Memory mimic aids reading

By Kimberly Patch, Technology Research News

The ease of any given task has a lot to do with how the brain carries it out.

Cognitive scientists have known for decades that human memory can be modeled using semantic networks -- sets of words like "cat" or "dog" whose connections depend on the strength of the associations between them. When the brain retrieves one piece of information it activates the retrieval of related pieces of information. This process of spreading activation enables us to easily retrieve related concepts.

Researchers from the Palo Alto Research Center (PARC) have devised software that leverages the way the brain models words to help speed the process of reading or skimming through digitized text. The software highlights portions of text in a way that makes it cognitively easier for the user to find what she is looking for.

The method could be used to make browsing Web search results, reading travel guides, and learning from textbooks easier, said Ed Chi, a research scientist at the Palo Alto Research Center.

The software, dubbed ScentHighlights, expands on a set of topics of interest supplied by a user to create a list of keywords tailored to the user's interests. "Users indicate their topics of interest by some method -- user profile, search keywords, clicking on words or index entries," said Chi. The system then adds related topics, he said. "The system uses all of the related concepts to create a list of important keywords that might be interesting to the user."

These keywords can be thought of as the pieces of semantic information that are likely to be activated in the user's memory, said Chi. The researchers' prototype highlights sentences that contain keywords in yellow, then highlights the user' search terms in pastel colors and related keywords in gray.

In one example, a reader is skimming through a biohazard book and has entered anthrax symptoms as a topic of interest. On one page several sentences contain the yellow highlight. Each instance of "anthrax" is highlighted light green, each instance of "symptoms" is highlighted light blue, and related keywords "nasal stuffiness", "twinges", "joints", and "fatigue" are highlighted gray.

The researchers' study showed that users performed a variety of information-finding tasks faster using the system than looking through a paper book. The next step is to perform eye tracking studies to understand how the software changes the user's eye movements while reading, said Chi. "We expect users to more quickly target relevant passages using ScentHighlights," he said.

ScentHighlights can be combined with a previously-developed indexing tool dubbed ScentIndex, which allows a user to query the index of an electronic book to get a page of index entries that are conceptually related to the query. When the user places the arrow over one of these entries, the highlighted paragraphs on that page appear in a text box.

The ultimate aim of the software is to reduce the cognitive stress of skimming for interesting content, said Chi. "It's very stressful to answer basic questions such as 'is this info available in the source document I'm reading?', 'which part of the document should I pay attention to?', 'what other related concepts are in this document that I should know about?'," he said. "I hope to show that this approach can alleviate these stresses."

The researchers' work falls under a new class of human-computer interface research that aims to first understand user behavior in terms of a scientific theory, and then, using that theory, find ways to improve user interfaces, according to Chi. "These types of research use back-end computation to model users [and use] the results to enhance user interaction," he said. "There is a strong theoretical validity to why it should work."

The software is ready for use in practical applications now, said Chi.

Chi's research colleagues were Lichan Hong, Michelle Gumbrecht and Stuart Card. They presented the work at the Intelligent User Interfaces Conference (IUI 2005) held in San Diego, January 9 to 12, 2005. The research was funded by Advanced Research Development Agency (ARDA) and the Palo Alto Research Center.

Timeline:   Now
Funding:   Corporate; Government
TRN Categories:  Human-Computer Interaction; Data Representation and Simulation
Story Type:   News
like to a new Related Elements:  Technical paper, "ScentHighlights: Highlighting Conceptually-Related Sentences during Reading," presented at the the Intelligent User Interfaces Conference (IUI 2005), San Diego, January 9-12, 2005


May 4/11, 2005

Page One

Memory mimic aids reading
Chip gauges cell reactions
View from the High Ground:
NYU's Nadrian Seeman

Surface tension drives nanomotor
Laser sniffs explosives
Nano pyramids boost fuel cells
Noisy snapshots show quantum weirdness


Research News Roundup
Research Watch blog

View from the High Ground Q&A
How It Works

RSS Feeds:
News  | Blog  | Books 

Ad links:
Buy an ad link


Ad links: Clear History

Buy an ad link

Home     Archive     Resources    Feeds     Offline Publications     Glossary
TRN Finder     Research Dir.    Events Dir.      Researchers     Bookshelf
   Contribute      Under Development     T-shirts etc.     Classifieds
Forum    Comments    Feedback     About TRN

© Copyright Technology Research News, LLC 2000-2006. All rights reserved.