Archive for the ‘Jaccard Similarity’ Category

…Locality Sensitive Hashing for Unstructured Data

Friday, May 9th, 2014

Practical Applications of Locality Sensitive Hashing for Unstructured Data by Jake Drew.

From the post:

The purpose of this article is to demonstrate how the practical Data Scientist can implement a Locality Sensitive Hashing system from start to finish in order to drastically reduce the search time typically required in high dimensional spaces when finding similar items. Locality Sensitive Hashing accomplishes this efficiency by exponentially reducing the amount of data required for storage when collecting features for comparison between similar item sets. In other words, Locality Sensitive Hashing successfully reduces a high dimensional feature space while still retaining a random permutation of relevant features which research has shown can be used between data sets to determine an accurate approximation of Jaccard similarity [2,3].

Complete with code and references no less!

How “similar” do two items need to be to count as the same item?

If two libraries own a physical copy of the same book, for some purposes they are distinct items but for annotations/reviews, you could treat them as one item.

If that sounds like a topic map-like question, your right!

What measures of similarity are you applying to what subjects?

Measuring Similarity in Large-scale Folksonomies [Users vs. Authorities]

Tuesday, September 11th, 2012

Measuring Similarity in Large-scale Folksonomies by Giovanni Quattrone, Emilio Ferrara, Pasquale De Meo, and Licia Capra.

Abstract:

Social (or folksonomic) tagging has become a very popular way to describe content within Web 2.0 websites. Unlike taxonomies, which overimpose a hierarchical categorisation of content, folksonomies enable end-users to freely create and choose the categories (in this case, tags) that best describe some content. However, as tags are informally defined, continually changing, and ungoverned, social tagging has often been criticised for lowering, rather than increasing, the efficiency of searching, due to the number of synonyms, homonyms, polysemy, as well as the heterogeneity of users and the noise they introduce. To address this issue, a variety of approaches have been proposed that recommend users what tags to use, both when labelling and when looking for resources.

As we illustrate in this paper, real world folksonomies are characterized by power law distributions of tags, over which commonly used similarity metrics, including the Jaccard coefficient and the cosine similarity, fail to compute. We thus propose a novel metric, specifically developed to capture similarity in large-scale folksonomies, that is based on a mutual reinforcement principle: that is, two tags are deemed similar if they have been associated to similar resources, and vice-versa two resources are deemed similar if they have been labelled by similar tags. We offer an efficient realisation of this similarity metric, and assess its quality experimentally, by comparing it against cosine similarity, on three large-scale datasets, namely Bibsonomy, MovieLens and CiteULike.

Studying language (tags) as used tells you about users.

Studying language as proscribed by an authority, tells you about that authority.

Which one is of interest to you?