IBM Research AI
Paper Graph
Welcome to PaperGraph

Papergraph is an online visual tool to understand the latest litterature in a given research community.

Upon searching a title or keyword, papergraph will search through Arxiv to find a selection of relevant papers.
For each paper, authors, citations and affiliations are extracted then connected to their respective papers.

This creates a network of papers, authors and affiliations, with node sizes depending on in-degree.

The most central authors are the most relevant to a given topic, as are commonly cited papers or affiliations.


You may search for full paper titles, keywords, partial or full author names and organizations.

Clicking on "show filters" next to Search will reveal additional search functionalities:

  • Date: Choose how far back in time your search goes, you may select one week, one month, one year etc.
  • Graph size: Select how many source papers will be used to generate the graph. WARNING: large networks may impact performance.
  • Affiliations: Restricts the search to a specific organization.
  • Categories: All papers are extracted from Arxiv's Computer Science categories, you may choose one or multiple categories to limit your search.

Advanced Search:

  • + : search term A AND term B, artificial+intelligence
  • | : search term A OR term B, intelligence|wisdom
  • Parentheses: Force an alternate order of operations, artificial+(intelligence|wisdom)
  • Double quotes: Allow inclusion of these symbols in a string: "machine+learning"

Viewing Modes

Network view: Displays networks of connected entities and trimming options. Full lists for papers, authors, citations and affiliations are available on the left, selecting an entity will show it's information on the right. Zoom by scrolling, move with click drag, double click to zoom out.

List view: Displays top lists of entities for the current search, selecting a colored entity will take you back to the network view. Lists of keywords are also generated based on paper titles, selecting one will launch a new search.

Flow view: Displays all available papers in the dataset, label sizes depend on the importance of a given paper, clicking a label will bring you to the network view and select the given paper.

Filtering in network view

Clicking one of the buttons above the network view will filter out irrelevant nodes, generate new connections and create a new network:

Show All
Shows full graph with all papers, authors and affiliations.

Only shows authors, authors with a common article are linked together.
X and Y co-author papers A and E together. Y and Z co-author papers A and C.

papers, papers that cite one another are connected.
A and B are co-cited by C and D. A and D are co-cited by E.

Only shows papers, two papers with a common citation are connected.
Only shows affiliations, connects affiliations with common papers.

Selection in network

When viewing a network, clicking on any node will reveal information about that node. For authors, their publication timeline is shown as well as their current organization. For papers, full author lists, abtracts and links are available. For citations, only titles and authors can be seen. For affiliations, a listed of connected authors are provided.

Questions and more information


Steven Ross

Owen Cornec

Mauro Martino


Top keywords:
Top 10 Papers:
Generated from citation scores of associated authors
Top 10 Authors:
Ranked by total number of citations
Top 10 Organizations:
Ranked by citation scores of associated authors
Time span:
Result size: