[
https://issues.apache.org/jira/browse/MAHOUT-242?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12803918#action_12803918
]
Zaki Rahaman commented on MAHOUT-242:
-------------------------------------
Thanks Drew. Some overdue feedback from testing this baby out...
My use case may differ from what's more typical, but basically I had 10MM+
semi-processed strings (lowercased, no punctuation, etc.) that I wanted to do
my analysis on, not a large document or a series of smaller documents. As a
result, I found that using the Lucene Analyzers in the NGramCollector to be
horribly ineffecient (the job was taking hours to complete 1% on my modest 7
node cluster). Instead I used some custom logic to generate the n-grams and
pass them to the collector. After this modification, the entire set of jobs ran
in a little less than an hour even on a larger dataset.
I know Robin mentioned something about an Ngram generator in the Bayes
classifier, I should check it out.
I guess I would advocate splitting the module as it exists into a couple of
different pieces:
1. A general purpose tool for tokenizing/analyzing a document/documents (or in
my case, 'documents' consisting of strings). I know there's some exisitng tools
for vectorizing text, and indices, so maybe this is something similar but the
basic idea would be to go from document --> analyzed text
2+3 (These can remain as is in the module or be split). N gram generator and
counter and LLR pieces.
Maybe this is all crazy talk, but it would seem to me refactoring/extracting
the pieces out like this would prove useful and allow for code reuse in other
modules (other NLP-type modules, elsewhere in the project).
> LLR Collocation Identifier
> --------------------------
>
> Key: MAHOUT-242
> URL: https://issues.apache.org/jira/browse/MAHOUT-242
> Project: Mahout
> Issue Type: New Feature
> Affects Versions: 0.3
> Reporter: Drew Farris
> Priority: Minor
> Attachments: MAHOUT-242.patch, MAHOUT-242.patch,
> mahout-colloc.tar.gz, mahout-colloc.tar.gz
>
>
> Identifies interesting Collocations in text using ngrams scored via the
> LogLikelihoodRatio calculation.
> As discussed in:
> *
> http://www.lucidimagination.com/search/document/d051123800ab6ce7/collocations_in_mahout#26634d6364c2c0d2
> *
> http://www.lucidimagination.com/search/document/b8d5bb0745eef6e8/n_grams_for_terms#f16fa54417697d8e
> Current form is a tar of a maven project that depends on mahout. Build as
> usual with 'mvn clean install', can be executed using:
> {noformat}
> mvn -e exec:java -Dexec.mainClass="org.apache.mahout.colloc.CollocDriver"
> -Dexec.args="--input src/test/resources/article --colloc target/colloc
> --output target/output -w"
> {noformat}
> Output will be placed in target/output and can be viewed nicely using:
> {noformat}
> sort -rn -k1 target/output/part-00000
> {noformat}
> Includes rudimentary unit tests. Please review and comment. Needs more work
> to get this into patch state and integrate with Robin's document vectorizer
> work in MAHOUT-237
> Some basic TODO/FIXME's include:
> * use mahout math's ObjectInt map implementation when available
> * make the analyzer configurable
> * better input validation + negative unit tests.
> * more flexible ways to generate units of analysis (n-1)grams.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
