Gradually adaptive recommendation based on semantic mapping of users′ interest correlations

Chen, Jian; Zhou, Xiaokang; Jin, Qun

doi:10.1002/dac.2835

Cited by 3 publications

(3 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…ere are many incomplete domains in migration learning that will be used in information recommendation systems [17]. Analyze users' information access history records, extract their whereabouts, label them according to semantic means, and finally search users for information [18]. For the generation of massive data information in the bus card, a comfortable bus route can be recommended to passengers according to their needs.…”

Section: Introductionmentioning

confidence: 99%

Application and Analysis of Artificial Intelligence in College Students’ Career Planning and Employment and Entrepreneurship Information Recommendation

Zhang

Zheng²

2022

Security and Communication Networks

View full text Add to dashboard Cite

With the arrival of graduation season, the number of graduates is expanding every year, and the employment rate of college students has become one of the data that colleges and universities pay attention to. To improve the employment rate of college students, we should first educate each college student in career planning, so that they can clearly understand their own positioning and future development direction. Career planning can guide students to understand their jobs and analyze their professional fields. Faced with massive employment information, it is also a big problem for college students to search and choose information. In this paper, the personalized recommendation system for entrepreneurship is described, and the basic principles of information recommendation are described. The basic information and personal interest points of college students are represented by feature vectors, which provide a favorable theoretical support for college students’ career planning and employment and entrepreneurship information recommendation. An information recommendation model under deep learning is formed. Finally, the performance of the model under the traditional algorithm and the optimized information recommendation model is evaluated, and the satisfaction of users to this system is scored so as to provide a convenient and quick information recommendation system for college students, thus indirectly improving the employment rate of graduates and providing corresponding solutions to the problem of difficult employment.

show abstract

Section: Introductionmentioning

confidence: 99%

Application and Analysis of Artificial Intelligence in College Students’ Career Planning and Employment and Entrepreneurship Information Recommendation

Zhang

Zheng²

2022

Security and Communication Networks

View full text Add to dashboard Cite

show abstract

“…Machine learning techniques including neural networks (Mostafa and Lam, 2000), Support Vector Machine (Lee and Choo, 2016), K-Nearest Neighbors and logistic regression (Caulkins et al , 2006; Zhang et al , 2007) are used to learn the mapping between the incoming set of documents relevant to user input and real numbers which represent the strength of user interests. The features of these documents are first extracted by using widely used techniques including information gain (Mitchell, 1997, Vu et al , 2017) and correlation coefficients (Chen et al , 2016). Li et al (2011) proposed a Rough Threshold Model to analyze and extract keywords from scientific publications.…”

Section: Literature Reviewmentioning

confidence: 99%

Expert profiling for collaborative innovation: big data perspective

Silva

Ma²

2017

IDD

View full text Add to dashboard Cite

Purpose Expert profiling plays an important role in expert finding for collaborative innovation in research social networking platforms. Dynamic changes in scientific knowledge have posed significant challenges on expert profiling. Current approaches mostly rely on knowledge of other experts, contents of static web pages or their behavior and thus overlook the insight of big social data generated through crowdsourcing in research social networks and scientific data sources. In light of this deficiency, this research proposes a big data-based approach that harnesses collective intelligence of crowd in (research) social networking platforms and scientific databases for expert profiling. Design/methodology/approach A big data analytics approach which uses crowdsourcing is designed and developed for expert profiling. The proposed approach interconnects big data sources covering publication data, project data and data from social networks (i.e. posts, updates and endorsements collected through the crowdsourcing). Large volume of structured data representing scientific knowledge is available in Web of Science, Scopus, CNKI and ACM digital library; they are considered as publication data in this research context. Project data are located at the databases hosted by funding agencies. The authors follow the Map-Reduce strategy to extract real-time data from all these sources. Two main steps, features mining and profile consolidation (the details of which are outlined in the manuscript), are followed to generate comprehensive user profiles. The major tasks included in features mining are processing of big data sources to extract representational features of profiles, entity-profile generation and social-profile generation through crowd-opinion mining. At the profile consolidation, two profiles, namely, entity-profile and social-profile, are conflated. Findings (1) The integration of crowdsourcing techniques with big research data analytics has improved high graded relevance of the constructed profiles. (2) A system to construct experts’ profiles based on proposed methods has been incorporated into an operational system called ScholarMate (www.scholarmate.com). Research limitations One shortcoming is currently we have conducted experiments using sampling strategy. In the future we will perform controlled experiments of large scale and field tests to validate and comprehensively evaluate our design artifacts. Practical implications The business implication of this research work is that the developed methods and the system can be applied to streamline human capital management in organizations. Originality/value The proposed approach interconnects opinions of crowds on one’s expertise with corresponding expertise demonstrated in scientific knowledge bases to construct comprehensive profiles. This is a novel approach which alleviates problems associated with existing methods. The authors’ team has developed an expert profiling system operational in ScholarMate research social network (www.scholarmate.com), which is a professional research social network that connects people to research with the aim of “innovating smarter” and was launched in 2007.

show abstract

“…Enlightened by the co-ranking algorithm in [5], we take into account the concept of user community and propose UCCC for user communities and contents based on the mutually reinforcing relationship between them. Quite a few researchers have noticed the important concept of communities in social networks, such as [24] and [25]. The proposed UCCC can evaluate UGC quality from a global view instead of the personalized view in [5].…”

Section: Introductionmentioning

confidence: 99%

User communities and contents co‐ranking for user‐generated content quality evaluation in social networks

Lin

Zhai

et al. 2014

Int J Communication

View full text Add to dashboard Cite

SUMMARYWith the massive popularity of social networks, more and more users can produce millions of user-generated contents (UGCs) daily. However, UGC quality is uneven, which has posed challenges to finding superior contents in such a large data set. In this paper, we present a new idea of UGC quality evaluation exploiting user communities, which are formed by users either in a friend circle or with similar interests in social networks. The intuition is that a user community can help evaluate the UGC quality better than a single user. Hence, we propose a new graph-theoretic user communities and contents co-ranking (UCCC) algorithm for UGC quality evaluation. UCCC evaluates UGCs and their related user communities simultaneously based on three different relationship networks: the network connecting UGCs, the network connecting user communities, and a third network that ties the two together. Contents and user communities are ranked following a co-ranking algorithm based on the assumption that there is a mutually reinforcing relationship between them. Experiments using real-world data have shown that UCCC outperforms competitive algorithms by a good margin in most cases and a user community is more useful than a single user for UGC quality evaluation.

show abstract

Gradually adaptive recommendation based on semantic mapping of users′ interest correlations

Cited by 3 publications

References 33 publications

Application and Analysis of Artificial Intelligence in College Students’ Career Planning and Employment and Entrepreneurship Information Recommendation

Application and Analysis of Artificial Intelligence in College Students’ Career Planning and Employment and Entrepreneurship Information Recommendation

Expert profiling for collaborative innovation: big data perspective

User communities and contents co‐ranking for user‐generated content quality evaluation in social networks

Contact Info

Product

Resources

About