Large Language Model Augmented Narrative Driven Recommendations

Large Language Model Augmented Narrative Driven Recommendations
复制标题

大语言模型增强叙事驱动的推荐

DOI:
10.1145/3604915
复制
发表时间:
2023
期刊:
ACM
影响因子:
--
通讯作者:
Zamani, Hamed
Zamani, Hamed
中科院分区:
--
文献类型:
--
作者:
Mysore, Sheshera;McCallum, Andrew;Zamani, Hamed

文献摘要

参考文献

被引文献

相似文献

叙事驱动推荐(Narrative-driven recommendations,NDR)提出了一种信息访问问题,即用户通过详细描述其偏好和上下文来请求推荐,例如,旅行者在请求推荐兴趣点的同时描述其喜欢/不喜欢和旅行情况。随着搜索和推荐系统中基于自然语言的会话界面的兴起,这些请求变得越来越重要。然而,NDR缺乏丰富的模型训练数据,目前的平台通常不支持这些请求。幸运的是,经典的用户-项目交互数据集包含丰富的文本数据,例如,评论,通常描述用户偏好和上下文-这可以用于引导NDR模型的训练。在这项工作中,我们探索使用大型语言模型(LLM)进行数据增强来训练NDR模型。我们使用LLM从用户-项目交互中创作合成叙事查询,并在合成查询和用户-项目交互数据上训练NDR检索模型。我们的实验表明,这是一个有效的策略,训练小参数检索模型,优于其他检索和LLM基线叙事驱动的推荐。
Narrative-driven recommendation (NDR) presents an information access problem where users solicit recommendations with verbose descriptions of their preferences and context, for example, travelers soliciting recommendations for points of interest while describing their likes/dislikes and travel circumstances. These requests are increasingly important with the rise of natural language-based conversational interfaces for search and recommendation systems. However, NDR lacks abundant training data for models, and current platforms commonly do not support these requests. Fortunately, classical user-item interaction datasets contain rich textual data, e.g., reviews, which often describe user preferences and context – this may be used to bootstrap training for NDR models. In this work, we explore using large language models (LLMs) for data augmentation to train NDR models. We use LLMs for authoring synthetic narrative queries from user-item interactions with few-shot prompting and train retrieval models for NDR on synthetic queries and user-item interaction data. Our experiments demonstrate that this is an effective strategy for training small-parameter retrieval models that outperform other retrieval and LLM baselines for narrative-driven recommendation.
AR-CF:在协同过滤中增强虚拟用户和项目以解决冷启动问题
DOI: 10.1145/3397271.3401038
发表时间: 2020
期刊: Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval
影响因子: --
作者:
Dong;Jihoo Kim;Duen Horng Chau;Sang
通讯作者: Sang
评估 Reddit 上叙事驱动的电影推荐
DOI: 10.1145/3301275.3302287
发表时间: 2019
期刊: Proceedings of the 24th International Conference on Intelligent User Interfaces
影响因子: --
作者:
Lukas Eberhard;Simon Walk;Lisa Posch;D. Helic
通讯作者: D. Helic
InPars-v2:大型语言模型作为信息检索的高效数据集生成器
DOI: 10.48550/arxiv.2301.01820
发表时间: 2023
期刊: ArXiv
影响因子: --
作者:
Vitor Jeronymo;L. Bonifacio;Hugo Abonizio;Marzieh Fadaee;R. Lotufo;Jakub Zavrel;Rodrigo Nogueira
通讯作者: Rodrigo Nogueira
DOI: 10.1561/1500000081
发表时间: 2023
期刊: Foundations and Trends® in Information Retrieval
影响因子: --
作者:
Zamani, Hamed;Trippas, Johanne R.;Dalton, Jeff;Radlinski, Filip
通讯作者: Radlinski, Filip
开始与搜索引擎对话 - 引发自然语言查询的界面
DOI: 10.1145/3406522.3446035
发表时间: 2021
期刊: Proceedings of the 2021 Conference on Human Information Interaction and Retrieval
影响因子: --
作者:
A. Papenmeier;Dagmar Kern;Daniel Hienert;A. Sliwa;Ahmet Aker;N. Fuhr
通讯作者: N. Fuhr