About this Research Topic
Search and recommendation are the primary methods for users to access information on the Web and are widely utilized in various scenarios and applications. Recently, there has been rapid development in generative language models, such as ChatGPT, which possess powerful semantic understanding and conversational capabilities, as well as a wealth of prior knowledge. Leveraging these capabilities, generative language models can be applied to various stages of the retrieval and recommendation pipeline, strengthening or replacing traditional solutions. This has led to a paradigm shift in the field of information retrieval, giving rise to generative retrieval and recommendation. These new paradigms aim to use powerful generative language models to retrieve documents or recommend items, and have garnered significant attention.
This Research Topic focuses on addressing the problems of integrating generative language models into search and recommendation systems. While driving search and recommendation into a generative paradigm, the following key issues need to be emphasized. Firstly, how to represent a document or item within a generative language model. Past research has shown that language models struggle to directly generate documents or items themselves, requiring the use of special identifiers as substitutes. Therefore, the selection of identifier types plays a crucial role in the entire generative retrieval and recommendation system. Secondly, how to effectively train generative retrieval and recommendation models. Generative training forms the foundation of generative language models, but when applied to information retrieval, which typically involves discriminative training, pure generative training may seem insufficient. Lastly, how to extend the generative paradigm to cover various scenarios in retrieval and recommendation, such as multimodal documents and products, is also a crucial issue that needs to be explored.
This Research Topic encompasses all applications of generative language models in search and recommendation, including but not limited to the following:
• Development of novel identifiers for generative search and recommendation
• Exploration of new training frameworks for generative search and recommendation
• Implementation of multimodal generative search and recommendation systems
• Utilization of generative language models across various stages of information retrieval, such as query understanding, ranking, data augmentation, evaluation, and more.
This Research Topic focuses on addressing the problems of integrating generative language models into search and recommendation systems. While driving search and recommendation into a generative paradigm, the following key issues need to be emphasized. Firstly, how to represent a document or item within a generative language model. Past research has shown that language models struggle to directly generate documents or items themselves, requiring the use of special identifiers as substitutes. Therefore, the selection of identifier types plays a crucial role in the entire generative retrieval and recommendation system. Secondly, how to effectively train generative retrieval and recommendation models. Generative training forms the foundation of generative language models, but when applied to information retrieval, which typically involves discriminative training, pure generative training may seem insufficient. Lastly, how to extend the generative paradigm to cover various scenarios in retrieval and recommendation, such as multimodal documents and products, is also a crucial issue that needs to be explored.
This Research Topic encompasses all applications of generative language models in search and recommendation, including but not limited to the following:
• Development of novel identifiers for generative search and recommendation
• Exploration of new training frameworks for generative search and recommendation
• Implementation of multimodal generative search and recommendation systems
• Utilization of generative language models across various stages of information retrieval, such as query understanding, ranking, data augmentation, evaluation, and more.
Keywords: Generative language models, Generative search, Generative recommendation, Generative information retrieval, Information retrieval
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
