3. Alternation patterns

Alternation patterns or analogies serve as a basis for all other scripts. Patterns describe an alternation between two forms from two different cells.

Input

For instance, when provided with a pair of forms ba ~ bab, qumin will abstract away the following pattern: Ø ~ b / ba_. Moreover, Qumin is able to generalize patterns that apply to different lexemes in a phonological meaningful way and then to select the best matching pattern.

Let us compute patterns for three cells on our Karelian dataset:

qumin action=patterns data=parakar/parakar.package.json cells="[abl.sg, gen.sg, part.sg]" hydra.run.dir=results/patterns

Tip

We used action=patterns to tell Qumin to compute patterns. In practice, this is not necessary as Qumin’s default action is to compute patterns.

Output

Qumin produces the following output folder:

  • cli.log: a log from the computation

  • metadata.json: the metadata descriptor

  • microclasses.txt: classes of lexemes that behave identically (i.e. that have the same patterns)

  • /patterns

    • /human_readable: readable outputs, for human eyes.

    • /machine_readable: machine readable outputs, for Qumin’s eyes.

The last two folders contain one file per pair of cells. Since we asked for 3 cells, it contains 3 files. Human readable outputs can be of a great help to understand morphological alternations. Let’s have a look at the beginning of one:

## Pattern u͜ɑ ⇌ ɑn / X+[-syll]_

Pairs of forms instantiating this pattern: 1315
Full pattern: u͜ɑ ⇌ ɑn / {b,bʲ,d,dʲ,dʲː,dː,d͡ʒ,e,eː,e͜i,e͜u,e͜y,f,fː,g,h,i,iː,i͜e,i͜u,i͜y,i͜æ,i︧eu︨,i︧ey︨,i︧æy︨,j,jː,k,kː,l,lʲ,lʲː,lː,m,mː,n,nʲ,nʲː,nː,o,oː,o͜i,o͜u,p,pʲ,pː,r,rʲ,rʲː,rː,s,sʲ,sʲː,sː,t,tʲ,tʲː,tː,t͡s,t͡ʃ,t͡ʃː,u,uː,u͜i,u͜o,u͜ɑ,u︧ou︨,u︧ɑu︨,y,yː,y͜i,y͜ø,y︧øy︨,z,zʲ,æ,æː,æ͜i,æ͜y,ø,øː,ø͜i,ø͜y,ŋ,ŋː,ɑ,ɑː,ɑ͜i,ɑ͜u,ʃ,ʃː,ʋ,ʋʲ,ʋː,ʒ}+{b,bʲ,d,dʲ,dʲː,dː,d͡ʒ,f,fː,g,h,j,jː,k,kː,l,lʲ,lʲː,lː,m,mː,n,nʲ,nʲː,nː,p,pʲ,pː,r,rʲ,rʲː,rː,s,sʲ,sʲː,sː,t,tʲ,tʲː,tː,t͡s,t͡ʃ,t͡ʃː,z,zʲ,ŋ,ŋː,ʃ,ʃː,ʋ,ʋʲ,ʋː,ʒ}_ <1315>
Examples:

| lexeme                   | part.sg                        | gen.sg                        |
|:-------------------------|:-------------------------------|:------------------------------|
| abai                     | ɑbɑju͜ɑ                        | ɑbɑjɑn                        |
| abeiččii                 | ɑbe͜it͡ʃːiju͜ɑ                 | ɑbe͜it͡ʃːijɑn                 |
| abiturientu              | ɑbiturʲi͜entu͜ɑ                | ɑbiturʲi͜entɑn                |
| abudengu                 | ɑbudʲeŋːu͜ɑ                    | ɑbudʲeŋːɑn                    |
| adru                     | ɑdru͜ɑ                         | ɑdrɑn                         |
| ad’jektiivu              | ɑdʲjektʲiːʋu͜ɑ                 | ɑdʲjektʲiːʋɑn                 |
| agan                     | ɑgɑnu͜ɑ                        | ɑgɑnɑn                        |

The file lists all patterns. For each patterns it contains following information:

  • Short, human readable representation of the pattern.

  • Expanded, internally used representation of the pattern.

  • Number of pairs of forms that instanciate the pattern

  • A table containing all these pairs of forms.

The patterns are sorted by frequency.

Tip

Have a look at the patterns reference to see all available options for pattern computations.