Vânia Vidal’s research while affiliated with Federal University of Ceará and other places

Large language models (LLMs) currently are the state of the art for pre-trained language models. LLMs have been applied to many tasks, including question and answering over Knowledge Graphs (KGs) and text-to-SPARQL, that is, the translation of Natural Language questions to SPARQL queries. With such motivation, this paper first describes preliminary experiments to evaluate the ability of ChatGPT to answer Natural Language questions over KGs. Based on these experiments, the paper introduces Auto-KGQA, an autonomous domain-independent framework based on LLMs for text-to-SPARQL. The framework selects fragments of the KG, which the LLM uses to translate the user’s Natural Language question to a SPARQL query on the KG. The paper describes preliminary experiments with Auto-KGQA with ChatGPT that indicate that the framework substantially reduced the number of tokens passed to ChatGPT without sacrificing performance. Finally, the paper includes an evaluation of Auto-KGQA in a publicly available benchmark, which showed that the framework is competitive, achieving an improvement of 13.2% in accuracy with respect to the state of the art and a reduction of 51.12% in the number of tokens passed to the LLM.

Um sistema de Grafo de Conhecimento Empresarial (Enterprise Knowledge Graph ou EKG) é um paradigma baseado em tecnologias da Web Semântica e Grafos de Conhecimento para integrar fontes de dados heterogêneas. Um EKG fornece uma visão ontológica e unificada, para que as aplicações tenham acesso integrado aos dados através da visão semântica. A exploração de dados na visão semântica requer ferramentas que apresentem grafos de forma compreensível, facilitando a interpretação e a tomada de decisão. Este artigo apresenta a ContextEKG Explorer1, uma ferramenta gráfica interativa para a exploração de dados na visão semântica de um EKG. A visão semântica é organizada em uma hierarquia de três níveis, simplificando a exploração das entidades em múltiplos contextos. A ferramenta oferece visualizações intuitivas das entidades e navegação fluida entre esses contextos.

An Enterprise Knowledge Graph (EKG) is a robust foundation for knowledge management, data integration, and advanced analytics across organizations. It achieves this by offering a semantic view that semantically integrates various data sources within an organization’s data lake. This paper introduces a novel data design pattern (DDP) aimed at constructing and managing the semantic view of an EKG. The proposed DDP logically organizes data into three hierarchical levels, facilitating the maintenance and the versatile exploration of the semantic view in various contexts. Furthermore, this paper details an interactive graphical interface developed to supports context-sensitive navigation of the semantic view, enhancing user interaction and resource utilization.

Neste artigo apresentamos LIght RDF Browser (LiRB), uma interface leve para a navegação interativa sobre Knowledge Graphs (KGs) RDF. LiRB apresenta uma interface Web baseada em texto, proporcionando uma experiência de navegação semelhante às páginas Web tradicionais, promovendo uma menor curva de aprendizagem aos usuários. Como diferenciais, LiRB utiliza consultas simples e rápidas, sendo ideal para a navegação de KGs massivos e/ou virtuais. Além disso, a ferramenta também permite a abstração de relacionamentos N-ários para representar a semântica de propriedades de relacionamentos, além da apresentação de consultas salvas. Por fim, LiRB também dá suporte ao recurso de exibição de timelines de eventos de recursos do KG.

This article presents the construction of a semantic knowledge graph of the open data of legal entities (SKG:CNPJ). SKG:CNPJ is obtained from the semantic integration of four data sources: RFB, IBGE, Correios, TCU and CEIS. The main objective of SKG:CNPJ is to provide a semantic layer, so that applications can have integrated access to the data source data through the Semantic Layer. The data and metadata of the SKG:CNPJ are made available on a Semantic Portal (Semantic-CNPJ) for query and visualization. The article also describes the use of SKG:CNPJ for construction of semantic queries.

Pervasive computing delivers services based on user needs through smart environments that incorporate and integrate everyday objects discreet and non-intrusive. Personal applications provide the data collected by sensors for Human Activity Recognition. The main limitation is that these activities need to be continuously segmented for HAR. Furthermore, a growing problem is related to the disambiguation of activities since some actions generated by the same sensors belong to different activities. This paper proposes a hybrid method, SeAct, which dynamically adjusts segment size, combining machine learning and semantic inference. Experiments with CAD-120 data sets and a state-of-the-art hybrid method improve recognition accuracy and precision.

Purpose Enterprise knowledge graphs (EKG) in resource description framework (RDF) consolidate and semantically integrate heterogeneous data sources into a comprehensive dataspace. However, to make an external relational data source accessible through an EKG, an RDF view of the underlying relational database, called an RDB2RDF view, must be created. The RDB2RDF view should be materialized in situations where live access to the data source is not possible, or the data source imposes restrictions on the type of query forms and the number of results. In this case, a mechanism for maintaining the materialized view data up-to-date is also required. The purpose of this paper is to address the problem of the efficient maintenance of externally materialized RDB2RDF views. Design/methodology/approach This paper proposes a formal framework for the incremental maintenance of externally materialized RDB2RDF views, in which the server computes and publishes changesets, indicating the difference between the two states of the view. The EKG system can then download the changesets and synchronize the externally materialized view. The changesets are computed based solely on the update and the source database state and require no access to the content of the view. Findings The central result of this paper shows that changesets computed according to the formal framework correctly maintain the externally materialized RDB2RDF view. The experiments indicate that the proposed strategy supports live synchronization of large RDB2RDF views and that the time taken to compute the changesets with the proposed approach was almost three orders of magnitude smaller than partial rematerialization and three orders of magnitude smaller than full rematerialization. Originality/value The main idea that differentiates the proposed approach from previous work on incremental view maintenance is to explore the object-preserving property of typical RDB2RDF views so that the solution can deal with views with duplicates. The algorithms for the incremental maintenance of relational views with duplicates published in the literature require querying the materialized view data to precisely compute the changesets. By contrast, the approach proposed in this paper requires no access to view data. This is important when the view is maintained externally, because accessing a remote data source may be too slow.

In this article, we present the MediBot. MediBot is a chatbot for querying drugs information. The presented system acted as a single access point for natural and simplified information retrieval of drugs, prices, and its risks. The chatbot has two modes of operation: Quick Response and Interactive modes. The first answers questions asked in natural language, while the second has three interactive tasks, namely Browser, Query, and Price Comparison. We present here the system architecture, the Linked Data Mashup’s construction process, and Chatbot MediBot’s activities modes, focusing on the new Price Comparison’s task. This task presents the best prices for medicines and their best potential substitutes extracted in real-time from the Web with the help of the information obtained from a linked data mashup.