Neo-Aramaic/Transducer

From LING073
Revision as of 12:20, 6 March 2018 by Cdalton2 (talk | contribs)

Jump to: navigation, search

The code for the transducer can be found in this Github repository.

Evaluation

Notes

Sadly, only 7 of our 127 tests pass at the moment. The reason for this is that the forms and the analyses in our transducer differ from the forms and analyses on our Grammar Documentation page in a variety of ways. Because of the templatic nature of Neo-Aramaic morphology, there were many things that made sense for us to do on the wiki page but are not reflected in the transducer. For example, canonically the root of the verb to open is the triconsonantal sequence ܦܬܚ, which can take on forms like ܦܵܬܹܚ 'he opens' or ܦܬܹܚܠܵܐ 'he opened'. At our current stage, we decided to hard-code the present-tense and past-tense lemmas of the verbs separately, such that ܦܵܬܹܚ analyzes as ܦܵܬ<vblex><pres><p3><m><sg> and ܦܬܹܚܠܵܐ analyzes as ܦܬܹܚ<vblex><past><p3><m><sg>. But unfortunately, these analyses are a bit different from the ones we scraped from our wiki page, which are ܦܬܚ<vblex><pres><p3><m><sg> and ܦܬܚ<vblex><past><p3><m><sg>. All of our tests involving verbs suffer from this issue, and the nouns have a similar problem: in our wiki page, the lemmas are the singular form of the noun with the schwa ending, but in our transducer, the lemmas don't include the final vowel so that either a schwa or an [a] can be added depending on whether the noun is singular or plural. In order to fix it, we will either have to edit the lemmas in our yaml file or change our transducer so that it's capable of inserting different vowels into the lemma based on whether it's past tense or present tense.