Cite4Me is a Web application that leverages Semantic Web technologies to provide a new perspective on search and retrieval of bibliographical data. The Web application presented in this work focuses on: (i) semantic recommendation of papers; (ii) novel semantic search & retrieval of papers; (iii) data interlinking of bibliographical data with related data sources from LOD; (iv) innovative user interface design; and (v) sentiment analysis of extracted paper citations. Finally, as this work also targets some educational aspects, our application provides an in-depth analysis of the data that guides a user on his research field.
The huge amount of Web data and resources, particularly in the academic area, calls for strategies to analyse and explore resources and data.
While scientific disciplines are very data- and knowledge-intensive, the lack of semantic tools hampers information management and decision making. This includes scientific data as well as unstructured academic publications as one of the key outcome of scientific work. This is due to information access o ered by digital library providers such as ACM Digital Library1 and Elsevier2 being mostly based on free text search and hierarchical classification3.
Thus, we present a novel Web application for exploratory search, retrieval and
visualization of scientific publications. Cite4Me aims at providing a single access point for
accessing papers and, therefore, assisting searchers on finding relevant topics, papers,
and unveiling new nomenclature more e ciently. For this, we use reference datasets
such as DBpedia4 to explore semantic relationships between scientific papers and user
queries. We also perform a topic coverage analysis to provide an overview of di erent
bibliographic datasets. Cite4Me is a Web application which exploits results of previous
research works [
Cite4Me implements semantic and co-occurrence-based methods to search and retrieve academic papers and suggest related work in a user-friendly interface that assists users in exploring relationships between authors, institutions, papers and query terms. Due to space restrictions, we present in this paper the most relevant features of Cite4Me to the Semantic Web field. 2.1
Search and Retrieval Cite4Me implements standard techniques, such as free text search, to search and retrieve scientific publications. In this section, we emphasize the semantic and exploratory search mechanisms.
Exploratory Search. The exploratory search or graph search component assists users to discover related work, people and institutions that are working on a specific topic. A crucial step to provide this type of search is the annotation of the publications’ content. For this, we used DBpedia Spotlight API5 for extracting entities, entity types and their categories. For instance, the categories of the extracted concepts are used to interlink publications through the topics they cover. In cases where two publications share the same category (dcterms:subject property), then a link between both publications is created. Figure 1 shows an example of topically related publications.
Semantic Search. The semantic search component of Cite4Me is similar to the explicit
semantic analysis (ESA) technique [
in the user query terms and the publications’ content are computed and ranked. The relatedness score is computed based on the tf-idf score for the entities found in the publications’ content. The ranking of the retrieved documents is based on the sum of the tf-idf scores of the matching concepts.
Figure 2 illustrates the semantic search functionality. Alongside the results of the semantic search a tag cloud shows the most prominent terms for a given user query. The tag cloud is updated while browsing through the list of results. The tags are selected based on the tf-idf score for the entities found in the abstract of the retrieved papers.
Paper recommendation. Another important feature of Cite4Me and which di erentiates it from similar tools is the semantic paper recommendation. Given a scientific publication, the tool recommends a related paper based on a score calculated according to direct and lateral relationships between the publication of interest and the remaining papers in our corpus.
To compute the relatedness score, we rely on previous work by Nunes et al. [
After computing the relatedness scored between enriched concepts, the paper
recommendation relies on an aggregated measure that takes into account the relatedness
inter-documents. Finally, we generate a ranked list of pairwise publications according
to the overall score (see [
Currently, Cite4Me is linked to a dataset (LAK Dataset6) which contains semistructured research publications from the ACM Digital Library (under a special license)
and other public datasets (see also [
This paper presented the application of previous works in the Semantic Web field within Cite4Me, a Web application that assists users in finding relevant scientific papers by exploring semantic relationships between them. For more information about the Cite4Me Web application please refer to http://www.cite4me.com.