[jira] [Resolved] (SPARK-28081) word2vec 'large' count value too low for very large corpora

["Sean Owen (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/SPARK-28081?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sean Owen resolved SPARK-28081.
-------------------------------
       Resolution: Fixed
    Fix Version/s: 2.4.4
                   2.3.4
                   3.0.0

Issue resolved by pull request 24893
[https://github.com/apache/spark/pull/24893]

> word2vec 'large' count value too low for very large corpora
> -----------------------------------------------------------
>
>                 Key: SPARK-28081
>                 URL: https://issues.apache.org/jira/browse/SPARK-28081
>             Project: Spark
>          Issue Type: Bug
>          Components: ML
>    Affects Versions: 2.4.3
>            Reporter: Sean Owen
>            Assignee: Sean Owen
>            Priority: Minor
>             Fix For: 3.0.0, 2.3.4, 2.4.4
>
>
> The word2vec implementation operates on word counts, and uses a hard-coded value of 1e9 to mean "a very large count, larger than any actual count". However this causes the logic to fail if, in fact, a large corpora has some words that really do occur more than this many times. We can probably improve the implementation to better handle very large counts in general.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org