Den 22.9.2011 kl. 14.31 skrev Francis Tyers:

>> Thank for your advices, I am a novice in computer linguistics, I have
>> to learn more about HFST, but  now I am   looking for a framework like
>> Two-Level Morphology, although is a "old" paradigm (80' 90'),  was
>> the first general model in the history of computational linguistics
>> for the analysis and generation of morphologically complex languages.

> HFST contains an implementation of two-level morphology. Kimmo (the guy
> who wrote PC-Kimmo) is the project leader :)
> 
>> For the moment I want to try PC-Kimmo, this  is a software  for the
>> implementation of the two-level model, the output when you parse a
>> english word, e.g. relationships, is:

Please do not use PC-Kimmo, unless you have a masochistic bend. Here is why:

PC-Kimmo was good for two things:
- Even Antworth wrote a nice (and still valid) intro to twol-c (two-level 
morphology)
- PC-Kimmo was nice in the days __before__ we had compilers for the twol rules.

Now we have compilers for the twol rules, be it the xerox tools like twolc 
(http://fsmbook.com) or hfst (and there is sfst, with another rule syntax).

So, we linguists should do what we can: Write morphophonological rules (in 
twolc and hfst-twolc), and let the compilers compile. The alternative is to 
write finite state tables for the automata, as explained in Antworth's book. 
Trust me, I did linguistics in the 80ies, and actually got some tables working. 
If you want to see the tables, you may use the option show-rules in twolc, and 
then be glad you did not have to write what you see there.

>> 
>> It is amazing, I dont know if HFST can do a tokenization and parsing
>> like that?
> Yes it can, and much more.

You may write your rules (the same rules) in PC-Kimmo or in HFST or Xerox's 
twolc. The difference is that only the latter two have compilers that make you 
able to run the rules.



Trond.


------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security
threats, fraudulent activity, and more. Splunk takes this data and makes
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2dcopy2
_______________________________________________
Apertium-stuff mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/apertium-stuff

Reply via email to