LIS 601: Week8
Study notes derived from the Garfield (1980) Article

General Concepts

1. New scientific fields are constantly emerging and old ones evolving.
2. Standard indexing practices tend make it difficult to keep abreast of research trends.
3. Garfield's "Research Fronts", in the SciSearch databases, are meant to address the deficiencies in traditional indexing practice through the use of citation analysis. (specifically, co-citation strengths  are used to cluster similar works to determine possible trends and emerging fields.)

Methods and Reasonings

1. Files are analysed to find those articles that are cited many times:
1. A threshold number is decided (17- or more times) to eliminate most works because clustering is computationally very expensive.
2. Co-citation link strengths are taken for the works that passed the threshold
1. About 800, 000 pairs identified
3. Co-citation used over bibliographic coupling because co-citation is time variant
1. Two works that are not bibliographically coupled might be brought together over time by researchers. It is through this bringing together that emerging fields might be synthesized.
4. Another threshold is applied:
1. Co-citation strengths must exeed some value, say 17 % (see formula)
2. Goal is to keep clusters created a manageable size.
5. Pairs that pass the threshold then undergo a single-link clustering method:
1. Works that are linked at strengths above a certain value are brought together to form clusters
6. Titles of the works in each cluster are then analysed for frequently occurring terms
7. Significant terms from the titles used to describe the cluster.
1. Cluster titles thus reflect the terminology of the new Research Front
2. Sometimes wierd titles are created however.

Garfield's Co-citation Link Strength Formula

X =  # of co-citations of A and B
Y = total # of citations of A + total # citations of  B

Link Strength (AB) = X/(Y-X)

Note how the strength is normalized by bringing into account the total number of citations for both A and B.