Connected Papers

What are Connected Papers?

Connected Papers is a visual tool for discovering literature, designed for researchers, students, and applied scientists. Once a user enters the title of a known paper, the platform generates a network diagram showing related studies, which helps to quickly understand the key papers, research branches, and development trends in a particular field.

It is not an ordinary citation tree, nor does it present search results arranged by keywords. The distance and strength of connections between papers shown in the graph reflect their similarity, which is usually determined based on mutual citations and the fact of being cited together; therefore, even if two papers do not cite each other directly, they can still be quite similar to one another.

Starting with Origin Paper

Users can search for papers using the title, DOI, author, or other identifiers, and select the original paper as the starting point for the map. The system then generates a set of papers that are most relevant to it, with the original paper highlighted in the center.

Origin Paper can affect the entire map. If the starting point is outdated, the topic is too broad, or it does not fully correspond to the research question, the results will also be inaccurate.

Complex problems can be addressed by using multiple starting points or by creating multiple diagrams for cross-verification.

Similarity map

Each node represents a paper, and the stronger the connection between nodes, the more similar the relationship between the two papers is in the academic network. Papers that are similar to each other form clusters, which help users to identify different methods, subfields, or research approaches.

The position in the map is not an indicator of the direction of citation, nor is it a measure of the quality of a paper. Clustering is influenced by the coverage of the database and by the algorithms used; therefore, visual distance cannot be presented as an academic conclusion.

Node size and color

In the diagram, nodes are usually displayed along with the number of citations and the year of publication, allowing researchers to observe both impact and temporal evolution at the same time. Users can easily identify earlier, classic works as well as more recent studies.

Citation counts are influenced by the field of study, the year in which the work was published, and the scope of the databases available. A low number of citations for a new paper does not necessarily mean it has low value; when compiling reviews, it is important to avoid focusing only on papers with high citation counts.

Prior Works

Prior Works helps identify the foundational papers, the sources of various methods, and the theoretical background in a research field, by taking into account the earlier studies that are frequently cited together in the papers shown in the diagram. It is useful for adding context to the historical overview provided in reviews.

The fact that early papers are frequently cited together does not mean that their conclusions remain valid; it is necessary to examine subsequent attempts to replicate them, the corrections made, and any controversies surrounding them.

Derivative Works

Derivative Works identifies studies that cite multiple papers from the dataset, thereby helping to understand how this field continues to develop and which subsequent papers integrate elements from various subfields.

Subsequent references may be used for purposes such as use, comparison, or criticism; it is not possible to determine the degree of support solely by looking at the number of relationships.

Multi-origin Graphs

Multi-origin allows the use of multiple papers as starting points, and is suitable for identifying related work among them once a set of core documents has been gathered. Compared to a single Origin, it helps to reduce the influence of any individual paper’s biases.

When there are significant differences in the topics of the papers, the common similarity network may be weak. It is recommended to choose a starting point based on a clear research question.

List View

In addition to the map view, users can switch to list view to browse papers by title, author, year, citations, and similarity. The list view is more suitable for filtering papers one by one, while the map view is better for understanding the overall structure.

Filters for selection

The platform offers filtering options such as year, allowing users to focus on recent research or to review historical papers. The filtering is applied only to the current set of data in the database; it does not mean that all databases are searched again.

Save the paper

Logged-in users can save papers of interest to create a personal reading list, and from there they can access Semantic Scholar, Google Scholar, or the publisher’s website.

The save function serves as a simple way to keep items for later use; it does not replace the attachment management, citation tools, and duplicate removal features provided by full-reference managers such as Zotero and EndNote.

Graph History

The platform retains a history of the diagrams, allowing users to return to the research topics they have previously worked on. For researchers who need to explore multiple possible directions, this is more convenient than having to search from scratch each time.

Share the map

The diagram page can be shared with supervisors, collaborators, or students for discussing the structure of the field and recommending readings. Before sharing, it should be clear that the diagram represents an exploratory view generated by an algorithm, and not a complete summary of the relevant information.

Data foundation

Connected Papers relies on paper identifiers and data from the Semantic Scholar academic network to generate its results. Semantic Scholar covers a large volume of interdisciplinary literature, but the metadata, citation information, and availability of full texts vary across different fields.

If the paper cannot be found, you can check the title, DOI, and its status on Semantic Scholar. For newly published articles, it may be necessary to wait for the data to be updated.

A map is not a reference tree.

Connected Papers clearly states on its page that papers are arranged based on their similarity, rather than in a hierarchical structure determined by direct citation relationships; strong connections indicate a similar structure in terms of mutual citations.

To accurately determine who has cited whom, it is necessary to consult the list of citations and references in the paper, rather than relying solely on visual connections.

How to use it in a literature review

The common approach is to first locate a highly relevant paper using a specialized database, then create a similarity diagram, examine the Prior and Derivative Works, save the candidate papers, and finally return to the database to conduct a full-text screening.

It is suitable as a snowballing and discovery enhancement tool, but it cannot replace Boolean search queries, multi-database coverage, duplicate removal, or inclusion/exclusion criteria.

How to find review papers

The large nodes in the diagram, the papers that integrate multiple branches at a later stage, and the list of Prior Works may contain reviews or key summaries. Users should verify further based on the publication type and abstract, rather than relying solely on their position in the diagram.

How to understand the research context

By observing nodes and clusters over time, it is possible to see how the initial foundational papers evolved into different topics, as well as where subsequent researches converged again. It is suitable for use in classroom lectures and to introduce newcomers to a new field.

Price and version comparison

Free version

The free version of Connected Papers costs nothing; it allows for the creation of 5 diagrams per month, and includes all the essential features such as similarity maps, prior works, derivative works, multi-origin information, paper saving, and graph history.

The free version does not have any functional limitations; the main restriction is the number of new diagrams that can be created per month. For users who occasionally explore literature, 5 diagrams are usually sufficient.

Academic Premium

The Academic plan is intended for academic institutions, non-profit organizations, and individual users; it offers an unlimited number of maps while retaining all functions. It is suitable for students and researchers who need to create a large number of research maps on a regular basis each month.

Business Premium

Business is intended for commercial and industrial users involved in profit-driven activities, and it also offers unlimited graphs as well as all functions. Academic has the same functional capabilities as Business; the main differences lie in the areas of use and the types of licenses.

Premium price

The official website offers the option to switch between annual and quarterly payments, but the specific amounts shown on the public page are dynamically displayed by the payment processing system; the values obtained in real time are not yet consistent. When making a purchase, one should select either Academic or Business option on the official pricing page, and the amount, taxes, and currency rates indicated on the payment page should be followed.

Do not consider the prices from third-party reviews as the current official prices. Nor should the wrong licensing type be chosen for personal academic use or commercial use.

Group team solution

The Academic Group and Business Group allow organizations to provide Premium access on a per-seat basis. Administrators purchase and assign seats; by default, an administrator account does not receive Premium access unless it is also added as a user seat.

Group purchasing is suitable for laboratories, schools, and R&D departments in companies; it is necessary to keep track of any changes in the number of users, renewals, members who leave the group, and ownership of the resources involved.

Payment and auto-renewal

Officially, various payment methods are supported, including credit cards, debit cards, PayPal, Alipay, and bank transfers; payments are processed through PayPro Global. Subscriptions are automatically renewed by default, but can be canceled at any time.

The bill may show PayPro Global instead of Connected Papers. It is necessary to verify the renewal period, taxes, and refund terms before making a purchase.

Scholarship program

Researchers who truly cannot afford the Premium cost can contact the authorities to request a discount or free Premium. Approval and the duration depend on individual cases decided by the authorities, and it is not an automatic entitlement.

Official API

Connected Papers has made the early access API available, along with official Python and JavaScript clients. Developers can request graphs using Semantic Scholar ShaID, check their remaining usage quota, and access graphs that were created within the last 30 days for free.

An API Key must be obtained by contacting the official authorities. The fact that the client software is open source does not mean that the API services are available indefinitely at no cost; the actual usage limits for commercial purposes need to be confirmed according to the relevant licenses.

API schema caching

If the graph associated with a particular paper has been accessed within the past 30 days, calling it again generally will not be counted as new usage. Developers can check the list of Free Access Papers to avoid redundant creation of such graphs.

Tutorial for Connected Papers

Complete a basic task.

1. Clarify the issue, time frame, location, source priority, and output format;
2. Upload materials for which usage rights are available or enter search queries in Connected Papers;
3. First, establish a framework starting from Origin Paper, and then supplement the evidence with similarity maps.
4. It is necessary to distinguish between factual information from the source, the author’s opinions, and AI-generated conclusions.
5. Check each item for dates, numbers, the original location, and any conflicting evidence;
6. The conclusions are manually revised, the verification time is recorded, and then they are published;

Create reusable professional workflows

1. Break down complex topics into four categories of questions: background, data, comparison, and conclusions;
2. The approach starts with Origin Paper, and fixed research steps are followed, including the creation of similarity maps as well as the determination of node size and color.
3. Give priority to using the official website, research papers, regulatory documents, and raw data;
4. A second person is assigned to review conclusions that are considered high-risk;
5. Save queries, evidence, versions, and unresolved issues;
6. Re-run after the data changes and update the conclusions;

API rate limiting

• The official client documentation states that, by default, each API key allows for the creation of up to 5 images per minute.
• When the system is busy or exceeds its capacity, the message OVERLOADED is returned, and retry attempts are made using exponential backoff times of 5, 10, 20, and 40 seconds;
• If a higher speed is required, contact the authorities; do not attempt to bypass the restrictions simultaneously.
• The application should also handle statuses such as BADID, NOTINDB, and OUTOFREQUESTS;

Official GitHub and open-source status

GitHub’s official repository provides the connectedpapers-py and connectedpapers-js clients under the MIT license, along with some auxiliary or branch projects. Developers can view the client calls and type definitions.

The full Connected Papers website, the similarity algorithms, the backend for graph construction, and academic databases are not open source; therefore, it should be labeled as \"core platform is closed source, API client is open source.\"

Is the full text provided?

Connected Papers mainly provides tools for finding papers and external access points; it is not responsible for bypassing paywalls. The availability of PDFs and full texts depends on Semantic Scholar, open-access options, or subscriptions from publishers or institutions.

Which users is it suitable for?

• Students and researchers who have just entered a new field of study;
• Those who need to quickly identify key papers and research clusters;
• A review team that conducts forward and backward citation tracking;
• Teachers who prepare reading lists for courses;
• Researchers who need to find common documents across multiple seed articles;
• Developers who generate paper similarity maps through APIs.

Product advantages

• The map interface is intuitive, making it suitable for quickly obtaining an overview of a particular field.
• Similarity networks can identify related papers that are not directly cited;
• Prior and Derivative Works distinguish between historical and subsequent research;
• The free version includes all features;
• Multi-start maps can reduce the bias of individual seeds;
• The official provides open-source Python and JavaScript clients.

Restrictions and Precautions

• The free version allows only 5 images per month, while the price for the Premium version needs to be checked on the dynamic pricing page.
• The map relies on SemanticScholar data, and as a result it may miss new papers, non-English publications, and citation records;
• Similarity is not a relationship of quality, causation, or direct reference;
• A systematic review must be conducted using professional databases and a thorough screening process;

Frequently Asked Questions

Is Connected Papers free?

The free version allows for the creation of 5 images per month and includes all the core features.

What are the lines in the Connected Papers diagram?

It primarily indicates the similarity between papers and their common citation structures, rather than being a simple direct citation tree.

What is the difference between Academic and Business?

Both offer an unlimited number of diagrams and the same set of functions; Academic is intended for academic, non-profit, and personal use, while Business is designed for commercial purposes.

Can I use an API?

Yes, the official team provides early access to APIs as well as clients in Python and JavaScript; an API key must be requested from the official team.

Can it replace Google Scholar?

It’s not possible. It is more suitable for exploring similar networks based on existing papers; for a comprehensive search using keywords, Google Scholar and specialized databases are still required.

Is Connected Papers open source?

The entire platform is not open source, but the official Python and JavaScript API clients are made available under the MIT license.