Modelo de arquitetura para uso de banco de dados híbridos adaptável a ferramentas de Processamento de Linguagem Natural (PLN)

• Ano de defesa: 2024
• Autor(a) principal: Araújo, Francisca Luzia Nogueira
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Banco de dados; Arquitetura híbrida; Processamento de linguagem natural; CNPQ: CIÊNCIAS EXATAS E DA TERRA: CIÊNCIA DA COMPUTAÇÃO
• Link de acesso: http://repositorio.ufc.br/handle/riufc/77011

Currently, databases have become omnipresent. Almost all IT applications are storing and retrieving information from databases. Storing data of different types is a significant challenge, and it may be necessary to use more than one type of database, making it more complex to obtain information about this data. In addition, the dynamics of modern organizations often deal with the need to reconcile opposing requirements provided by databases of different types, such as relational databases (or SQL) and non-relational databases (or NoSQL). Therefore, non-expert users who need to interact with heterogeneous data lack a means by which they can access databases transparently. On the other hand, Natural Language Processing or NLP enables communication between people and machines through techniques that allow the interpretation of natural language used by humans through a computational device. This paper presents an architecture model of a system adaptable to NLP tools capable of translating queries in natural language to formal database query language and, after translation, allowing the execution of queries on databases stored in hybrid, local, or distributed databases. Aimed at enabling adjustments to the proposed architecture, due to the evolution of the current state of the art, it was designed to enable additions of new databases, new algorithms, and/or new natural language translation tools to formal query language of databases, as well as allowing adaptations to recognize new languages in input queries. The strategy used was the creation of modules with well-defined and separated functionalities from others, where to add a new translation tool to the proposal, only one module needs to be modified, for example. To ensure the adaptability of the proposal, the source code was made available and tests were conducted on a cluster of computers, with the possibility of implementation also in a cloud computing services infrastructure; moreover, users can make adjustments to also support big data