INESC-ID

About: INESC-ID is a nonprofit organization based out in Lisbon, Portugal. It is known for research contribution in the topics: Computer science & Context (language use). The organization has 932 authors who have published 2618 publications receiving 37658 citations.

Detecting Audio Events for Semantic Video Search

Abstract: This paper describes our work on audio event detection, one of our tasks in the European project VIDIVIDEO. Preliminary experiments with a small corpus of sound effects have shown the potential of this type of corpus for training purposes. This paper describes our experiments with SVM classifiers, and different features, using a 290-hour corpus of sound effects, which allowed us to build detectors for almost 50 semantic concepts. Although the performance of these detectors on the development set is quite good (achieving an average F-measure of 0.87), preliminary experiments on documentaries and films showed that the task is much harder in real-life videos, which so often include overlapping audio events. Index Terms: event detection, audio segmentation

NavTap: a long term study with excluded blind users

Abstract: NavTap is a navigational method that enables blind users to input text in a mobile device by reducing the associated cognitive load.In this paper, we present studies that go beyond a laboratorial setting, exploring the methods' effectiveness and learnability as well as its influence on the users' daily lives. Eight blind users participated in designing the prototype (3 weeks) while five took part in the studies along 16 more weeks. Results gathered in controlled weekly sessions and real life usage logs enabled us to better understand NavTap's advantages and limitations. The method revealed itself both as easy to learn and improve. Indeed, users were able to better control their mobile devices to send SMS and use other tasks that require text input such as managing a phonebook, from day one, in real-life settings.While individual user profiles play an important role in determining their evolution, even less capable users (with age-induced impairments or cognitive difficulties), were able to perform the assigned tasks (sms, directory) both in the laboratory and in everyday use, showing continuous improvement to their skills. According to interviews, none were able to input text before. Nav-Tap dramatically changed their relation with mobile devices and noticeably improved their social interaction capabilities.

OSARE: Opportunistic Speculation in Actively REplicated Transactional Systems

Abstract: In this work we present OSARE, an active replication protocol for transactional systems that combines the usage of Optimistic Atomic Broadcast with a speculative concurrency control mechanism in order to overlap transaction processing and replica synchronization. OSARE biases the speculative serialization of transactions towards an order aligned with the optimistic message delivery order. However, due to the lock-free nature of its concurrency control algorithm, at high concurrency levels, namely when the probability of mismatches between optimistic and final deliveries is higher, OSARE explores additional alternative transaction serialization orders in a lightweight and opportunistic fashion. A simulation study we carried out in the context of Software Transactional Memory systems shows that OSARE achieves robust performance also in scenarios characterized by non-minimal likelihood of reorder between optimistic and final deliveries, providing remarkable speed-up with respect to state of the art speculative replication protocols.

“I can feel it too!”: Emergent empathic reactions between synthetic characters

Abstract: Empathy is often seen as the capacity to perceive, understand and experience others' emotions. This concept has been incorporated in virtual agents to achieve better believability, social interaction and user engagement. However, this has been mostly done to achieve empathic relations with the users. Instead, in this article we focus on empathy between synthetic characters and propose an analytical approach that consists in a generic computational model of empathy, supported by recent neuropsychological studies. The proposed model of empathy was implemented into an affective agent architecture. To evaluate the implementation a small scenario was defined and we asked a group of users to visualize it with the empathy model and another group to visualize it without the model. The results obtained confirmed that our model was capable of producing significant effects in the perception of the emergent empathic responses.