CiteSeer Digital Library Research Articles

Bibliometric analysis of CiteSeer data for countries

This article describes the results of our analysis of the data from the CiteSeer digital library. First, we examined the data from the point of view of source top-level Internet domains from which the data were collected. Second, we measured country shares in publications indexed by CiteSeer and compared them to those based on mainstream bibliographic data from the Web of Science and Scopus. And third, we concentrated on analyzing publications and their citations aggregated by countries. This way, we generated rankings of the most influential countries in computer science using several non-recursive as well as recursive methods such as citation counts or PageRank. We conclude that even if East Asian countries are underrepresented in CiteSeer, its data may well be used along with other conventional bibliographic databases for comparing the computer science research productivity and performance of countries.

Mining citation information from CiteSeer data

The CiteSeer digital library is a useful source of bibliographic information. It allows for retrieving citations, co-authorships, addresses, and affiliations of authors and publications. In spite of this, it has been relatively rarely used for automated citation analyses. This article describes our findings after extensively mining from the CiteSeer data. We explored citations between authors and determined rankings of influential scientists using various evaluation methods including citation and in-degree counts, HITS, PageRank, and its variations based on both the citation and collaboration graphs. We compare the resulting rankings with lists of computer science award winners and find out that award recipients are almost always ranked high. We conclude that CiteSeer is a valuable, yet not fully appreciated, repository of citation data and is appropriate for testing novel bibliometric methods.

Learning author-topic models from text corpora

We propose an unsupervised learning technique for extracting information about authors and topics from large text collections. We model documents as if they were generated by a two-stage stochastic process. An author is represented by a probability distribution over topics, and each topic is represented as a probability distribution over words. The probability distribution over topics in a multi-author paper is a mixture of the distributions associated with the authors. The topic-word and author-topic distributions are learned from data in an unsupervised manner using a Markov chain Monte Carlo algorithm. We apply the methodology to three large text corpora: 150,000 abstracts from the CiteSeer digital library, 1740 papers from the Neural Information Processing Systems (NIPS) Conferences, and 121,000 emails from the Enron corporation. We discuss in detail the interpretation of the results discovered by the system including specific topic and author models, ranking of authors by topic and topics by author, parsing of abstracts by topics and authors, and detection of unusual papers by specific authors. Experiments based on perplexity scores for test documents and precision-recall for document retrieval are used to illustrate systematic differences between the proposed author-topic model and a number of alternatives. Extensions to the model, allowing for example, generalizations of the notion of an author, are also briefly discussed.