Using the NASA Thesaurus to Support the Indexing of Streaming Media Gail Hodge Information International Associates, Inc. Janet Ormes & Patrick Healey.

Using the NASA Thesaurus to Support the Indexing of

Streaming MediaGail Hodge

Information International Associates, Inc.

Janet Ormes & Patrick HealeyNASA Goddard Space Flight Center Library

Historic Context • The Library has collected and circulated the

Center’s colloquia on audio or video since 1967• A catalog of these holdings have been posted on

the Library’s web site since 2001• Patrons required to come to the Library, resulting

in limited accessibility of recorded colloquia • Streaming Media Center Project began in 2001 as

part of the Library’s response to Knowledge Management initiatives

Introducing the GSFC Media Center

Streaming Media • Streaming media

– Video that is encoded for delivery across the internet/intranet

• Encoding – Computer processing of video to a format for web

casting • Web casting

– The act of delivering audio and video content across the internet/intranet

– Can be delivered live or on-demand

The Goddard Library Streaming Media Center

• The Streaming Media Center is now available from the Library website (http://library.gsfc.nasa.gov)

• Can be included in personalized portals • Library has collected >350 hours of video

– >100 hours indexed

• Currently broadcasting 2 hours daily for the Earth Observing Systems Knowledge Management Pilot

Access Issues • Current Needs

– Need to know the overall topic of the video

– More likely to remember the topic, presenter, date or series

• Permanent Access– Less likely that users will remember the video’s

metadata

– More likely that users will want specific information

– Terminology may change over time

Indexing Video Content

• Video indexing is similar to a back-of-the book index for specific information

• Entering a keyword leads you to the specific location of the subject

Features of Selected Software• Compares recognized speech with stored

default terminology

• Uses speaker inflection to identify meaningful intervals

• Indexing and Search components included

Incorporation of NASA Thesaurus

• Added specific scientific terminology • Incorporated terms and their NTs, RTs and

UF/USE relationships • Used text of Astrophysics Data System to provide

terms in grammatical structures• Provides query expansion and improves relevancy

Query Expansion“Saturn Moons”

+ Ios+ Triton

Or“Scatha Satellite”

+ P78-2 Satellite

Query Expansion (Illustrated)

Sample Search (aurora) on same one hour lecture entitled “Jupiter’s Aurora”. One file was indexed using the NASA thesaurus, the other was indexed using a more basic scientific word list.

GREATER overall relevance understanding

Ignores IRRELEVANT content (Speech Recognition Error)

MORE relevant content found (2M+ VS 20 Sec’s)

Benefits

Relevance Interval Creation• Relevance Interval Creation links related

concepts within media files, which drives Relevance Intervals

• External knowledge from the thesaurus improves the accuracy of the Creation process because the explicit knowledge in text is incomplete

Relevance Interval (Illustrated)

Sample Search (aurora) on same one hour lecture entitled “Jupiter’s Aurora”. One file was indexed using the NASA thesaurus, the other was indexed using a more basic scientific word list.

GREATER overall relevance understanding

Ignores IRRELEVANT content (Speech Recognition Error)

MORE relevant content found (2M+ VS 20 Sec’s)

Benefits

Benefits• Identify relevant pieces of content within a

longer video• Stream more relevant, specific information

intervals to users• Minimize manual processing• Ultimately improve reuse of information

and increase opportunities for knowledge sharing