Date: Tue, 14 Jul 92 04:53:41 -0400 From: lojbab@grebyn.com (Logical Language Group) Message-Id: <9207140853.AA23244@daily.grebyn.com> Subject: your rafsi proposals, especially drata vs drani Cc: cowan@snark.thyrsus.com Content-Length: 4853 Lines: 138 I believe all of your rafsi recommendations but one are likely to be followed. I can't offer much justification or hope for dra switch to drata. You say that there is not enough usage to prove much, but as you can see in this message, we have an awful lot of proposed and used lujvo using certain words - that is why I said a long time ago that if you felt a word deserved a rafsi, a good half-dozen lujvo might be a minimum in order to make it happen. In the absence of better usage data, I have to use what we've got, overriding the statistics only when I can make a good common sense argument (as I'm now doing in stripping many of the culture and element words of their rafsi, when it makes it possible to give them to words that will be more widely used). The alternative to a statistical approach is our culturally biased instincts and aesthetics, and I just don't trust them. You can blame yourself for my increased trust in our usage statistics, anyway: 1) your translations have covered a wide range of topics, so as not to skew the data towards one field (and indeed perhaps turning some of the previous skewings from JCB's data to noise); 2) your bent towards dikyjvo means that the quality of your typical lujvo is far more trustworthy in statistical analysis of how the words will be used than that of the typical Lojbanist (or worse - old Loglanist). I may not want to formally commit to dikyjvo, but I recognize that they often give a good first cut as to what a reasonable lujvo should be. Now to drani. As the following Lojban usage data shows, there is more use of drani both overall and in final position; and even with the Eaton data added in, there are only 2 lujvo with drata in final position. CCV and CVV rafsi have to be preferred for gismu concentrated in the final position. dat is perfectly fine for the heavy majority of usages for drata in first position. Neither rai/ra'i nor dai/da'i nor dan/ran are easily available to drani. In addition, it is a convenient parallel in that srera has a CCV too: sre. Similarly, da'a is well used by damba, and ra'a by srana; the latter has other competitors that would beat out drata if it were deemed less important. It turns out that even if dra could be pried free from drani, darxi has a much better claim on it than drata. Indeed, it came fairly close in the statistics, and will almost certainly grab either dai or da'i from their current holders. Supporting data: Actual Lojban text drani 1st pos 4 draselri'a correct + 2nd-place-of + causes 4 drari'a correct + causes 8 dragugvi'e correct + country + guest 1 drabai correct + force 1 dratra correct + form 4 drapre correct + person 3 drasratse correct + vertical + sits mid pos 2 sevdrari'a dreamer + correct + causes final pos 3 kavdrani captures + correct 1 jbidra nearby + correct 1 kazdrani quality-of + correct 2 srasirdra vertical + directly + correct drata 1st pos 1 datygugvi'e other-than + country + guest 4 datklu other-than + culture 6 datpre other-than + person 1 dratyzu'e other-than + purposeful-agent final pos 8 mordrata pattern + other-than TLI Eaton lujvo proposals drani 1st pos drani +zbasu drani +fengu drani +punji mid pos cliva +bartu +drani +pluta rarna +drani +pajni jikca +drani +zukte final pos nu +cenba +drani bartu +ka +drani xlali +drani temci +drani drata 1st pos drata +se +stuzi drata +claxu drata +zukte drata +dunda drata +sevzi +xrani drata +snada +djica drata +stuzi +bajra drata +claxu +cinmo drata +stuzi +minde drata +sevzi +punli +zbasu mid pos nu +drata +raktu +cinmo nu +drata +claxu djica +drata +prenu +se +ponse final pos mutce +drata Meanwhile The worst totally unsolvable problem so far appears to be kansa, with a 'uncovered' score of 19 in final position, but competing against katna with an enormous 43 final score. Neither has any alternative to ka'a, so katna must win. No solution short of changing a gismu, which is not an option at this point. By comparison, the worst problem with the current rafsi assignments using the old statistics had only an 'uncovered' score of 8. To give you some idea why the usage statistics are causing so much change, I have a list of more than 25 words that have WORSE than 19 'uncovered scores' using the current rafsi assignments. I consider these to be must-fixes (barda/bra is only a 16 uncovered score by comparison), and 25 words means at least 50 rafsi assignment changes (more likely 100+) which alone would constitute much of the percentage change, which has crept back up to 15% again. (Byu the way, the numbers in the left column are word frequencies.)