Brazil

Popular mobile applications receive millions of user reviews.
These reviews contain relevant information, such as problem reports and improvement suggestions. The reviews information is a valuable knowledge source for software requirements engineering since the analysis of the reviews feedback helps to make strategic decisions in order to improve the app quality. However, due to the large volume of texts, the manual extraction of the relevant information is an impracticable task. In this paper, we investigate and compare textual representation models for app reviews classification. We discuss different aspects and approaches for the reviews representation, analyzing from the classic Bag-of-Words models to the most recent state-of-the-art Pre-trained Neural Language models. Our findings show that the classic Bag-of-Words model, combined with a careful analysis of text pre-processing techniques, is still a competitive model. However, pre-trained neural language models showed to be more advantageous since it obtains good classification performance, provides significant dimensionality reduction, and deals more adequately with semantic proximity between the reviews&#39; texts, especially the multilingual neural language models.

BRACIS 2020

From Bag-of-Words to Pre-trained Neural Language Models: Improving Automatic Classification of App Reviews for Requirements Engineering

automatic classification of app reviews

neural language models for app reviews

machine learning for app reviews

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.

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible. The experience for on-site participants would closely approximate a normal pre-COVID *ACL conference, with 5-6 thematically organized parallel sessions and live Q/A and interactive discussion immediately after the talks. Presentations by virtual participants will be equitably interleaved with those of on-site participants, projected on the auditorium screens as if on-site, and also followed immediately by live Q/A and interactive discussion at a time during reasonable waking hours for the virtual presenter. For all participants, on-site and virtual, who are unable to attend a session due to either time-zone issues or because they are participating in another session live, talk recordings and slides will be available online at a minimum after the live presentation (and in many cases before as well), and questions may be submitted in advance on session-specific discussion boards and answered live in session with the usual visual aids if desired.

<iframe style="width:700px;height:400px" src="https://online.fliphtml5.com/ebtyf/ceby/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe>

Please Note: The EMNLP registration system is not currently connected to the underline site as we are still in the process of building out EMNLP 2021. You will receive access instructions from underline the week of November 1st. 

Access is given only to EMNLP upon registration, if you have not registered please do so [here](https://2021.emnlp.org/registration).

Registered attendees will receive access the week of November 1st.

EMNLP 2021

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible.

Popular mobile applications receive millions of user reviews.
These reviews contain relevant information, such as problem reports and improvement suggestions. The reviews information is a valuable knowledge source for software requirements engineering since the analysis of the reviews feedback helps to make strategic decisions in order to improve the app quality. However, due to the large volume of texts, the manual extraction of the relevant information is an impracticable task. In this paper, we investigate and compare textual representation models for app reviews classification. We discuss different aspects and approaches for the reviews representation, analyzing from the classic Bag-of-Words models to the most recent state-of-the-art Pre-trained Neural Language models. Our findings show that the classic Bag-of-Words model, combined with a careful analysis of text pre-processing techniques, is still a competitive model. However, pre-trained neural language models showed to be more advantageous since it obtains good classification performance, provides significant dimensionality reduction, and deals more adequately with semantic proximity between the reviews' texts, especially the multilingual neural language models.

technical paper

Stock market forecasting has been a quite popular challenge in machine learning research. Recently, studies have been using deep learning techniques, such as Convolutional Neural Networks (CNN), to perform regressions in prices or classification in the trading signal as an operation indication. 
However, these works have not yet achieved a satisfactory financial result. In this paper, we propose a financially profitable stock market method that uses a new approach called Multichannel CNN Trading Classifier (MCNN-TC).
The model was evaluated using data from the Brazilian stock market. The results indicate a promising financial trading performance compared to the Buy and Hold strategy and good classification metrics.

Stock Trading Classifier with Multichannel Convolutional Neural Network

Trading in the stock market always comes with the challenge of deciding
the best action to take on each time step. The problem is intensified by the
theory that it is not possible to predict stock market time series as all information
related to the stock price is already contained in it. In this work we propose a
novel model called Discrete Wavelet Transform Gated Recurrent Unit Network
(DWT-GRU). The model learns from the data to choose between buying, holding
and selling, and when to execute them. The proposed model was compared to
other recurrent neural networks, with and without wavelets preprocessing, and
the buy and hold strategy. The results shown that the DWT-GRU outperformed
all the set baselines in the analysed stocks of the Brazilian stock market.

Gated Recurrent Unit Networks and Discrete Wavelet Transforms Applied to Forecasting and Trading in the Stock Market

Heuristic search is the most common approach to solve classical planning tasks optimally. OpSeq is an innovative approach to solve planning tasks that decomposes the planning task into a master problem and a subproblem. The master problem generates an assignment of integer counts for each task operator, and the subproblem verifies if a plan satisfying these counts exists. If a plan does not exist, OpSeq learns a new constraint to inform the master problem. OpSeq solves the subproblem using a SAT solver. In this dissertation, we propose a new solver OpSearch: an A*-based approach to sequence operator counts that uses the frontier of the search to learn a constraint on failure. 
We show that OpSearch solves the subproblem better than OpSeq. It solves more planning tasks, scales better, and learns constraints more informative than OpSeq. We prove that OpSearch only learns constraints that maintain all optimal solutions. OpSearch extends the research on an entirely new method to solve planning tasks. Also, our solver opens many new lines of research based on decomposition, such as fast and stronger anytime lower bounds, new methods for agile planning, and new approaches to solve diverse planning.

Sequencing Operator Counts with State-Space Search

In this work, a hybrid partitional clustering algorithm between K-Means and Group Search Optimization is employed to deal with automatic disease segmentation in leaf images, as part of an automatic plant disease classification system: the GSO-KM. An unsupervised image clustering approach is adopted to deal with the problem. GSO-KM is compared to five algorithms from data clustering literature through twelve images, which present variant degrees of disease, using four evaluation metrics. Experimental analysis showed that GSO-KM is able to solve the image segmentation problem properly for the evaluated scenario.

Image Clustering Based on a Hybrid Group Search Optimization and K-Means Approach for Automatic Plant Disease Segmentation

In this paper, a hybrid Otsu and improved Particle Swarm Optimization (PSO) algorithm is presented to deal with multilevel color image thresholding problem, named APSOW. In APSOW, the historical information represented by the local best solutions found so far by PSO population are permuted among the current population, using a randomized greedy process. APSOW also implements a weedout operator to prune the worst individuals from the population. The proposed APSOW is compared to other hybrid EAs and Otsu approaches from literature (include standard PSO model) through twelve benchmark color image problems, showing its potential and robustness.

A Multilevel Thresholding Approach Based on Improved Particle Swarm Optimization for Color Image Segmentation

Recipe sharing websites have become very popular in the past years, allowing individuals to use such systems in an attempt to find a desired recipe. But sometimes finding recipes which best fit the user’s wishes, while still satisfying his food restrictions, may become a very time consuming and difficult task. In this work, we propose a recipe multi-label classification approach as part of a recipe recommendation system for people with food restrictions, in an attempt to automatically identify whether an input recipe or list of ingredients fits into one or more food restrictions, satisfying both user’s expectations and needs. The experimental evaluation includes two approaches for feature selection, as a manner to reduce the computational costs for the proposed system.

A Cooking Recipe Multi-Label Classification Approach for Food Restriction Identification

Electricity consumption is growing rapidly worldwide. Renewable
energy resources, such as solar energy, play a crucial role in this scenario,
contributing to satisfy demand sustainability. Although the share of
Photovoltaic (PV) power generation has increased in the past years, PV
systems are quite sensitive to climatic and meteorological conditions, leading
to undesirable power production variability. In order to improve energy grid
stability, reliability, and management, accurate forecasting models that relate
operational conditions to power output are needed. In this work we evaluate
the performance of regression methods applied to forecast short term (next
day) energy production of a PV Plant. Specifically, we consider five regression
methods and different configurations of feature sets. Our results suggest that
MLP and SVR provide the best forecasting results, in general. Also, although
features based on different solar irradiance levels play a key role in predicting
power generation, the use of additional features can improve prediction results.

Comparative Study of Photovoltaic Power Forecasting Methods

A utilização de técnicas de aprendizado de máquina popularizou-se nos últimos anos nas mais diversas áreas e aplicações. No entanto, a aplicação de tais técnicas ainda depende de um profissional especializado, que execute a sequência de tarefas necessárias para sua execução. Na área de banco de dados, o SQL democratizou a utilização de sistemas gerenciadores de bancos de dados (SGBD) para o usuário final através de uma linguagem declarativa simples, que se assemelha à linguagem natural. Neste artigo, é proposta a linguagem Clustering Visualization Query Language (CVQL), que estende o SQL para a realização de agrupamentos e visualização destes resultados. Agrupamento é uma técnica utilizada para dividir os dados em grupos que compartilham características similares. A visualização destes agrupamentos em diferentes formas é uma funcionalidade essencial para o usuário final analisar os grupos gerados. O CVQL foi implementado utilizando o SGBD MySQL e a biblioteca scikit-learn. Para apresentar o sistema, é adotado um exemplo de consulta com apenas 8 linhas sobre uma base de dados real de casos de Covid-19 nos EUA. Para o processamento da sequência de tarefas executadas pela consulta seriam necessárias 140 linhas de código na linguagem Python, que demonstra a utilidade do sistema.

A Clustering Visualization Query Language

The DBSCAN algorithm is a traditional density-based clustering
method. This algorithm allows to identify groups of different topological for-
mats and also to discard isolated noises. The DBSCAN usually presents good
results, however, it performs several calculations of distances in the clustering
process, causing low efficiency, and its application in large datasets is not re-
commended. This work presents a new sampling method that makes it possible
to apply DBSCAN to a reduced set of examples, in order to approximate the ori-
ginal result of the algorithm applied to the entire dataset. Therefore, the method
makes it possible to execute large datasets with results similar to DBSCAN on
entire dataset, however, with greater computational efficiency.

A biased sampling method for applying DBSCAN

We describe an intelligent triage system for emergency rooms; the system interacts with patients and classifies them by priority level and with respect to medical specialty. The system consists of a conversational interface, coupled with sensors and a physical robot-like platform, and classifiers that operate on symptoms and measurements so as to select a medical specialty and to output a priority level. Tests with human subjects demonstrated that our Healthbot system was well received and in fact preferred to alternatives. Tests have also shown that the classifiers reached accuracy consistent with a doctor’s output.

From Bag-of-Words to Pre-trained Neural Language Models: Improving Automatic Classification of App Reviews for Requirements Engineering

Downloads

Next from BRACIS 2020

Stock Trading Classifier with Multichannel Convolutional Neural Network

Similar lecture

End-to-end Neural Information Status Classification

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

From Bag-of-Words to Pre-trained Neural Language Models: Improving Automatic Classification of App Reviews for Requirements Engineering

.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

Stock Trading Classifier with Multichannel Convolutional Neural Network

Similar lecture

End-to-end Neural Information Status Classification

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads