Brazil

WebMedia 2020

Media Content Analysis I

Promoted by the Brazilian Computer Society (SBC), the Brazilian Symposium on Multimedia and Web (WebMedia) is the main event of the theme in Brazil and an excellent opportunity for scientific and technical exchange among students, researchers and professionals in the areas of Multimedia, Hypermedia and Web. This year we will be celebrating to 26th edition of WebMedia, which will be held online from november 30th to December 4th .

Webmedia 2020’s programming includes:

– Technical sessions for the presentation of scientific papers
– National and international keynotes
– Panel with companies
– Workshops
– Short Courses
– Open community meetings

The topics of interest for submitting papers are:

– Web and Social Networks
– Multimedia
– Digital TV, Ubiquitous and Mobile Computing

Besides the main track, WebMedia 2020 will have in parallel the Tools and Applications Workshop, works Tools and Applications Workshop, the Workshop on Undergraduate Research Work and the Workshop The Future of Interactive.

The WebMedia 2020 organization is under the responsibility of the Programa de Pós-graduação Doutorado em Ciência da Computação – Associação UFMA/UFPI (DCCMAPI), do Programa de Pós-graduação em Ciência da Computação/ CCET (PPGCC), do Laboratory of Advanced Web Systems/UFMA (LAWS) e do Laboratório de Sistemas Multimídia/UFMA (TELEMÍDIA).).

#PartiuWebMedia2020 #WebMedia2020

technical paper

Due to easy access and low cost, social media online news consumption has increased significantly for the last decade. Despite their benefits, some social media allow anyone to post news with intense spreading power, which amplifies an old problem: the dissemination of Fake News. In the face of this scenario, several machine learning-based methods to automatically detect Fake News (MLFN) have been proposed. All of them require Fake News datasets to train and evaluate their detection models. Although recent MLFN were designed to consider data regarding the news propagation on social media, most of the few available Fake News datasets do not contain this kind of data. Hence, comparing the performances amid those recent MLFN and the others is restricted to a very limited number of datasets. Moreover, all existing datasets with propagation data do not contain news in Portuguese, which impairs the evaluation of the MLFN in this language. Thus, this work proposes FakeNewsSetGen, a process that builds Fake News datasets that contain news propagation data and support comparison amid the state-of-the-art MLFN. FakeNewsSetGen’s software engineering process was guided to include all kind of data required by the existing MLFN. In order to illustrate FakeNewsSetGen’s viability and adequacy, a case study was carried out. It encompassed the implementation of a FakeNewsSetGen prototype and the application of this prototype to create a dataset called FakeNewsSet, with news in Portuguese. Five MLFN with different kind of data requirements (two of them demanding news propagation data) were applied to FakeNewsSet and compared, demonstrating the potential use of both the proposed process and the created dataset.

FakeNewsSetGen: a Process to Build Datasets that Support Comparison Among Fake News Detection Methods.

Many recent works have successfully applied some types of Convolutional Neural Networks (CNNs) to reduce the noticeable distortion resulting from the lossy JPEG compression technique. Most of them are built upon the processing made on the spatial domain. In this work, we propose a JPEG image decoder that is purely based on the frequency-to-frequency domain: it reads the quantized DCTcoefficients received from a low-quality image bitstream and, using a deep learning-based model, predicts the missing coefficients in order to recompose the same image with enhanced quality. In experiments with two datasets, our best model was able to improve from images with quantized DCT coefficients corresponding to quality Factor (QF) of 10 to enhanced quality images with QFslightly higher than 20.

Decoder-Side Quality Enhancement for JPEG Images Using a Deep Learning-Based Prediction Model for Quantized DCT Coefficients

Face recognition systems are present in many modern solutions and thousands of applications in our daily lives. However, current solutions are not easily scalable, especially when it comes to the addition of new targeted people. We propose a cluster-matching-based approach for face recognition in video. In our approach, we use unsupervised learning to cluster the faces present in both the dataset and targeted videos selected for face recognition. Moreover, we design a cluster matching heuristic to associate clusters in both sets that is also capable of identifying when a face belongs to a non-registered person. Our method has achieved a recall of 99.435% and a precision of 99.131% in the task of video face recognition. Besides performing face recognition, it can also be used to determine the video segments where each person is present.

A Cluster-Matching-Based Method for Video Face Recognition

A major concern when building Recommender Systems (RSs) for recruitment domains is ensuring actual matching opportunities for all candidates and job vacancies on the system as fast as possible. Indeed, long periods waiting for matchings imply losses of business opportunities, causing financial damages for both candidates and companies. We refer to these scenarios where candidates or job vacancies suffer from the absence of matching opportunities as the Matching Scarcity Problem (MaSP). First, we formalize the MaSP itself. Second, we proposed and evaluated strategies to identify automatically candidates and job vacancies suffering from MaSP. Finally, we proposed five heuristic strategies to mitigate MaSP. Our strategies consist of introducing changes in curricula and job descriptions to approximate candidates suffering from MaSP to jobs semantically related to them, and vice versa. The best strategy was able to reduce up to 50% the number of both CVs and jobs suffering from MaSP before applying it in our sample.

Mitigating Matching Scarcity in Recruitment Recommendation Domains

About 50 million people around the world have Alzheimer's disease. The main characteristic of the disease is the deterioration of cognitive function, and as a result of the fact that the disease has no cure or treatment. This article presents a systematic mapping of the literature to identify articles that use technological devices to assist patients with Alzheimer's disease, in order to show an overview of the literature, identifying gaps and research opportunities. The systematic mapping study was conducted out in six databases, including articles from January 2015 to July 2020. An initial search resulted in 12,177 articles, of which 28 were selected after applying the filter criteria. The main results obtained are: (1) most articles (25.71%, 7/28) reported solutions used in devices with sensor development; (2) most technologies (22.06%, 6/28) were focused on monitoring; (3) most papers (82.14%, 23/28) address solutions for dementia, citing Alzheimer's disease indirectly; (4) more than half of the studies aimed at monitoring patients (54%, 15/28); (5) most studies (82.14%, 23/28) performed an experiment with patients with the disease. The results corroborate the need for technological devices to help caregivers monitor patients, in addition to evidence of research opportunities in palliative care and support for daily activities.

Technologies applied in the care of patients with Alzheimer’s disease: a systematic review

The popularization of Location-based social networks (LBSNs) in last years has provided a lot of improvements in several Recommender Systems to the task of points-of-interest (PoI) recommendation. In this paper, we provide an updated view of the PoI recommendation, identifying relevant efforts, results, contributions, and limitations. Through a systematic mapping, we selected 73 relevant papers published in the last three years (2017, 2018, and 2019) in the main vehicles of the area (e.g., RecSys, VLDB, SIGIR, WWW, TKDE, etc). As major limitations, first, we identified that these works prioritize accuracy over other quality dimensions, despite the consensus in the RS community that accuracy is not enough to assess the practical effectiveness of RSs. Further, we found a low intersection of metrics and datasets used in these works, along with a large number of metrics used in a few distinct studies. These observations show restrictions for reproducibility and straightforward comparison of results in the area. Finally, we highlight as a promising future work the in-depth exploitation of textual data, since just a few of the evaluated papers marginally use this rich data source.

A Survey on Point-of-Interest Recommendation in Location-based Social Networks

The current advances have recognized the online recommendation task as a Multi-Armed Bandit (MAB) problem and proposed several algorithms by combining it with Collaborative Filtering (CF). In general, their effectiveness is related to their practical ability to handle the trade-off between selecting the most relevant item (exploitation) or choose an unexpected item to add more knowledge to the system (exploration). However, we observed that a naive failure has been introduced in the main CF bandit models, which have intrinsically reduced itself to pure exploitation of the information known a prior by the system (e.g., items’ popularity) or a pure exploration of the items in the first user’s interaction. Conversely, we hypothesize this trade-off should be considered since the first user’s interactions to improve the system’s effectiveness. Thus, we propose a new methodology to measure the impact of this problem. This methodology has two main phases. In phase 1, traditional non-personalized methods resemble pure-exploitation and pure-exploration to feed the learning of all MAB models. Then, analyzing the MAB effectiveness in phase 2, we may determine the best approach to be considered in the first user’s interactions. Moreover, we also propose and add a new strategy that combines exploration and exploitation in phase 1. Indeed, our experimental results show that the combination of exp-exp at first user’s interactions improves the system’s effectiveness without adding more effort to users.

The impact of first recommendations based on exploration or exploitation approaches in recommender systems’ learning

Location-Based Social Networks (LBSNs) have become important tools for people interested in exploring new places. And, similar to traditional recommendation domains, handling the trade-off between accuracy and diversity is a major challenge to provide useful recommendations. However, this domain adds an equally relevant dimension to this challenge: the geographical distance between users and each point-of-interest (POI). Besides understanding how the characteristics of services offered by each POI fit the user needs, realizing how far users are willing to move to fulfill these needs is of paramount relevance. Moreover, the users present distinct levels of interest in diversification. In this paper, we propose a strategy to provide POI recommendations linearly combining categorical and geographical diversifications in a personalized way. Indeed, our strategy is able to prioritize quality dimensions that better suit the personalized needs of each user, with gains up to 10% when compared with unpersonalized versions.

Combining complementary diversification models for personalized POI recommendations

More than just a common habit or pastime of people, listening to music is often used as a way to facilitate listeners to achieve a certain mental state or to perform specific activities. Among the most common activities performed when listening to music is relaxation. According to findings from the area of music therapy, a good musical choice for relaxation needs personalization. However, until now, many studies in the area of Music Information Retrieval (MIR) have focused on only identifying general characteristics of music according to certain contexts. Thus, this paper focuses on expanding the findings of music therapy, considering aspects of musicology and a large amount of data. To this end, we: (i) explore more than 60 thousand handcrafted playlists of almost 5 thousand users; (ii) create a user preference model based on high and low-level features of the songs and (iii) analyze the differences between the users' preferences. The results of our analysis suggest that personalization plays a key role in the playlists created by the users. With this work, we want to pave the way and encourage the improvement of context-aware music recommender systems or other MIR concepts like user behavior analysis.

Media Content Analysis I

Next from WebMedia 2020

FakeNewsSetGen: a Process to Build Datasets that Support Comparison Among Fake News Detection Methods.

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES