The content delves into the evolving landscape of Software Engineering, focusing on Requirements Engineering in the era of AI integration. It discusses challenges, solutions, and examples related to integrating NLP and generative AI into software systems. The chapter aims to engage students, faculty, and industry researchers in discussions about text data-centric tasks relevant to RE.
The chapter emphasizes the importance of data requirements for AI-centric software solutions. It highlights challenges encountered during data collection, annotation, processing, and validation stages. Various techniques such as transfer learning, prompting, self-training, automated labeling, hybrid techniques, generative agents, domain adaptation, rationale-augmented learning, adversarial prompting are discussed with illustrative examples.
Key points include addressing class imbalance through resampling techniques and artificial data generation. The importance of representativeness in training data is highlighted along with strategies to mitigate societal bias. Subjectivity in annotations is addressed through clear guidelines. Techniques like hybrid models and generative agents are recommended for handling complex tasks. Guidelines for data confidentiality and compliance-related concerns are provided.
The chapter concludes by outlining an end-to-end pipeline for managing text data in an education feedback analysis system before model training and after deployment readiness.
To Another Language
from source content
arxiv.org
Дополнительные вопросы