|
Post by Hrafn Loftsson on Oct 10, 2009 20:06:04 GMT -5
Here you can discuss and ask questions about Assignment II.
Regards, Hrafn.
|
|
thors
New Member
CS Dweeb
Posts: 23
|
Post by thors on Oct 11, 2009 20:29:04 GMT -5
I _think_ there's a í/ý error in one of the icelandic texts - „hún týndi hausana ...“
Tíndi - picked up, týndi - lost.
Right ?
Also, why the mangled english translation below each icelandic text ?
|
|
|
Post by Hrafn Loftsson on Oct 12, 2009 5:08:16 GMT -5
Yes, the spelling error has been corrected.
The English word-by-word translation is just to give the non-Icelandic speakers an indication of the meaning of the sentence. The translation can therefore help them to figure out the word class for each word.
|
|
nik
New Member
Posts: 8
|
Post by nik on Oct 15, 2009 5:31:15 GMT -5
|
|
|
Post by tihomir on Oct 15, 2009 13:42:47 GMT -5
Hi, Nik. You can read on Wikipedia about the (middle) voice, but a typical example of the usage of the active, middle and passive voice in Icelandic would be: Ég lokaði glugganum. I closed the window. This is active voice. The grammatical subject ("I") performs the action. Glugginn lokaðist. The window was/has (been) closed. This is middle voice (the verb ends in -st) and this sentence usually means that we do not know who closed the window (maybe it was the wind…). Even if we know that Jón closed it, we would not put "af Jóni" ("by John") at the end of this sentence. Glugganum var lokað. The window was closed. This is passive voice. The grammatical subject (the window) is being acted upon. Here you can add "af Jóni" at the end. Here the verb "loka" is in the past participle form (neuter) and is preceded by a form of "vera" (to be). Of course there are exceptions!!! Maybe the native speakers in the class can add something... BÍN (http://bin.arnastofnun.is/) is very useful for this exercise. Middle voice is called miðmynd in Icelandic. Active voice is germynd. Past participle is Lýsingarháttur þátíðar. If you are using the internet through a university network, you should also be able to get free access to snara.is/ which is a great collection of dictionaries.
|
|
|
Post by Hrafn Loftsson on Oct 19, 2009 16:02:15 GMT -5
I received the following question from one of the international students in the class:
"I have absolutely no idea how I determine the following property: Category a–does not govern case, u–exclamation, o–governs accusative, þ–governs dative, e–governs genitive Do you have a hint?"
My answer:
This is indeed difficult for non-native Icelandic speakers.
In general:
1) an adverb in Icelandic is "aa" 2) an interjection (like "yes", "no") is "au" 3) a preposition is either "ao", "aþ", "ae". Which of these is correct in the given context is difficult to decide for a non-native speaker unless the following nominal (e.g. adjective, pronoun or noun) is unambiguously one of the three cases.
You should be able to guess if you have an adverb or a preposition. However, given a preposition, it is difficult for you to guess which one is correct and that is ok :-)
|
|
thors
New Member
CS Dweeb
Posts: 23
|
Post by thors on Oct 19, 2009 16:09:38 GMT -5
@thiomir The second example („Glugginn lokaðist“) would stand correctly as „The window closed [shut].“. was/has been indicates an (inter)action while the clean cut case implies some kind of omnipotence (the window closed shut all by itself) If there's anyone having problems running the IceNLP tools, note that it may help a lot to be located in the same directory as the script/class files being run.
|
|
thors
New Member
CS Dweeb
Posts: 23
|
Post by thors on Oct 19, 2009 16:11:22 GMT -5
Oh, and if anyone of our foreign students is having problems determining the properties of the Icelandic texts....
Have heart. Even us locals have problems (I for one don't even know what half of the english tag words stand for...)
|
|
|
Post by Hrafn Loftsson on Oct 25, 2009 16:13:47 GMT -5
The grades for this assignment are now available in MySchool.
Three comments:
1) When computing the accuracy of a tagger, you cannot disregard the punctuation characters because they are also tokens. However, you could compute two accuracy numbers, one with punctuation characters and another one without.
2) The number of errors made by IceTagger for the given Icelandic sentences were actually six but not five as I had stated.
3) When computing the accuracy of TriTagger in exercise III, some students mixed up the ratio of unknown words and total tagging accuracy. The total tagging accuracy is just the ratio of the number of tags produced by a tagger divided by number of correct tags produced by the tagger. That number is often quite dependent on the ratio of unknown words (because unknown words are more difficult to tag), but the ratio of unknown words is not equal to tagging accuracy.
|
|