Many new service providers and hardware manufacturers (Roku, Boxee, Google TV, Apple TV, Samsung, LG amongst them) are vying to bring consumers all manner of multimedia content from a diverse set of sources. Multimedia - video, audio and images - is computationally harder for devices to deal with as compared to plain old text documents. Regardless media (content) consumers are looking for relief from the barrage of content that comes their way constantly. Multimedia search offers a partial solution to the consumers' problems.Imagine that you heard of an interesting movie on your commute home. That evening when you sit down in front of the TV, you want to investigate the movie that you heard about. Ordinarily you would open up your laptop, search for the movie, read some reviews and add the movie to your Netflix queue (if you liked what you read). If the movie wasn't on Netflix, you might buy and download it from Amazon or iTunes. Then you would make sure that all your equipment was correctly connected. Finally, you would sit down to enjoy the movie.
The above scenario is a lot more appealing than waiting 2-3 days for a Netflix DVD or running to the local Blockbuster only to find out that the store doesn't have the movie on hand. However, the coming generation of Smart TVs will do better - if the TVs have a well implemented multimedia search engine. With a Smart TV, it ought to be possible to do away with the laptop and simply search on the TV, read the reviews, subscribe/rent/buy the movie from Netflix/Amazon/iTunes and start seeing it within a couple of minutes. All without fidgeting with multiple gadgets or interfaces. Multimedia search is key to making this work.
A multimedia search engine needs to worry about many different properties of the content that it searches including:
- Professionally produced vs user generated
- Live vs catalog / archived
- Video vs audio vs static images
- Short vs TV-show length (~20 to ~45 minutes) vs movie length
- User bookmarked/queued vs other
- Payment model: subscribe vs rent vs buy vs free (or Ad supported)
- Source: Netflix vs Hulu vs Cable/Satellite etc
In search engine literature, search engine quality is determined by the ranking, freshness, comprehensiveness and presentation of the results. Lets look at each of these in turn with an eye towards the context - that of a lay user wanting to quickly find consumable content on a TV.
Ranking:While many of the traditional measures of ranking such as "mean reciprocal rank" likely apply equally well to multimedia search, professionally produced content probably has more reliable signals to algorithmically determine ranking. This is because most professionally produced content has associated structured meta data and (likely) good quality anchor text by way of reviews and citations. As such a multimedia search engine is better off trusting signals from professional content over those from user generated content, and thus ranking professional content higher.
Freshness: In watching TV many consumers likely think of live TV as a category onto itself. This is especially true for Sports events, breaking news and shows airing for the very first time. A multimedia search should have special provisions for ingesting live TV guides and crawling web-sites of TV stations, TV networks, Sports teams and News sites such that search engines results are kept "almost live" fresh.
Comprehensiveness: A multimedia search should vie for comprehensiveness, but it needs to prioritize multimedia content over plain old text. As such in crawling web-sites, a good multimedia search engine would seek out sites which have video, audio and images (in that order) over text.
Presentation: Given the properties discussed above, presenting the results in straight linear form as web search engines currently do, hardly seems appropriate. I am not going to cover detailed user-interface/interaction-design here, but it seems beneficial to call out some properties of the search results in a visually appealing fashion. Perhaps live content, content from the users bookmarks (or queue) and professionally produced content could be surfaced separately from all other content. Further, content could be marked as video, audio or image/picture so that users know what to expect before they click. Finally, the length of the material (or the number of images) along with the payment model (subscribe/rent/buy) and source (Netflix/Hulu/Cable/Satellite) could also be surfaced - just so that the user can make a more informed decision regarding the appropriateness of the results vis-a-vis their intent, before clicking.
Several people that I have talked to are skeptical about search as the interface to discovering TV content - especially given that web search results seem to be less and less satisfying. Personally I believe that there are are a handful of use cases (including the one described above), where a well executed multimedia search engine would come in handy.
0 comments:
Post a Comment