Prototype Details
Prototype:
Event matching: linking culture-related events from website data to learning content
Researchers:
Bernhard Göschlberger /Fabian Dopler
Expertise:
Data Analytics / App Development
Studio:
Data Science / MoKnowLab
Event Matching
Linking culture-related events from website data with learning content
Background
The company Hublz GmbH develops digital scavenger hunts for users interested in culture. These scavenger hunts are to be enriched (automatically) with suitable cultural events, both spatially and temporally, and presented to the user as a recommendation.
Method
In the course of the FFG innovation check for the company Hublz GmbH, a prototype was developed with which it is possible to query culture-related event data from several (a) structured as well as (b) unstructured data sources.
Together with the client, a sample data set was labeled with respect to the target group assignment.
Based on labeled data, different supervised learning approaches were investigated and a corresponding classifier was trained. Using Jupyter Notebook, a minimalistic web prototype was developed for experimentation.
Goal
The goal was to assign events from publicly available sources as precisely as possible to the appropriate target groups.
Twelve target groups were defined by the client. Examples include:
- F1 – Families with children under 6
- F3 – Families with young people between 10 and 14
- J1 – young people between 14 and 20
- TH – adults who like to attend theater performances
- BA – adults who like to go to dance performances
Which Innovations Does the Prototype Offer?
- Data output basis for economic exploitation by the customer
- Higher interest of the end users in target group specific advertised events.
- Target group classification scheme
- Classification based on unstructured text
