ArticleRank

Overview

ArticleRank has been derived from PageRank to measure the influence of journal articles.

• J. Li, P. Willett, ArticleRank: a PageRank-based Alternative to Numbers of Citations for Analysing Citation Networks (2009)

Concepts

ArticleRank

Like links between webpages, citations between articles (e.g., books or reports) indicate authority and quality. It is generally assumed that the more citations an article receives, the greater its perceived impact within its research domain.

However, not all articles are equally important. Hence, this approach based on PageRank was proposed to rank articles.

ArticleRank retains the basic PageRank methodology while making some modifications. When an article passes its rank among its forward links, it does not divide the rank equally by the out-degree of that article, but by the sum of the out-degree of that article and the average out-degree of all articles. The rank of article u after one iteration is:

where Bu is the backlink set of u, d is the damping factor. This change in the denominator reduces the bias that makes articles with few out-links seem to contribute more to their forward links.

Considerations

The implementation uses a normalized variant where scores sum to 1. The base rank (1 - d) is replaced by (1 - d) / n, where n is the total number of nodes. Additionally, ranks from dangling nodes (nodes with no outgoing edges) are redistributed equally to all nodes.

In comparison with WWW, some features have to be considered for citation networks, such as:

• An article cannot cite itself, i.e., there is no self-loop in the network.
• Mutual citations are not allowed; an article cannot be both a forward link and a backlink at the same time.
• Citations in a published article are fixed, meaning its forward links remain static.

Example Graph

GQL
INSERT (book1:book {_id: "book1"}), (book2:book {_id: "book2"}),
       (book3:book {_id: "book3"}), (book4:book {_id: "book4"}),
       (book5:book {_id: "book5"}), (book6:book {_id: "book6"}),
       (book7:book {_id: "book7"}),
       (book1)-[:cite]->(book4), (book1)-[:cite]->(book5),
       (book2)-[:cite]->(book4), (book3)-[:cite]->(book4),
       (book4)-[:cite]->(book5), (book4)-[:cite]->(book6)

Parameters

Run Mode

Returns:

ArticleRank for all nodes:

GQL
CALL algo.articlerank({
  damping: 0.8,
  maxIterations: 50,
  order: "desc"
}) YIELD nodeId, score

Result:

Stream Mode

Returns the same columns as run mode, streamed for memory efficiency.

GQL
CALL algo.articlerank.stream({
  order: "desc",
  limit: 3
}) YIELD nodeId, score
RETURN nodeId, score

Result:

Stats Mode

Returns:

GQL
CALL algo.articlerank.stats() YIELD nodeCount, minScore, maxScore, avgScore

Result:

Write Mode

Computes results and writes them back to node properties. The write configuration is passed as a second argument map.

Write parameters:

Writable columns:

Returns:

GQL
CALL algo.articlerank.write({damping: 0.85}, {
  db: {
    property: "ar_score"
  }
}) YIELD task_id, nodesWritten, computeTimeMs, writeTimeMs