Desarrollo y evaluación de herramientas basadas en Procesamiento de Lenguaje Natural e Inteligencia Artificial para el tratamiento de especificaciones técnicas y no técnicas con el fin de mejorar la calidad, extraer información y consolidar el conocimiento

Permanent URI for this collection

https://repositorio.uai.edu.ar/handle/123456789/3412

Browse

Browsing Desarrollo y evaluación de herramientas basadas en Procesamiento de Lenguaje Natural e Inteligencia Artificial para el tratamiento de especificaciones técnicas y no técnicas con el fin de mejorar la calidad, extraer información y consolidar el conocimiento by Author "Delle Ville, Juliana"
  • Item
    Implementing Accuracy for Responsible AI in Newsrooms
    (Even3, Brazil, 2024-8-9) Quintanilla Portugal, Roxana ; Delle Ville, Juliana ; Antonelli, Ruben Leandro
    This paper explores the intersection between software development and journalism, highlighting the fundamental importance of implementing nonfunctional requirements to achieve a balance between immediacy and accuracy. In software development, requirements encompass user needs and demand precise and continuous updates to meet time-to-market demands. In journalism, audience engagement drives the need for precise news coverage, especially with the growth of artificial intelligence (AI). Non-functional requirements (NFRs) have gained relevance, emphasizing the need for effective balance. This work investigates the integration of technologies such as Named Entity Recognition (NER) and topic modeling into news updating processes as means to enhance both efficiency and precision. Additionally, this strategy, beneficial in requirements management and applicable across domains, is explored. The article is structured to delve into the imperatives of news writing, the proposed strategy, potential applications, and directions for future research.
  • Item
    Multi-word Entity Extraction and Rich Relationship Identification to Derive Conceptual Models from Natural Language Specifications
    (Even3, Brazil., 2024-8-9) Maltempo, Giuliana ; Delle Ville, Juliana ; Cecconato, Santiago ; Pellegrino, Federico ; Distante, Damiano ; Antonelli, Ruben Leandro
    Requirements engineering is a critical phase in software development. Errors in requirements specifications may become costly problems later on; therefore, such errors should be found and corrected early in the engineering process. Describing requirements in natural language is propitious for both the domain experts and the software development team. However, natural language may give rise to diverse interpretations as a consequence of the different backgrounds of the two participants involved. It is therefore necessary to provide guidance on the specification of unambiguous requirements. In previous work, we have advanced the notion of kernel sentences as an appropriate structure for the specification of knowledge. We have also discussed conceptual models as a useful technique to summarize specifications so that all participants have a concise overview of the domain. To achieve consistent and coherent specifications, we presented a two-step method: first compliance with kernel format is checked, and then a conceptual model is derived to summarize the knowledge gathered. This paper extends the conceptual model previously derived from kernel sentences by identifying multi-word entities and establishing various new relationships among entities. This is intended to help achieve better quality specifications. We also describe a prototype that uses natural language processing and artificial intelligence tools to support the method. Finally, we present the results of a preliminary evaluation of our method, which show a promising applicability.