To design and develop entity detection for the domain of economic news and to exploit it for the realization of trend mining tools for economic news content.
The envisaged approach is to develop a framework that can accommodate at least the following three parameters: media format (text – multimedia), data type (full text only - numerical economic indicators), and chronological perspective (contemporary content – historical archives). The tools should be of interest to news channels and platforms, and to digitized historical news archives. For the former, the tools should support facetted semantic content interpretation across media types and integration with numerical data types that are in use as indicator of economic trends. For the latter, the tools should deliver pertinent content scholars in economics and history and help them refine their models.
- Economic barometer based on the integration of textual and numerical data.
- Timeline visualization techniques for historical news archives.
Description of work
There is an intricate collaboration with the industrial participants, because they are essential in providing the data and user requirements for the analysis and cross-media mining functionality as basis for the development of new content services for their domain (Persgroep), and for the experimentation with entity detection and visualization of trends (Teezir). For EUR there is a dual added value in setting up collaboration between computational economics and experts in entity detection. First of all because it will stimulate the innovative coupling of data mining technology to text and media mining, and secondly because models for the generation and transfer of economic insights can now be checked against multifaceted historical data sets.
- Task 3.1 Development of training corpora (EUR/Persgroep),
- Task 3.2 Identification relevant economic indicators and model for integration with text mining (EUR),
- Task 3.3 Tool development and usability testing (EUR/Teezir /Persgroep)