Reopen.

Hmmm. OK, maybe a split is best. I've created a "rare" 截断/さいだん entry. I'll close this and reopen briefly for any possible discussion.

@@ -9,4 +8,0 @@
-</k_ele>
-<k_ele>
-<keb>截断</keb>
-<ke_inf>&rK;</ke_inf>
@@ -23,4 +18,0 @@
-</r_ele>
-<r_ele>
-<reb>さいだん</reb>
-<re_restr>截断</re_restr>

仙人掌 is currently tagged as [rK] for the reading シャボテン on entry 1387180. シャボテン has two surface forms: シャボテン and 仙人掌. Can we say with confidence that 仙人掌 is truly [rK] for シャボテン? It probably is (since 仙人掌 can also be read as サボテン and せんにんしょう), but strictly speaking I don't think we have any way to measure these usages. This talk about surface forms being rare for particular readings makes me kind of uneasy. I think it's safer to think of [rK] as being strictly "for the entry" rather than also "for the readings."

I agree that duplicating / splitting entries is a good practice when it enhances clarity. Unlike the サボテン entry (which is quite a bit busier), I personally don't see the benefit for this 切断 entry; I think the separation would just make it more disorganized. If it were up to me, my preference would be to keep the entry as it currently is.

All the kokugos make reference to cloth and paper for 裁断. I don't think a merge would be appropriate. It would also mean having to tag 截断 as an irregular kanji form.

> I thought the [rK] tags signify that surface forms are rare for their respective entries rather than for particular readings.
That's correct. But this is precisely why I suggest splitting out さいだん. By tagging a surface form as rK, we're marking it as "rare" for all reading and senses that apply to that form. It's for this reason that we don't tag a form as rK if it's above the 3% threshold for just a particular sense.

> I don't think splitting the entry adds any new information that can't already be inferred from the existing entry. 
Perhaps, but I don't think the entry in its current state is consistent with the definition of rK. We essentially have two words here (せつだん and さいだん), only one of which is rarely written as 截断. We split off はおうじゅ from サボテン for the same reason.

> The conventional wisdom with databases is to avoid data duplication whenever possible.
For years, this was more or less JMdict's policy, with even the slightest overlap being used as justification for merging forms. But it resulted in some very messy and hard-to-read entries, and we've since undone a lot of these merges. We permit some degree of duplication for the sake of clarity and readability.

Daijs has this to say in its entry for 截(せつ):
「截」を「サイ」と読むのは「裁」などとの混同による。

Sense [1] of 裁断 seems to have a specific cloth / paper nuance
which isn't mentioned in the せつだん entries. But since the サイ
reading of 截 seems to originate from 裁, I don't think 截断
would be out-of-place in the 裁断(さいだん) entry.

And just to confuse things, there is 裁断/さいだん/cutting (1296100 sense 1).  If we are moving さいだん from this entry, perhaps it should go in there?

I thought the [rK] tags signify that surface forms are rare for their respective entries rather than for particular readings. That's why we don't put [rK] tags on forms which have unique sense restrictions.

Outside of specific cases such as this one (in which the reading restriction is provided by the refs), I don't think we can adhere to that standard. For example, if in actuality 90% of the さいだん usages belonged to 截断 (rather than 100%), we probably wouldn't split 截断・さいだん into a separate entry. In general I don't know how we'd even go about measuring those sorts of usage ratios.

I don't think splitting the entry adds any new information that can't already be inferred from the existing entry. The conventional wisdom with databases is to avoid data duplication whenever possible.

Indeed, there's no difference in meaning. But 截断 is not an rK form for さいだん, only せつだん. I propose keeping 截断 on this entry, tagged as rK, and splitting out さいだん into a separate entry where 截断 is not tagged as rK. This would also allow us to tag さいだん as "rare".

The kokugos (daijr/s, shinmeikai) describe さいだん as 慣用読み and merely redirect to せつだん. Doesn't seem to have any distinct meaning from せつだん.

What is at stake by keeping さいだん in this entry? It's not clear to me how this could cause confusion.

By splitting the reading into a separate entry, I think we'd be making that information more difficult to discover. As it is right now, the info can be summarized into a single table:
https://gist.github.com/stephenmk/7da3f54216bfe5018706651e13efba2d

截断 is only rK for せつだん. For さいだん, it's the only kanji form. I think さいだん should be a separate entry.

Adding sK/sk forms.

@@ -12,0 +13,4 @@
+</k_ele>
+<k_ele>
+<keb>接断</keb>
+<ke_inf>&sK;</ke_inf>

Google N-gram Corpus Counts
╭─ーーーー─┬───────────┬───────╮
│ 切断　　 │ 1,573,771 │ 99.9% │
│ 截断　　 │     1,015 │  0.1% │ 🡠 daijr/s, etc.
│ 接断　　 │       425 │  0.0% │ 🡠 drop and hide (no refs)
│ せつだん │     2,167 │  N/A  │
│ さいだん │     2,955 │  N/A  │
╰─ーーーー─┴───────────┴───────╯

HiddenForm 接断[iK]

@@ -12,4 +12 @@
-</k_ele>
-<k_ele>
-<keb>接断</keb>
-<ke_inf>&iK;</ke_inf>
+<ke_inf>&rK;</ke_inf>

@@ -29,0 +30 @@
+<pos>&vt;</pos>

daij, shinmeikai, my ime

@@ -22,0 +23,4 @@
+<r_ele>
+<reb>さいだん</reb>
+<re_restr>截断</re_restr>
+</r_ele>

merge

not in any of my sources; i suspect it is [iK].  probably gets enough non-edict web hits to be worth recording though

@@ -11,0 +12,4 @@
+</k_ele>
+<k_ele>
+<keb>接断</keb>
+<ke_inf>&iK;</ke_inf>

Koj, Daijr, Wadoku (all merge).

Merging 1567110.

@@ -8,0 +9,3 @@
+</k_ele>
+<k_ele>
+<keb>截断</keb>