[HIVE-9557] create UDF to measure strings similarity using Cosine Similarity algo

• Type: Improvement
• Status: Reopened
• Priority: Major
• Resolution: Unresolved
• Component/s: UDF
• Labels: CosineSimilarity SimilarityMetric UDF
• Assignee: Alexander Pivovarov
• Reporter: Alexander Pivovarov
• Created: 03/Feb/15 05:15
• Updated: 07/Jul/15 06:56

Description

algo description http://en.wikipedia.org/wiki/Cosine_similarity

--one word different, total 2 words
str_sim_cosine('Test String1', 'Test String2') = (2 - 1) / 2 = 0.5f

reference implementation:
https://github.com/Simmetrics/simmetrics/blob/master/src/uk/ac/shef/wit/simmetrics/similaritymetrics/CosineSimilarity.java