| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 6.7 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A Fujifilm atua no setor de saúde, na produção e venda de hardware de imagem médica
(Imagiologia, Radiologia, Endoscopia) e soluções de software integradas. Com a venda
dessas soluções para clínicas e hospitais, a empresa oferece suporte técnico por meio de
uma plataforma de tickets onde permite a troca de mensagens para solucionar questões
como, problemas de hospedagem, bugs, comportamentos suspeitos do programa, pedidos
de funcionalidades e esclarecimentos.
No entanto, a Fujifilm identificou que a resolução de tickets depende excessivamente do
conhecimento técnico individual dos agentes de suporte, o que torna o processo lento e
altamente dependente da sua experiência prévia. Isto afeta a eficiência e a consistência no
atendimento ao cliente. Assim, surgiu a necessidade de uma solução integrada que auxilie a
resolução de novos tickets, aproveitando o vasto histórico de tickets resolvidos, acumulado
ao longo dos anos.
Para atender a essa demanda, foi desenvolvido uma solução automatizado que classifica os
tickets com o uso de técnicas de Processamento de Linguagem Natural e Machine Learning,
para a identificação de tickets similares, em conjunto com um gerador de recomendações
de resolução de ticket por meio de ferramentas de inteligência artificial, como a OpenAI
API. Esta solução é disponibilizada através de uma API para facilitar a integração com
outras plataformas e inclui um mecanismo de reforço automático que aprimora o modelo de
classificação.
A proposta foi em grande parte realizada, tendo resultados promissores, embora algumas
limitações tenham sido encontradas, indicando que é necessário mais trabalho futuro para
aprimorar a solução e atender plenamente aos requisitos de qualidade esperados.
Fujifilm operates in the healthcare sector, producing and selling medical imaging hardware (Imaging, Radiology, Endoscopy) and integrated software solutions. With the sale of these solutions to clinics and hospitals, the company provides technical support through a ticketing platform that allows a message exchange to resolve issues such as hosting problems, bugs, suspicious program behaviors, feature requests, and clarifications. However, Fujifilm identified that ticket resolution relies excessively on the individual technical knowledge of support agents, making the process slow and highly dependent on their prior experience. This affects efficiency and consistency in customer service. Thus, the need arose for an integrated solution to assist in the resolution of new tickets by leveraging the vast historical data of resolved tickets accumulated over the years. To meet this demand, an automated solution was developed that classifies tickets using Natural Language Processing and Machine Learning techniques to identify similar tickets, along with a ticket resolution recommendation generator utilizing artificial intelligence tools such as the OpenAI API. This solution is made available through an API to facilitate integration with other platforms and includes an automatic reinforcement mechanism that enhances the classification model. The proposal has been largely implemented, yielding promising results, although some limitations have been encountered, indicating that further work is needed to enhance the solution and fully meet the expected quality requirements.
Fujifilm operates in the healthcare sector, producing and selling medical imaging hardware (Imaging, Radiology, Endoscopy) and integrated software solutions. With the sale of these solutions to clinics and hospitals, the company provides technical support through a ticketing platform that allows a message exchange to resolve issues such as hosting problems, bugs, suspicious program behaviors, feature requests, and clarifications. However, Fujifilm identified that ticket resolution relies excessively on the individual technical knowledge of support agents, making the process slow and highly dependent on their prior experience. This affects efficiency and consistency in customer service. Thus, the need arose for an integrated solution to assist in the resolution of new tickets by leveraging the vast historical data of resolved tickets accumulated over the years. To meet this demand, an automated solution was developed that classifies tickets using Natural Language Processing and Machine Learning techniques to identify similar tickets, along with a ticket resolution recommendation generator utilizing artificial intelligence tools such as the OpenAI API. This solution is made available through an API to facilitate integration with other platforms and includes an automatic reinforcement mechanism that enhances the classification model. The proposal has been largely implemented, yielding promising results, although some limitations have been encountered, indicating that further work is needed to enhance the solution and fully meet the expected quality requirements.
Description
Keywords
Classificação Recomendação PLN IA Texto Ticket