[
https://issues.apache.org/jira/browse/LUCENE-3842?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13221932#comment-13221932
]
Robert Muir commented on LUCENE-3842:
-------------------------------------
{quote}
I looked at the patch but I don't fully get what it does. Looks like a
combination of state sequence unions, am I right?
{quote}
Well the conversion should ultimately be more useful for the suggester to
intersect with the FST than a tokenstream, because a tokenstream is like a
word-level automaton, if dogs is a synonym for dog, then we have:
smelly dog|dogs(positionIncrement=0).
So for our intersection, we would prefer it to be a deterministic at
'character' (byte) level instead. So the conversion should produce an automaton
of: smelly dog(s?) in regex notation... this is easier to work with.
at index time its useful too, because in the FST we only care about all the
possible byte strings, so this should be easier to enumerate than a tokenstream
(especially if you consider multiword synonyms, decompounded terms etc where
some span across many).
> Analyzing Suggester
> -------------------
>
> Key: LUCENE-3842
> URL: https://issues.apache.org/jira/browse/LUCENE-3842
> Project: Lucene - Java
> Issue Type: New Feature
> Components: modules/spellchecker
> Affects Versions: 3.6, 4.0
> Reporter: Robert Muir
> Attachments: LUCENE-3842-TokenStream_to_Automaton.patch,
> LUCENE-3842.patch
>
>
> Since we added shortest-path wFSA search in LUCENE-3714, and generified the
> comparator in LUCENE-3801,
> I think we should look at implementing suggesters that have more capabilities
> than just basic prefix matching.
> In particular I think the most flexible approach is to integrate with
> Analyzer at both build and query time,
> such that we build a wFST with:
> input: analyzed text such as ghost0christmas0past <-- byte 0 here is an
> optional token separator
> output: surface form such as "the ghost of christmas past"
> weight: the weight of the suggestion
> we make an FST with PairOutputs<weight,output>, but only do the shortest path
> operation on the weight side (like
> the test in LUCENE-3801), at the same time accumulating the output (surface
> form), which will be the actual suggestion.
> This allows a lot of flexibility:
> * Using even standardanalyzer means you can offer suggestions that ignore
> stopwords, e.g. if you type in "ghost of chr...",
> it will suggest "the ghost of christmas past"
> * we can add support for synonyms/wdf/etc at both index and query time (there
> are tradeoffs here, and this is not implemented!)
> * this is a basis for more complicated suggesters such as Japanese
> suggesters, where the analyzed form is in fact the reading,
> so we would add a TokenFilter that copies ReadingAttribute into term text
> to support that...
> * other general things like offering suggestions that are more "fuzzy" like
> using a plural stemmer or ignoring accents or whatever.
> According to my benchmarks, suggestions are still very fast with the
> prototype (e.g. ~ 100,000 QPS), and the FST size does not
> explode (its short of twice that of a regular wFST, but this is still far
> smaller than TST or JaSpell, etc).
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
