دانلود کتاب دسترسی به اطلاعات در عصر هوش مصنوعی مولد

Preface In recent years, Generative Artificial Intelligence (GenAI) has emerged as a groundbreaking technology that promises to revolutionize many industries and people’s personal and professional lives. This book discusses GenAI and its role in information access, or more broadly, information interaction, both now and for decades to come. It is well known that information is the lifeblood of decision- making and action, and being able to help people find, understand, and use information is a central tenet of helping them attain their goals. The intersection of GenAI and information access is often referred to as Generative Information Retrieval (GenIR). We use “GenAI” and “GenIR”, and variants thereof, where appropriate throughout the book, depending on the context. The book is targeted to graduate students as well as advanced undergraduates and researchers interested in issues of information retrieval, access, and interactions, as well as applications of GenAI in various informational contexts. While some of the parts assume prior background in IR or AI, most others do not, making this book suitable for adoption in various classes as a primary source or as a supplementary material for a wide variety of curricula and training programs. The role of GenAI in information access is complex and dynamic, with many dimensions. To address this, following our brief introduction to GenAI and GenIR (Chap. 1), we divide the remainder of the book into eight chapters, each targeting a different dimension or sub-topic. The chapters are described in more detail below. Each chapter is written by subject matter experts in the field and dives deep into the dimension at hand, presenting a general overview, a summary of the latest advances in that area, and a path forward. Chapter 2: Foundations of Generative Information Retrieval—This chapter, authored by Qingyao Ai, Jingtao Zhan, and Yiqun Liu, delves into the core principles of GenIR, exploring the synthesis of results, validation and attribution processes, the phenomenon of hallucination in Large Language Models (LLMs), and the expansion beyond text to include multimedia and multimodal interactions with LLMs, Retrieval-Augmented Generation (RAG), and corpus understanding. Chapter 3: Interactions with Generative Information Retrieval Systems— Authored by Mohammad Aliannejadim Jacek Gwizdka and Hamed Zamani, this chapter investigates the dynamics of user engagement with GenIR systems. It covers intent understanding, the art of querying, crafting new experiences, the intricacies of conversational systems, the role of agency and automation, the craft of prompt engineering, and the importance of explanations. Chapter 4: Adapting Generative Information Retrieval Systems to Users, Tasks, and Scenarios—Johanne Trippas, Damiano Spina, and Falk Scholer dis- cuss research on tailoring GenIR systems to fit individual users, specific tasks, and diverse scenarios. The chapter covers personalization, specialization, hybrid inference, and task understanding, reflecting the adaptability of GenIR systems. Chapter 5: Improving Generative Information Retrieval Systems Based on User Feedback—Qingyao Ai, Zhicheng Dou, and Min Zhang take the helm in this chapter, focusing on the iterative improvement of GenIR systems through user feedback. It examines alignment, the integration of humans in the loop, continual learning, Reinforcement Learning from Human Feedback (RLHF), and the nuances of implicit and explicit feedback. Chapter 6: Generative Information Retrieval Evaluation—Marwah Alaofi, Negar Arabzadeh, Charles Clarke, and Mark Sanderson provide a comprehensive overview of the evaluation metrics, methodologies, and the importance of repro- ducibility in GenIR systems. They also explore the connection to foundational IR evaluation frameworks such as Cranfield and TREC. Chapter 7: Sociotechnical Implications of Generative Artificial Intelligence for Information Access—Bhaskar Mitra, Henriette Cramer, and Olya Gurevich discuss the societal implications of GenIR. They present an overview of systemic risks from the usage of GenAI in IR (e.g., negative information ecosystem impact, safety, and power dynamics), impact of evaluation methods, and ecosystem incen- tives around misuse. Chapter 8: Recommendation in the Era of Generative Artificial Intelli- gence—Wenjie Wang, Yongfeng Zhang, and Tat-Seng Chua explore the realm of recommendations within GenIR. They delve into recommender systems, personal- ized recommendations, explainable recommendations, and the integration of LLMs with recommendation systems. Chapter 9: Designing for the Future of Information Access with Generative Information Retrieval—Vanessa Murdock, Chia-Jung Lee, and William Hersh envision the future of information access through the lens of GenIR. They discuss new experiences such as proactive information access, emerging business models, applications beyond traditional information retrieval, the ubiquity of GenIR, and its application in specialized domains such as healthcare. These chapter brief summaries provide a glimpse into the comprehensive cover- age of GenIR topics that the book will offer. Each chapter promises to contribute valuable insights into the rapidly evolving field of GenAI, and GenIR in particular.