[HIVEMALL-146] Implement yet another UDF to generate n-grams from a list of words

• Type: New Feature
• Status: Closed
• Priority: Major
• Resolution: Done
• Fix Version/s: 0.5.0
• Assignee: Takuya Kitazawa
• Reporter: Takuya Kitazawa
• Created: 02/Oct/17 04:21
• Updated: 09/Nov/17 03:33
• Resolved: 04/Oct/17 03:17

Description

Hive has ngrams() function to obtain n-grams of a list of words: https://cwiki.apache.org/confluence/display/Hive/StatisticsAndDataMining#StatisticsAndDataMining-ngrams()andcontext_ngrams():N-gramfrequencyestimation

While the existing function returns "estimated" top-k list of frequent n-grams, NLP applications sometimes need to get "exact" list of n-grams which include all of 1-, 2-, ..., n-grams. To give an example, for an input ["machine", "learning"], we might expect to get the following result: ["machine", "learning", "machine learning"].

Hence, this ticket requests to implement yet another UDF something like ngrams(). Implementation could be similar to getNgrams() in the Stanford CoreNLP library: https://github.com/stanfordnlp/CoreNLP/blob/d6318a0cb06dba635550477bc843952cc5a5f868/src/edu/stanford/nlp/util/StringUtils.java#L2132-L2142

Issue Links

• links to: GitHub Pull Request #118