Scenario 2

Scenario 2: Multimodal Brand Monitoring

This use-case scenario focuses on monitoring the reputation of particular brands based on multimodal analysis of brand-related data from social networks and websites. The steps of this scenario are the following:

1. A crawling – parsing procedure gets data from predefined sources (e.g. twitter, facebook, etc) and a detection procedure decides which data refer to particular brands of interest
2. All data are stored in the database and a subset of these is periodically being used to train supervised sentiment recognition models. Towards this end, an annotation procedure is performed, as described in the respective deliverable.
3. The vast majority of data are analyzed by the trained models to extract sentiment estimates
4. A set of aggregated statistics is presented to the end user. Temporal dependencies are also provided – when available