Scienti c conferences play a crucial role in the eld of Computer Science by promoting the cross-pollination of ideas and technologies, fostering new collaborations, shaping scienti c communities, and connecting research e orts from academia and industry. However, current systems for analysing research data do not provide a good representation of conferences. Speci cally, these solutions do not allow to track research trends, to compare conferences in similar elds, and to analyse the involvement of industrial sectors. In order to address these limitations, we developed the AIDA Dashboard, a tool for exploring and making sense of scienti c conferences which integrates statistical analysis, semantic technologies, and visual analytics.
Scienti c conferences play a crucial role in the eld of Computer Science by promoting the cross-pollination of ideas and technologies, fostering new collaborations, shaping scienti c communities, and connecting research e orts from academia and industry. For this reason, every signi cant eld is usually associated with multiple conferences that help de ning its challenges and paradigms and to coordinate the e ort of all the interested stakeholders.
Therefore, understanding and monitoring Computer Science conferences is an important task for editors, researchers, companies, research policy makers and other users working in this space. Several applications and services already provide a wide variety of functionalities to support the exploration of research data and produce various kinds of analytics. These include Microsoft Academic Graph, Semantic Scholar, Scopus, Web of Science, OpenCitations, and many others. However, these systems tend to neglect conferences and o er only a very Copyright c 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). limited set of relevant analytics, such as the number of papers or citations. In the rst instance, they do not allow users to examine the trends of the relevant research topics. It is thus di cult to assess what are the research challenges that a conference is actually addressing and how its focus changed in time. Secondly, there is poor support in comparing conferences to determine the best performing ones in speci c elds. For instance, we would like to know which are the main conferences in Semantic Web, how they compare in terms of average citations or other metrics, and how their performance changes in the last few years. A third limitation is that current systems do not report any analytics about the industry involvement. Conversely, it can be argued that conferences are the premium public venues in which industry and academia interact, hence monitoring these dynamics is critical for assessing a conference.
In order to address these limitations, we developed the AIDA Dashboard,
a tool for exploring and making sense of scienti c conferences which integrates
statistical analysis, semantic technologies, and visual analytics. The AIDA
Dashboard was developed in collaboration with Springer Nature for assisting editors
in assessing conferences, but it also supports several other use cases. It introduces
three novel features that state-of-the-art systems are currently lacking. First, it
associates to conferences a very granular representation of their topics from the
Computer Science Ontology (CSO)[
The AIDA Dashboard is a web application that allows users to visualize several kind of analytics about a speci c conference (see Figure 1). The backend is developed in Python, while the frontend is in HTML5 and Javascript.
The AIDA Dashboard builds on the Academia/Industry DynAmics [
support AIDA. AIDA was generated by integrating several knowledge graphs and bibliographic corpora, including Microsoft Academic Graph (MAG), Dimensions, DBpedia, CSO, and the Global Research Identi er Database (GRID).
The research papers were annotated with CSO topics using the CSO
Classi er [
Each research article was also linked to the industrial sectors described in INDUSO by mapping the a liations of the authors to their DBpedia entities, which in turn are mapped to INDUSO. For instance, an article that was written by authors who have Toyota as a liation would be associated to the industrial sector Automotive.
AIDA is available at http://w3id.org/aida under the CC-BY 4.0 license.
It was recently used for supporting the generation of adavanced analytics about
research dynamics and forecasting the impact of research topics on industry [
Dashboard is the rst step in allowing users to access AIDA through a userfriendly but comprehensive interface.
In order to support the AIDA Dashboard, we pre-computed a full set of analytics for each conference from AIDA-KG and store it in a JSON le that will be loaded by the web interface. This solution allows AIDA Dashboard to be extremely scalable, since for a given conference it needs to query the server only once, to retrieve its associated le. Every other operation is handled by the front-end.
AIDA Dashboard is highly scalable and allows to browse the di erent facets of a conference according to seven tabs: Overview, Citation Analysis, Organizations, Authors, Topics, Similar Conferences, and Industry.
Figure 1 shows the Overview tab. This is the main view of a conference that provides introductory information about its performance, the main authors and organization, and the conference rank in its main elds in terms of average citations for paper during the last ve years.
The Citation Analysis tab reports the evolution in time of several citationbased metrics such as the impact factor and the average citations for paper. It also shows the evolution of the rank and the percentile of the conference in di erent elds. For instance, the Conference on Neural Information Processing Systems (NeurIPS) is currently the second conference in terms of average citations in Neural Network, the third in Machine Learning, and the twelfth in Arti cial Intelligence. This visualization is typically used by Springer Nature editors to assess the performance of conferences within di erent communities and to identify emerging conferences.
The Organizations and Authors tabs show several analytics about the main institutions and researchers active in the conference. Organizations can be ltered according to their type (academia or industry) and are associated with their number of publications, citations, and average citations for paper. The researchers are associated with similar analytics, but also with their H-index and H5-index, in order to quickly identify high impact researchers. Editors use this information to understand the quality of researchers and organizations attracted by the conferences. This is particularly important for assessing relatively young conferences that may not have developed yet a strong citation record.
The Topic tab allows users to analyse the topic trends in time. Speci cally it shows two selections of topics: frequent topics and ngerprint topics. The rst is the set of topics which appear more frequently in the conference. The second is the set of most distinctive topics of the conference. It is obtained by computing the di erence between the topic distribution of the conference and the one of the full dataset. Preliminary analyses revealed that this second set is usually able to better represent the topics considered central to the conference.
The Similar Conferences tab compares the conference under analysis with all the other conferences in the same elds according to their number of publications, citations, and average citations for paper. The user can contextualise the comparison to di erent elds. For example, ISWC can be compared with all the other conferences in the elds of Semantic Web, Internet, or Computer Science.
Finally, the Industry tab reports the percentage of publications and citations from academia, industry, and collaborative e orts as well as the industrial sectors analysis. The latter shows the percentage of produced publications and citations received by companies in di erent sectors. For instance, the main industrial sectors of ISWC are Computing and IT, Information Technology, Management, Telecommunication, and Health Care. 3
The current version of AIDA Dashboard already provides an array of interesting functionalities, many of which go beyond what is available in other current tools. Nevertheless, we are still at a relatively early stage and we are planning to introduce new ones. As rst step, we plan to add a geographical tab for analysing the distribution of countries active in a conference. We also want to expand the set of entities that could be analysed by the dashboard, producing similar analytics also for journals, organizations, and scienti c communities. Finally, we plan to perform a comprehensive user study with editors and researchers from di erent communities in order to assess the system and collect useful feedback. For such a purpose, to generalize the presented dashboard we only need to replace the CSO ontology with others within the domain under study. As such, we have already started working with the MeSH ontology within the bio-informatics domain to have our dashboard working in that domain as well.