As broadband usage is increasing, many websites are increasing the audiovisual content available to attract and retain Internet navigators. Unfortunately multimedia, such as images, mp3 podcasts and video, presents a plethora of visibility problems for traditional search engines.
Search engine indexing and retrieval technology is tuned for text-based content, such as html files, not multimedia (also known as rich media). Over time, search engines have become very good at finding text in binary document types, such as word processor files (.doc, .odt), some pdfs and even, to a limited degree, animated Flash files. Multimedia files, such as video, still present their own set of problems that are considered in this article.
A Picture is Worth a Thousand Words: Indexing Image Files
As Danny Sullivan’s 2003 article on Multi-media search engines attests, most major search engines, such as Google, began addressing media files with image search. So what primary clues do search engines and thus, search engine optimization (SEO) specialists, have to identify an image?
- Image file name (starting with the extension!)
- Image alt (alternative) text
- Text information in the html file around the embedded image
Some image types, such as .gif files, allow meta text about the image to be saved with the image itself. It is unlikely that this meta information is often specified nor useful. You have probably already come across meta information, information about information, if you have ever authored an html document or updated a word processor document’s properties. Html documents may contain information about the document, such as the language, description and keywords, encapsulated in html meta tags.
While there have also been many valid projects to detect an image’s subject by reviewing pixels, such in face recognition (or filtering out nude images), search engines need to balance what is possible with what is realistic (consider the breadth in the gradation of human skin color). Search engines like scalable solutions, solutions which can be applied to the entire web rather than to a simple laboratory setting.
As anyone who has used image search can attest, the process is far from foolproof. Automatic algorithmic analysis of web page text commentary around an embedded image may work much of the time, but can create some very perplexing results on other occasions – there is no guarantee that the editorial text content on a web page is really related to the images which appear on a given page.
Sound Check: Indexing Audio Files
With audio media, search engines have basic information similar to that with images:
- Audio file name
- Text information in an html file around the embedded audio file
- Standby message text (using the object tag standby attribute)
- Tagged meta information included in the audio file
Meta information is supported in the mp3 format as ID3 tags. Other audio formats with meta tag support include ogg, flac, mpc (musepack) and ape (monkey audio). As meta information is better supported in audio files, it is easier for search engines to truly identify the contents of an audio file, but only to the extent the audio file creator accurately specified meta tag information when creating the file.
Occasionally, audio providers will include a text transcript of the audio content. We highly recommend this practice. Consider usability: Internet users on dial-up connections can download text files much more efficiently than audio files. Search Engines have much richer information to associate with the audio file.
Roll the Cameras: Indexing Video Files
With the great increase in the number of video content and video viewers on the web, search engines have turned their attention to video search, such as Yahoo! Video, and as exemplified by Google Video and YouTube, video sharing (hosting) as well. Along with the big names in search, relatively new names are trying to make a splash in the video search business. One, Blinkx, has even released a guide to video seo (a tip of the hat to Andrey Golub for pointing this out).
The Blinkx SEO guide discusses initial video indexing efforts, such as Altavista, encompassing the same basic techniques as used in image and audio indexing:
- Video file name
- Text information in the html file around the embedded video (including transcripts)
- Some meta information from the video file itself, such as creation date
In essence, a good first start but subject to the same limitations as image indexing which relies on inferring an object’s contents.
Cut: Second Generation Video Search, According to Blinkx
Blinkx aims to differentiate itself from what it calls first generation video indexing by considering information culled from an analysis of the video itself: speech recognition, visual analysis and recognition and video optical character recognition (OCR) to allow software to listen to, watch and read the text appearing on the video content itself.
By text, keep in mind that many video formats support subtitle tracks, (both standard and hearing impaired, known as closed-captioning in North America). In some cases, the subtitles are stored in a textual format that could be indexed directly. In other cases, the subtitles are saved as images that need to undergo transcription via OCR, with all of the potential inaccuracies that entails.
Blinkx makes a particular note regarding products which maintain meta data during video conversion. Their recommended products include:
So after considering multimedia search indexing issues, what can marketing professionals and webmasters do to improve the visibility of rich media in search engines? In a related article, Antezeta considers the top 8 video search engine optimization tips.
- 8 Ways to Optimize Video for Search Engine Visibility
- Now there are 6 ways to keep website content out of search engines
- Search Engine Crawlers: Who’s visiting my site and why?
- Search engine support of rel=”” link attributes – cheat sheet