Real-time Video Recommendation Exploration

Huang, Yuanding; Cui, Bin; Jiang, Jie; Hong, Kunqian; Zhang, Wenyu; Xie, Yiran

doi:10.1145/2882903.2903743

Cited by 57 publications

(22 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We compare our ssRec to two state-of-the-art baselines, CTT [17] and UCD [36]. CTT fuses collaborative filtering, type and temporal factor together to generate recommendation over streams.…”

Section: B Evaluation Methodologymentioning

confidence: 99%

“…Algorithm 1 illustrates the general framework for the KNN query over CPPse-index. Given an incoming social item v, our algorithm performs KNN query by three important steps: (1) compute the hash values based on the entity-category pairs contained in v, by which a set of extended signature trees are located (Lines 5-6); (2) generate pseudo-query based on the item and each located extended signature tree (Line 7); (3) select and rank the top-k relevant users (Lines [13][14][15][16][17][18][19][20][21][22]. We maintain a max-heap U k with size k as our output ranked list.…”

Section: The Relevance Between Them Can Be Computed By Plugging Statimentioning

confidence: 99%

“…However, existing diversity-based approaches handle the user preferences as static, which ignores the temporal evolution of social users' preference. Recent recommendation approaches have been proposed to capture the user preferences over streams [8], [9], [17], [21]. They mainly focus on how to extend the traditional recommendation techniques such as matrix factorization [15] to streaming environments by applying them to media data with the support of efficient stream processing.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Online Social Media Recommendation Over Streams

Zhou

Qin

et al. 2019

2019 IEEE 35th International Conference on Data Engineering (ICDE)

View full text Add to dashboard Cite

As one of the most popular services over online communities, the social recommendation has attracted increasing research efforts recently. Among all the recommendation tasks, an important one is social item recommendation over high speed social media streams. Existing streaming recommendation techniques are not effective for handling social users with diverse interests. Meanwhile, approaches for recommending items to a particular user are not efficient when applied to a huge number of users over high speed streams. In this paper, we propose a novel framework for the social recommendation over streaming environments. Specifically, we first propose a novel Bi-Layer Hidden Markov Model (BiHMM) that adaptively captures the behaviors of social users and their interactions with influential official accounts to predict their long-term and short-term interests. Then, we design a new probabilistic entity matching scheme for effectively identifying the relevance score of a streaming item to a user. Following that, we propose a novel indexing scheme called CPPse-index for improving the efficiency of our solution. Extensive experiments are conducted to prove the high performance of our approach in terms of the recommendation quality and time cost.

show abstract

Section: B Evaluation Methodologymentioning

confidence: 99%

Section: The Relevance Between Them Can Be Computed By Plugging Statimentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Online Social Media Recommendation Over Streams

Zhou

Qin

et al. 2019

2019 IEEE 35th International Conference on Data Engineering (ICDE)

View full text Add to dashboard Cite

show abstract

“…Collaborative filtering (CF) is widely used in recommender systems, which models user interests by exploring user-item interactions with the assumption that people with similar interests tend to make similar choices. For example, Huang et al [15] developed a realtime matrix factorization based CF algorithm with an adjustable online updating strategy for video recommendation. However, CF methods suffer from the problem of cold start and data sparsity [7,25,31].…”

Section: Video Recommendationmentioning

confidence: 99%

“…As shown in Figure 1, different users have different interest groups and the interest groups of one user may also be diverse. However, existing methods commonly focus on learning the user interest representation directly from micro-video features without grouping [4,8,15,25]. In this way, the large groups with a majority of historical micro-videos dominate the user interests, and the micro-videos in small groups are rarely recommended.…”

Section: Introductionmentioning

confidence: 99%

What Aspect Do You Like

Jiang¹,

Wang

Wei³

et al. 2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

Online micro-video recommender systems aim to address the information explosion of micro-videos and make the personalized recommendation for users. However, the existing methods still have some limitations in learning representative user interests, since the multi-scale time effects, user interest group modeling, and false positive interactions are not taken into consideration. In view of this, we propose an end-to-end Multi-scale Time-aware user Interest modeling Network (MTIN). In particular, we first present an interest group routing algorithm to generate fine-grained user interest groups based on user's interaction sequence. Afterwards, to explore multi-scale time effects on user interests, we design a time-aware mask network and distill multiple temporal information by several parallel temporal masks. And then an interest mask network is introduced to aggregate fine-grained interest groups and generate the final user interest representation. At last, in the prediction unit, the user representation and micro-video candidates are fed into a deep neural network (DNN) for predictions. To demonstrate the effectiveness of our method, we conduct experiments on two publicly available datasets, and the experimental results demonstrate that our proposed model achieves substantial gains over the state-of-the-art methods.

show abstract