Whats Next for Search?

At this point in the history of the Internet, it seems that most of the development of Web search is behind us—there is not much more that Bing, Google, Yahoo, and their ilk can do to improve, other than to squeeze a bit more performance when searching the ever-growing trove of website data for matching keywords. So what’s next?

It turns out there is quite a bit going on at the search engine companies to expand Internet searching beyond text-based keyword searching. Here’s a summary:

Image and Video Searching

All the major search engines have enabled users to search for images and videos, but they have relied completely on metadata and tagging. Metadata is information about the image or video, such as the size or length, who created it, when it was created, and so on. Tagging is more free-form text descriptions of the item and often depends on media consumers to take the time and effort to tag these files. Neither metadata nor tagging account for the actual visual data in the images and videos themselves. So, if you want to search for images of Charlie Chaplin, search engines will return only those images for which “Charlie Chaplin” appears in the metadata or tags. Images that do not have “Charlie Chaplin” in the metadata or tagging are not returned. The search engine can’t tell from the image data itself whether Charlie Chaplin is in the image or not. Similarly, if you have an image of a person and don’t know who it is, the search engines can’t help you.

There was some work a few years ago on incorporating speech recognition into video search. The idea was to automatically transcribe the speech in the audio track of a video and make the transcribed speech part of the metadata. But speech recognition still doesn’t work that well, and it is useless, for example, for clips from silent films or other content that doesn’t rely on spoken words.

So a lot of work remains in this area—don’t look for any major breakthroughs until supporting technologies such as facial recognition, computer vision, and artificial intelligence become more mature. In the meantime, if you have image or video content that you want users to find by searching, make sure you describe it accurately in the metadata and tagging, using as many different keywords as you can to describe the content.

Music/Audio Search

Like image and video search, music search today depends strongly on metadata and tagging. It’s a bit more straightforward because users typically will search for music by artist, title, album, or lyrics. But what if you heard part of a song on the radio, and you have no idea who recorded it, what the title is, or what the lyrics were (if it had any), but you remember the tune? Well, unlike video and image search, there has been much more progress in this area, and commercial products are available that “listen” to music or to your humming or singing and can identify the song. Google’s Sound Search and other services such as Midomi provide just this functionality. It’s not a stretch to see how this could be extended to non-music audio files, such as radio programs, podcasts, and other spoken-word recordings.

Metadata and tagging are still an important part of audio search, and if you have podcasts or audio clips on the Web, do yourself a favor and make sure you tag them properly and completely to maximize the hit rate. But search-by-sound is here and will continue to mature, providing another avenue for locating audio resources on the Internet.

Text and keyword search will continue to be the foundation of Internet searching for the foreseeable future, but other search technologies are here and maturing or are on the horizon. So keep talking with your web developer about the latest in SEO strategies to stay on top of the search results.