- From: Bernhard Ortner <bernhardortner@gmx.net>
- Date: Tue, 21 Jul 2015 14:22:56 +0200
- To: public-sentiment@w3.org
Dear All As I have proposed it in the meeting, here are some information for those people that missed the call today. Current Projects (will be updated in the wiki) My main project currently is Fupol [1], a policy decision making tool, that facilitates to collect various opinions over different social media (SM) channels, e.g. facebook, twitter, rss... or even newspaper articles. Furthermore it uses categorization to summarize those social media streams and to provide a top down view on currently discussed topics. My background: I'm currently focused on the Information Retrieval of LD, i.e. data harmonizing and integration and text mining. As I wrote it above, in Fupol we (my company and I) integrated various SM sources and ran several (domain specific) natural language processing "algorithms" such as Latent Dirichlet allocation (LDA) or Non-negative matrix factorization (NMF). My Master thesis had its main focus on detecting LD structures on the LOD cloud that can be used as anonymization pattern (will be finished this summer). Furthermore I'm also a part of W3C's RSP group [2] where the group focuses on standardizing RDF Streams. Consequently my next step is to analyze posts on a "high level" for example apply sentiment analysis to (streamed) micro posts. Use Case: I want to apply sentiment analysis at a RSP Processor, e.g. C-Sparql. Two main issues have to be concerned: 1) The algorithm has to finish within a defined delta (a frequently used delta is 1 second) 2) limited space consumption of the underlying algorithm Consequently current approaches may not be applicable, because they violate these constraints. Best Bernhard [1] http://fupol.eu/en [2] https://www.w3.org/community/rsp/
Received on Tuesday, 21 July 2015 12:23:27 UTC