[
https://issues.apache.org/jira/browse/JOSHUA-273?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15339419#comment-15339419
]
ASF GitHub Bot commented on JOSHUA-273:
---------------------------------------
Github user mjpost commented on the pull request:
https://github.com/apache/incubator-joshua/commit/6d2213a20b74432fc7cb131c732f7507b74053e9#commitcomment-17933088
In src/main/java/org/apache/joshua/decoder/StructuredTranslation.java:
In src/main/java/org/apache/joshua/decoder/StructuredTranslation.java on
line 57:
Hi Felix — I guess I probably don't. I put this in so as to have the
convenience functions of querying the maps (that are in FeatureVector), and
didn't realize why you guys had stuck with just a HashMap. I'll revert it.
In general, this relates to JOSHUA-273: I want to remove all the output
code from KBestExtractor and place it in StructuredTranslation (and then, in
turn, to collapse Translation and StructuredTranslation). There are at least
two ways to do this; do you have a preference?
- StructuredTranslationFactory could be called to create the necessary
entries in StructuredTranslation, which would then just be stored as strings.
This approach makes StructuredTranslation basically just a container that is
operated on.
- StructuredTranslation can do its own computation, formatting the
"outputFormat" line from the config file. This view makes StructuredTranslation
itself a bit smarter.
matt
> Joshua API
> ----------
>
> Key: JOSHUA-273
> URL: https://issues.apache.org/jira/browse/JOSHUA-273
> Project: Joshua
> Issue Type: Improvement
> Reporter: Matt Post
> Fix For: 7
>
>
> We have a lot of work to do to clean up the decoder's internal object
> pipeline in order to create a nice, clean API.
> (This is just a stub for this issue; I will return soon with a better
> description and roadmap. Others feel free to edit, as well).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
