Shallow Sentiment Analysis for Brussels
Labresearch by Tom De Smedt
PhD student in computational creativity, CLiPS, Universiteit Antwerpen
http://organisms.be
Shallow sentiment analysis for "Brussels"
---------------------------------------------------------------------
What are people saying about Brussels? Can we derive a useful starting point, or noteworthy problems in the state of Brussels as it is from these expressions?
With "sentiment" is meant feelings, opinions, personal and subjective expressions.
With "shallow" is meant that none of these results are representative.
They originate from an internet mining algorithm with the general purpose of providing creative inspiration, new angles of thought. The software (currently in its testing stage) is developed as part of my PhD research in computational creativity. The goal of this research is to develop novel software tools that allow people to be more creative based on real-world / real-time data, and opinions from others.
The software is built on MBSP (Memory-based Shallow Parser) developed at the Computational Linguistics and Psycholinguistics group at the University of Antwerp. The visualizations are realized in NodeBox (http://nodebox.net), developed at the Experimental Media Group at Sint Lucas Antwerpen.
WE ... BRUSSELS
---------------------------------------------------------------------
In the first exercise, expressions in the form of "We love / hate / visited Brussels" are mined from the internet. The winner, "We think Brussels" is undoubtedly just the start of a sentence (like "We think Brussels is lovely") and must therefore be disregarded. This leaves us with top candidates such as "adore" and "hate", of which "hate" is the winner by a narrow margin. We could deduct that people have strong (and opposing) feelings about Brussels.
BRUSSELS IS ...
---------------------------------------------------------------------
This exercise mines adjectives related to Brussels from the internet. Again, the winners ("best", "further" and "famous") are likely just the start of an expression (such as "Brussels is best known as the home of the EU" or "Brussels is famous for X or Y") and should therefore be disregarded. This leaves us with interesting winners like "sumptuous" and "boring". To anyone who has been to Brussels but doesn't known about the all cosy little districts in detail, "boring" comes as no surprise. Many areas of Brussels could indeed be described as non-specific, gray, or boring. Contrarily, Paris or London are almost never described as boring in general terms. Why is this? This could be a first challenge to tackle.
BRUSSELS IS ... THAN ...
---------------------------------------------------------------------
In this exercise the internet is mined for sentences that compare the concept of "Brussels" to other concepts. The winner is quite surprising: "Brussels is prettier than I thought / expected". This might indicate that visitors who take their time to delve into Brussels undergo a change of opinion (from boring to pretty). There seems to be a potential that does not easily surface.
BRUSSELS COLORS
---------------------------------------------------------------------
In this exercise, web pages that contain both the word "Brussels" and a color (also as a word) are counted. The winner is "blue", by a narrow margin to "red" and "black". Not a very heartening conclusion. Also, the predominant color shades mentioned are "cool" and "light". My personal opinion is that these are not the right colors for Brussels. It might be worthwhile to market other, more refreshing color combinations in relation to Brussels.
BRUSSELS ANALOGIES
---------------------------------------------------------------------
Finally, based on the "properties" of Brussels (e.g. easy, boring, chaotic, beautiful, quick, proud, ...) the software is able to deduct analogies. For example, given how people describe Brussels and how various flowers look, what flower would Brussels be - taking into account the amount of authors that share the same opinion?
- Flower: iris or lily (similarities on boring and chaotic)
- Animal: toad or chicken (similarities on boring, beautiful and proud)
- Building: prison or school
- Person: Arnold Schwarzenegger
Note that the results are still experimental at best.
More information on the analogy software can be found at:
- http://nodebox.net/code/index.php/Perception
- http://nodebox.net/code/index.php/Perception_comparisons
