1190 Vienna | AustriaPhone +43 (1) 320 35 55-0
E-Mail office[AT]modul.ac.at
From September 3-5 the Triple-I [1] conference 2008 (Innovations Conference for Knowledge Management, New Media Technology and Semantic Systems) took place in Graz, Austria. As in previous years, the conference was held in a relaxed yet productive and highly interactive atmosphere - this time with 444 official delegates, eight of them from the IDIOM [2] and RAVEN [3] project teams.
Professor Arno Scharl [4] presented the US Election 2008 Web Monitor [5] at the "Professors4Industry" track on Wednesday and participated in an expert panel on the impact of Web 2.0 technologies on the media industry on Thursday (together with Waltraud Wiedermann from APA DeFacto, Markus Lipp from Kleine Zeitung Digital and Werner Haas from Joanneum Research).
[6]On Friday, MODUL University Researcher Stefan Gindl [7] presented a novel approach for Sentiment Detection, called Building Tagged Linguistic Unit Databases for Sentiment Detection (Link [8]). This approach refrains from the usage of the traditional, simple tagged dictionary (also called opinion lexicon), which mainly contains binary sentiment information for terms. He proposed the usage of a database containing so-called Tagged Linguistic Units, which do not only hold information about the sentiment of a term but also context information.
This context information (e.g., part-of-speech tags, named entities or geo-locations) helps to improve the detection of the right sentiment for whole sentences as well as the sentiment expressed towards a certain entity in a sentence (e.g., a person). Despite this improvement the method will not suffer from performance loss, as many complex methods do (e.g., full parsing of the text, retrieving all relations of terms in a sentence among each other).
In the track on Knowledge Acquisition from the Social Web, the amount of presentations covering folksonomies [9] was remarkable. One goal of the presented work is to find ways to support ontology learning from folksonomies, along with automatic detection of relationships between tags, such as the presentation by Andreas Hotho [10], while other approaches looked at approaches to support users using folksonomies [11], or to find metrics for folksonomies. It appears that this topic will stay for a while, as some of the presented papers presented road maps rather than finished projects.
Last but not least the winners of the Triplify Challenge [12] were awarded - have a look at the winning projects!
Contact
Reinhard Fischer, MODUL University Vienna, Department of New Media Technology, Am Kahlenberg 1, 1190 Vienna, Austria reinhard.fischer@modul.ac.at [13] | www.modul.ac.at [14]
Links:
[1] http://triple-i.tugraz.at/
[2] http://www.idiom.at/
[3] http://www.modul.ac.at/nmt/raven
[4] http://www.modul.ac.at/scharl
[5] http://www.ecoresearch.net/election2008
[6] http://www.modul.ac.at/triple-i-panel-discussion
[7] http://www.modul.ac.at/gindl
[8] http://www.modul.ac.at/building-tagged-linguistic-unit-databases-sentiment-detection
[9] http://en.wikipedia.org/wiki/Folksonomies
[10] http://www.kde.cs.uni-kassel.de/hotho
[11] http://www.tagcare.com/
[12] http://triplify.org/Challenge
[13] mailto:reinhard.fischer@modul.ac.at
[14] http://www.modul.ac.at
Comments