Wednesday, February 4, 2015

Clip and Save: Cataloging and metadata in large video collections

Tonight's class lecture included some interesting thoughts on video cataloging, particularly on how granular it needs to be.  Is it useful to catalog, for example, the entire NBC news broadcast from February 3, 2015, as if it were a single, monolithic item?  Perhaps our users really just want to watch that 2 1/2 minute segment on Harper Lee's new novel.

Thinking along exactly these lines, NBC has digitized and indexed 12,000 separate stories from its archives, drawing from both current material and much older content--as far back as the 1920's. NBC now sells this digitized content to libraries, universities, and K-12 schools (there are both higher education and K-12 versions) in the form of a streaming media database called NBC Learn. Each clip has a full set of descriptive metadata, and what is more interesting, this metadata has been converted into a complete set of MARC records available (readable XML version here) to subscribing libraries if they want to make every individual clip available through their online catalog.  While one can certainly debate the value of adding thousands of records for short video clips to the library catalog, it is interesting that the records are available.

VAST Academic Video, a similar-sized video clip collection from Alexander Street Press, which features more scholarly, independent, and/or historical material, likewise fully indexes every clip in the collection and provides MARC records for 80% or more of them. VAST also supplies its index data to the major library discovery products and will soon be available through Google searches.

Comparing the two databases side-by-side, it's easy to see that VAST far outstrips NBC Learn in terms of specific indexed fields; on the other hand, it should be noted that the keyword fields in NBC Learn are comprehensively populated and also individually hot-linked to perform an instant keyword search of the entire database.  When converted to MARC 21 format, VAST's records are more complete and standard in content, and the subjects map to LCSH. NBC Learn's records are shorter, and the keywords map to 653 - Index Term-Uncontrolled, resulting in keyword-searchable records, but no controlled subjects. The end result is two good, but quite different, approaches to indexing massive amounts of short video content.

NBC Learn VAST Academic Video Online
Title  Title 
Translated Title
Transcript Transcript
Citation (MLA, APA, Chicago) Citation (MLA, APA, Chicago)
Source Publisher
Producer Institution
Creator Author
Director
Narrator
Air/Publish Date Release Date
Event Date Date Recorded
Resource Type Format
Content Type
Clip Length Duration
Copyright Copyright Message
Copyright Date Original Release Date
Description Abstract/Summary
Release Notes
Related links
Keywords Subject
Discipline
Field of Interest
Specialized Area of Interest
Topic/Theme
Health Subject
Anthropologist/Ethnographer
Client/Patient Age
Cultural Group Discussed
Featured Speaker
Language Of Edition
Original Language
Person Discussed
Place Discussed
Place Published / Released
Series
Series Number
Therapist
Therapist/Physcologist Gender
Therapist/Psychologist Race/Ethnicity

1 comment: