Resources

Theme Recognition in action –  Part 1

Since our announcement, a couple of weeks ago, we have received a lot of interest and numerous questions on how it works on a real case.

So, let’s take an example:

The BPI (Banque Publique d’Investissement) who, through our QIO (Questions>Insights>Outcomes) is asking us:
“What are the trends and how do compare those trends between Silicon Valley and the French Ecosystem for 6 topics: Artificial Intelligence, IoT, Transportation, Drones, Smartcity and Security.”

1/ The Encyclopedic recognition of Themes

Netvibes is using a personalized version of both DBPedia and DBPedia Spotlight, both open-sourced.

DBPedia structures the unstructured nature of Wikipedia to make it queryable. DBPedia Spotlights then annotate every single piece of content within the corpus aggregated in Cloudview (from Exalead, our sister brand at Dassault Systemes).

The aggregation process, to create and feed the corpus, is made of our own crawlers which will gather in permanence, content from personalized vertical libraries for Mainstream, Tech, Biz and Investment news. This way, we solely analyze noise-free content which is a fondamental requirement to get trustworthy content onto which we can build intelligence.

Immediate benefits of this Theme Recognition process:

a. Up-to-date encyclopedic annotations.
Corpuses are annotated with the biggest, most reliable and updated encyclopedia: Wikipedia.

b. Accurate disambiguation.
Using very rich and up-to-date profiling from Wikipedia means more accuracy the type of entity spotted.

c. Crowdsourced, international ontology.
Using customly improved ontologies open doors to more powerful filtering and smarter theme recognition.

2/ Drilling down to Insights

Let see it in action:

a. As an example, let say we want to find Insights on the latest Artificial Intelligence (AI) technics in the US within the News and Tech press in the past 3 months.

After querying the News and Tech press libraries over the period of time, Netvibes will return the following Top Themes for AI: 

 

Clearly, and unsurprisingly, Machine Learning is the top subject within AI.

b. Now let’s drill down on Machine Learning…

c. Artificial Neural Network (ANN) looks definitely intriguing… let’s dig into it 

d. Human Brain? as a Theme for AI?

e. Cognitive psychology! (mental processes (as perception, thinking, learning, and memory) now gives us 2 articles to read, straight within Netvibes:

They are both here (enjoy. really interesting):

Bottom line.

In the past 3 months, in the News and Tech US presses, within “AI”, “Machine Learning” is the top subject where “Artificial Neural Network” looks at miming the “Human Brain” like with “Cognitive Psychology”, illustrated by 2 articles on bottleneck (wired) and mind reading technics (foxnews).

Pretty insightful, no? (oh, and plenty of other Insights can be found in this example…)

To be continued on Part 2…

Rendezvous:

Want to watch Netvibes in Action? we’re presenting at the next BPI event on Nov 8th… 

Join us here

deck in english, show in french

 

Previous Post Next Post

No Comments

Leave a Reply