We are incredibly excited to announce today the debut of the new 2.0 version of the Television Explorer, part of the GDELT Summary platform and using data from the Internet Archive's Television News Archive! You can go ahead and get started or read below for all of the new additions to the Explorer platform.
Some of the biggest changes from the old system to the new Television Explorer platform are described in more detail below.
Vastly Expanded Catalog Of 163 Stations
The new Television Explorer now searches all 163 searchable television stations that the Internet Archive's Television News Archive has monitored since 2009, including CSPAN, international stations like Al Jazeera, BBC News and DW and Spanish language material from Univision and Telemundo (though not all stations have been monitored for the full 2009-present period). We are excited to see the kinds of analyses possible with this vastly expanded monitoring catalog.
Switch From Sentences To Airtime
The original Television Explorer reported how many sentences of the closed captioning transcript of each broadcast matched your search. This made comparing certain stations more difficult, since the average length of a closed caption sentence varies across stations and was also confusing to many users more accustomed to thinking in terms of airtime when working with broadcast data sources.
Thus, the new Television Explorer breaks each broadcast into a sequence of 15 second clips and searches for your keywords/phrases within a given clip. Results timelines report the percent of 15 second clips monitored from a station in a given day that contained your keyword, meaning results now represent the percent of monitored airtime from each station focusing on your topic (at the resolution of 15 second clips). A single mention of your keyword in a given 15 second clip will count that entire clip as matching, so 10 matching clips does not mean that 10 x 15 seconds of airtime focused on your search, just that 10 distinct clips mentioned your keyword somewhere in that 15 seconds. Note that phrases that span across two clips are counted towards only the first clip to prevent double counting.
Redesigned Context Searches
We have completely redesigned how we implement the "context" keyword searches, eliminating the edge case overcounting that could occur in certain circumstances under the old system. When using the new "context" operator, the system searches for your keyword both in the matching 15 second clip and in the 15 second clips immediately before and after it. Thus, a search for "trump AND (context:"russia" or context:"putin")" will identify all of the clips that mention "trump" and then search those clips and the clips immediately before and after them for either "russia" or "putin". In effect, when searching for multiple standard keywords, they must all appear in the same 15 second clip together, while "context" queries effectively search a 45 second interval.
It is recommended that when searching for multiple keywords, you should designate one as your main keyword and the others as context keywords.
Advanced users have asked for access to the underlying reporting statistics on the total number of seconds of airtime the Internet Archive has monitored from each station on a given day/hour so they can ensure that major findings in their results are legitimate, rather than monitoring artifacts (for example that a sharp decrease in discussion of Trump on a given station is real and not a result of an outage in that station's data for a few days).
You can now view an instant normalization timeline for any query by running your search and then clicking on the "Export" button at the top right of the volume timeline visualization and selecting "View Normalization Graph" from the popup that appears. This will open a new browser window that reports the total number of 15 second clips monitored each day/hour from each station searched by your query.
If you need to know the precise shows monitored from each station on a given day, you can also download the new daily Inventory Tables.
GDELT Summary Integration
The original Television Explorer was a standalone tool built initially as a proof-of-concept prototype and thus had an interface that was fairly clunky, was not mobile friendly and did not adhere to modern design principles. The new Television Explorer interface has been integrated into the GDELT Summary platform, ensuring it is fully mobile optimized and leverages all of the work we are doing in user experience optimization, while the underlying search platform is part of the GDELT 2.0 API suite (the TV 2.0 API), leveraging our new search infrastructure.