Laboratory Integration and Testing of the INGENIOUS Social Media App (LIT#8 & LIT#17)
The first and second rounds of the Laboratory and Integration Tests (LITs) of the Social Media Application, which is being developed by our partners from CERTH, took place on the 30th of October 2020 and the 28th of May 2021. Both of them consisted of three parts. First, the presentation of the Social Media Application framework (Figure 1) and the progress on its development. Secondly, a live demonstration and, finally, an open discussion, where the attendees could ask for clarifications or leave some comments.
In the following paragraph, the progress between each Laboratory Test is presented according to the targeted workflow (Figure 1). Until the first round of LITs and LIT#8, the Twitter Crawler was integrated under the GDPR and Twitter Policy while the end users provided the search criteria for each use case. Additionally, the verification and the localisation (in English) analysis were implemented as well as a supplementary service for detecting nudity in Twitter images. Finally, the Annotation Tool Interface (Figure 2) was implemented and integrated. The above tools were demonstrated live (Figure 3) and the scenario in which they were tested refers to a terror attack in multiple locations.
2nd Round of LITs
The main objective in the second round of the Laboratory and Integration Test and LIT#17, was the Social Media App (SMA) integration with the Fusion Engine and the COP. The scenario in which it was tested refers to a fire in a residential area in which SMA aims to detect and visualize (by the COP) tweets relevant to the above scenario. To tackle that, a machine learning algorithm (Figure 4) has been trained with the annotation data collected by end users, while a rule-based approach has been implemented for the Localisation in the Greek language (Figure 5).
Last but not least, an alternative visualisation integrated to the Annotation Tool User Interface can be seen in Figure 6. The heat map indicates areas of high activity based on the quantity of retrieved tweets.
In the end, an extended discussion took place about clarifications and open issues. In fact, the component was successfully demonstrated in its base functionality and further improvements were discussed. Additionally, the thorough presentation and explanation of technical parts helped end-users to suggest future improvements and efficient solutions to the first responders.