Moses project 2023-01-30 to 2023-02-03
to : Martin Holmes
Minutes: 390
Following the work started last week, I implemented a set of templates for the orthography generation on Monday. I also started an XSpec file for the XSLT lib, because there are a number of fairly gnarly character-replacement functions that happen at the end of the process, which need thorough testing. We also had a discussion about some of the schwa-insertion rules, which are partially dependent on the morpheme structure (meaning they should happen early in the process) and partially dependent on the resulting phonological segment sequence (meaning they should happen late). We may be able to handle this dichotomy by having multiple passes in different modes which retain the hyph structure until we can finally dispense with it.
On Tuesday we revisited this and discovered that so far, only a single pass is actually required; implemented a few more templates.
On Wednesday, coded a couple more templates.
On Thursday, completed the initial set of morpheme-matching templates, leaving three which raised errors and three which relate to nasal suffixes, which are a bit different. One turned out to be a typo in the spec document, so we’re down to two problems so far.