Loading...
Publication
TransformaĆ§Ć£o de Comandos de Voz em Pesquisas Estruturadas
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 5.28 MB | Adobe PDF |
Advisor(s)
Abstract(s)
Com o aumento da interaĆ§Ć£o entre seres humanos e computadores e o progressivo avanƧo das
tecnologias de inteligĆŖncia artificial (IA), surge a necessidade de sistemas mais eficientes que
proporcionem uma experiĆŖncia natural e adaptĆ”vel aos utilizadores.
Esta dissertaĆ§Ć£o explora o desenvolvimento de uma aplicaĆ§Ć£o que converte texto livre
(potencialmente proveniente de comandos de voz) em consultas estruturadas no formato
GraphQL para o contexto de pontos de interesse (POI).
A soluĆ§Ć£o proposta engloba uma aplicaĆ§Ć£o de modelos linguĆsticos avanƧados (LLMs) e
integraĆ§Ć£o com serviƧos em nuvem (AWS), visando garantir escalabilidade e precisĆ£o na
execuĆ§Ć£o do sistema.
A tarefa envolve examinar e comparar diferentes modelos de IA disponĆveis para ajustar o
modelo escolhido de acordo com as necessidades do projeto por meio de tƩcnicas de ajustes
finos personalizados (fine-tuning).
A soluĆ§Ć£o criada Ć© capaz de se integrar flexivelmente com Interface de ProgramaĆ§Ć£o de
AplicaƧƵes (APIs) em GraphQL para permitir consultas avanƧadas com base em texto livre. AlƩm
disso o sistema foi validado por meio de avaliaƧƵes prƔticas e de usabilidade, para mostrar sua
viabilidade tĆ©cnica como a sua eficiĆŖncia no processamento das solicitaƧƵes.
Os resultados obtidos evidenciam o potencial desta abordagem para aprimorar a interaĆ§Ć£o
entre humanos e computadores em diferentes cenƔrios, indicando caminhos para pesquisas
futuras que ampliem o modelo para outras Ć”reas e contextos de aplicaĆ§Ć£o.
With the increase in interaction between humans and computers and the progressive advancement of artificial intelligence (AI) technologies, there is a need for more efficient systems that provide a natural and adaptable experience for users. This dissertation explores the development of an application that converts free text (potentially from voice commands) into structured queries in GraphQL format for the context of points of interest (POI). The proposed solution encompasses the application of advanced linguistic models (LLMs) and integration with cloud services (AWS), aiming to ensure scalability and accuracy in the system's execution. The task involves examining and comparing different AI models available to adjust the chosen model according to the project's needs through customized fine-tuning techniques. The solution created is capable of flexibly integrating with Application Programming Interfaces (APIs) in GraphQL to enable advanced queries based on free text. In addition, the system was validated through practical and usability evaluations to demonstrate its technical feasibility and efficiency in processing requests. The results obtained highlight the potential of this approach to improve human-computer interaction in different scenarios, indicating avenues for future research that extend the model to other areas and application contexts.
With the increase in interaction between humans and computers and the progressive advancement of artificial intelligence (AI) technologies, there is a need for more efficient systems that provide a natural and adaptable experience for users. This dissertation explores the development of an application that converts free text (potentially from voice commands) into structured queries in GraphQL format for the context of points of interest (POI). The proposed solution encompasses the application of advanced linguistic models (LLMs) and integration with cloud services (AWS), aiming to ensure scalability and accuracy in the system's execution. The task involves examining and comparing different AI models available to adjust the chosen model according to the project's needs through customized fine-tuning techniques. The solution created is capable of flexibly integrating with Application Programming Interfaces (APIs) in GraphQL to enable advanced queries based on free text. In addition, the system was validated through practical and usability evaluations to demonstrate its technical feasibility and efficiency in processing requests. The results obtained highlight the potential of this approach to improve human-computer interaction in different scenarios, indicating avenues for future research that extend the model to other areas and application contexts.
Description
Keywords
Artificial Intelligence Natural Language Processing GraphQL Cloud Computing Points of Interest Large-Scale Language Models InteligĆŖncia Artificial Processamento de Linguagem Natural ComputaĆ§Ć£o em Nuvem Pontos de Interesse Modelos de Linguagem de Grande Escala