[
https://issues.apache.org/jira/browse/TIKA-5029?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18137590#comment-18137590
]
ASF GitHub Bot commented on TIKA-5029:
--------------------------------------
tballison closed pull request #3524: TIKA-5029: extract MP4 text subtitle
tracks as embedded WebVTT
URL: https://github.com/apache/tika/pull/3524
> Mine octet-stream in commoncrawl to improve detection
> -----------------------------------------------------
>
> Key: TIKA-5029
> URL: https://issues.apache.org/jira/browse/TIKA-5029
> Project: Tika
> Issue Type: Task
> Reporter: Tim Allison
> Priority: Minor
>
> With the advent of agents, we can have point them at files that were
> identified as application/octet-stream in commoncrawl and look for patterns.
> The mime_detected field in commoncrawl derives from Tika, so we know at least
> that an earlier version of Tika couldn't identify these files.
> Just looking at the 10k octet-streams in our regression corpus has found some
> nice improvements.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)