Difference between revisions of "Bikol/Transducer"

From LING073
Jump to: navigation, search
(→‎Notes)
Line 1: Line 1:
 
*[https://github.swarthmore.edu/eortici1/ling073-bcl GitHub for Bikol Transducer]
 
*[https://github.swarthmore.edu/eortici1/ling073-bcl GitHub for Bikol Transducer]
  
−
== Notes ==
+
== Analyser Evaluation ==
 +
=== Notes ===
 
41 of 59 tests pass.
 
41 of 59 tests pass.
  
Line 10: Line 11:
 
<i>Note: By adding sa, nin, and kan (all prepositions) to the transducer, the coverage grew drastically from 0.46% to 10.81%.</i>
 
<i>Note: By adding sa, nin, and kan (all prepositions) to the transducer, the coverage grew drastically from 0.46% to 10.81%.</i>
  
−
== Evaluation ==
+
=== Evaluation ===
 
*Total number of stems in transducer:
 
*Total number of stems in transducer:
 
**From Grammar Documentation: <b>34</b>
 
**From Grammar Documentation: <b>34</b>
Line 40: Line 41:
 
**sarong
 
**sarong
  
 +
== Generator Evaluation ==
 +
=== Initial evaluation of morphological generation ===
  
 
[[Category:Bikol]] [[Category:Sp17_Transducers]]
 
[[Category:Bikol]] [[Category:Sp17_Transducers]]

Revision as of 18:15, 20 February 2017

Analyser Evaluation

Notes

41 of 59 tests pass.

The forms that do not pass include all forms for the Future and Progressive tenses, two Past tense forms, and all Plural Verbs and Plural Adjectives. All of these forms do not pass in my morphological analyzer because they require reduplication and/or infixation in the middle of the base of the word. Since these morphological phenomena are difficult to implement, I do not have those passing.

However, all forms that only require prefixes or suffixes pass.

Note: By adding sa, nin, and kan (all prepositions) to the transducer, the coverage grew drastically from 0.46% to 10.81%.

Evaluation

  • Total number of stems in transducer:
    • From Grammar Documentation: 34
    • Total, including the three added common words: 37
  • Current coverage over combined corpus: 10.81%
  • Number of tests that pass in:
    • bcl.yaml: 41/59 pass
    • commonwords.yaml: 3/20 pass
  • Current list of top unknown words covered by aq-covtest:
    • na
    • an
    • asin
    • mga
    • ni
    • esus
    • si
    • siya
    • ka
    • mo
    • niya
    • ko
    • ki
    • man
    • ta
    • saka
    • an
    • sinda
    • tawo
    • sarong

Generator Evaluation

Initial evaluation of morphological generation