Multi-Level Fine-Grained Interactions for Collaborative Filtering

Abstract

In recent years, review-based collaborative filtering (CF) has been extensively studied, which is an combination between natural language processing (NLP) and recommender systems. The core pattern behind CF is to first model user and item, and then adopts a relatively primitive interaction between them for personalized recommendation. This pattern is very similar to the issue of sequence matching in NLP, where sequence 1 and sequence 2 are matched with a fine-grained interaction leading to a better result. Therefore, there is a tremendous room for further improvement in current review-based CF to release the power of fine-grained interaction. To this end, we treat the user review set and item review set as two sequences, and design a multi-level matching attention layer for fine-grained interaction. In addition, we devise the aspect-level and review-level attention to measure the contribution of each review. Extensive experiments on 24 public datasets show that the proposed model consistently outperforms the state-of-theart approaches. More importantly, by selecting the relevant reviews according to the aspect attention score and review attention score, we can observe which specific item aspects that user mainly concerned and which item characteristic highly matched with the user preference, in which the recommendation interpretability can be enhanced.

Deep learning techniques for recommender systems based on collaborative filtering

“…Thus, exploring a sequential architecture comes as a natural and reasonable choice to learn data dynamics, especially when data representations tend to be sparse. Bellini et al (2017), Chae et al (2019), He et al (2019), Hu et al (2019), Jhamb et al (2018), Lee et al (2017Lee et al ( , 2018, Liang et al (2018), Liu et al (2017), Nisha and Mohan (2019), Song et al (2019), Wang, Chen, et al (2019), Wang et al (2020) Convolutional neural network (CNN) 9 Chen, Cai, et al (2019), Da Costa and Dolog (2019), Hyun et al (2018), Liu et al (2017Liu et al ( , 2019, Wang, Chen, et al (2019), Zhang, Cheng, and Ren (2019), Zhang, Yao, et al (2017), Zheng et al (2017) Generative adversarial network (GAN) 3 Chae et al 2019, Lee et al (2017), Wang, Chen, et al (2019) Graph neural network (GNN) 2 Wu, Hong, et al (2019), Zheng et al (2018) Multilayer perceptron (MLP) 20 Bai et al (2017), Cao et al, 2018, C. Chen et al (2020, L. Chen, Zheng, et al (2018), W. Chen, Cai, et al (2019) , Zhou et al (2019) Neural attention 13 (Cao et al (2018), L. Chen, Zheng, et al, 2018, Chin et al, 2018, W. Fan et al (2019, Feng & Zeng, 2019, Jhamb et al (2018…”

“…Feng and Zeng (2019) interpret the review‐based CF as a sequence matching problem as in the Natural Language Processing (NLP) field. User and item sequences are matched through a fine‐grained interaction leading to a better result (Feng & Zeng, 2019).…”

“…Feng and Zeng (2019) interpret the review‐based CF as a sequence matching problem as in the Natural Language Processing (NLP) field. User and item sequences are matched through a fine‐grained interaction leading to a better result (Feng & Zeng, 2019). To do so, the authors propose an approach based on the BERT technique (a short for Bidirectional Encoder Representations from Transformers) (Devlin et al, 2018) to encode user/item text reviews and a multi‐contextual layer attention mechanism to capture complex dynamic relations in multi‐level, aspect‐level and review‐level from the input data.…”

Course Recommendation System based on Natural Language Processing

“…If the cosine distance is big, it indicates that the consumer is probable to experience the movie. otherwise, we're possibly to avoid the item from the recommendation [17].…”

Development and Utilization of Aesthetic Education Resources Using Collaborative Filtering Model

“…e number of students using the internet is rapidly increasing, and users download and post educational materials there. e availability of educational resources grows exponentially in such a relatively unrestricted environment, leading to "information overload" [7]. People are simultaneously finding it more and more challenging to quickly and accurately locate the information they require.…”