[ddj] new user-friendly interface for gdelt global online news text and visual search apis

kalev leetaru kalev.leetaru5 at gmail.com
Wed Nov 29 15:10:15 UTC 2017


A number of you have asked about a more user friendly interface to the open
GDELT online news fulltext search APIs, and we just released GDELT Summary
yesterday that should address many of your comments and feedback:


https://blog.gdeltproject.org/announcing-gdelt-summary/


It is essentially a non-technical user-friendly wrapper around the APIs,
allowing them to more readily be used by non-data journalists in the
newsroom and as a quick way to rapidly iterate a query to find the best
search terms/filters and then get back the embed or JSON URL for a given
visualization (volume timeline, tone chart, source map, image list, etc).
(This also allows non-data SME journalists to refine a query and then pass
it over to the data side of the house for them to instantly access the URL
for the JSON output of the API for 15 minute live feeds of content, iframe
embed, etc).

In essence, you have a simple search interface that lets you search in
English across the live English machine translations GDELT produces of all
online news coverage it monitors in 65 languages. Thus, a terror attack in
Egypt or a bridge collapse in China can be seen through the eyes of local
and global press through local voices in local languages. (Pair with Google
Translate for basic gist translations or flag articles of particular
interest to send to colleagues with that language specialty).

We've also made it very easy in this new interface to perform deep
learning-powered visual search of global online news imagery. Several
special query terms allow you to search the 10,000 objects and activities
visually recognized from each image, the two million topics recognized from
the caption of each image on the page and anywhere it appeared elsewhere on
the web across the majority of languages it appeared in, an estimate of its
popularity via a reverse Google Images search, search of the embedded
metadata and OCR'd text from the image background, etc.

You can also combine text+image searches, asking for articles that mention
"climate change" in the body text and which also includes a photograph of
protesters, for example, for reporting on climate protests. Or see what
kind of social sharing imagery is being used to illustrate a world leader's
visit to a given country or a current topic of interest.

Reach out directly to me at kalev.leetaru5 at gmail.com with any
feedback/comments/etc, as we're particularly interested in how these tools
can be used by data journalists to create live alerts, trend analyses,
assist in deep diving local coverage in other parts of the world and in
other languages, explore the visual narratives of the news, etc.


Kalev
https://www.gdeltproject.org/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.okfn.org/pipermail/data-driven-journalism/attachments/20171129/f7478116/attachment-0002.html>


More information about the data-driven-journalism mailing list