Brazil

ICD coding from electronic clinical records is a manual, timeconsuming and expensive process. Code assignment is, however, an important task for billing purposes and database organization. While many works have studied the problem of automated ICD coding from free text using machine learning techniques, most use records in the English language, especially from the MIMIC-III public dataset. This work presents results for a dataset with Brazilian Portuguese clinical notes. We develop and optimize a Logistic Regression model, a Convolutional Neural Network (CNN), a Gated Recurrent Unit Neural Network and a CNN with Attention (CNN-Att) for prediction of diagnosis ICD codes. We also report our results for the MIMIC-III dataset, which outperform previous work among models of the same families, as well as the state of the art. Compared to MIMIC-III, the Brazilian Portuguese dataset contains far fewer words per document, when only discharge summaries are used. We experiment concatenating additional documents available in this dataset, achieving a great boost in performance. The CNN-Att model achieves
the best results on both datasets, with micro-averaged F1 score of 0.537 on MIMIC-III and 0.485 on our dataset with additional documents.


BRACIS 2020

Predicting Multiple ICD-10 Codes from Brazilian Portuguese Clinical Notes

clinical notes

multi-label classification

icd coding

neural networks

natural language processing

The 9th Brazilian Conference on Intelligent Systems (BRACIS), the 16h Encontro Nacional de Inteligência Artificial e Computacional (ENIAC), and the 12th Concurso de Teses e Dissertações em Inteligência Artificial e Computacional (CTDIAC) was held from 20 to 23 October, 2020.

BRACIS originated from the combination of the two most important scientific events in Brazil in Artificial Intelligence (AI) and Computational Intelligence (CI): the Brazilian Symposium on Artificial Intelligence - SBIA (22 editions), and the Brazilian Symposium on Neural Networks - SBRN (13 editions). ENIAC is a national forum for researchers, practitioners, educators and students to present and discuss innovations, trends, experiences, developments and work in progress in the fields of Artificial and Computational Intelligence. CTDIAC is a satellite event whose goals are to disseminate and award the best dissertations and theses (MSc and PhDs concluded in Brazilian universities from June 2018 to May 2020) in Artificial and Computational Intelligence areas.

In this year, the Symposium on Knowledge Discovery, Mining and Learning (KDMiLe) will be held in conjunction with BRACIS 2020. This sysposium aims at integrating researchers, practitioners, developers, students and users to present their research results, to discuss ideas, and to exchange techniques, tools, and practical experiences – related to the Data Mining and Machine Learning areas.

technical paper

In this presentation, we will talk about the use of a semi-supervised technique called Randon Walk in Feature Sample Networks and the application of this technique in sentiment analysis of Portuguese tweets. We also present a change in the classification mechanism and the results found.

Semi-Supervised Sentiment Analysis of Portuguese Tweets with Random Walk in Feature Sample Networks

Automatic License Plate Recognition (ALPR) systems are used in many real-world applications, such as road traffic monitoring and traffic law enforcement, and the use of deep learning can result in efficient methods. In this work, we present an ALPR system efficient for edge computing, using a combination of MobileNet-SSD for vehicle detection, Tiny YOLOv3 for license plate detection and OCR-net for character recognition. This method was evaluated in two datasets on a NVIDIA Jetson TX2 system, obtaining 96.87% of accuracy and 8 FPS of framerate in a public real-world scenario dataset and achieving 90.56% of accuracy and 11 FPS of framerate in a private dataset of traffic monitoring images, considering the recognition of at least six characters. It is faster than related works with similar deep learning approaches, that achieved at most 2 FPS, and slightly inferior in accuracy, with less than 10% of difference in the worst scenario. This shows the proposed method is well balanced between accuracy and speed, thus, suitable for embedded devices.

A Robust Automatic License Plate Recognition System for Embedded Devices

Phonetic analysis of speech, in general, requires the alignment of audio samples to its phonetic transcription. This task could be performed manually for a couple of files, but as the corpus grows large it becomes unfeasibly time-consuming, which emphasizes the need for computational tools that perform such speech-phonemes forced alignment automatically. Therefore, due to the scarce availability of phonetic alignment tools for Brazilian Portuguese (BP), this work describes the evolution process towards creating a free phonetic alignment tool for BP using Kaldi, a toolkit that has been the state of the art for open-source speech recognition. Five acoustic models were trained with Kaldi and tested in phonetic alignment, where the evaluation took place in terms of the phone boundary metric. The results show that its performance is similar to some Kaldi-based aligners for other languages and superior to an outdated phonetic aligner for BP based on HTK toolkit.


Towards a Free, Forced Phonetic Aligner for Brazilian Portuguese Using Kaldi Tools

An efficient way to deal with classification problems is by using Fuzzy Rule-Based Classification Systems. A key point in this kind of classifier is the Fuzzy Reasoning Method (FRM). This mechanism is responsible for performing the classification of examples into predefined classes. There are different generalizations of the Choquet integral in the literature that are applied in the FRM. This paper presents an initial study of a new FRM that is ensemble-based, which combines different generalizations in a more effective final classifier. We have constructed two distinct ensemble decision-making methods considering the majority and weighted voting approaches. The experimental results demonstrate that the performance of the proposed methods is statistically equivalent compared to the state-of-art generalizations.

A Fuzzy Reasoning Method based on ensembles of generalizations of the Choquet integral

In social simulation, agents play different roles with different behaviors. Each has its goals and can cooperate or not with others and consume resources. Sometimes it is necessary to regulate these agents with mechanisms like norms that explain action limitations and how the society works for the agents. To explain how the world works for the agents, in different domains, ontologies have been used to provide human knowledge. In this sense, a more special kind of ontology is built with the legal information of a society: legal ontologies. These legal ontologies operate with concepts and data collected from human society and this can be provided to the agents. With this type of information, the agents can be assisted in monitoring and evaluating actions, comparing their beliefs with existing laws. Platforms into programming multi-agent systems provide an environment to develop the social simulations, but not always provided a connection with ontologies resources. This study proposes the AgentDevLaw middleware that integrates platforms of agents’ simulation with an existent legal ontology. The application of the software component is exemplified and tested with simulation scenarios in two agent’s environments: JaCaMo and JADE.

AgentDevLaw: A Middleware Architecture for Integrating Legal Ontologies and Multi-Agent Systems

Novelty detection is an important topic in data stream classification, as it is responsible for identifying the emergence of new concepts, new patterns, and outliers. It becomes necessary when the true label of an instance is not available right after its classification. The time between the classification of an instance and the arrival of its true label is called latency. This is a common scenario in data streams applications. However, most classification algorithms do not consider such a problem and assume that there will be no latency. On the other hand, a few methods in the literature cope with the existence of infinite latency and novelty detection in data streams. In this work, however, we focus on the scenario where the true labels will be available to the system after a certain time, called intermediate latency. Such a scenario is present in the stock market and weather datasets. Moreover, aiming for more flexible learning to deal with the uncertainties inherent in data streams, we consider the use of fuzzy set theory concepts. Therefore, we propose a method for classification and novelty detection in data streams called Fuzzy Classifier with Novelty Detection for data streams (FuzzCND). Our method uses an ensemble of fuzzy decision trees to perform the classification of new instances and applies the concepts of fuzzy set theory to detect possible novelties. The experiments showed that our approach is promising in dealing with the emergence of new concepts in data streams and inaccuracies in the data.

A Fuzzy Approach For Classification and Novelty Detection in Data Streams Under Intermediate Latency

Presentation will be about the paper entitled FT-BlinGui: a fuzzy-based wearable device system to avoid visually impaired collision in real time. There are several technological applications and computational techniques for detection and prevention of obstacles. To provide a more efficient perception of obstacles in dynamic scenarios, the fuzzy systems embedded in technological applications have been presented as an efficient approach. Was investigated the applicability of the fuzzy time series for the forecasting collision with obstacles. The time series analyzed was obtained through a wearable device that detects, in real time, obstacles for the visually impaired. Simulation results demonstrate that FT-BlinGui can provide a collision-avoidance alerting system for wearable devices as a result of Fuzzy Time Series forecasting.



FT-BlinGui: a fuzzy-based wearable device system to avoid visually impaired collision in real time

Aggregation functions are employed to combine inputs that are typically interpreted as degrees of membership in fuzzy sets, degrees of preference, strength of evidence or support of a hypothesis. The behavior of each aggregation function can be associated to the area of distributive justice, which refers to fairness in the way data are distributed. Two important theories of distributive justice commonly considered are the utilitarianism and egalitarianism, but some alternative theories, as the sufficientarianism, has gained attention recently. This paper presents three different versions of sufficientarianism as aggregation functions: the weak, axiological and strong sufficientarianism. Additionally, we introduce new logical properties related to the sufficientarianism and show which of these new aggregation functions satisfy them. In particular, we prove the axiological aggregation functions introduced here satisfy an important logical property whereas the weak sufficientarian functions do
not. On the other hand, we also show the strong sufficientarian aggregation
functions may lead to the loss of logical properties.

Aggregation with Weak, Axiological and Strong Sufficientarian Functions

An ensemble method should cleverly combine a group of base classifiers to yield an improved classifier. The majority vote is an example of a methodology used to combine classifiers in an ensemble method. In this paper, we propose to combine classifiers using an associative memory model. Precisely, we introduce ensemble methods based on recurrent correlation associative memories (RCAMs) for binary classification problems. We show that an RCAM-based ensemble classifier can be viewed as a majority vote classifier whose weights depend on the similarity between the base classifiers and the resulting ensemble method. More precisely, the RCAM-based ensemble combines the classifiers using a recurrent consult and vote scheme. Furthermore, computational experiments confirm the potential application of the RCAM-based ensemble method for binary classification problems.


Ensemble of Binary Classifiers Combined Using Recurrent Correlation Associative Memories

Link prediction is an online social network (OSN) analysis task whose objective is to identify pairs of non-connected nodes that have a high probability of getting connected in the near future. Recently proposed link prediction methods consider topological data from OSN past states (i.e., snapshots that depict the structure of the network at certain moments in the past). Although those past states-based methods retrieve information that describes how the network's topology was at the events of link emergence (i.e., moments when the existing edges were created), they do not take into account contextual data concerning those events. Hence, they take the chance to disregard information about the circumstances that may have influenced the appearance of old edges, and that could be useful to predict the creation of new ones. To remedy this issue, this work extends a past states-based method so that it retrieves both topological and contextual data from the events of edge emergence and combines them to predict links. The extended method presented promising results on experimental data. Overall, it overcame the original method in five different scenarios from five co-authorship OSN frequently used for link prediction method evaluation.

Predicting Multiple ICD-10 Codes from Brazilian Portuguese Clinical Notes

Downloads

Next from BRACIS 2020

Semi-Supervised Sentiment Analysis of Portuguese Tweets with Random Walk in Feature Sample Networks

Similar lecture

Experimenting Sentence Split-and-Rephrase Using Part-of-Speech Labels

Predicting Multiple ICD-10 Codes from Brazilian Portuguese Clinical Notes

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from BRACIS 2020

Semi-Supervised Sentiment Analysis of Portuguese Tweets with Random Walk in Feature Sample Networks

Similar lecture

Experimenting Sentence Split-and-Rephrase Using Part-of-Speech Labels

Downloads