Thursday, June 14, 2007

Silobreaker "Truthiness" – Middle East

We’ve had a lot of positive response on the “truthiness” report that was posted last week about the upcoming US Presidential Election 2008. So much that we will try to make these reports regular features on our blog.

Silobreaker provides more insightful results than traditional search engines by returning not just lists of headlines, but a fuller picture of people, places, companies, organizations, keyphrases, etc that are related or associated with each search result. This is a simple way of providing quick glances of instant “truthiness” that otherwise may require days if not weeks of manual research to work out.

Last week I took a look at what keywords each presidential candidate in the US was associated with and the differences between Democrats and Republicans.

Here is another way that Silobreaker can slice up the news flow. Let’s take a look at one of the most talked-about organizations in the Middle East right now – Hamas.

Q: Are there differences in the tonality of the reporting about Hamas between news sources from different countries?

I looked at sources from 5 different origins; US, Jewish, Arabic/Muslim, UK, and Chinese; their reporting around Hamas and especially what keyphrases and organizations that were associated to Hamas in such reporting (top of a list means the strongest association).

As I pointed out last week, this is not scientific research and may come with some obvious flaws (after all it took me all of 10 minutes). Real political commentators or media analysts may have very different views, but I guess one of the beauties with these truthiness reports is that they don’t take any views. They find what they find without any bias and leave the interpretation to others. Users may also find that these search results trigger questions and ideas, otherwise not thought of, leading to more comprehensive research through Silobreaker’s many other tools and features (or elsewhere).

At a first glance at the lists below, many keywords seem to be the same regardless of the origin of the source (possibly because the keywords represent the issues -but to which there are often two sides). At a closer look, however, the lists suggest more than subtle differences both in terms of tonality and the order in which certain keywords are placed on the lists. The result also suggests that sources from origins not immediately involved have a more emotionally detached and diplomatic tone in their reporting.

For instance, “Terrorist Group” and “Terrorist” are the most strongly associated keywords to Hamas amongst US and Jewish sources, while such terms come much further down the list for Arabic/Muslim sources (16th place) and hardly at all for Chinese sources (29th place).


In some more detail
Roughly 3,000 unique news articles were found in Silobreaker from the sources above in which Hamas was mentioned. All articles were in English. Blogs were not included in the analysis above.

Please note that the advanced search functionality that I used in Silobreaker to obtain the results above is not yet deployed in the beta version. Advanced search is only available in our subscription service so far, but is due to be released in the beta soon.

The relationships or associations that Silobreaker extracts and visualizes are not pre-determined or manually configured in any way. Silobreaker deduces relationships between terms or entities statistically by computing several measurements, such as co-occurrences, text-proximity, spelling variations etc in the information flow. This enables Silobreaker to offer more insightful search results than just traditional lists of headlines (often too many) by showing people, companies, places, keyphrases, etc that are related or associated with the search results. This in turn enables the user to move away from linear analysis to multi-dimensional exploration and discovery.

Check it out at http://beta.silobreaker.com/

Kristofer Mansson, CEO

Monday, June 4, 2007

Silobreaker "Truthiness"

At Silobreaker we talk a lot about “intelligent search” and providing more insightful results than traditional search engines. But what do we really mean by that and how does it translate into practical use?

Let me show you a simple example of how Silobreaker can be used for analytical purposes. This is not groundbreaking scientific research and may come with some obvious flaws, but it provides a quick glance of instant “truthiness” that otherwise may require days if not weeks of manual research (it took me about 20 minutes).

Let’s take look at what’s going in the upcoming presidential election in the US (covered quite extensively in Silobreaker here).

Q: Can Silobreaker analytically process the huge volumes of news around this topic to work out what issues are debated, which issues each candidate is associated with and what differences there are between the Democrats and the Republicans?

A: The simple answer is; yes it can!

By searching all news articles in Silobreaker for each of the candidates during the month of May and especially what keyphrases each candidate is associated with in such news flow, a picture starts to appear. I also searched on Al Gore and George W Bush to get similar perspectives for one Democrat and one Republican who are not in the race for the White House.

Real political commentators may interpret the news for the month of May very differently, but the Silobreaker analytics suggest that:
  • While Health Care is the most commonly associated keyphrase with the Democratic candidates, Abortion is the equivalent keyword associated with Republican Candidates (in fact it is the top keyword for every single Republican candidate).

  • Global Warming/Climate Change and references to War on TerrororWar in Iraq make the top 10 list for most Democratic candidates, but such keyphrases are noticeable only by their absence for most of the Republicans candidates.

  • Immigration is another keyword primarily associated with most Republicans, while Tax has similar associations in both camps.

  • While some observers seem to think that Al Gore may still announce his candidacy, the Silobreaker analysis suggest that he has some catching up to do in terms of policies – except for environmental issues, of course, which almost entirely populate his top 10 list.

  • The day-to-day issues for the current President, George W Bush, seems to have little in common with the policy-platform on which most of his fellow Republicans think they will replace him.

  • Judging by the percentage share each Democratic candidate received of the total news coverage for all Democratic candidates, John Edwards, Hillary Clinton, and Barack Obama are way ahead of the rest of their Democratic peers.

  • The equivalent percentages for the Republicans, i.e. each candidate’s share of all Republican candidates’ news coverage, suggest that John McCain, Rudy Giuliani and Mitt Romney have taken a lead in the Republican camp (in terms of news coverage – good and bad) .


In some more detail
Roughly 8,000 unique news articles were found in Silobreaker for the month of May in which one or several candidates were mentioned by name. All articles were in English from some 1,000 sources around the world. Blogs were not included in the analysis above.

The relationships or associations that Silobreaker extracts and visualizes are not pre-determined or manually configured in any way. Silobreaker deduces relationships between terms or entities statistically by computing several measurements, such as co-occurrences, text-proximity, spelling variations etc in the information flow. This enables Silobreaker to offer more insightful search results than just traditional lists of headlines (often too many) by showing people, companies, places, keyphrases, etc that are related or associated with the search results. This in turn enables the user to move away from linear analysis to multi-dimensional exploration and discovery.

Check it out at http://beta.silobreaker.com


Kristofer Mansson, CEO