View in EDS

A scientific citation recommendation model integrating network and text representations.

Saved in:
Bibliographic Details
Title: A scientific citation recommendation model integrating network and text representations.
Authors: Qiu, Tianshuang1 (AUTHOR), Yu, Chuanming1 (AUTHOR) yucm@zuel.edu.cn, Zhong, Yunci1 (AUTHOR), An, Lu2 (AUTHOR), Li, Gang2 (AUTHOR)
Source: Scientometrics. Nov2021, Vol. 126 Issue 11, p9199-9221. 23p.
Subjects: Logistic regression analysis, Acquisition of data, Information networks, Citation networks
Abstract: The number of scientific papers is increasing in the rapid growth. How to make paper acquisition efficient and provide effective citation recommendation is essential for researchers. Although the application of scientific citation recommendation has shown great improvements, the in-depth mining and fusion of various types of information has been ignored. In this paper, we propose a scientific citation recommendation model integrating network and text representation (SCR-NTR), which comprises data acquisition, feature representation, feature fusion and link prediction. We compare the network representation and text representation, respectively, and select the models performing best in the pre-experiment as the sub-models of SCR-NTR. The method of vector concatenate fusion is employed to fuse two kinds of information, and the logistic regression classifier is selected to carry out the link prediction. The extensive experiments reveal that our model can effectively improve the performance on citation recommendation. In addition, the effect of different fusion methods and different classifiers are investigated, and qualitative analysis is conducted to further verify the effectiveness of SCR-NTR. The experimental results show that leveraging both network and text representation can enhance the recommendation performance, and the heterogenous network representation learning can capture richer semantic information of the given network than the homogeneous one. [ABSTRACT FROM AUTHOR]
Copyright of Scientometrics is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:The number of scientific papers is increasing in the rapid growth. How to make paper acquisition efficient and provide effective citation recommendation is essential for researchers. Although the application of scientific citation recommendation has shown great improvements, the in-depth mining and fusion of various types of information has been ignored. In this paper, we propose a scientific citation recommendation model integrating network and text representation (SCR-NTR), which comprises data acquisition, feature representation, feature fusion and link prediction. We compare the network representation and text representation, respectively, and select the models performing best in the pre-experiment as the sub-models of SCR-NTR. The method of vector concatenate fusion is employed to fuse two kinds of information, and the logistic regression classifier is selected to carry out the link prediction. The extensive experiments reveal that our model can effectively improve the performance on citation recommendation. In addition, the effect of different fusion methods and different classifiers are investigated, and qualitative analysis is conducted to further verify the effectiveness of SCR-NTR. The experimental results show that leveraging both network and text representation can enhance the recommendation performance, and the heterogenous network representation learning can capture richer semantic information of the given network than the homogeneous one. [ABSTRACT FROM AUTHOR]
ISSN:01389130
DOI:10.1007/s11192-021-04161-0