Literature discovery and visualisation through citation networks
When anyone is going to start any kind of research or simply want to make an analysis to find out how a particular discipline, theory or idea is doing, one of the first steps is always focused on the search for information. Most of us have a natural tendency to use citation indexes, which allow us to know the impact that a previous work has had on the scientific community. However, nowadays there are many other tools that can be very useful for the discovery of new scientific literature.
Among these tools are those that allow specific searches by author, topic or field of research, as well as those that use data analysis techniques to identify patterns and visualise information in a clearer and more intuitive way. These applications have proven to be very practical for researchers, as they allow them to access a large amount of information quickly and efficiently, discovering relevant documents much more easily than using citation indexes.
In addition, most of them have explored a territory hitherto ignored by the big players in the industry, who continue to be anchored in presenting the results of queries in long, tedious and unattractive lists of authors and/or titles: the visualisation of information. These tools, by presenting the relationships between documents in a visual and interactive way, facilitate the rapid understanding of the fields of study, thus favouring the identification of connections between documents that would otherwise be much more difficult to find.
Obviously, all this has two further advantages: it saves time, as the process of searching and analysing related documents is much faster, and it encourages collaboration by facilitating the generation of new ideas and approaches based on evidence: invisible collaboration, which is that which occurs between researchers who, without knowing each other or working together, collaborate, through their research, in continuous advances within a common area of knowledge.
In this entry I do not intend to make an exhaustive analysis of all of them. In fact, it is possible that I will not include all of them either. What I intend is that interested people (mainly my students) have a starting point from which to locate tools, presented in a unified way, that allow them to advance in their research in the most satisfactory way possible and, if possible, saving time. At the same time, I might be able to get them to stop using wikipedia as a fundamental source of information 🙂
Connected Papers
This is a free tool that allows you to both discover and explore related scientific articles. These relationships are provided visually, which makes it much easier to understand what you see. After an initial query, by means of the title of a publication, its DOI, its arXiv identifier or the URLs of Semantic Scholar or Pubmed, it generates an interactive graphic that shows the relationship between the documents based on the citations and bibliographic references.
In this graph, which is nothing more than a set of nodes connected by arcs representing their relationships, the articles are ordered according to similarity, which is the first fundamental element in this application. Otherwise, only documents that are directly related to each other by means of citations would be displayed. However, by applying this method it is possible to relate documents that are not directly cited to each other but are thematically connected. For this, co-citation and bibliographic coupling are used as similarity metrics (the more common references X documents have, the more likely it is that they deal with the same thing, so the distance between them in the graph should be smaller).
In my opinion, the second important element of this tool is the management of the timeliness of the information. In order to have precise information about how up-to-date the publications in the graph are, different colour codings are used to specify which articles are relatively up-to-date in relation to the document that initiated the search.
URL: https://www.connectedpapers.com/
Litmaps
It is possibly one of the free tools that currently offers the most options. In fact, there are so many that they are beyond the scope of this first approach. Once registered, it allows:
- Seed: a map of relationships between papers is generated based on the citations and the citations of these citations, in which the article used as the origin of this map is named like this option: seed. It serves to locate the article originating from this graph within the overview of the discipline in question. Within the graph, the nodes that appear to the left of the seed are those cited in the article, while those that appear to the right are those that cite the article. The size of the nodes may depend on the number of citations received, the number of references to the article, the relevance on the map or the “momentum” (citation count adjusted to the date of publication).
- Discover: allows you to enter one or more documents as a reference. From the analysis of this information, it generates a map that will be surrounded by one or two outer circles. The nodes in the first outer circle correspond to articles that can be inferred to complement those previously referenced on the basis of immediate citations only. The second circle, which can only be accessed in the paid version, will show the nodes of the suggestions from the analysis of the citations of the citations.
- Map: if several articles have been highlighted in the previous phase, in this mode it will show all the nodes of those articles related to each other, in the form of a citation map. This map can also be created from the bibliographic references stored in any bibliographic reference manager after importing records.
Iris.ai
Two types of analysis can be carried out using this tool:
- Exploring: the search starts with a URL that is used as a sample document. From it, the relationship with other resources is shown, but instead of showing this relationship by means of a graph, this time thematic clusters are used.
- Focusing: requires the introduction of a series of terms, by way of a title, a brief description of the central problem of the study and a collection of bookmarks previously stored in the system. This is followed by a series of questions to refine the dataset to work with and, finally, a map of related and proposed documents is generated.
The main shortcomings of this tool:
- Possibly the use of clusters is not the most successful visualisation option to show relationships between documents.
- The usability of the system leaves much to be desired.
- It is a paid software. It only allows a 10-day trial.
- At no point is it clear which repositories are used or which can be filtered.
URL: https://iris.ai/
Open Knowledge Maps
Created by the organisation of the same name under the auspices of the European Commission. The tool generates knowledge maps based primarily on the input of keywords.
The maps are created from the 100 most relevant documents for the search terms. The relevance is obtained from criteria based on similarity. Once this is done, a knowledge map is created based on the metadata (using the BASE search engine for this). The papers that have more words in common are grouped, if the search has been done on BASE, forming nodes that use the metaphor of size to indicate those that have more associated documents. If the search has been done on PubMed, citations are used as an element to define the size of the nodes.
URL: https://openknowledgemaps.org/
ResearchRabbit
Possibly one of the most successful tools in the boom in artificial intelligence-based applications. The search can be initiated from a title, DOI, Pubmed identifier or from a Zotero collection (by the way, it has a very good integration with this manager) or from a reference in bibtex or RIS formats. From there, the magic begins: list of similar works from the analysis of the words of the initial document and the authors, analysis of citations received, exploration of other works by the same authors, list of suggested authors… Possibly the best of the free tools currently available.
URL: https://www.researchrabbit.ai/
Inciteful
It presents two search modes. On the one hand, it shows a network generated from the analysis of a single document, which will generate a map with those that are most relevant according to the citations received and, on the other hand, it shows the connection that exists between two different documents. Like ResearchRabbit, it can be connected to a Zotero collection via a plugin.
Carrot2
The initial search is done by keyword or on PubMed documents and, what is most striking, is that it is the only one of the tools described here that allows the search to be limited by language and/or country. It does not stand out for being a tool for discovering new documents, but rather for being an application that allows thematic clusters to be made on the basis of an article or a set of documents retrieved from a keyword search. In this sense, it has more limitations as an exploratory tool for a discipline.
URL: https://search.carrot2.org/
