Abstract

Abstract
Social networks are becoming a wide repository of information, some of which may be of interest for general audiences. In this study we investigate which features may be extracted from single posts propagated throughout a social network, and that are indicative of its relevance, from a journalistic perspective. We then test these features with a set of supervised learning algorithms in order to evaluate our hypothesis. The main results indicate that if a text fragment is pointed out as being interesting, meaningful for the majority of people, reliable and with a wide scope, then it is more likely to be considered as relevant. This approach also presents promising results when validated with several well-known learning algorithms.

Abstract
According to the World Health Organization, 85% of the world population suffers from back pain, which accounts for over 50% of physical incapacity, permanent or temporary, among individuals in working-age. In most situations, this is caused by an incorrect posture, which causes changes in the spine structure. This paper proposes an instrumented vest for postural reeducation to address this issue. The vest has a set of inertial measurement unit (IMU) sensors strategically placed to provide an accurate characterization of the spine profile. The sensor readings are classified by a central processing unit. In case of an incorrect posture, users are alerted by an audio signal and through vibration. The wearable system works in stand-alone mode, but can also communicate with external systems through an API. Two applications were developed to communicate with the device through this API, one intended to run on a desktop computer and the other one for Android devices. These applications monitor spine profiles in real time and notify the user of incorrect postures, among other functionalities. The device prototype and the applications have been tested by 10 individuals in two different settings, first without any kind of feedback and then with feedback enabled. The tests demonstrate the usability, accuracy and robustness of the system, proving its high level of reliability in classifying postures and effectiveness for postural reeducation. In the future, the system is expected to be used as a platform for a serious game, to promote posture reeducation in a real world scenario.

Abstract
Despite a clear improvement of search and retrieval temporal applications, current search engines are still mostly unaware of the temporal dimension. Indeed, in most cases, systems are limited to offering the user the chance to restrict the search to a particular time period or to simply rely on an explicitly specified time span. If the user is not explicit in his/her search intents (e.g., "philip seymour hoffman'') search engines may likely fail to present an overall historic perspective of the topic. In most such cases, they are limited to retrieving the most recent results. One possible solution to this shortcoming is to understand the different time periods of the query. In this context, most state-of-the-art methodologies consider any occurrence of temporal expressions in web documents and other web data as equally relevant to an implicit time sensitive query. To approach this problem in a more adequate manner, we propose in this paper the detection of relevant temporal expressions to the query. Unlike previous metadata and query log-based approaches, we show how to achieve this goal based on information extracted from document content. However, instead of simply focusing on the detection of the most obvious date we are also interested in retrieving the set of dates that are relevant to the query. Towards this goal, we define a general similarity measure that makes use of co-occurrences of words and years based on corpus statistics and a classification methodology that is able to identify the set of top relevant dates for a given implicit time sensitive query, while filtering out the non-relevant ones. Through extensive experimental evaluation, we mean to demonstrate that our approach offers promising results in the field of temporal information retrieval (T-IR), as demonstrated by the experiments conducted over several baselines on web corpora collections.

Abstract
This paper describes a methodology for Diabetic Retinopathy detection from eye fundus images using a generalization of the Bag-of-Visual-Words (BoVW) method. We formulate the BoVW as two neural networks that can be trained jointly. Unlike the BoVW, our model is able to learn how to perform feature extraction, feature encoding and classification guided by the classification error. The model achieves 0.97 Area Under the Curve (AUC) on the DR2 dataset while the standard BoVW approach achieves 0.94 AUC. Also, it performs at the same level of the state-of-the-art on the Messidor dataset with 0.90 AUC.

Abstract