=Paper=
{{Paper
|id=Vol-2180/paper-66
|storemode=property
|title=The Smart Book Recommender: An Ontology-Driven Application for Recommending Editorial
Products
|pdfUrl=https://ceur-ws.org/Vol-2180/paper-66.pdf
|volume=Vol-2180
|authors=Thiviyan Thanapalasingam,Francesco Osborne,Aliaksandr Birukou,Enrico Motta
|dblpUrl=https://dblp.org/rec/conf/semweb/Thanapalasingam18a
}}
==The Smart Book Recommender: An Ontology-Driven Application for Recommending Editorial
Products==
https://ceur-ws.org/Vol-2180/paper-66.pdf
The Smart Book Recommender: An Ontology-Driven
Application for Recommending Editorial Products
Thiviyan Thanapalasingam1, Francesco Osborne1, Aliaksandr Birukou2,
Enrico Motta1
1
Knowledge Media Institute, The Open University, MK7 6AA, Milton Keynes, UK
{ thiviyan.thanapalasingam, francesco.osborne, enrico.motta }@open.ac.uk
2
Springer-Verlag GmbH, Tiergartenstrasse 17, 69121 Heidelberg, Germany
aliaksandr.birukou@springer.com
Abstract. Promoting books and journals to the relevant research communities is
an important task for major academic publishers. Unfortunately, identifying
which are the best editorial products to market at a certain academic venue is a
time-consuming and error-prone process. Here we present the Smart Book
Recommender (SBR), an ontology-based recommender that supports the
Springer Nature editorial team in selecting the editorial products to market at
specific venues. SBR provides an interactive visualisation for analysing the
topics characterizing conference series and books. It builds on a dataset of 27K
books, journals, and conference proceedings annotated with topics from the
Computer Science Ontology, a large-scale ontology of research areas. A user
study showed that SBR is able to produce useful recommendations for both
editors and researchers.
Keywords: Recommender System, Scholarly Data, Scholarly Ontologies, Data
Mining, Conference Proceedings, Metadata.
1 Introduction
Major academic publishers, such as Springer Nature, often attend academic conferences
and present their most recent books and journals relevant to the conference topics. The
purpose of such stands is twofold: the conference participants can browse the most
recent research and the authors get additional exposure for their work.
Typically, the selection of products is performed by publishing editors who possess
domain specific knowledge and years of experience in publishing books and journals
on the conference topics. The process usually requires browsing through a vast
catalogue of editorial products on multiple platforms 1,2 . It is a time-consuming and
error-prone practice since it is easy to miss some important publications. Furthermore,
the constant emergence of new research areas over time poses the challenge of keeping
up to date with the research dynamics.
The ongoing collaboration between Springer Nature (SN) and the Knowledge Media
Institute (KMi) of The Open University has given rise to a number of semantic
1
https://link.springer.com
2
https://www.springer.com
technologies to support publishing activities [1]. These solutions include the Smart
Topic Miner3, a tool for classifying proceedings according to a large-scale Computer
Science Ontology, and the Smart Topic API, a web service for assigning semantic
topics to research papers [2].
The most recent product of this collaboration is the Smart Book Recommender
(SBR) [3], an ontology-based recommender system designed for suggesting books,
journals, and proceedings for specific Computer Science (CS) venues. This demo paper
is complementary to the one accepted in the ISWC 2018 In-Use track [3] and focuses
on the main functionalities of the system. A demo of SBR is available at
http://rexplore.kmi.open.ac.uk/SBR_demo/.
Figure 1: SBR interface showing the search tools (top) and one of the recommended items for
the selected conference (bottom)
2 Smart Book Recommender
The Smart Book Recommender (SBR) is a web application that takes as input a
conference series (e.g., “International Semantic Web Conference”) and returns the
books, journals, and conference proceedings that are characterized by a set of similar
research topics (e.g., the "Handbook of Semantic Web Technologies”). The frontend
was implemented using HTML5 and JavaScript, while the backend was developed in
Python and queries a MariaDB database.
SBR builds on a dataset of 27K books, journals, and conference proceedings
published by Springer Nature that we annotated with topics from the Computer Science
Ontology (CSO) 4 using the Smart Topic API [2]. CSO is a large-scale ontology of
research topics which includes about 26K research topics and 226K relationships [4].
It was automatically generated using the Klink-2 algorithm [5] on a dataset of 16
million publications, mainly in the field of Computer Science. SBR produces
3
A demo of STM is available at http://rexplore.kmi.open.ac.uk/STM_demo/
4
https://w3id.org/cso/
recommendations for a specific conference series by suggesting the items associated
with a similar distribution of semantic topics. It does so by computing the cosine
similarity between the vectors of topics associated with the conference proceedings and
all the other editorial products. For a more comprehensive discussion of the SBR
implementation, please refer to [3].
2.1 User Interface
Figure 1 shows the interface of the SBR web application. The users can select a
conference series by using either the full conference name or its abbreviated form (e.g.,
“International Semantic Web Conference” or “ISWC”). After the conference series has
been selected, a list of recommended editorial products is loaded via an AJAX request.
The results are displayed as cards and sorted in descending order of similarity. Each
card summarises an editorial product using its title, publication year, fifteen most
significant topics, and the overall similarity score with the input conference. SBR also
provides hyperlinks to the corresponding Springer Link catalogue pages, where more
information can be found.
Figure 2: Subset of an interactive graph showing the similarities and dissimilarities in the topic
coverage of ISWC (2012 – 2016 proceedings) and Open Data for Education (book).
SBR allows users to filter recommendations according to i) publication types (books,
journals, or proceedings), ii) publication years, and iii) authors and editors of the books.
The publishing editors can also inspect the topic taxonomy of the suggested products,
rate the results, and export the list of recommendation, either as a JSON or CSV file.
This document is reviewed and sent to the Exhibit Department which takes care of
dispatching the publications to the conference.
Figure 2 shows the interactive visualisation offered by SBR for comparing the topics
of the input conference series with the ones of a recommended item. This interface
intends to provide an intuitive explanation for the similarity score by representing items
as taxonomies of topics. It displays in red and blue the topics that appear only in one of
the two items and in green the shared ones. The users can also change the granularity
of the representation by opting to show only topics that appear in a minimum number
of chapters.
2.2 User Study
We conducted a user study for evaluating SBR involving 7 SN publishing editors and
7 researchers5. These users rated as relevant 76.1% of the first ten recommended items
(i.e., precision@10). They also reported that SBR was “intuitive” and “easy to pick up”
and that its interface was “simple” and “well-organised”. In particular, editors and the
researchers yielded respectively an average SUS score6 of 77.1±15.2 and 80.3±11.3,
which converts in a percentile rank of about 75%.
The users also suggested some additional functionalities that we are working to
implement, such as the ability to change the topic-based representation of the input
conference by adding and removing significant research areas, and the ability to take
download statistics into consideration.
3 Conclusion
In this demo paper we presented our work on the Smart Book Recommender, an
ontology-driven recommendation system designed for assisting the Springer Nature
publishing editors in selecting books to be marketed at conferences. The semantic
representation of research conferences used by SBR was considered very helpful both
by editors and researchers, since it allows the users to understand why a certain item
was recommended and to compare different products.
As future work, we plan to cover other scientific domains, such as Engineering and
Life Science.
References
1. Osborne, F., Salatino, A., Birukou, A., Thanapalasingam, T., Motta, E.: Supporting
Springer Nature Editors by means of Semantic Technologies. In International Semantic
Web Conference 2017 Industry Track. (2017)
2. Osborne, F., Salatino, A., Birukou, A., Motta, E.: Automatic Classification of Springer
Nature Proceedings with Smart Topic Miner. In International Semantic Web Conference
2016 Application Track. (2016)
3. Thanapalasingam, T., Osborne, F., Birukou, A., Motta, E.: Ontology-Based
Recommendation of Editorial Products. In International Semantic Web Conference 2018
In-use Track. (2018). Pre-print available on https://goo.gl/osLBQz.
4. Salatino, A.A., Thanapalasingam, T., Mannocci, A., Osborne, F. and Motta, E.: The
Computer Science Ontology: A Large-Scale Taxonomy of Research Areas. In International
Semantic Web Conference 2018 Resource Track. (2018). Pre-print available on
https://goo.gl/oZ1eT4.
5. Osborne, F., Motta, E.: Klink-2: integrating multiple web sources to generate semantic
topic networks. The Semantic Web-ISWC 2015, pp. 408-424. Springer. (2015)
5
The study is reported in [3]. The relevant data are available at
https://figshare.com/articles/Smart_Book_Recommender_Evaluation_Data/6087032/2
6
https://www.usability.gov/how-to-and-tools/methods/system-usability-scale.html