Research Summary of Recommendation System Based on

Knowledge Graph
Xiaoxi He Xinsheng Ke
Beijing Jiaotong University Beijing Jiaotong University
19120609@bjtu.edu.cn xske@bjtu.edu.cn
ABSTRACT 1.1 Application of Recommender System
This article will systematically review the knowledge graph-based In recent years, the recommendation method using additional in-
recommendation system from three aspects: knowledge graph, rec- formation in the form of knowledge graph has attracted scholars’
ommendation system, and application of knowledge graph and attention. Knowledge graph ng was originally used to improve
recommendation system. Specifically, it investigates and analyzes the performance of search system. It describes a large number of
related articles in recent years, and summarizes and elaborates on relationships between entities. It adopts the network structure,
the concepts, construction, and research of knowledge graphs, as which can enhance the user’s cognition and explanation of the
well as collaborative filtering recommendation systems and hybrid relationship between the user and the project, so as to improve the
recommendation systems. On this basis, the application scenarios accuracy of the recommendation. In the recommendation system
of the recommendation system based on the knowledge graph are based on knowledge graph the user information and items collected
discussed and analyzed. The article summarizes the research work are usually expressed in the form of knowledge graph.
on the recommendation system of knowledge graphs in recent
years, and also proposes detailed analysis and prospects based on
the application status of different fields. 1.2 Recommended Steps of Recommender
System
CCS CONCEPTS Usually, the recommender system first represents the collected user
• Applied computing; • Operations research; • Forecasting; information, project information, user historical behavior and other
data or some related external data. Then, a recommendation system
KEYWORDS is designed, which uses knowledge graph to generate recommen-
dation. This recommendation system usually includes two steps:
Information management, Knowledge Graph, Recommendation the construction of knowledge graph and the generation of rec-
System, Collaborative Filtering ommendation using knowledge graph. Based on this, this paper
ACM Reference Format: analyzes and discusses the literature review of knowledge graph
Xiaoxi He and Xinsheng Ke. 2021. Research Summary of Recommendation recommendation system.
System Based on Knowledge Graph. In The 2021 3rd International Conference
on Big Data Engineering (BDE 2021), May 29–31, 2021, Shanghai, China. ACM,
New York, NY, USA, 6 pages. https://doi.org/10.1145/3468920.3468935 2 KNOWLEDGE GRAPH
2.1 Concept Research
1 INTRODUCTION With the continuous progress and development of artificial in-
With the explosive growth of Internet and real information, more telligence related technology, knowledge graphing has gradually
and more online service providers provide users with recommenda- become an important branch of artificial intelligence. Knowledge
tion services for goods, music, movies, etc. Recommender system engineering has been successfully applied in the big data environ-
can meet the personalized needs of users and bring great business ment. However, knowledge graphing is a branch of knowledge
value to online service providers. At the same time, the research engineering. There are many representative general knowledge
of recommendation methods and systems has promoted the rapid graphs at home and abroad, and there are usually two kinds of data
development of preference mining, big data processing, decision sources: the large-scale knowledge base represented by Wikipedia,
support and other related theories and practices, and its academic and the gradual accumulation by crowdsourcing.
value has also been widely concerned. In essence, knowledge graph can be regarded as the knowledge
base of semantic network. Semantic network uses graph structure
to store information structurally. On the basis of semantic network,
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed retained the use of graph structure to represent the characteristics
for profit or commercial advantage and that copies bear this notice and the full citation of information in semantic network, and expressed the entities in-
on the first page. Copyrights for components of this work owned by others than ACM volved in data and the relationship between entities in a structured
must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,
to post on servers or to redistribute to lists, requires prior specific permission and/or a form. With the help of automatic construction technology, open
fee. Request permissions from permissions@acm.org. domain knowledge graphing represented by Yago, DBpedia, etc. As
BDE 2021, May 29–31, 2021, Shanghai, China far as the scope and direction of the graph are concerned, it can
© 2021 Association for Computing Machinery.
ACM ISBN 978-1-4503-8942-6/21/05. . . $15.00 be divided into two categories: general domain knowledge graph
https://doi.org/10.1145/3468920.3468935 and vertical domain knowledge graph. General domain knowledge

104
BDE 2021, May 29–31, 2021, Shanghai, China Xiaoxi He and Xinsheng Ke

graphing is usually used in search and other scenarios, which re- and the secret of knowledge graphing technology and its relation-
quires comprehensive knowledge, but the accuracy of knowledge ship with related disciplines are gradually revealed; 3) the main
is relatively low. Representative general domain knowledge graphs challenges and key problems of knowledge graphing construction
include: Zhishi, cn-dbpedia, Sogou zhicube and Baidu Zhixin, as technology are summarized Question.
well as foreign freebase, Yago and DBpedia, such as freebase, Yago Yuan Xuping [5] focused on the key link of knowledge extraction
and DBpedia. The vertical domain knowledge graph is mainly used in knowledge graph, proposed to use deep learning algorithm to
in intelligent problem-solving, knowledge reasoning and decision- solve the key problem of knowledge extraction in knowledge graph,
making. At present, the vertical domain knowledge graph mainly and introduced graphic database to provide a new idea for the draw-
involves the direction of tourism, transportation and medicine. ing of knowledge graph. By introducing the deep learning algorithm
Jiao Xiaojing, Wang Lancheng [1] pointed out that there are of machine learning, the framework of business domain knowledge
still some problems in the current related research, such as unclear graphing based on deep learning is built, and the deep belief net-
concept and unclear subject orientation, which need to be sorted work algorithm is used to automatically extract the knowledge units
out and clarified. Through literature review and concept compar- contained in the unstructured business text and the relationship
ison, this paper defines the concept category of knowledge map, between the units.
clarifies the relationship between knowledge map and knowledge
map, defines the subject orientation of knowledge map, and points
out that knowledge map and knowledge map overlap, but are not
identical. It is an information visualization method, mainly used 2.3 Application Research
to reveal the structural characteristics of scientific knowledge in Since 2012, papers on knowledge graphing have emerged in an
scientometrics. This conclusion provides a reference for the further endless stream, and the research in various fields is also increasing.
study of knowledge map. The essence of knowledge graphing can be analyzed from different
Huang Hengqi, Yu Juan et al [2] think that the connotation of perspectives. If you look at it from the perspective of the world
knowledge graph is not clear enough. Moreover, due to the incom- wide web, it’s like creating hypertext links and data links.
pleteness of the documents, the utilization and reuse rate of the As an intelligent and efficient method of knowledge organization,
existing knowledge graph are not high. Therefore, their research knowledge graphing can help users to query the required informa-
gives the definition of knowledge graph and analyzes the rela- tion quickly and accurately. Li Tao and Wang sichen[6] reviewed
tionship between knowledge graph and ontology. Ontology is the the research contents of scholars, scientific research institutions or
pattern layer and logical basis of knowledge map, and knowledge enterprises on knowledge mapping, introduced the development
graph is the instantiation of ontology. The research results can be and construction methods of knowledge mapping, including the ori-
used as the basis of knowledge graphing research, promote the gin, development and final formation of the concept of knowledge
faster development and wider application of knowledge graping, mapping. Both of them look forward to the future development
and point out the future research direction of knowledge graphing. direction and challenges of knowledge mapping. Although many
systems have been applied in many knowledge fields, their basic
theory and application technology still need further research.
2.2 Construction Research Wang Yongchao, Luo Shengwen, et al. [7] thought that knowl-
Qin Changjiang [3] systematically summarized the theoretical basis edge graph is a structured semantic web knowledge base, which
and technical methods of knowledge graphing, and elaborated the describes concepts and their relationships in the form of symbols.
relationship between knowledge graphing and knowledge graphing, Because knowledge graph is often used for interactive knowledge
information visualization and knowledge management. Using the discovery and reasoning, the visualization of knowledge graph has
co-occurrence analysis methods of scientometrics, cluster analysis, attracted a lot of attention. Therefore, in the literature, scholars
factor analysis, multi-dimensional scale analysis and social network summarized the related work of knowledge mapping visualization,
analysis technology, the research focuses and main research fields introduced the visualization expression of knowledge graphing,
in different periods in the history of science are analyzed, and a analyzed the applicable types of knowledge graphing, described
comprehensive knowledge map of the relationship between the the challenges and visualization methods of large-scale knowledge
two disciplines is constructed. It is innovative and scientific for him graphing visualization, heterogeneous network visualization analy-
to apply the method of empirical analysis knowledge graphing to sis method, introduced the application cases of knowledge graphing
the research of the development history of Chinese humanities. visualization, and predicted the future development direction and
Duan Hong [4] started with the definition and technical frame- choice.
work of knowledge graphing, and comprehensively analyzed the Chai Bo and Zhang Qiang [8] applied the retrieval method and
key technologies involved in the construction of knowledge graph- device to the relational data table of power terminal equipment
ing from bottom to top. 1) This paper expounds the definition through the construction of knowledge graph. According to the
and connotation of knowledge graphing, and gives the technical relational data table, the knowledge graph concept layer and the
framework of constructing knowledge graphing. According to the knowledge graph instance layer are constructed; combined with the
abstract degree of input knowledge materials, knowledge graphing knowledge graph concept layer and the knowledge graph instance
is divided into three levels: information extraction layer, knowledge layer, the knowledge graph is formed. By transforming table data
fusion layer and knowledge processing layer; 2) the research status into concept layer and instance layer, the relationship between
of key technologies involved in each level is classified and explained, relational data can be more clearly reflected, and the accuracy of

105
Research Summary of Recommendation System Based on Knowledge Graph BDE 2021, May 29–31, 2021, Shanghai, China

user requirement understanding and resource description can be 3.3 Hybrid Recommendation System
improved. Shen Peng and Li Tao [14] used an improved hybrid method, which
fully considered and utilized the advantages of collaborative filter-
3 RECOMMENDER SYSTEM ing algorithm and content attribute filtering. The algorithm dis-
In the era of information overload, recommender system helps cussed is different from the previous methods in this field, because
users to find the information they need from a large amount of it includes a novel method to find the similar content between two
information, which attracts users’ attention. Recommender system things. An analysis is included to prove the new method and how it
is a mature technology and has been used in many fields. But rec- provides practical recommendations. Wang Chuanlong, Shao Yabin
ommender system is widely used at the same time, there are also [15] collaborative filtering technology is one of the most valuable
some problems. For example, how to solve the problem of recom- core technologies in recommender system. It can deeply tap the
mendation for new users, the amount of user data in the system potential interests of users and make more reasonable recommen-
is too small, how to recommend different types of information for dations to users. However, cold start, data sparsity, scalability and
users, and the timeliness of items that users need to recommend. other problems still restrict the application of this technology in the
actual recommender system. Aiming at the problems of cold start
3.1 Recommendation Technology and data sparsity, this paper proposes a hybrid collaborative filter-
Recommendation technology based on knowledge spectrum can ing recommendation model based on nearest neighbor propagation
be divided into recommendation technology based on ontology clustering.
and recommendation technology based on linked open data (LOD).
Ontology based recommendation technology mainly uses the de- 3.4 Application of Recommender System
tailed description of concept level in ontology to express entity With the wide application of recommendation system, more and
features more accurately through the detailed classification de- more researchers focus on recommendation algorithm and system
scription of concepts, so as to mine the deep information in facts. application, and write a large number of excellent papers. The
Thanapalasingam [9] built an ontology based scientific publication relatively typical papers are shown in Table 1
recommendation system, which solved the problems of topic rep-
resentation and scientific publication recommendation. Although 3.5 Shortcomings of Existing Recommendation
ontology based recommendation technology has achieved good System
results in some areas, the first problem of using ontology based rec-
Although the research of recommendation algorithm has lasted
ommendation technology is the construction of ontology. Alnazer
for many years, and there are many related papers and technolo-
et al. [10] used ontology technology to analyze user preferences,
gies, there are still some common problems. For example, when
established nutrition and health oriented user profiles, and used
using recommendation algorithm to achieve high-quality recom-
the profiles to organize relevant knowledge and recommend health-
mendation, the network model involved is often complex, which
ier food for users. Kethavarapu et al. [11] dynamically generated
makes the training model and algorithm require a large amount
the ontology according to the position data on the job website,
of time and space costs. In addition, if you want to get a better
and then recommended the position information for job seekers
recommendation effect, you need to collect a lot of sample data,
through reordering and similarity algorithm.
and a good training model needs a lot of sample data. At the same
time, the traditional recommendation algorithm has the problems
3.2 Collaborative Filtering Recommendation
of sparse data, poor scalability and cold start, which makes the
System current recommendation system unable to meet the needs of users.
Sun Guangfu, Wu Le et al. [12] proposed a modeling method of All kinds of problems make recommendation system in trouble.
temporal behavior between users (products). Based on this method,
we can find the neighbor set which has the greatest impact on the 4 APPLICATION OF RECOMMENDATION
current user (product). The verification results on two real datasets SYSTEM BASED ON KNOWLEDGE GRAPH
show that the proposed sequential recommendation system can
more effectively predict the actual score of users and improve the 4.1 Why Apply Knowledge Mapping in
accuracy of recommendation compared with the traditional recom- Recommendation System
mendation system based on social network information and tag As a multidisciplinary field, recommender system includes data
information. Chen Biyi, Huang Ling, et al. [13] proposed a collabora- mining, natural language processing, machine learning and so on.
tive filtering recommendation system combining explicit feedback Similarly, the development of various disciplines will bring new
and implicit feedback. Firstly, the implicit user / item vector is vitality to the recommendation system. However, although many
trained by weighted low rank approximation. Then the baseline scholars have made great efforts to achieve accurate and personal-
evaluation is introduced and supplemented by the implicit user / ized recommendation, there are still problems of data sparsity and
Project vector. Through the combination of explicit and implicit cold start, which affect the recommendation effect. In recent years,
user / item vectors, the user’s preference for item prediction is knowledge graphing as a way to generate recommendation based
obtained. The feasibility and effectiveness of the proposed recom- on edge information has come into the view of scholars, which
mendation system are verified by comparing with several typical has caused one of the hot topics to improve the accuracy of rec-
algorithms on standard datasets. ommendation system. The recommendation algorithm based on

106
BDE 2021, May 29–31, 2021, Shanghai, China Xiaoxi He and Xinsheng Ke

Table 1: Papers related to the application of recommender system

Authors Research Contents

Ren Yonggong, Shi Jiaxin, et al. [16] They proposed the algorithm of mining the feature relationship between items and
building the relationship attribute fusion collaborative filtering between items, which
solved the problem of cold start in recommendation system, and improved the
accuracy and diversity of item recommendation.
Zhao Nan and PI Wenchao [17] They proposed a video recommendation system for multi-dimensional feature
analysis and filtering, which can effectively solve the problem of large data set size
and high sparsity.
Xia pila, Li Yanmin et al. [18] This paper systematically introduced the concept, functions, application fields and
frontier topics of recommender system, that is, how to integrate various types of user
feedback and user preference information to build a more reliable recommender
system.
Wu Xiyu [19] He proposed a collaborative filtering recommendation system based on knowledge
graph representation learning, which replaces neighborhood set in collaborative
filtering recommendation system and semantic neighborhood set in knowledge graph
representation learning, and improves recommendation performance.

knowledge map can not only alleviate the data sparsity and cold DBpedia knowledge graph to find recommendation entities in the
start, but also improve the interpretation performance and enrich knowledge graph through propagation activation algorithm.
the recommendation results, which is of great significance. The core idea of recommendation technology based on open
Most recommendation models based on knowledge spectrum are linked data is to apply the rich semantic information in the data
based on existing recommendation models. Through the knowledge layer of knowledge graph to the existing recommendation methods,
graphing project, the structured knowledge of users and other enti- which is used to calculate the user interest and the similarity be-
ties can be described more precisely, and more in-depth information tween the contents, and give the recommendation results. Lu et al.
can be obtained through knowledge reasoning and added to the [23] built a world-class travel knowledge graph based on DBpedia
recommendation model, thus enriching the diversity of recommen- and wikidata, so as to describe tourist attractions in a finer seman-
dation results. At present, the recommendation algorithm combined tic granularity, and improve the recommendation effect for tourist
with knowledge graph has made progress in the fields of automatic attractions. However, the semantic annotation of tourist attractions
question answering system and personalized recommendation. using strings is not ideal. Wenige et al. [24] applied the recommen-
dation method based on open link data to information retrieval and
4.2 Related Applications recommendation in digital library, using flexible similarity calcu-
In recent years, the combination of knowledge spectrum and rec- lation method and constraint based recommendation method to
ommendation system has achieved good results in academia and enhance the existing search function in the system. Based on the
industry. The original business purpose of knowledge scope is to open link data, chicaiza et al. [25] combined the characteristics of
enhance Google’s search results, but with the deepening of re- each user (such as user identity) to make recommendations, so as to
search, people find that it is also very useful in intelligent Q & A. get more suitable results for users. Because the construction of open
Knowledge map can re clean and combine the data from different link data is less difficult, it is often constructed in an automated
sources and components. The new data is just the most directly way, and the construction cost is low, so it is widely used.
related data of the recommendation system, and then make top-N Huang Wenbin and Zhang Weikai [26] aimed at the research
recommendation for the target users. in the field of recommender system, constructed the knowledge
Passant [20] established a music recommendation model by cal- graph based on CO citation analysis. Using the web of science
culating the semantic distance between entities in the knowledge database as the data source, this paper extracts the data to study
graph. Hassan [21] studied the hybrid recommendation system the recommender system, generates the author co citation matrix
based on ontology and collaborative filtering. Firstly, for two given and transforms it into Pearson correlation coefficient matrix, then
entities, it uses ontology based semantic similarity reasoning to carries out factor analysis, cluster analysis and multi-dimensional
determine the association network between the targets through the scale analysis, and constructs the knowledge graph of recommender
relationship between the entities, and calculates the similarity of system research field. The analysis shows that the research of rec-
different entities according to the weight; secondly, it calculates the ommender system is in a period of rapid development, and the
similarity between the contents by combining with collaborative number and scope of related scholars are expanding. The recom-
filtering algorithm; finally, it recommends the similarity to the users mender system based on collaborative filtering is the most core
by comprehensively considering the similarity between the entities research content. Personalized recommendation, content-based rec-
and the content N contents with the highest similarity. Cheekula ommender system and data mining based recommender system are
et al. [22] proposed to use the hierarchical category information in the research hotspots in this field.

107
Research Summary of Recommendation System Based on Knowledge Graph BDE 2021, May 29–31, 2021, Shanghai, China

Yao Sheng[27] mainly studied the technology of introducing ignoring the description of features from a more generalized di-
knowledge graphing and deep learning into the recommendation mension, which makes the description of user characteristics too
model. By introducing a text summarization algorithm based on "accurate". To a certain extent, the generalization ability of the
deep learning, he used the coding network in the algorithm to ex- recommendation method is not good, and it can only give recom-
tract the information in the text, decoded the network to obtain mendation results for some specific data, and the migration ability
the keyword sentence information of the text, and proposed a col- is weak, there are some limitations in application.
laborative filtering recommendation model based on deep learning
text summarization algorithm. By introducing text information 5 CONCLUSION
into collaborative filtering algorithm, the problem of data sparsity This paper starts with the research status of knowledge graphing,
is alleviated. On this basis, a new knowledge graph representa- summarizes and expounds the concept, construction and applica-
tion learning algorithm is introduced into collaborative filtering tion of knowledge graphing. On this basis, it summarizes the re-
algorithm to extract structured information from knowledge graph. search status of recommender system at home and abroad, including
Zhou Jing, Sun Ximin and others [28] studied the comprehensive recommendation technology, collaborative filtering recommenda-
clustering algorithm, SVD decomposition algorithm, collaborative tion system, hybrid recommendation system, and recommendation
concern recommendation system based on products, collabora- application. Finally, it summarizes and analyzes the research re-
tive concern recommendation system based on users and product sults of the application of the recommendation system under the
similarity algorithm, and proposed an intelligent recommendation knowledge graph.
system based on knowledge map data application combined with Due to knowledge graph can only play a better recommendation
recommendation strategy. It is proved that the data application effect when it reaches a certain scale, there are still many problems
based on knowledge graphing can be implemented and deployed, to be solved before using knowledge graph for recommendation.
and can fully meet the needs of users to obtain information on There are still many problems to be solved. Typical problems include
the Internet platform. Chaizhiwei and Xie min [29] generated a multi-source information fusion, user feature mining and potential
knowledge system represented by RDF triples based on the domain demand analysis in knowledge graphing, multi-dimensional recom-
knowledge description file. The knowledge graph generated by mendation, spatiotemporal reasoning, user privacy protection and
the invention not only retains the semantic and logical reasoning so on. In addition, in recent years, deep learning, representation
ability of the triples, but also can make use of the efficient graph learning, location awareness and other technologies also provide
computing ability of the attribute graph database to lay a solid many possible improvement directions for the recommendation
foundation for the further application of the knowledge graph. system based on knowledge graph, which is worthy of further
Knowledge graphing technology is the research hotspot of li- research.
brary science. According to this, Wu Rong and Duan Hongtao [30]
studied the library recommendation system based on knowledge ACKNOWLEDGMENTS
graphing, and compared it with the traditional collaborative fil- The design of this paper is successfully completed under the careful
tering algorithm based on ALS. The results show that the MKR guidance and strict requirements of Professor Ke Xinsheng. From
recommendation system based on knowledge graph can achieve the topic selection to the specific writing process, the first draft
good book recommendation effect and help the construction of and final draft of the paper are inseparable from the hard work and
personalized library recommendation system in the future. sweat of Professor. Here, I would like to express my deep gratitude
Dong Shengqi [31] proposed a recommendation method based and high respect to the professor.
on knowledge graph, which integrates movie knowledge into tradi-
tional recommendation system. By analyzing movie entities and REFERENCES
the relationship between entities, it completes the construction of [1] Jiao Xiaojing, Wang Lancheng. Concept Discrimination and Discipline Orienta-
movie knowledge graph. The knowledge graph based on Shanxi tion of Knowledge graph [J]. Library and Information Work, 2015 (15): 5-11
[2] Huang Hengqi, Yu Juan, Liaoxiao, et al. A review of Knowledge graph Research
Mobile IPTV is constructed by graph database. The experimental [J]. Computer System Application, 2019, V.28 (06): 3-14
results show that the knowledge graph driven recommendation [3] Qin Changjiang River. Theory and Practice of Knowledge graph Construction
method improves the recall rate, accuracy and coverage rate, which [M]. Intellectual Property Press, 2010
[4] Duan Hong. Summary of Knowledge graphing Technology [J]. Computer Re-
provides method support and technical support for IPTV personal- search and Development, 2013 (03)
ized recommendation, effectively improves the user reach rate and [5] Yuan Xuping. Construction of Business Domain Knowledge graphing Based on
Deep Learning [D]. East China Normal University, 2015
accuracy rate, and improves the IPTV operation effect. [6] Li Tao, Wang sichen, Li Huakang. Development and Construction of Knowledge
graphing [J]. Journal of Nanjing University of Technology, 2017, 41 (001): 22-34
[7] Wang Yongchao, Luo Shengwen, Yang Yingbao, et al. Overview of Knowledge
4.3 Shortcomings of Existing Methods graphing Visualization [J]. Journal of Computer Aided Design and Graphics,
2019 (10)
With the deepening of the research on the recommendation method [8] Chai Bo, Zhang Qiang, Song Bochuan, et al. A Knowledge graph Construction,
based on knowledge map, many scholars have done a lot of ex- Retrieval Method and Device: CN111309868a [P]. 2020
[9] Bordes A, Usunier N, Thanapalasingam, et al. Translating Embeddings for Mod-
ploration on it, which provides a new idea for the implementa- eling Multi- Relational Data, 2013[C]
tion of recommendation technology, and the proposed method has [10] Al Nazer, Kapanipathi P, Doran D, et al. Entity Recommendations Using Hierar-
achieved quite good results in some specific application scenar- chical Knowledge Bases[J]. 2015
[11] Kethavarapu, Al- Nazer A, Al- Bukhitan S, et al. Health, Food and User’s Profile
ios. However, most of the existing methods focus on the use of Ontologies for Personalized Information retrieval[J]. Procedia Computer Science,
knowledge mapping to describe user characteristics more precisely, 2015, 52:1071-1076

108
BDE 2021, May 29–31, 2021, Shanghai, China Xiaoxi He and Xinsheng Ke

[12] Sun Guangfu, Wu Le, Liu Qi, et al. Collaborative Filtering Recommendation [21] Bordes A, Hassan, Collobert R, et al. Learning Structured Embeddings of Knowl-
Algorithm Based on Temporal Behavior [J]. Acta Sinica Sinica, 2013 (11): 2721- edge Bases: Aaai Conference on Artificial Intelligence, 2012b[C]
2733 [22] Cheekula. FoodWiki: a Mobile App Examines Side Effects of Food Additives Via
[13] Chen Bi Yi, Huang Ling, Wang Chang Dong, et al. Collaborative Filtering Rec- Semantic Web[J]. Journal of medical Systems, 2015, 40(2):41
ommendation Algorithm Based on Explicit Feedback and Implicit Feedback [J]. [23] Pan JZ, Lu. RDFS(FA): Connecting RDF(S) and OWL DL. IEEE Trans-
Acta Sinica Sinica, 2020, 000 (003): 794-805 actions on Knowledge and Data Engineering, 2007, 19(2): 192–206. [doi:
[14] Shen Peng, Li Tao. Application of hybrid collaborative filtering algorithm in 10.1109/TKDE.2007.37]
Recommendation System [J]. Computer technology and development, 2019, 029 [24] Wenige . WordNet: An Electronic Lexical Database. Cambridge: MIT Press, 2018
(003): 69-71,76 [25] Chicaiza, Liu XR, Li N, et al. Using Content to Evaluate Semantic Similarity on
[15] Wang Chuanlong, Shao Yabin. Hybrid Recommendation System based on neigh- HowNet. Proceedings of 2012 Eighth International Conference on Computational
bor propagation clustering [J]. Journal of Xihua University: Natural Science Intelligence and Security. Guangzhou, China. 2013
Edition, 2020 (2): 1-7 [26] Huang Wenbin, Zhang Weikai, Xu Yang. Knowledge graphing of Recommender
[16] Ren Yonggong, Shi Jiaxin, Zhang Zhipeng. Item cold start Recommendation System Research Based on CO citation analysis [J]. Modern intelligence, 2015,
Algorithm Based on Relation Mining and Collaborative Filtering [J]. Pattern 35 (11): 5-14 + 19
Recognition and Artificial Intelligence, 2020 (1): 75-85 [27] Yao Sheng. Design and Implementation of Recommendation System Based on
[17] Zhao Nan, PI Wenchao, Xu Changqiao. A video Recommendation Algorithm for Knowledge graph Representation Learning [D]. Southeast University
Multidimensional Feature Analysis and Filtering [J]. Computer Science, 2020, [28] Zhou Jing, sun Ximin, Yu Xiaokun, et al. Knowledge graph and Data Application
047 (004): 103-107 - Intelligent Recommendation [j]. Telecommunication Science, 2019, V.35 (08):
[18] Luo Kaqi, Xia pila, et al. Recommender System: Technology, Evaluation and 171-178
Efficient Algorithm [M]. China Machine Press, 2015 [29] Chai Zhiwei, Xie min. An Automatic Construction Method and System of Knowl-
[19] Wu Xiyu, Chen Qimai, Liu Hai, et al. Collaborative Filtering Recommendation edge graph Based on Description Fefinition: cn110083709a [P]. 2019. (in Chinese)
Algorithm Based on Knowledge graph Representation Learning [J]. Computer [30] Wu Rong, Duan Hongtao. Research on the Application of Knowledge graph in
Engineering, 2018, 044 (002): 226-232 Library Recommendation System [J]. Automation Application, 2020 (08): 74-76 +
[20] Passant, Xue G, Yang Q, et al. Transferring Naive Bayes Classifiers for Text 79
Classification, 2010[C] [31] Dong Shengqi. Research on Recommendation method and Application Driven by
Knowledge graph [J]. Shanxi Science and Technology, 2020, 035 (002): 53-57,63

109