Search

The digital sector is rapidly growing in importance and the services offered therein have become crucial to the ever-increasing number of Internet users. Online services, frequently offered for free to end users, are centered around the collection and the analysis of data. The accumulation of large amounts of information contributes to the centralization of market power in the hands of a few large operators, hindering the contestability of the market. An accurate control by antitrust authorities of data-driven mergers can constitute a suitable tool to, on the one hand, prevent the formation or the strengthening of dominant positions based on the acquisition of large datasets and, on the other hand, protect users' welfare. Data-driven mergers, in fact, can negatively impact the services involved with regard to aspects like quality, by lowering the level of data protection offered, and the possibility of choice for users, by allowing a higher personalization. By examining the European Commission's decisions on data-driven mergers, this article aims at identifying the advances and the criticalities characterizing this field. Although the European institution has consistently recognized the competitive advantage linked to data, it often ignored the effects of data-driven acquisitions on the welfare of end users. Only two decisions recognized data protection and users' choice between different services as important competitive parameters, however they remained isolated cases. As a matter of fact, in its recent decision on the Google-Fitbit merger the institution maintained a rigid separation between data protection and competition law. Moreover, the possible anti-competitive effects of the acquisition on end-users' welfare were not considered. We believe that this approach is problematic, especially in cases that, like Google-Fitbit, involve companies active in the market for smart wearable devices. With the advance of the IoT, in fact, smart objects will be able to collect even more information and the line between offline and online will become thinner. In this context, the protection by antitrust authorities of consumers' welfare, including quality and a wide range of choice, will be crucial to maintain an effective competition.

Il modello prosodico implementato nell'attuale versione di FESTIVAL in italiano è basato interamente su semplici regole fonetico/acustiche. In questo lavoro vengono descritte le metodologie e gli esperimenti realizzati per la creazione del nuovo modello prosodico "data-driven" della nuova versione di FESTIVAL. I parametri di questo nuovo modello sono stati appresi mediante tecniche automatiche di modellizzazione prosodica che si basano totalmente sulle informazioni contenute nei corpora vocali utilizzati in fase di sviluppo. Al fine di valutare le proprietà e l'efficacia di questo modello, è stato realizzato un primo esperimento di stima dello stesso, messo a confronto con il modello originale "rule-based", mediante un test percettivo. I risultati di questa prima valutazione vengono qui presentati brevemente.