Module talk:sw-conj/archives/1

Preliminary discussion
, I thought I'd move the discussion here to flesh out Swahili (and more generally, Bantu) conjugation. We do currently have conjugation tables for Swahili, but they are terrible: unwieldy, poorly designed, and at times incorrect. There has been no effort to fix them since Robert Ullmann's days. What would be great is if we could create a module that could be used as a model upon which one could relatively easily create conjugation modules for other Bantu languages, much like how I copied MOD:ny-headword to make MOD:wmw-headword. The biggest outstanding problem is deciding what to show, because the possibilities are nearly endless: for a transitive verb, all the combinations of subject and object concords could potentially be employed, but showing them doesn't give much use (but possibly they could be in autocollapsed subtables within the table). We could opt to show stems for each of the tenses/moods, or give examples (e.g. only show 1st pers. sing.), but that would not display the fact that vowels may collapse according to predictable rules in certain forms. We might also consider listing those rules at the bottom of the table in prose. These problems have not really been addressed satisfactorily for any language on Wiktionary, given how European languages don't suffer from this; one example is Turkish (example), where many forms are just not mentioned at all and four separate tables have to be opened. The idea of segregating by positive and negative forms might serve us well here, but I always found separate tables aesthetically questionable. What are your thoughts on solving these issues? —Μετάknowledge discuss/deeds 07:53, 26 February 2017 (UTC)
 * Sorry, I needed to go to sleep. The question of what to show is an interesting one. I am something of an inclusionalist, so I'll always fight for more instead of less. First of all, what do you think of Code's Mod:zu-verbs? I'd like to avoid something like the basque tables (T:eu-conj-nornori-table, T:eu-conj-nor-table). I like the appearance of the old Swahili tables more than I like Code's Zulu tables, but I don't like everything about them. I think the idea of nested table makes a fair bit of sense to be honest with the header of the table showing a few example forms. I think the first step should be to make an example table of the most forms we intend to show so we know what forms will need to be generated. Does that make sense? — JohnC5 17:41, 26 February 2017 (UTC)
 * No worries, I had to sleep as well. I think that her Zulu tables are problematic in multiple ways; for example, they link to forms that to the best of my knowledge do not exist independently, and they ignore some of the modal inflections entirely. I'm not planning to touch Zulu, since she's been a bit possessive of it, but I don't think it makes for a good model. But I really don't know what the most forms we want to show even is. Not showing forms with the object would give us a big advantage, namely that the table wouldn't have to be given the verb's transitivity, which would make rollout of the tables considerably slower. Bear in mind that inflected forms can be created even if tables don't link to them. —Μετάknowledge discuss/deeds 18:24, 26 February 2017 (UTC)
 * Sorry, I've been busy all day. I was thinking that we could have the tables not have the objects by default, but you could add something to make it transitive (or intransitive just to mark that it had been done). That way, you can roll it out agnostically, and people can go in later and specify if they want. That sound acceptable? — JohnC5 04:14, 27 February 2017 (UTC)
 * That's actually a great idea I hadn't considered. I'll need to compile all possible inflections we might want to put in a table so you can use it to actually build this thing, but that might take me a few days (I'm rather busy as well, and procrastinating like mad on a couple projects). In the mean time, one more consideration is what to do with the various forms created by means of verbal extensions (applicative, passive, causative, and the like). I think that Wyang's is an excellent model for this, although in terms of the execution, I'd prefer not to use inscrutable codes. I am inclined to see them as derivational, rather than inflectional, but not all linguists choose to delineate them thus; if derivational,  could be placed in a 'Derived terms' section, but if inflectional, it would have to be somehow integrated into the conjugation template. —Μετάknowledge discuss/deeds 04:33, 27 February 2017 (UTC)
 * No need to rush. I am a master procrastinator myself. is very impressive but, as you say, inscrutable. I'll say that I do not (yet) possess the knowledge to make such determinations concerning Swahili. Do these verbal formations inflect fully or only partially? If the causative, say, would require a full set of TAM forms, then they should be separately lemmatized as derivations. If they inflect defectively, then maybe they should be inflections. — JohnC5 06:26, 27 February 2017 (UTC)
 * , they do inflect fully, and as I said, my bias is toward treating them as derivational. They also often have additional or specialised meanings — but then again, so can reflexive verbs, which are decidedly not derivational (they simply put a reflexive morpheme in the object slot). We'll probably want to have specialised templates for those forms that link to an explanatory appendix, although replicating that for each language will be quite a lot of effort, and I'm hoping that we can develop standards that can be ported from one Bantu language to another. —Μετάknowledge discuss/deeds 06:46, 27 February 2017 (UTC)
 * Sounds good to me! WT:Wiktionarians say you live in America, but you spell specialised all Commonwealth-y. Get your story straight! :P — JohnC5 07:05, 27 February 2017 (UTC)


 * I have created what should be an exhaustive description of verbal conjugation for Standard Swahili here. Not considering the relative forms, which are somewhat uncommon outside of literary use nowadays, there are upwards of ten thousand valid inflected forms for a normal transitive verb, which has got to be a strong argument for a radically different approach from, say, Latin. What are your thoughts now? —Μετάknowledge discuss/deeds 03:48, 7 March 2017 (UTC)
 * Eeek, ten thousand? What's it like if we don't have the object? — JohnC5 04:01, 7 March 2017 (UTC)
 * A lot less, but still thousands. There will have to be some telling rather than showing, but where is the optimal cutoff? —Μετάknowledge discuss/deeds 04:08, 7 March 2017 (UTC)
 * I'm good up to 500 hundred, I'd say, though there will still need to be subtables. — JohnC5 04:10, 7 March 2017 (UTC)
 * So I guess that subpage should be cleaned up into an appendix page that the conjugation table can link to; that way, whatever doesn't seem worth covering in full in the table can still be explained. What further information (if any) would you like in building this? I want to cross-check everything before deployment on entries, but what I've written up is likely to change little, if at all (except for the copular verb). —Μετάknowledge discuss/deeds 04:16, 7 March 2017 (UTC)
 * What I need is an example table! :P And I will certainly help with the layout of the table. — JohnC5 04:43, 7 March 2017 (UTC)
 * I just don't seem to be able to make the wikicode for even a table. I think User:Metaknowledge/Swahili verbs should be sufficient to make it, though. —Μετάknowledge discuss/deeds 04:57, 7 March 2017 (UTC)
 * Could you make a big old list? Then I'll make a table for it, then you critique the table? — JohnC5 05:04, 7 March 2017 (UTC)
 * Uhhh... a list of everything? That's definitely not doäble by hand, and I can try in Python, but even that will take me a while (I am pretty bad at coding in general). Or is that not what you meant? —Μετάknowledge discuss/deeds 05:10, 7 March 2017 (UTC)
 * I meant just of the forms we would want, but ok. I'll get working on this. Did you list the phonological changes that occur? — JohnC5 05:17, 7 March 2017 (UTC)
 * Sorry, I'm really rather useless beyond the linguistic aspects of it. There are very few phonological changes, and I believe I got them all, but it may not be very clear — do ask if you have questions. —Μετάknowledge discuss/deeds 05:20, 7 March 2017 (UTC)
 * No worries! I'll start sloooowly working on this. — JohnC5 05:33, 7 March 2017 (UTC)
 * No rush, I was certainly slow enough in getting this part done. But it would be good to have it finalised within the next couple months, so I can run it by a friend of mine who's a L1 speaker of Swahili in person. —Μετάknowledge discuss/deeds 05:53, 7 March 2017 (UTC)

Ok, I've been trying to work through this mentally, but there seem to be some rules omitted. Several questions: what are concords 15-18? I'm having trouble finding reference to them. Also, If I look at the table for say (assuming it is correct), just in the general affirmative indicative, what leads to 2p mwaleta and not maleta or mleta? There seem to be many such cases where added vowels or glides appear. Similarly, why concord 8 vyaleta and not vileta? Thanks! — JohnC5 02:50, 18 March 2017 (UTC)
 * All these issues are treated in my writeup (which I am still expanding to cover some more literary stuff, by the way). The concords for classes 15-18 are given explicitly there, and I also explain the collapse in the gnomic (which the table weirdly calls the "general affirmative indicative") by saying the following: "Note that the gnomic, because it starts with a vowel, causes a collapse of the subject concord. The positive pronominal forms are 1s n-, 2s w-, 3s Ø-, 1p tw-, 2p mw-, 3p w-. The class concords collapse as they do to form the syllabic onset in the inflected forms of ." —Μετάknowledge discuss/deeds 05:13, 18 March 2017 (UTC)
 * Ah gnomic = general in that table. Cool. Also, I saw the values for 15-18 there, but I do find references to these classes in other Swahili documentation. What's going on there? Are those obscure or underused? — JohnC5 05:30, 18 March 2017 (UTC)
 * No, they just have very few lexical items in them. 15 is the verbal infinitive and the last three are the locative classes; there's more on this at Appendix:Swahili noun classes. —Μετάknowledge discuss/deeds 05:43, 18 March 2017 (UTC)
 * Thank you so much for your edits to the page! It looks vastly better in every respect, and is well on its way to becoming an appendix page. I have just written User:Metaknowledge/Swahili verbs/derivation; you don't need to bother with it now, but it will be necessary to build MOD:sw-derivations (the last step in this process, I think). —Μετάknowledge discuss/deeds 07:09, 18 March 2017 (UTC)
 * No problem. The process of redoing it has massively clarified my understanding of the topic. For the moment, are we going to ignore the objects entirely? Shall I try to make a grand unified table for you to look at? I'd like to no have to split the singular and plural as sw-conj-aff does, but I'm not sure how yet. Also, what's the deal with the 1pl. imperative? — JohnC5 07:35, 18 March 2017 (UTC)
 * I don't know whether to ignore objects — we've got to leave out something (a lot of somethings!), but at least an exemplar would be good. A grand unified might be helpful, but since I can't even figure out how this could be best presented, all I can really do is check it for accuracy! I'd ignore the existing templates, which are just so awful. Splitting positive and negative into separate templates was just bizarre. And I'm not sure what the problem with the plural imperative is; I clarified the wording a bit. —Μετάknowledge discuss/deeds 16:28, 18 March 2017 (UTC)
 * As far as I can tell, the writeup does not describe the 1pl. imperative which appears to be tu- -e or something like that. The writeup also makes it sound like the 1pl. would take the ending -eni as it is plural, but this does not appear to be the case. I am conflicted on the issue of splitting positive and negative. They would divide nicely into subtables and they do differ in their layouts, viz. the positive has TAM that the negative lacks and vice-versa. What other things could we put in collapsable subtables? — JohnC5 17:42, 18 March 2017 (UTC)
 * There is no 1pl imperative; there is only a singular (2s) and plural (2p). You could use the subjunctive to arrive at the same result (tu- -e), but that would still just be a subjunctive. As for positive vs. negative, we may not want to show all those TAM markers anyway... or they could go in autocollapsed subtables. —Μετάknowledge discuss/deeds 17:49, 18 March 2017 (UTC)
 * Ok, so sw-conj-aff is wrong on the 1pl. imperative front. Makes sense. It's using a 1pl. optative/hortatory subjunctive as an imperative like Latin does. Ok, I'm going to make a set of abbreviated tables to see what you think of them and decide which to use. These won't be the full production versions until later. — JohnC5 17:53, 18 March 2017 (UTC)
 * Check out User:Metaknowledge/Swahili verbs/table tests. I think I may need to split singular and plural though. Thoughts? — JohnC5 21:22, 18 March 2017 (UTC)
 * The general layout is good, and it certainly makes this feel less overwhelming. Now for a bunch of criticisms that you are not to take personally: A great deal of forms are mising — is this intentional? We'll need some sort of a special space, probably at the top, to keep the infinitive, imperative, and habitual forms. The organisation of the classes is mostly good, but 17 is ku-, not *ki- (and thus identical in all forms to class 15). The subjunctive is a separate TAM on equal footing with the others, and cannot be combined with them; I have clarified this in the writeup. The habitual really doesn't need its own line, because it only has one form for all classes, and it can go up top. Grouping various TAM as "indicative", "conditional", etc doesn't seem particularly useful, and getting rid of those might make the table a bit leaner. —Μετάknowledge discuss/deeds 21:39, 18 March 2017 (UTC)
 * While I've only half followed the discussion, the table looks good in its overall setup. What might be nice, though, is if the bare stem of the various moods/tenses is shown even when the table is collapsed. Expanding the table would then show this stem with inflections. —CodeCat 21:47, 18 March 2017 (UTC)
 * I certainly agree. I will also need some help with other aspects of the layout style—as a colorblind person, I don't trust my intuitions. I also need to change the style so it looks more like the PIE or Proto-Celtic tables. Any help you can provide would be much appreciated. — JohnC5 21:57, 18 March 2017 (UTC)
 * I don't know much about Swahili verbs though, or really anything at all, other than the semblances to Zulu. And I'm not all that well versed in Zulu verbs either, the current table is missing stuff but I don't really know for sure what's actually missing. —CodeCat 22:00, 18 March 2017 (UTC)
 * I'm literally just asking about html markup. — JohnC5 22:09, 18 March 2017 (UTC)
 * Thanks, I don't take those personally at all. I don't like the table as it stands, so it would be hard for me to be insulted by its deficiencies. I omitted forms because I ran out of patience in that current run. Could you please list all of the subtables you'd like, the names you'd use, and the order in which they should appear? Once that is done, I rejigger the table accordingly. — JohnC5 21:48, 18 March 2017 (UTC)
 * Choosing names is hard, and I'm not happy with all of my choices... many Anglophone grammars opt to name the tenses merely by calling them the "KA tense", for example. Also, I concur with CC's suggestion. —Μετάknowledge discuss/deeds 21:53, 18 March 2017 (UTC)

Ok, the new table is up for consideration. It's not complete by any means, but just an idea. Please suggest colors and style stuff as well. — JohnC5 04:37, 19 March 2017 (UTC)
 * I quite like the table. Colours and styles are fine, I guess, although personally I like warmer colours, but no biggie. What is a cause for concern is that the relative forms are all incorrect (apart from the ones that happen to be correct like a stopped clock twice a day). It seems that someone thought the relative affix is simply "o" as "o" is added to every form. It's added to the correct place within the verb, but it should be ye for class 1 (singular of M-WA) (including 1st, 2nd and 3rd person singular), o for class 2 (plural of M-WA) (including 1st, 2nd and 3rd person plural), o for class 3 (singular of M-MI), yo for class 4 (plural of M-MI), lo for class 5 (singular of JI-MA), yo for class 6 (plural of JI-MA), cho for class 7 (singular of KI-VI), vyo for class 8 (plural of KI-VI), yo for class 9 (singular of N/N), zo for class 10 (plural of N/N and U/N), o for class 11 (singular of U/N), o for class 14 (U), ko for class 15 (ku), po for 16 (pa), ko for 17 (ku) and mo for class 18 (mu). I would have simply changed it myself but it's more than a simple substitution and requires adding more coding and since I am not sure I'm correctly replying or signing off here, that should really be done by someone else. In any case - this should be regarded as high priority as, as it stands, there is incorrect information in the tables. Janadume (talk) 16:01, 2 September 2017 (UTC)
 * Thank you for taking a look at this and contributing to Swahili entries around here! The relative forms are not actually incorrect, but instead reflect the fact that showing all possible forms would make the table unwieldy in size. The relative morpheme need not match the subject morpheme in noun class, because I could say "vitabu alivyosema" (the books which they read). I suppose that for intransitive verbs, most such mismatches would be quite rare, although you still get e.g. "alipokufa" (when they died). Of course, all of these forms are not too common in modern colloquial Swahili, which tends to avoid the relative forms as much as possible. Your suggestion of showing only forms of the -o- of reference that match the subject morpheme is not actually a bad idea, but either way, we end up glossing over the actual complexity (with the exception of readers who look at the linked-to appendix). —Μετάknowledge discuss/deeds 17:28, 2 September 2017 (UTC)
 * Thanks for your question. I hope you and Meta can work this out; in a tech industry standpoint, Meta is the product owner, and I am merely implementation guy. — JohnC5 21:12, 2 September 2017 (UTC)

error with monosyllabic verbs in the negative present
hi, i believe that monosyllabic verbs like -la, -nywa, and -ja should drop the infinitive 'ku' in the negative present, so for example the 1st-person singular present positive conjugation of nywa should be sinywi. but with sw-conj, it doesn't remove the 'ku' for that cell and shows sikunywi instead. it looks like there are still some uses of 'sikunywi' online, but not nearly as many as 'sinywi'. i'm going to try to fix this issue in the next few days but am open to feedback as to whether we should still keep the 'sikunywi' forms as alternates. thanks, --Habst (talk) 23:58, 24 May 2019 (UTC)
 * , you're right, it's sinywi in Standard Swahili and the module should definitely be changed (for space, I don't think we should include dialectal or nonstandard forms in the table). Thank you for going over it carefully! —Μετάknowledge discuss/deeds 02:40, 25 May 2019 (UTC)

Way too big
The table generated by this module is just huge, my browser hangs for a couple of seconds every time I dare visit a page including a Swahili verb. Could this not be trimmed? Swahili is agglutinative, such that a list of “selected forms” would be much more appropriate, as is the case with for example Korean verbs. MuDavid 栘𩿠 (talk) 03:47, 9 March 2022 (UTC)


 * Here. Jodi1729 (talk) 04:59, 26 November 2022 (UTC)
 * I know. MuDavid 栘𩿠 (talk) 04:31, 27 November 2022 (UTC)