Fachgebiet Psycholinguistics

Philipp Blandfort

Office: DFKI, Room 3.20

E-mail: Philipp.Blandfort[at]dfki.de

Curriculum Vitae


Education

since 2016

Doctoral Student, Center for Cognitive Science (TU Kaiserslautern)
Psycholinguistics and Language Development Group (TU Kaiserslautern) and Smart Data & Knowledge Services (DFKI)

2015
M.Sc. in Mathematics, TU Kaiserslautern

2013
B.Sc. in Mathematics, TU Kaiserslautern


Academic Employment

2013-2015
Research Assistant, digipen GmbH

2013
Tutor (Übungsleiter), TU Kaiserslautern, Mathematics Department

2010-2011
Tutor (Übungsleiter), TU Kaiserslautern, Mathematics Department


Research Interests

My overarching research goals are to better understand perspectives in multimedia communication and to make ways of interpretation more machine accessible.
This general theme motivates a bunch of questions from several fields:

  • What is a perspective and how to describe it in a way that is accessible to machine learning models? And why should we actually care about perspectives?
  • How to model variations in online communication?
  • How does linguistic style relate to perspectives? And what is style anyway?
  • How can we automatically generate data from a given perspective or with a particular style? For example, how to generate negative captions about images?
  • What is characteristic for ways of interpretation of political extremists and can such information be used to automatically detect ideologies?
  • What causes changes in people's perspectives?
  • How do perspectives relate to emotions, personality or other psychological factors?

For the practical part I typically work with image and text data from social media platforms such as twitter.
My toolkit mostly contains machine learning methods. Some relevant buzzwords here would be:

  • Deep learning
  • Natural language processing
  • Recurrent neural networks
  • Crowdsourcing


Publications

2017

Captioning in the Wild: How People Caption Images on Flickr
*Philipp Blandfort, *Tushar Karayil, Damian Borth, Andreas Dengel (2017, forthcoming)
Multimodal Understanding of Social, Affective and Subjective Attributes, An ACM MM'17 Workshop

2016


Generating Affective Captions using Concept And Syntax Transition Networks
*Tushar Karayil, *Philipp Blandfort, Damian Borth, Andreas Dengel (2016)
MM '16 Proceedings of the 2016 ACM on Multimedia Conference

Introducing Concept And Syntax Transition Networks for Image Captioning
*Philipp Blandfort, *Tushar Karayil, Damian Borth, Andreas Dengel (2016)
ICMR '16 Proceedings of the 2016 ACM on International Conference on Multimedia Retrieval


* denotes equal contribution

Zum Seitenanfang