Similarity measures in scientometric research: The Jaccard index versus Salton's cosine formula

By L. Hamers, Yves Hemeryck, Guido Herweyers, Marc Janssen, Hans Keters, R. Rousseau, A. Vanhoutte. 1989

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Evaluation & Benchmarks

Citations in the corpus

SIGMa: simple greedy matching for aligning large knowledge bases (2012, methodology): “In order to handle the varying lengths of strings, we use the Jaccard similarity coefficient between the sets of words, a metric often used in information retrieval and other data mining fields [20, 14].”
Towards Robustifying NLI Models Against Lexical Dataset Biases (2020, methodology): “We ﬁrst sort the samples in the MNLI matched validation set using Jaccard distance (Hamers et al., 1989) and choose the samples with the smallest distance (highest overlapping).”
Offensive Language and Hate Speech Detection for Danish (2019, methodology): “We used a Jaccard index [19] to assess the similarity of our annotations.”
How to normalize cooccurrence data? An analysis of some well-known similarity measures (2009, unknown): “Theoretical analyses of relations between different measures can be found in the work of Egghe (2008) and Hamers et al. (1989).”
An experimental study of the intrinsic stability of random forest variable importance measures (2016, methodology): “Jaccard index is widely used in the literature of stability evaluation, which calculates the similarity between pairs of feature rankings concerning the aspect of overlap [38].”
A Survey of Link Prediction in Complex Networks (2016, unknown): “This measure is closely related to the Jaccard index, and some works have shown that, in most practical situations, the Salton index yields a value that is approximately twice the Jaccard index [Hamers et al. 1989].”
International collaboration in science and the formation of a core group (2008, background): “Since the distributions are not expected to be normal, it has been suggested that the cosine instead of the Pearson correlation is the proper measure for normalization (Ahlgren et al., 2003; Boyack et al., 2005; cf. Hamers et al., 1989).”
The Second DIHARD Diarization Challenge: Dataset, task, and baselines (2019, methodology): “JER is based on the Jaccard similarity index [19, 20], a metric commonly used to evaluate the output of image segmentation systems, which is defined as the ratio between the sizes of the intersections and unions of two sets of segments.”
International collaboration in science : a new dynamic for knowledge creation (2004, unknown): “In order to normalize for the size effects among countries, we also used the cosinee between the vectors for individual countries as input to the visualization usingg Pajek network analysi s software (HAMERS et al., 1989).”