DTV and search and recommender systems - TvTechnology

DTV and search and recommender systems

The 500-channel universe envisioned in the mid-’90s may not have arrived, but there are easily more channels and programs available today than are humanly possible to manage. The answer to this quandary begs implementation.
Author:
Publish date:

The 500-channel universe envisioned in the mid-’90s may not have arrived, but there are easily more channels and programs available today than are humanly possible to manage. Electronic Program Guides (EPGs) attempt to help the viewer sift through all this content, but are rudimentary at best. Their best implementation is to enable a simple, one-button record on DVRs.

But if you can’t find a program, you can’t watch it or record it. EPGs do provide some program information. PSIP tables, such as the EIT (Event Information Table) and ETT (Extended Text Table), contain lots of descriptive program data. Title, actors and plot information are all available. Programs can be listed by genre or title on a program/channel grid. But if the presented information doesn’t contain a reference to something you are looking for, that content remains lost.

There are two features that can solve this dilemma: program search and recommender systems. All the data to enable these capabilities already exist in a DTV transmission.

Search capabilities

Existing content categorization is of little help if the viewer wants to know if a program with a particular actress is available in a broadcast, VOD or PPV offering. This type of information is often present in the EPG, so, why can’t this data be used to enable a keyword search through programming?

Every television has some form of microprocessor on board. It could easily be used to implement a program search function by way of a little coding.

The information necessary to enable this capability is already available. Keywords could be extracted from the EPG. A list of actors, subject and other search terms could be developed by analyzing PSIP data. A list of the characteristics of available programs could be built and presented in a window. Then, highlighting a term, in lieu of a complete alphanumeric input, could initiate a program search, ending in a list of programs meeting the criteria.

Recommender systems

Taking this a step further, just as a Web browser keeps a history, a DTV could as well. This could raise privacy issues, so, those implementations using a backchannel (two-way communication) would need a way to insure that this information is not accessible to a service provider without viewer permission.

By performing a statistical analysis on viewing selections and previous searches, and using this information in a new search, the probability of locating a show of interest would be increased. An advanced feature could automate such background searches. When something of potential interest is found, an alert could be sent to the viewer along with viewing and recording options.

Current options

Right now, the only applications that can perform a programming search based on descriptive criteria are on the Web. Yahoo video search (http://tv.yahoo.com/) offers the ability to find content based on keyword entries.

The Web application, however, is typically isolated from a DTV, STB or DVR. It can’t send an e-mail reminder that a show will be on tonight, and there is no way to run a constant search in the background.

Metadata is the key

Enabling program search and recommender features will require broadcasters to pay close attention to data entered into the EIT and ETT PSIP tables. The ability of such features would directly depend on this information. This means that the broadcasters have to transmit content descriptive metadata.

Determining exactly how to describe a program will become a science. Some principles and techniques now being developed by Internet search marketers may be applicable to TV programming.

For these features to work on large program libraries, the metadata will need to be automatically generated. Systems that use speech recognition and closed-captioned analysis to produce descriptive metadata are just now emerging. Expect intelligent search and recommender systems to use this metadata to track user preferences and suggest an evening’s entertainment.

An ironic twist to this scenario could find broadcasters tweaking dialog and scene composition to provide a favorable aggregation of keywords by an automated metadata extraction system. By finely tuning these content parameters, broadcasters could increase the number of program search hits on their content.

Here tomorrow

As more programming becomes available through digital storage and emerging delivery channels, the number of shows and clips offered will become virtually infinite. Program search and recommender systems will be required so viewers can experience this vast reservoir of entertainment and educational content.