Wednesday, August 25, 2004

HP Speechbot Search Engine Parses Multimedia Files

I came across an old search engine that seems to have been missed during the past few years bu thas some pretty awesome technology, Speechbot, HP Lab's version of a multimedia search engine. The cool part is that it parses the audio portion of multimedia files and renders audio into text. The bad part is they're only indexing publicly accessible audio files from sites such as NPR, and American Radio Works. So unless you are a real National Public Radio buff and can't get enough of Click -n- Clack (the Car Talk guys) then you're out of luck.

Pros

• Allows the actual content of a multimedia file to be indexed as if it were text and searchable. This is a vast improvement over relying on the author & title info stored in most multimedia files or even worse the file name and text that is near the file itself for deriving the meaning of the multimedia file.
• Far superior to AltaVista's multimedia search functionality and Lycos' MP3 search feature.
• Not only can you search for the files but you can view the transcripts and play the file after you've found the one you were looking for.
• Links to original multimedia files are kept in tact.
• Very Quick!

Cons

• Lack of depth to what's currently indexed (probably due to copyright laws)
• Unless you're an NPR nut or a Dr. Laura follower, you're probably not going to find anything entertaining.

Multimedia Search Humor
Searches for "50 cent" or "in da club" don't return anything meaningful.

Further Reading
Read the research paper that explains the technology used in Speechbot for youself.

Posted By Jason Dowdell at 12:06 PM
Permanent Link: HP Speechbot Search Engine Parses Multimedia Files | Comments (0)

Post a Comment

Name:
Email:
WebSite:
Comment:
Remember Me
Subscribe