User:Popcorndude/N-Tape usecases

From Apertium
Jump to navigation Jump to search

I've been working on a library for manipulating transducers with arbitrary numbers of tapes. This page is to collect uses of such transducers.

Feel free to add more.

Historical Ling

Modeling sound changes

Old English → Middle English

<firespeaker> so like, if you think about OE→ME→ModE changes
<popcorndude> well, as of today I can probably apply such rules
<popcorndude> (though I can't compile them yet)
<firespeaker> gɔːs>i:gɔːsi
<firespeaker> gɔːs>i:gœːsi
<firespeaker> gœːs>i:gœːsə
<firespeaker> gœːs:gœːs
<firespeaker> geːs:geːs
<firespeaker> giːs:giːs
<firespeaker> I think you get from OE to ModE with these stages, but you'd want to check the details
<firespeaker> some of them refer to :SR, I don't think any need to refer to UR: though
<firespeaker> (the last few stages can be either, it's pretty straightforward at that point)
<popcorndude> if you had a segmentation tape you could write a rule for reanalysis
<firespeaker> right
<firespeaker> the UR:SR forms I gave are the two sides of twol
<firespeaker> UR has the segmentation
<firespeaker> it'd be good to have a morph tape too

Arabic loanwords in Kyrgyz

<firespeaker> popcorndude: remind me later to tell you about ɑw→oː coalescence and reharmonisation in Kyrgyz words of Arabic origin
<firespeaker> (I'm a bit busy now)
<popcorndude> ok
<firespeaker> short set of data: *molla → moldo, *hoʤa → qoʤo, but *sawda → soːdɑ and *ʃa`ula → ʃoːlɑ
<firespeaker> and unborrowed data: *tɑw → toː, etc.
<firespeaker> I have some thoughts on how to model it that would benefit from an n-tape transducer
<firespeaker> oh and another important pieces of unborrowed data: *sawlɑn → soːron
<firespeaker> *piece
<firespeaker> my thoughts are near the end of the paragraph
<firespeaker> another good example: Kazakh /ɑ/ fronting between palatals versus the phonologisation of what appears to have been exactly the same thing in Tatar


<firespeaker> popcorndude: another use case for n-level transducers: having a morphological level available for transcriber transducer rules

Also having a segmentation tape. Or a tape corresponding to each line of a gloss.