion, Echo offers a graph showing the distribution of sentiments on the same subject across all documents.


“It allows you to quickly check how that specific opinion matches the opinions registered by the community of writers in your chosen set,” suggests Hailpern. “And you can do that for any opinion that you encounter.”


A third aspect of the system, the Echo Locator, reflects patterns across all the documents being investigated. You can browse the entire set, see key words and sort documents by the degree to which they contain opinions (on any specific topic or overall) that are either representative or out-of-line as compared to the others.


Pilot study


Echo is grounded research in social analytics, sentiment analysis and the economics of attention being undertaken in HP’s Social Computing Research Group, of which Huberman is director.


The tool took Hailpern, an expert in user interface design, just 3 months to build last summer, after which he tested it with a number of HP employees.


“We piloted it with people in Legal and in Marketing on sets of up to 300 documents,” Hailpern recalls. “They felt it was both faster and more accurate than skimming through such a vast group of articles and that it gave them a clear sense of the scope of opinions expressed throughout.  One person said, ‘it’s almost like cheating.’”


Next, he and Huberman plan to develop a demonstrator that invited participants can populate with their own sets of documents.

“There’s a lot that we can learn by seeing how people use it,’ he says. “Really, we’re just scratching the surface of this new approach of using context in the crowd to position opinions.”


For a demonstration of how Echo works, watch this HP Labs video:




For more on the research context behind Echo and the methods used in its creation, download Hailpern and Huberman’s research white paper: Echo: The Editor’s Wisdom with the Elegance of a Magazine


0 Kudos
About the Author


Nov 29 - Dec 1
Discover 2016 London
Learn how to thrive in a world of digital transformation at our biggest event of the year, Discover 2016 London, November 29 - December 1.
Read more
Each Month in 2016
Software Expert Days - 2016
Join us online to talk directly with our Software experts during online Expert Days. Find information here about past, current, and upcoming Expert Da...
Read more
View all