Advances and challenges in conversational recommender systems: A survey

Huang

Proceedings of the Web Conference 2021

et al. 2021

Self Cite

332

119

Knowledge graph (KG) plays an increasingly important role in recommender systems. A recent technical trend is to develop endto-end models founded on graph neural networks (GNNs). However, existing GNN-based models are coarse-grained in relational modeling, failing to (1) identify user-item relation at a fine-grained level of intents, and (2) exploit relation dependencies to preserve the semantics of long-range connectivity.In this study, we explore intents behind a user-item interaction by using auxiliary item knowledge, and propose a new model, Knowledge Graph-based Intent Network (KGIN). Technically, we model each intent as an attentive combination of KG relations, encouraging the independence of different intents for better model capability and interpretability. Furthermore, we devise a new information aggregation scheme for GNN, which recursively integrates the relation sequences of long-range connectivity (i.e., relational paths). This scheme allows us to distill useful information about user intents and encode them into the representations of users and items. Experimental results on three benchmark datasets show that, KGIN achieves significant improvements over the state-ofthe-art methods like KGAT [41], and CKAN [47]. Further analyses show that KGIN offers interpretable explanations for predictions by identifying influential intents and relational paths. The implementations are available at https://github.com/ huangtinglin/Knowledge_Graph_based_Intent_Network.

Section: Related Workmentioning

confidence: 99%

Learning Intents behind Interactions with Knowledge Graph for Recommendation

Huang

Proceedings of the Web Conference 2021

et al. 2021

Self Cite

332

119

“…• Interactive recommendation, which focuses on improving the recommendation based on the interaction history, i.e., previous recommendations and corresponding user feedback [52,61,73]. • Conversational recommender system (CRS), which is a kind of interactive recommender system that further utilizes the abundant features of items to efficiently and flexibly identify user preferences [12,21,23,27].…”

Section: Social Networkmentioning

confidence: 99%

“…To demonstrate the efficacy and advantage of KuaiRec, we leverage it to conduct the evaluation of the CRS, which has caught the eyes of research community recently due to its great potential [12]. Existing solutions for evaluation of CRSs usually use the user simulation techniques based on the sparsely-observed data.…”

Section: Social Networkmentioning

confidence: 99%

KuaiRec: A Fully-observed Dataset and Insights for Evaluating Recommender Systems

Gao,

Li,

Lei

et al. 2022

Preprint

Self Cite

Recommender systems are usually developed and evaluated on the historical user-item logs. However, most offline recommendation datasets are highly sparse and contain various biases, which hampers the evaluation of recommendation policies. Existing efforts aim to improve the data quality by collecting users' preferences on randomly selected items (e.g., Yahoo! [36] and Coat [47]). However, they still suffer from the high variance issue caused by the sparsely observed data. To fundamentally solve the problem, we present KuaiRec, a fully-observed dataset collected from the social video-sharing mobile App, Kuaishou. The feedback of 1, 411 users on almost all of the 3, 327 videos is explicitly observed. To the best of our knowledge, this is the first real-world fully-observed dataset with millions of user-item interactions in recommendation.To demonstrate the advantage of KuaiRec, we leverage it to explore the key questions in evaluating conversational recommender systems. The experimental results show that two factors in traditional partially-observed data -the data density and the exposure bias -greatly affect the evaluation results. This entails the significance of our fully-observed data in researching many directions in recommender systems, e.g., the unbiased recommendation, interactive/conversational recommendation, and evaluation. We release the dataset and the pipeline implementation for evaluation at https://chongminggao.github.io/KuaiRec/.

“…Traditional static recommender systems that primarily predict a user's preference based on historical data (e.g., click history, ratings) have inherent disadvantages in handling some practical scenarios, such as when a user's preference drifts over time or when the recommendation is highly context-dependent [21]. With the emergence of intelligent conversational assistants such as Amazon Alexa and Google Assistant, conversational recommender systems (CRSs) that can elicit the dynamic preferences of users and take actions based on their current needs through multi-turn interactions have a strong potential to improve different aspects of recommender systems [17] and therefore CRSs have recently seen a growing research interest.…”

Section: Crss and Lmsmentioning

confidence: 99%

“…With the prevalence of language-based intelligent assistants such as Amazon Alexa and Google Assistant, conversational recommender systems (CRSs) have attracted growing attention as they can dynamically elicit users' preferences and incrementally adapt recommendations based on user feedback [17,21]. As one of the most crucial foundations of CRSs, Natural Language Processing (NLP) has witnessed several breakthroughs in the past few years, and the use of pretrained transformer-based language models (LMs) for downstream tasks is one of them [36].…”

Section: Introductionmentioning

confidence: 99%

Unintended Bias in Language Model-driven Conversational Recommendation

Shen¹,

Li²,

Bouadjenek³

et al. 2022

Preprint

Conversational Recommendation Systems (CRSs) have recently started to leverage pretrained language models (LM) such as BERT for their ability to semantically interpret a wide range of preference statement variations. However, pretrained LMs are well-known to be prone to intrinsic biases in their training data, which may be exacerbated by biases embedded in domain-specific language data (e.g., user reviews) used to fine-tune LMs for CRSs. We study a recently introduced LM-driven recommendation backbone (termed LMRec) of a CRS to investigate how unintended bias -i.e., language variations such as name references or indirect indicators of sexual orientation or location that should not affect recommendations -manifests in significantly shifted price and category distributions of restaurant recommendations. The alarming results we observe strongly indicate that LMRec has learned to reinforce harmful stereotypes through its recommendations. For example, offhand mention of names associated with the black community significantly lowers the price distribution of recommended restaurants, while offhand mentions of common male-associated names lead to an increase in recommended alcohol-serving establishments. These and many related results presented in this work raise a red flag that advances in the language handling capability of LM-driven CRSs do not come without significant challenges related to mitigating unintended bias in future deployed CRS assistants with a potential reach of hundreds of millions of end users.