Archive December 20, 2018

Semantic Frame Extractor online

By in Uncategorized

A new Penelope Component, called Semantic Frame Extractor is now online!

Frame semantics is commonly used as a methodology for representing the meaning of linguistic utterances. While semantic frames have successfully been formalised on a large scale, it is still a major challenge to automatically extract them from raw text. This Penelope component overcomes this challenge by using precision language processing techniques. Concretely, the component takes a sentence (or a list of texts) and a frame of interest (e.g. ‘Causation’) as input and returns all instances of this frame, and its frame elements, that occur in the sentence (or list of texts). The language processing part of the semantic frame extractor has been developed within the Fluid Construction Grammar (FCG) framework. 

The OpenAPI specification of the component is available at https://app.swaggerhub.com/apis/EHAI/Semantic-Frame-Extractor-API/1.0.0. As all components, it can be used form any programming languages are via Penelope interfaces such as the Penelope Workbench.

Post-doctoral position available

By in Jobs

We are looking to hire a post-doctoral researcher (100%) to join the Penelope team in Brussels. You will have a leading role in shaping the Penelope platform, which consists of tools and techniques for mapping opinions expressed in online (social) media. Responsibilities include the design, implementation and integration of novel tools (in particular NLP-related), steering the development of Penelope within the project and (if desired) supervision of students.

Requirements and application: The candidate should have experience in NLP and machine learning. Very good spoken and written command of English is required. The starting date is flexible, depending on the availability of the candidate. Women are especially encouraged to apply. The initial contract will be for 1 year, and can possibly be extended. Please send your application by e-mail to ehai@ai.vub.ac.be including your CV, transcripts and certificates, a cover letter, a statement of research interests, your PhD thesis, as well as the contact details of 3 references before 15 January 2019.

As an employee of the Vrije Universiteit Brussel you will work in a dynamic, diverse and multilingual environment. Our green campus is located in the centre of Brussels, the lively capital of Belgium and Europe. Depending on your experience and academic merits you will receive a salary according to the official pay scales. Hospitalisation insurance and free use of public transport for travel to and from work are standard employment benefits. Both campuses have extensive sporting facilities which are at your disposal and a nursery is within walking distance. More information is available at www.vub.ac.be/en/jobs.
Contact: For more information, please contact Dr. Katrien Beuls (katrien@ai.vub.ac.be).

Inspiration day of the Flemish Radio and Television (VRT)

By in Uncategorized

The main ideas of the ODYCCEUS project were introduced to a delegation of the Flemish Radio and Television (VRT) at the inspiration day organised at the VUB AI Lab. Paul Van Eecke and Katrien Beuls demonstrated the semantic frame extractor, which detects semantic frames in news paper articles and tries to fill in their slots. An example of such a semantic frame is the Causation frame, with the verb “cause” that evokes the frame and phrases such as “climate change” and “rising sea levels” as the cause and effect slots.  You can read more about a first version of the semantic frame extractor in the online web demo.

A further follow-up meeting is planned with the VRT News editorial team to discuss a potential collaboration.

Workbench piloted at JRC in Ispra

By in Uncategorized

The Penelope workbench was presented and piloted today at the EU Commission’s Joint Research Centre (JRC) in Ispra during the third Resonances Summer School, this year devoted to the theme of “Big Data”. The pilot took place in a session on “New tools for social media and fake news”, in which Luc Steels gave a talk and Michael Anslow (Sony CSL Paris) demonstrated the workbench. The Resonances Summer School is an initiative of the Sci-Art project, which is joining art and science to support EU policy making.

Guest lecture at VUB-SMIT

By in Uncategorized

Katrien Beuls gave an invited talk at the VUB centre for Studies in Media, Innovation and Technology (SMIT) where she introduced the ODYCCEUS project. She talked about the goals of the project to create conceptual and social maps from online media and use these maps to make social discourse and social dynamics visible. The semantic frame extractor was also demonstrated, as it was used as an augmented reading tool for a newspaper article about climate change.

First pilot study of the Penelope workbench

By in Uncategorized

The Penelope workbench, which is the first Penelope interface that was released to the public and allows social scientists to build pipelines with available Penelope components, has been piloted at the first ODYCCEUS conference in Leipzig this week. The results of the pilot study show that the participants rated the workbench as useful (8/10) and state that it met their expectations (7.5/10). The workbench can be tested on https://penelope.vub.be/workbench.

Penelope at Databeers Brussels

By in Events Tag ,

Katrien Beuls (VUB AI Lab) gave a talk about opinion tracking in online news media at the latest Databeers event in Brussels. She introduced her work on close and distant reading techniques for text analysis and promoted the Penelope platform as an open-source community effort to join scraping, NLP and statistical services in a responsive web application that allows social scientists and media researchers to find patterns in large amounts of text, but also to get a deeper analysis of a selected text, in terms of its argumentation structure and links to external knowledge sources.