Scholars Recommendation System Based on academic knowledge graph - - PowerPoint PPT Presentation

Scholars Recommendation System

Group member: Chengyongxiao Wei, Xianze Wu and Hanyi Sun

Based on academic knowledge graph

CONTENTS

n PART ONE n PART TWO n PART THREE n PART FOUR

Project Description Method Result Conclusion

CONTENTS

n PART ONE n PART TWO n PART THREE n PART FOUR

Project Description Method Result Conclusion

Existing Work: Recommendation in Baidu

Motivation

• Similar recommendation system is needed in academic field!
• For students who want to find supervisors or scholars who have just entered a particular field.
• Existing work: Co-authors recommendation in Google Scholar
• Limitation: only utilizes co-author information

Motivation

• More significant information is in knowledge graph.

research fields, affiliations, conferences, etc.

• We aim at an real-time scholar recommendation

system based on AceKG, which recommends

• thers according to the scholar being searched.
• Based on those information, latent relationships

between scholars could be dug out.

CONTENTS

n PART ONE n PART TWO n PART THREE n PART FOUR

Project Description Method Result Conclusion

Methods

• .
• ..
• .

For a given scholar, recommend authors have close connection with him/her

Recommendation based on cooperation network author A author B

Direct cooperation: A and B are coauthors “Cross-author cooperation”: A and B don’t collaborate, but both of them cooperate with C.

author A author C author B Paper 1 Paper 1 Paper 2

For a given author, recommend important scholars on his/her research fields.

Recommendation based on research fields

• Significant scholars and their works benefit users to get familiar with this field.
• Focus on scholars who are important in several fields or critical in a field.
• Obtain significant scholars by important papers.

For a given author, recommend similar scholars in the same affiliation.

Recommendation based on affiliation

• Express similarity between scholars by the overlap
• f research fields
• A variant of TF-IDF algorithm:

!"#$ = &'( )*+,-. /0 121-.( 3) 03-45 0 &'( 6/624 )*+,-. /0 121-.( 78"

$ = log(

)*+,-. /0 (&ℎ/42.( 3) > )*+,-. /0 (&ℎ/42.( 3) > 2)5 "3-45 0) 03-45 0 ∈ ": 6/1 10 03-45( /0 6ℎ- D3E-) 2*6ℎ/. (&ℎ/42. & ∈ >: 43(6 /0 &2)53526-(

Recommend scholars who have published papers in same conference

Recommendation based on conferences

author paper paper paper conference author paper

…

paper paper paper paper

… …

conference

Utilize Author è Paper è Conference è Paper è Author information in academic knowledge graph

CONTENTS

n PART ONE n PART TWO n PART THREE n PART FOUR

Project Description Method Result Conclusion

Case study:

with direct cooperation without “cross-author cooperation” with direct cooperation and “cross-author cooperation”

Demo

• Prof. Xinbing Wang: http://acemap.sjtu.edu.cn/authorTmp/page?AuthorID=7E0DFF97
• Prof. Luoyi Fu: http://acemap.sjtu.edu.cn/authorTmp/page?AuthorID=80008266
• Prof. Xiaohua Tian: http://acemap.sjtu.edu.cn/authorTmp/page?AuthorID=80A899B4

CONTENTS

n PART ONE n PART TWO n PART THREE n PART FOUR

Project Description Methods Result Conclusion

Conclusion

• In our project, we designed a real-time recommendation system based on academic knowledge graph, which

recommend scholars in three dimensions: cooperation network, research fields and affiliations. Besides, we finished offline recommendation based on conference.

• Task Division:
• Chengxiaoyong Wei:

recommendation algorithm based on cooperation network, apply algorithms to Acemap

• Xianze Wu:

recommendation algorithm based on research fields and affiliation, apply algorithms to Acemap, UI design

• Hanyi Sun:

recommendation algorithm based on conference, UI design

Thanks