Tuesday, October 30, 2007

Multimedia Search

Presented by Ran Hock, Online Strategies

Ran formerly worked for DIALOG. He now has his own business to teach people how to use the Internet more effectively and efficiently. He says that he’s trained more than 14,000 people in over 30 countries.

Trivia: His son is the lead singer in a rock group, the Explosion.

Ran spoke in sound bites!

Some Facts/Observations


Use different tools to search for different types of media.
Use more than one engine. This is always good advice for searching.
Indexing is a problem, but in some cases it’s getting better.
Search qualifiers are often quite limited.
Use the Big Services -- Google, Yahoo!, Ask, Live, Exalead, and AOL.

Trends in the Big Engines

Integration of multimedia results with regular Web search results.
Google’s Universal Search - a bold step backwards. Altavista did it in 2002, only without images. Universal search has maps, images.
Ask.com does a better job of integrating images, and includes news images.
Yahoo! provides images with results.
Exalead does great job of integrating images in results. Links to multimedia on right side.

Image Search Engines


Indexing problem: For most images, there are few words on web pages that are definitely descriptive of the image.
Tagging could theoretically help. But not for a billion images, not with any consistency, not without high spam risks.
When searching, start with not more than two search terms.

Major Image Search Sites


Google has high numbers, but often low relevance.
Yahoo has lower numbers, but often offers higher relevance. Includes audio in addition to images and video.
Ask.com often has very low numbers and very high relevance.
LiveSearch frequently has low numbers. Its relevancy varied.
Exalead often has low numbers; relevancy varies.
Picsearch has lower numbers, but often higher relevance.
For all, there often is very little overlap among the first 10-20 items retrieved.

Search Features

Google - content type (any, news, faces), size, file type, coloration, site/domain, safe searching
Yahoo- size, coloration, site/domain, safe search
Ask - suggestions, size, file type, coloration, related video
Live Search - size, link to video, safe search
Exalead - size, file type, coloration, layout
Picsearch - size, images-animations, coloration
Coloration searching can be useful for locating line drawings.
Limit to .gov domain to find images in the public domain.

Images - Searching Flickr

Can search everyone’s photos, your photos, groups, Flickr members, or by location.
Can browse by month, category (groups from multiple photographers, sets from individual photographers)
Can search for group names
Advanced Search, Click on Search box to get to advanced search page, where you can limit results to Creative Commons-licensed photos. (This is how I locate photos for the DMHS PodSquad podcasts and for my instructional presentations and handouts. It's time for all of us to learn about Creative Commons licensing!)

Searching for Audio

Audio found on web pages is what is usually retrieved -- i.e., podcasts and music

Audio from Web Pages - Search Engines

- Old Timers - Altavista and AllTheWeb (identical results since both owned by Yahoo. These two have better results than Yahoo! Audio Search, however.)
- Yahoo! Audio Search
- Exalead
- AOL Audio Search (music only)
- Internet Archive

Yahoo! Audio Search


Over 50 million audio files
Search by subject, artist, album, etc.
Options to limit by materials, such as podcasts, the web, and online music stores
Refine by song title, artist name, album title, and lyrics

Exalead Audio Search

Do a regular web search and narrow it to audio.
Results are nicely organized.

Internet Archive

200,000 audio recordings, mostly music
44,000 recordings of live concerts and tons of old-time radio programs are available.
Open Source Audio, primarily Creative Commons material, is provided.
Can narrow search by 12 categories

Searching for Podcasts

Yahoo! Audio Search - incorporates capability of greatly enhanced metadata using RSS enclosures
Do an audio search and narrow results to Podcasts
Podscope - speech-to-text using voice detection software. Must look for synonyms sometimes. Example: MRSA vs. staph
Everyzing - speech to text
Podcast.net over 10,000 podcasts

Searching for Music

Sites to search include iTunes, eMusic, Rhapsody, Napster, etc.

Video Searching

More and more searchability due to tech. apps. such as voice recognition and enhanced metadata using RSS enclosures
Both big and specialty search engines
Google, Yahoo!, Live, Exalead, AOL are big players. Google, Yahoo! and AOL allow you to upload your own video.

Google Video

Archived TV programs, educational videos, and personal productions
Viewing some videos may require a fee. However, this may be going away because Google is offering refunds.
All content is stored on either Google Video or YouTube. No videos are retrieved from other sites.
Google is backing away from transcripts.
Advanced search for language, duration, domain, genre

Yahoo! Video


Gathers video by crawling and from metadata provided directly by suppliers.
Advanced search offers format, size, duration, domain, and safe search.

AOL Video

Some content is offered for purchase.
Contains legitimate clips from news services.
Can narrow results by popularity, ratings, and most recent.

YouTube

Searchability - terms automatically ANDed, can use OR and a minus sign for NOT
Title, description, and tags are searched
Can narrow by categories.

TVEyes

Radio and TV search - real-time monitoring
Aimed at the corporate world, which wants to know what media is saying about their companies.
Fee-based but some free initial searching is available.
Indexes audio feeds.

Shadow TV

Fee-based TV monitoring/ news clipping service
Monitors over 120 stations including all major US networks and cable stations.
Closed captioning is searchable and readable.
Has a 4-yr. archive
Search by keyword and some Boolean operators.

BlinkxTV

This free service searches primarily TV with some radio.
Has 18 million hours of video.
Creates transcripts.
Can search keywords and use Boolean operators.
Can rank results by date/relevance.
Conceptual searching
Has “wallet” feature that allow you to embed a search on your web page.
Advanced search offers exact phrase and ability to restrict by content providers.

Summary

Be aware of the range of tools.
Using more than one tool will pay.
Look for search quality to improve as new technologies are implemented.

www.onstrat.com/mmsearch.html has links to all sites covered during presentation.

ran@onstrat.com
www.onstrat.com

No comments: