Jiawei Chen

Jiawei Chen

Auto-generated profile, based on publicly available information

Jiawei Chen

School of Information Management and Engineering, Shanghai University of Finance and Economics, Shanghai 200433, China

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 December 2024 in Information Systems Research
Verified
Journal Article
Closed access
JCJiawei Chen
Jiawei Chen
LHLuo He
Luo He
HLHongyan Liu
Hongyan Liu
YYYinghui (Catherine) Yang
Yinghui (Catherine) Yang
XBXuan Bi
Xuan Bi
On short video sharing platforms, users often choose background music for their videos. In this paper, we study the problem of background music recommendation for short videos on short video sharing platforms. In our recommendation setting, the item (music) is not recommended directly to the user, but to the video created by the user. When making music recommendations for videos, we consider three important players: users, videos, and music. We define a unique background music recommendation problem and design a novel background music recommendation model to address the problem. We propose a model based on the deep learning framework to effectively address the distinctive three-way relationships among users, videos, and music. Our model considers not only of the conventional user–music alignment, but also the alignment between videos and music. To evaluate our model, we conduct comprehensive experiments on real-world data collected from one of the most popular short video sharing platforms. Our proposed model significantly outperforms other existing models in recommendation performance. The superiority of our proposed model remains consistent across various scenarios, including cold-start recommendations, data sets with varying density levels, and data sets spanning diverse video categories.

1 June 2021 in Information Systems Research
Verified
Journal Article
Closed access
JCJiawei Chen
Jiawei Chen
YYYinghui (Catherine) Yang
Yinghui (Catherine) Yang
HLHongyan Liu
Hongyan Liu
In recent years, more and more platforms where both buyers and sellers can write reviews for each other have emerged. These bilateral reviews are important information sources in the decision-making process of both buyers and sellers. In this study, we develop a comprehensive relational topic modeling approach to analyze bilateral reviews for better online transaction prediction. The prediction results will enable the platform to increase the chance that the buyer and seller reach a transaction by presenting buyers with offerings that are more likely to lead to a transaction. Within the framework of the relational topic model, we embed a topic structure with both shared and corpus-specific topics to better handle text corpora generated from different sources. Our model facilitates the extraction of the appropriate topic structure from different document collections that helps enhance the transaction prediction performance. Comprehensive experiments conducted on real-world data sets collected from sharing economy platforms demonstrate that our new model significantly outperforms other alternatives. The robust results obtained from multiple sets of comparisons demonstrate the value of bilateral reviews if they are processed properly. Our approach can be applied to many platforms where bilateral reviews are available.

10 December 2019 in ACM Transactions on Management Information Systems
Verified
Journal Article
Open access
JCJiawei Chen
Jiawei Chen
HLHongyan Liu
Hongyan Liu
YYYinghui (Catherine) Yang
Yinghui (Catherine) Yang
JHJun He
Jun He
Consumers increasingly make informed buying decisions based on reading online reviews for products and services. Due to the large volume of available online reviews, consumers hardly have the time and patience to read them all. This article aims to select a compact set of high-quality reviews that can cover a specific set of product features and related consumer sentiments. Selecting such a subset of reviews can significantly save the time spent on reading reviews while preserving the information needed. A unique review selection problem is defined and modeled as a bi-objective combinatorial optimization problem, which is then transformed into a minimum-cost set cover problem that is NP-complete. Several approximation algorithms are then designed, which can sustain performance guarantees in polynomial time. Our effective selection algorithms can also be upgraded to handle dynamic situations. Comprehensive experiments conducted on twelve real datasets demonstrate that the proposed algorithms significantly outperform benchmark methods by generating a more compact review set with much lower computational cost. The number of reviews selected is much smaller compared to the quantity of all available reviews, and the selection efficiency is deeply increased by accelerating strategies, making it very practical to adopt the methods in real-world online applications.

24 July 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
JCJiawei Chen
Jiawei Chen
MWMengnan Wu
Mengnan Wu
Nowadays, digital platforms have increasingly relied on recommendation systems to connect users with relevant products and services. However, these systems often operate in environments vulnerable to major external shocks. Such shocks can reshape user preferences, item characteristics, and interaction patterns, creating a critical challenge for platforms seeking to maintain effective recommendation. In such settings, interactions accumulated before the shock contain useful collaborative signals but may also include outdated information, whereas interactions observed afterward reflect emerging behaviors but are often too sparse for recommendation. How to selectively reuse historical knowledge while adapting to post-shock changes remains insufficiently studied. To address this research gap, we formulate the cross-phase recommendation problem and propose an adaptive graph neural network framework for resilient recommendation under external shocks, which adapts historical knowledge at both the representation and interaction levels. Specifically, we develop a personalized knowledge transfer module that learns user- and item-specific gates to determine how much historical information should be retained and adapted after the shock. We further design a personalized graph augmentation module that selectively incorporates informative historical interactions into the sparse graph after the shock. Comprehensive experiments on a real-world dataset demonstrate that our proposed framework outperforms other alternatives. Its superiority remains consistent across cities, evaluation metrics, and test months. The temporal results also show that it recovers rapidly from the initial performance decline and sustains strong recommendation performance over an extended period.

10 November 2018 in SSRN Electronic Journal
Verified
Preprint
Open access
PDF file
JCJiawei Chen
Jiawei Chen
HLHongyan Liu
Hongyan Liu
JHJun He
Jun He
In recent years, live streaming has experienced rapid growth and become a new social media with wide influence, ranking top in attracting internet traffic. On live streaming platforms, plenty of anchors simultaneously broadcast live streams on different topics. How to recommend appropriate anchors to viewers and make them stay are very important for the platforms. Different from video-sharing platforms such as YouTube, one of the major characteristics of live streaming is real-time user participation, which plays an important role in analyzing viewer's watching preferences and has not been well studied in existing research on recommendation systems. Therefore, in this paper, we study how to take real-time interactions between viewers and anchors into consideration to make a better live stream recommendation. To this end, we develop an interaction-aware recommendation framework, combining matrix factorization and deep learning models to infer viewer preferences. The matrix factorization component can effectively learn viewer preferences at the anchor level, and the deep learning component is innovatively introduced to learn viewer preferences for real-time interactive behaviors at the live stream level. This interaction-aware recommendation framework captures comprehensive relationships between real-time user participation and viewer's watching duration, which helps enhance real-time anchor recommendation performances. Comprehensive experiments conducted on a real-world dataset demonstrate that our new model significantly outperforms baselines. Interpretable viewer preferences for real-time interactive behaviors are helpful for a more profound understanding of viewer's behaviors.

12 July 2018 in SSRN Electronic Journal
Verified
Preprint
Open access
PDF file
JCJiawei Chen
Jiawei Chen
YYYinghui (Catherine) Yang
Yinghui (Catherine) Yang
HLHongyan Liu
Hongyan Liu
For sharing economy platforms such as Airbnb, how to improve transaction success rate is a vital issue. We develop an integrated modeling framework that analyzes item descriptions and online user reviews for transaction success prediction. Different from using unilateral reviews in traditional online markets, we incorporate bilateral reviews common in sharing economy to predict results of potential transactions, which has not been addressed in the literature before. A new relational topic model is developed to capture the process of leveraging textual information to predict transaction results. Within the framework of the relational topic model, we embed a topic structure with both shared and private topics to better handle text corpora generated from different sources. This new relational topic model framework facilitates the extraction of the appropriate topic structure from different document collections which helps enhance transaction success prediction performances. Comprehensive experiments conducted on a real-world dataset based on Airbnb demonstrate that our new model significantly outperforms other alternatives.