Alexandra Moreira

Semi-automatic Follow-Up of Graduates

Many institutions needs to gather and store information about people connected to their activities. An automated system to accomplish this task needs to overcome several challenges: collect information on textual and semi-structured textual databases, execute textual inferences, identify individuals, integrate distributed information, etc. Multiagent systems may help solve these problems gradually and scalably. This article describes the design and implementation of a Multiagent system that collects public information about graduates gathered from a mailing list.

Sentiment Analysis Applied to Hotels Evaluation

Websites evaluating products and services are becoming quite common. The large number of evaluations form a substantial corpus that can be used to train and test sentiment analysis tools. The analyzes produced by these tools allow companies and institutions in general to make important decisions that may be vital to the institution’s future. This paper describes an implementation of the Naive Bayes algorithm for the polarity analysis of the reviews from Rio de Janeiro hotel services, reporting the development and difficulties of the data extraction, processing and analysis methods of a corpus with 69076 comments. The results show that the tool is suitable for detecting feelings of positive and negative polarity, but does not present satisfactory results for neutral polarity.

Development of a Brazilian Portuguese Hotel’s Reviews Corpus

The provision of voluntary textual information mediated by the Internet, and particularly by Web 2.0, provided an opportunity for the creation of large linguistic corpora. These corpora can serve as a fundamental resource for the development of applications focused on natural language, especially those using deep learning techniques that require big datasets. One type of application that benefits from these resources is the ones that perform sentiment analysis. This article describes the creation of corpus aimed to support sentiment analysis applications. It consists of reviews hotels located in the Brazilian capitals and the Federal District, written in Brazilian Portuguese language. The reviews that make up the corpus have been taken from TripAdvisor and have undergone normalization and POS tagging. The primary goal is to make it available to the community to be used in machine learning tasks geared toward natural language.

A Deep Learning Approach for Sentiment Analysis Applied to Hotel’s Reviews

Sentiment Analysis is an active area of research and has presented promising results. There are several approaches for modeling that are capable of performing classifications with good accuracy. However, there is no approach that performs well in all contexts, and the nature of the corpus used can exert a great influence. This paper describes a research that presents a convolutional neural network approach to the Sentiment Analysis Applied to Hotel’s Reviews, and performs a comparison with models previously executed on the same corpus.

An Approach for Semantically Enriching Volunteered Geographic Data with Linked Data

Volunteered geographic information (VGI), which pertains to geographic information voluntarily collected and shared, represents a paradigm shift in the way geographic information is created and shared. However, there are still hurdles in properly using such information. One alternative to improve the use of VGI is to use semantic enrichment. Semantic enrichment is a potential way of mitigating several issues that plague VGI such as low data quality, unreliability, and difficulty in use and recovery, among others. The present study discusses the possibility of semantically enriching volunteered geographic data using Linked Data and presents a simplified algorithm to automate this process.

Automatic Creation of Ontology Using a Lexical Database: An Application for the Energy Sector

The development of ontologies is a task that demands understanding of a domain and use of computational tools to develop them. One of the main difficulties is to establish the appropriate meaning of each term. There are several lexical bases that can help in this task, but the problem is how doing this automatically. This paper presents a technique for the automatic ontology development based on lexical databases available on the Web. The intention is that the whole process is to take place with a minimum of human intervention. The technique was able to generate the correct hierarchy for 64 % of the terms in a case study for the electricity sector. The tool was applied in the electrical power domain, but the goal is that the tool can be used for any domain.

Integrating a Metadata Editor with Hyperbolic Tree to Improve Data Access in Spatial Data Infrastructures

The storage and management of metadata is an essential task in systems that deal with spatial data. However, to accomplish this task it is necessary to count with a set of tools that operate in integrated mode. This paper describes the development of edpMGB, a metadata editor for the MGB profile. The editor is open-source and is being released in the cloud via Web following the Software as a Service (SaaS) model, so it can be accessed from any site on the Internet, requiring only a Web browser. The editor was developed on the scope of the research and development project (R&D) Geoportal Cemig SDI-based corporate GIS (Geoportal Cemig - SIG corporativo baseado em IDE), whose objective is the implementation of a corporate SDI for the Minas Gerais Power Company (Companhia Energetica de Minas Gerais - Cemig). Furthermore, this work proposes the use of a hyperbolic tree to aid indexing metadata, facilitating its recovery. After indexing, the user can navigate through the tree nodes and perform searches over the metadata related to the search words.