RESUMO
Creativity is an inherently human skill, and thus one of the goals of Artificial Intelligence. Specifically, linguistic computational creativity deals with the autonomous generation of linguistically-creative artefacts. Here, we present four types of text that can be tackled in this scope-poetry, humour, riddles, and headlines-and overview computational systems developed for their generation in Portuguese. Adopted approaches are described and illustrated with generated examples, and the key role of underlying computational linguistic resources is highlighted. The future of such systems is further discussed together with the exploration of neural approaches for text generation. While overviewing such systems, we hope to disseminate the area among the community of the computational processing of the Portuguese language.
RESUMO
Electronic Medical Records (EMRs) are written in an unstructured way, often using natural language. Information Extraction (IE) may be used for acquiring knowledge from such texts, including the automatic recognition of meaningful entities, through models for Named Entity Recognition (NER). However, while most work on the previous was made for English, this experience aimed at testing different methods in Portuguese text, more precisely, on the domain of Neurology, and take some conclusions. This paper comprised the comparison between Conditional Random Fields (CRF), bidirectional Long Short-term Memory - Conditional Random Fields (BiLSTM-CRF) and a BiLSTM-CRF with residual learning connections, using not only Portuguese texts from medical journals but also texts from the Coimbra Hospital and Universitary Centre (CHUC) Neurology Service. Furthermore, the performances of BiLSTM-CRF models using word embeddings (WEs) trained with clinical text and WEs trained with general language texts were compared. Deep learning models achieved F1-Scores of nearly 83% and 75%, respectively for relaxed and strict evaluation, on texts extracted from the medical journal. For texts collected from the Hospital, the same achieved F1-Scores of nearly 71% and 62%. This work concludes that deep learning models outperform the shallow learning models and that in-domain WEs get better results than general language WEs, even when the latter are trained with much more text than the former. Furthermore, the results show that it is possible to extract information from Hospital clinical texts with models trained with clinical cases extracted from medical journals, and thus openly available. Nevertheless, such results still require a healthcare technician to check if the information is well extracted.
Assuntos
Aprendizado Profundo , Processamento de Linguagem Natural , Neurologia , Humanos , Idioma , Publicações Periódicas como Assunto , PortugalRESUMO
The dependability of systems and networks has been the target of research for many years now. In the 1970s, what is now known as the top conference on dependability-The IEEE/IFIP International Conference on Dependable Systems and Networks (DSN)-emerged gathering international researchers and sparking the interest of the scientific community. Although it started in niche systems, nowadays dependability is viewed as highly important in most computer systems. The goal of this work is to analyze the research published in the proceedings of well-established dependability conferences (i.e., DSN, International Symposium on Software Reliability Engineering (ISSRE), International Symposium on Reliable Distributed Systems (SRDS), European Dependable Computing Conference (EDCC), Latin-American Symposium on Dependable Computing (LADC), Pacific Rim International Symposium on Dependable Computing (PRDC)), while using Natural Language Processing (NLP) and namely the Latent Dirichlet Allocation (LDA) algorithm to identify active, collapsing, ephemeral, and new lines of research in the dependability field. Results show a strong emphasis on terms, like 'security', despite the general focus of the conferences in dependability and new trends that are related with 'machine learning' and 'blockchain'. We used the PRDC conference as a use case, which showed similarity with the overall set of conferences, although we also found specific terms, like 'cyber-physical', being popular at PRDC and not in the overall dataset.
RESUMO
Background: To effectively combat the rising incidence of syphilis, the Brazilian Ministry of Health (MoH) created a National Rapid Response to Syphilis with actions aimed at bolstering epidemiological surveillance of acquired, congenital syphilis, and syphilis during pregnancy complemented with communication activities to raise population awareness and to increase uptake of testing that targeted mass media outlets from November 2018 to March 2019 throughout Brazil, and mainly areas with high rates of syphilis. This study analyzes the volume and quality of online news content on syphilis in Brazil between 2015 and 2019 and examines its effect on testing. Methods: The collection and processing of online news were automated by means of a proprietary digital health ecosystem established for the study. We applied text data mining techniques to online news to extract patterns from categories of text. The presence and combination of such categories in collected texts determined the quality of news that were analyzed to classify them as high-, medium-and low-quality news. We examined the correlation between the quality of news and the volume of syphilis testing using Spearman's Rank Correlation Coefficient. Results: 1,049 web pages were collected using a Google Search API, of which 630 were categorized as earned media. We observed a steady increase in the number of news on syphilis in 2015 (n = 18), 2016 (n = 26), and 2017 (n = 42), with a substantial rise in the number of news in 2018 (n = 107) and 2019 (n = 437), although the relative proportion of high-quality news remained consistently high (77.6 and 70.5% respectively) and in line with similar years. We found a correlation between news quality and syphilis testing performed in primary health care with an increase of 82.32, 78.13, and 73.20%, respectively, in the three types of treponemal tests used to confirm an infection. Conclusion: Effective communication strategies that lead to dissemination of high quality of information are important to increase uptake of public health policy actions.
Assuntos
Sífilis Congênita , Sífilis , Feminino , Humanos , Gravidez , Brasil/epidemiologia , Saúde Pública , Sífilis/epidemiologia , Sífilis Congênita/epidemiologiaRESUMO
This study analyzes online news disseminated throughout the pre-, during-, and post-intervention periods of the "Syphilis No!" Project, which was developed in Brazil between November 2018 and March 2019. We investigated the influence of sentiment aspects of news to explore their possible relationships with syphilis testing data in response to the syphilis epidemic in Brazil. A dictionary-based technique (VADER) was chosen to perform sentiment analysis considering the Brazilian Portuguese language. Finally, the data collected were used in statistical tests to obtain other indicators, such as correlation and distribution analysis. Of the 627 news items, 198 (31.58%) were classified as a sentiment of security (TP2; stands for the news type 2), whereas 429 (68.42%) were classified as sentiments that instilled vulnerability (TP3; stands for the news type 3). The correlation between the number of syphilis tests and the number of news types TP2 and TP3 was verified from (i) 2015 to 2017 and (ii) 2018 to 2019. For the TP2 type news, in all periods, the p-values were greater than 0.05, thus generating inconclusive results. From 2015 to 2017, there was an ρ = 0.33 correlation between TP3 news and testing data (p-value = 0.04); the years 2018 and 2019 presented a ρ = 0.67 correlation between TP3 news and the number of syphilis tests performed per month, with p-value = 0.0003. In addition, Granger's test was performed between TP3 news and syphilis testing, which resulted in a p-value = 0.002, thus indicating the existence of Granger causality between these time series. By applying natural language processing to sentiment and informational content analysis of public health campaigns, it was found that the most substantial increase in testing was strongly related to attitude-inducing content (TP3).