As an anonymous user, you can only add new data. If you would like to also modify existing data, please create an account and indicate your languages on your user page.

Help talk:DefinedMeaning

From OmegaWiki
(Redirected from Talk:DefinedMeaning)
Jump to: navigation, search

Contents

Words and translations[edit]

Why the word must be translated ? And what if there is no such word in the language of the translation ? I think that the more important thing is to explain the original word, isn't it ? luna 20:03, 16 March 2006 (CET)

It is the combination of the word and the definition that must be translated. When you translate "horse" into Dutch it becomes "paard". The expression "paard" has multiple meanings, it is therefore important to translate bothe the definition and the word itself in order to know what is exactly meant.
When there is no corresponing frase in the "target" language, you will look for a word, a phrase that provides the best fit. It may mean that it only serves in translation but that it has to be used with care. In this case it is not a "native" or "indiginous" meaning and consequently a flag must be set to indicate that it works best in one way to this language. GerardM 16:49, 19 March 2006 (CET)
I still don't really understand the relation between Help:Expression, Help:Definition and Help:DefinedMeaning.
[...]
I think that actually a good example could be very cool.
luna 20:01, 19 March 2006 (CET)
I would say that here, expression = spelling (I would have called it spelling myself).
Definition is DefinedMeaning. The only difference between these 2 may be that DefinedMeaning is the concept itself (the fact that something has a definition), and Definition is the actual sentence that is used. (I see you're French, so if you want, we can discuss it in French).
Let's take colour for example. "colour" the word itself, with 6 letters is the expression. It has several definitions/DefinedMeaning (7 noun definitions and 5 verb definitions). There is also another expression in the same language that is linked to the same meanings: "color". In another language, it would be "couleur", and each 7 noun defs and 5 verb defs should be translated in French (and maybe they are not all definitions of the expression "couleur").
Now, some words can not be translated, as for example that obscure Dutch word: wadlopen which means (if I'm correct) walking in the water at low tide. In these cases, the definition can be translated in French, but there is not expression for that meaning.
I hope I'm right :) Kipcool 23:06, 23 March 2006 (CET)

On DefinedMeaning and its implementation[edit]

I would like to share some thoughts on the implementation of the database on which OmegaWiki is going to be based on. Although I am a very recent member of this site, I dare to write this note because I believe that ideas on the OmegaWiki engine have only a relatively short time window for discussion before they are considered approved and written in the marble (provided this hasn't happened yet). Let me know if it should be placed somewhere else.

In DefinedMeaning I find some contradictory definitions for a DefinedMeaning. In the preamble it is said that it is a table in a database. In the "Definition" section it is defined as "the combination of a word (so called Expression) and a definition in one particular language that describe a concept", which looks more like an entry in this table. The rest of the page does not clarify the real nature of DefinedMeaning, only how it should be treated. I tried to read other explanations in Talk:DefinedMeaning, in the talk section of other users on this site and on other source through the Internet, but I must say that none of these entirely clarified this concept to me. So, I will try to explain here what I have understood, or dreamt, about the subject, by means of an example. I hope that some member of the commission will tell me whether I am very far from the current implementation or not.

Let us assume that a random user, like me, wants to enter the Italian word conto in the OmegaWiki. With a dictionary at hand I first understand that this word has at least two meanings. So, I make a new entry in the words-it table, which could be:

          words-it --> 13 conto

Then I go to the meaning table meanings, and enter two meanings. These meanings should in general be more elaborated, but let us keep things simple, for the time being:

          meanings --> 45 it="un calcolo, l'atto di calcolare"
                       46 it="un registro di debiti e di crediti"

As you see, each meaning is identified by an integer number, 45 and 46 in this case. The fact that the table has a column for storing an explanation of this meaning in Italian is accessory, and, to tell you the truth, you do not need to understand these explanations to understand the functioning. Last, I go to a relation table and enter a relation between word 13 in words-it and meanings 45 and 46 in meanings:

          acceptations-it --> 13 45
                              13 46

Good. A simple query system on this database will now be able to tell me that conto has at least two meanings, and that their explanation in Italian is un calcolo, ... and un registro .... I am happy and I go to bed. But an English user, in the same evening, feels compelled to enter the word account. So, he follows the same strategy and enters:

          words-en --> 19 account

Then, he is magically informed that he is interested by meaning 46 (more on this later), so he goes to table meanings and writes:

          meanings --> 46 it="un registro ..." en="a record of debit and credit entries"
                       47 en="a statement explaining one's conduct"

(the meaning of 46 in Italian was, obviously, already there; he just added the English meaning). Then, of course, he must add:

          acceptations-en --> 19 46
                              19 47

Last, his French friend decides it is time to insert calcul:

          words-fr --> 35 calcul

          meanings --> 45 it="un calcolo, ..." fr="opération ou ensemble d'opérations"
                       48 fr="concrétion pierreuse qui se forme ..."

          acceptations-fr --> 35 45
                              35 48

This scheme has several advantages. First, with only this amount of data inserted, a query on conto (IT) will be able to return something like this:

          conto: 1) un calcolo, l'atto di calcolare (fr=calcul);
                 2) un registro di debiti e di crediti (en=account);

The fact that explanations are in Italian is not significant. If all meanings were also translated into Japanese, you could get an Italian vocabulary for Japaneses with translations in English and French. Actual strings could also be held somewhere else, for instance, all Italian strings in a meanings-it table linked to meanings.

Second, note that none of the three users had to know the exact correspondence of his words into foreign languages: as long as he is able to understand the explanations in the meanings table, he will be able to contribute (and the underlying engine will automatically discover links for him). This scheme could also be used to write a dictionary of synonyms for just one language

The fact that words (or, better, expressions) are linked through a table of meanings is very important. As you can see, if N languages are involved on the same level, the overall cost of data insertion and bookkeeping scales as N. This is in contrast to traditional two-languages dictionaries, where the cost for a full translation environment scales as N2.

In a personal page (User:Kipcool/all_about_DefinedMeanings) user Kipcool tries to express his understanding of a DefinedMeaning, which is not far from mine. The first and the second one are particularly interesting. How do we know when a meaning is already inserted in the meanings table? Note that this would be a problem even if all meanings were expressed in just one language, because there is no obvious way to browse through them.

All in all, I think that the DefinedMeaning concept is far from being understood by new users, and a more detailed DefinedMeaning page is absolutely required. --Bettelli 09:42, 22 March 2006 (CET)

First of all all Expressions belong to a Language. Therefore in an expression the language is implied. The consequence is that the definition for that expression will have to be in the same language. To translate completely into an other language, you will have to translate the definition as exact as possible as well as provide an Expression that is the translation.
When a new word is added, it is clear that a DefinedMeaning can exist for the concept. The consequence is that we will often have to merge new DMs into one. The result will be one DM that is richer in translations. GerardM 15:08, 23 March 2006 (CET)
I believe it is important to insist on this point, because it will be central to the development of OmegaWiki; I also find very strange that no one is talking about these issues here. Maybe this thread should be moved to a more visible place?
I agree with you that all expressions should be associated to a language. However, I do not see why "the consequence is that the definition for that expression will have to be in the same language"; this is no consequence at all, it is just, as I believe to understand, a design choice (and not very good, in my opinion). I can think of more than one reason why it should not be this way. Suppose I want to introduce the 28 or so known Etruscan words in OmegaWiki. Then, I will not be able to write the description in Etruscan ... Also, what if for some reason I want to introduce words from a dialect, but, at least initially, I do not want to introduce explanations in the same dialect? I think that an Italian dialect with definitions in Italian (then automatically linked to definitions in other main languages) would be valuable enough (in some sense, even more valuable than definitions in the original dialect). Won't I be allowed to introduce them?
Then, you write that "the consequence is that we will often have to merge new DMs into one". What does it mean? From the definition you give in the initial page a DefinedMeaning is an (Expression, description) pair. How can you merge two such pairs? How can you blend the two Expressions into one? In my opinion, Expressions can only be linked (through abstract meanings) and not merged.
But, most important, I asked whether my interpretation fits your idea of DefinedMeaning, and I can see no hint to an answer on this. What I want to understand is wether the very important relational layer behind OmegaWiki is considered to be fixed without further discussions. Because, if it is still open, I think it should be discussed better and with more details, and I would definitely like to participate to this discussion. --Bettelli 20:41, 23 March 2006 (CET)
You do not understand; an Expression and a definition in the same language are what makes a DefinedMeaning. This DefinedMeaning is translated consequently the definition is an integral part. When you cannot define the meaning of an Etruscan word, it has an isolated place in the database. When it can be defined, chances are that it can be associated with an existing DefinedMeaning. That will suffice.
When you want to add translations (=expressions) but no definitions to a DefinedMeaning, then you don't. It means that you agree with the definition but do not bother to translate the definition. Fine.
You can merge DefinedMeanings when you agree that both definitions define the same thing. The trick will be to choose which DefinedMeaning is to be merged into the other.
The relational layer is fixed. Several experts have expressed their opinion that it can work. But more importantly there was no other mechanism that would allow us to do the same thing. Now discussion to get a better understanding is encouraged. When you can prove that the concept of DefinedMeaning is wrong, we have a problem. GerardM 23:08, 23 March 2006 (CET)
I think you're speaking about the same thing, that is: what Bettelli is talking about is (almost) what will be in OmegaWiki. The hardest part here is to explain how it works so that everybody can understand it. A good start may be to add concrete examples to DefinedMeaning. Kipcool 23:44, 23 March 2006 (CET)

What I understand on the topic[edit]

I'm a little lost, is it like this?:

                          English                  French

                                                       ___________  synonyme
                                                      |
 Expressions                king                 roi == reine
                              |                   |      |    
    coupling ---------------->|                   |      |
                              |     Translation   |      |
 DefinedMeaning        he who governs  ====     ce qui gouverne


--Demos- 13:03, 25 March 2006 (CET)

This is how I believed (and still believe) it should be (although reine is the feminine of roi) --Bettelli 16:50, 25 March 2006 (CET)
Not at all. I will get you some diagrams on line that describe it better :) GerardM 14:17, 25 March 2006 (CET)
I have an ERD that is based on the tarball. I will upload this and explain a bit when I can upload it. Currently the size of files that can be uploaded is too limited .. I have asked Erik to ammend this.. Thanks, GerardM 17:12, 25 March 2006 (CET)

I've been wondering...[edit]

I have read aplenty of all these discussions already, and I had, in fact, a little question. I don't know if it is really a big thing, but I'd like to hear your ideas.

Just, for a moment, think in Vietnamese. You add the word tôi. The DefinedMeaning corresponding to this would be (in English) something like: "the speaker when he/she refers to him/herself". In English this would be I, so you do the things that need to be done that the system sees I as a spelling of "the speaker...". All right. You could go to bed, now, but you're enthusiastic so you add another word.

Anh. This means I when a big brother speaks to his little brothers or sisters. So you invent a good DefinedMeaning (something like: "the speaker when he refers to himself speaking to his little brothers and/or sisters") and I is, again, a good translation, so you do what is necessary. You're tired, you go to bed.

Now there's an English speaking person browsing the dictionary, wondering what I means. If I have this all right, she/he will now get two definitions:

  1. the speaker when he/she refers to him/herself
  2. the speaker when he refers to himself speaking to his little brothers and/or sisters

No he/she will be thinking: "Okay, the first one's all right. But what's the second definition all about? The first one catches it all already, doesn't it?"

This is just one of the many examples that could be found. Am I forgetting something? Do I have something wrong?

David 14:56, 3 April 2006 (CEST)

Of course, the definition 2 can be said I in English, but I doesn't mean only the def 2, it's a broader term (kind of an hypernym I would say). That is as if you attached the word animal to the definition of an horse. So, in my opinion, we should not link the def 2 as a definition of I. Now, what I don't know is how/where should we specify that Anh can be translated as I in English.
Maybe saying that I is an hypernym of def 2 would be ok? Then, the software would propose hypernyms as translations when there is no exact translation...? Kipcool 16:48, 3 April 2006 (CEST)
When a DefinedMeaning has a translation that does not fit its definition, it will get a flag set to indicate that the translation does best fit the DefinedMeaning but has to be used with caution. The Vietamese word Anh would have its own DefinedMeaning. It would translate to the English I just like tôi would. These DefinedMeanings will eventually not be shown for those people who do not have languages selected where the translation has this flag set to off. GerardM 18:07, 3 April 2006 (CEST)
Ah, I see. You've thought about everything. :-) It's getting more and more complex, but then languages are complex, aren't they? David 12:48, 4 April 2006 (CEST)
Naturally flagging a possible translation as "no good fit" is not sufficcient, I'm afraid. How would one deal with the case that anh must be used in the context where it applies, and tôi would be a outright wrong translation in this context? I do not know whether or not that's true for anh/tôi, but of course there are such cases, as well as more lenient ones. This requires collective (at least pairwise) flagging of the words of one language with respect to each of their possible common translations into the other language. Likely but not necessarily this relation holds for translations into other languages as well. -- Purodha Blissenbach 03:23, 10 July 2006 (CEST)
If you use an example only to say that you do not know if it applies, it is hard to accept it as a valid example. David does speak Vietnamese and he was happy with the answer .... GerardM 07:26, 10 July 2006 (CEST)

Who decides the definition?[edit]

Who decides the definition part of the defined meaning? If it is bad, how can it be changed - who decides on that, where, and when? // Habj 19:42, 2 May 2006 (CEST)

The best way is to have a great definition. The original language pair of an Expression and a Definition make up the DefinedMeaning. To appreciate if a definition is good or not can be best decided by people who know that language because semantic drift may cause the unease with a definition. When the definition is not good in any language, a new definition needs to be agreed on and, this definition needs to be verified against all translations of the Expression and yes, it needs to be translated to all languages. GerardM 21:06, 2 May 2006 (CEST)

Have I got it?[edit]

Sometimes I think I understand the concept clearly, then I have doubts. I will restate what I think a DefinedMeaning is here, and hopefully get my mistakes corrected.


A DefinedMeaning is a combination of an Help:Expression and a Help:Definition; an Expression being a given graphic form (Spelling) associated with a given Language; and a Definition being the concept expressed by this Expression restated in one or several sentences.

To a DefinedMeaning may be attached Translated Definitions, that is a translation adhering as closely as possible to the concept as described by the First Definition. Ideally all Translated Definitions should be translations directly from the First Definition to minimize semantic drift, though most likely in practice this will not always be the case.

To a DefinedMeaning may be attached other Expressions, which may be in any Language including the Language associated with the DefinedMeaning. There may be several Expressions from the same Language (i.e.,synonyms). Each such match between an Expression and a Definition may be marked as being a good match or not; i.e, if the Expression adequately covers the given concept semantically and/or stylistically.

A DefinedMeaning may be started from any Language.

If the match between a secondary Expression and a Translated Definition for the same Language is good, this combination may be considered a DefinedMeaning as well, and there is no need to add another DefinedMeaning to express that particular concept for the Expression again. However, the Expression might very well express other concepts and need to be associated with other Definitions to form other DefinedMeanings.

If the match is not good, it is not clear whether or not they should be considered as a DefinedMeaning or not.

DefinedMeanings may be related to other DefinedMeanings, and the relationship may be classified as f ex narrower term, broader term, part of theme etc. The exact set of classifications that will be used is not yet finalized.


Any comments and clarifications are gratefully accepted.--Sannab 22:17, 3 May 2006 (CEST)

This is a great summary .. Thank you :)
I think that when an Expression does not fit the DefinedMeaning well, there should be a Definition and therefore a DefinedMeaning that does it justice. GerardM 22:44, 3 May 2006 (CEST)
Now suppose someone removes all Expressions and Definitions... has the DefinedMeaning then disappeared? You can't view it any more, but is it still present in the database? —Vildricianus 23:07, 3 May 2006 (CEST)

A DefinedMeaning will not exist without its Definition and Expression. Delete it, and the DefinedMeaning will be deleted as well. GerardM 23:29, 3 May 2006 (CEST)

Perfect! Will the ability to delete Definitions be restricted to sysops? —Vildricianus 19:43, 4 May 2006 (CEST)
I have no clue how this will be done. I think something more elaborate may be needed. Something like a certain babel level combined with a sysop ? GerardM 20:04, 4 May 2006 (CEST)

Keeping semantic drift under control[edit]

As a freshman on this wiki, I might have got something wrong, but anyway I'll give it a try. I was wondering about keeping the semantic drift under control. It might be useful to be able to track where the chain starts from, for example by showing that the original Defined Meaning was in Italian, then someone translated the pair to English, and someone else from English to French, and a fourth from French to Spanish. This might trigger an invitation to people who know Italian and Spanish to actually verify, and maybe cast a vote on, whether the Defined Meaning in Spanish still corresponds to the starting point in Italian; it will also help people who know more than one language to choose the starting point that is nearest to the original, or at least as a reminder to compare it for reference. Sergio.ballestrero 18:14, 8 May 2006 (CEST)

There is as yet not software support for showing the original DefinedMeaning, but it is a top priority to implement this. Hopefully it will be implemented soon.--Sannab 19:25, 8 May 2006 (CEST)

what about pronunciation ?[edit]

Will it exist a mechanism to link some Expressions which represents pronounciation with their respective spelling ?

luna 20:59, 9 May 2006 (CEST)

The answer to this one is that it will be linked not to the Expression but to the SynTrans table.. An Expression can have multiple ways of pronouncing them. This does not apply at the SynTrans level.
There is a direct link to the Expression from the SynTrans. :) GerardM 23:35, 9 May 2006 (CEST)
I don't really understand what you mean. What is exactly the SynTrans table and how can it apply to pronounciations ?
luna 20:02, 18 July 2006 (CEST)


Source language[edit]

This article states even in bold that a defined meaning can be generated from any language but upon giving me my edit rights Gerard told me that the English definition is the canon that must be translated as litterally as possible. I am sorry but this does not compute. Those two statement are simply contradictory. If for whatever reasons defined meanings can only be given in English (yet?) that should be clearly stated

Jcwf 16:43, 11 May 2006 (CEST)

The current text is: In the future a DefinedMeaning may be started from any Language, and eventually the First Expression and First Definition will be marked or shown aside on the edit page. For the moment all DM's are in English'... Mind the words in the future and for the moment. Since 14/7 it is now possible at add DM, but it is not yet possible to mark a Fist Defintion or a First Expression. So for the moment it is still advisable to have a First defintion in English. HenkvD 20:14, 18 July 2006 (CEST)
Actually Henk, I added those words after making my comment. I would not mind an update of the current status though. IMHO this page must reflect current status.

Jcwf 152.1.193.141 22:57, 1 August 2006 (CEST)

you mean literal ?[edit]

I wonder about GerardM insisting for the definitions to be translated literally. Who can explain why it should be so ?

wiktionary : literal

  1. Exactly as stated; read or understood without additional interpretation; according to the letter or verbal expression; real; not figurative or metaphorical.
The literal translation is "hands full of bananas" but it means empty-handed.
[...]

I guess there is misinterpretation somewhere. How can you intend that we translate definitions literally, that is more or less word-for-word ? In most cases, a literal translation :

  1. sounds weird in the target language, even when mended to be grammatically correct, for instance "how do you go?" (for "comment vas-tu ?"),
  2. is semantically wrong, for instance "comment fais-tu" (for "how do you do?").

So I really wonder what's the goal, and if we should take seriously Gerard's insistance.

  • Either you really mean "literal", then I don't understand how the whole thing could work.
  • Or what you mean is that the definition should be translated by itself, as it is, not through the expression.

About that point, see user:spir/Notion.
--spir 09:32, 22 September 2008 (EDT)

Connotations, level of language, degree of specialisation, idiomaticity and degree of accuracy[edit]

An important point to understand is that, even in a given language, the correspondence between words and beings (or acts or attributes etc.) is not one to one. An expression implies more that the designation of the being, especially, it can include a connotation, a level of language, a degree of specialisation, indicate a regionalism etc..

  • "to get worse", "to worsen" and "to deteriorate" have the same meaning, but don't pertain to the same level of language.
    • Yes, we know that this is a missing feature...
  • A "beast" is biologically the same being as an "animal", but the connotation (the message I express by the choice of my word) is not the same.
    • If the definition is properly written, it should indicate something about the connotation. (Beast: a wild and agressive animal?)
  • To "breast-feed" is not that different from to "suckle" but the former is only for women, and the second for non-human female mammalians. (specialisation of words)
    • We usually indicate this in the definition. We can also add the information using the classes and annotations.
  • In English and French we have the Przewalski's Horse (cheval de Przewalski). This expression says that it's a kind of horse. But in Mongolian, the word tahi designating them as nothing to do with the word aduu designating farmed horses, which is also very different from the word mor designating ridden horses. Biologically speaking, the Przewalski's Horse is a particular subspecie: Equus ferus przewalskii, different from the usual Equus ferus caballus, both part of the Equus ferus specie. This means that the word horse can be use in a subspecific (in the biological meaning of the term: concerning one subspecie) acceptation, or in a broader one, designating any member of the specie. In zoology latin, the word "equus" (horse) is also used for the whole genus, which includes zebras. (degree of specialisation)
    • Each corresponds to a separate DefinedMeaning. This is what we do already. The latin words for the order, species and subspecies are further indicated in the annotations (see for example the annotations for DefinedMeaning:hyena_(736876)).
  • In Mongolian, the word unaga designates a less than one year old foal. Such a defined meaning doesn't correspond exactly to any English word or expression. If I have to translate it, I'll probably give a full translation (paraphrase) for the first occurrence: "less than one year old foal", but if the term appears 10 times in the text, I'll use the much more idiomatic "foal", which is less precise, for the 9 other occurrences. (degrees of idiomaticity and accuracy)
    • Already implemented with the checkbox "identical meaning".

All these facts should be recorded in the database, with precise links between defined meanings, and it's not the case. For instance, there should be a field to indicate that the subject of "to breast-feed" has to be a woman, and another field containing a link to "to suckle" as a correspondent word for other subjects. The defined meaning "to suckle" would in turn contain a field saying that its subject has to be a non-human female mammalian, and a link to "to breast-feed" as a correspondent word for other subjects. And so forth for the other relationships between words (connotation, level of language, degree of specialisation etc.). This implies that OmegaWike's idea of "class" should be

  1. clarified: should it be a class of the word (signifiant) or of the being (signifié)? A horse, defined as an animal, is clearly not a "specie" (one of that article's present classes), but the word "horse" signifies an animal according to its specie (while, for instance, the word "beast" signifies an animal according to its sub-human despicable nature). A specie is a class of animals, not a single animal. I don't understand what is supposed to be the difference between "class" and "theme"?
    • Class is what should be used. Putting an element in a class allows to add annotations to that elements, where the annotations are specific to the class (for example, putting the class "animal" allows to add the annotations about order, species, subspecies, ...).
    • Theme comes from the import of GEMET (a thesaurus with some relations between meanings), but we don't use it in practice.
  2. narrowed. Saying that a horse is a mammalian would be more useful than saying it's an animal (which would then become useless).

Both "less than one year old foal" and just "foal" should appear as English expressions for the "unaga" defined meaning, but with the attributes "not idiomatic, accurate" for the former, "idiomatic, approximative", for the latter.

    • No: "less than one year old foal" is the definition and "foal" is the translation that is to be marked as "not identical meaning".

All this information would be useful for human users and crucial for automatic translators. --Fiable.biz 16:09, 26 August 2009 (UTC)

    • The two dots comments are from me. --Kipcool 07:41, 3 September 2009 (UTC)
The problem with definitions and annotations is that they are not easily readable by a software. It's not enough to explain in a human-readable way the difference and relation between "to breast-feed" and "to suckle", Przewalski's Horse and "horse", "unaga" and "foal". This information should be computer-readable. See the discussion about multiple-inheritance hierarchy below. --Fiable.biz 02:34, 6 September 2009 (UTC)

What an automatic translator will need[edit]

If we want this dictionary to be usable by an open source automatic translator, we need of course information about connotations, level of language, regionalism, degree of specialisation, grammar, but it's still not enough. For instance, it's unusual to speak with a table. If one does so, he's regarding the table as a person. For a human reader, this is obvious and need not to be written, but for an automatic translator, this kind of things are essential. If I say in French "Catherine a cassé la table. Il faut parler avec elle." ("Catherine broke the table. We must speak with her."), the program has to understand that the complement of parler with the preposition avec has to be a person. Catherine is a female first name, so that Catherine is a woman, thus a person, but table is not a person. So the one to talk to is Catherine and not table, so elle has to be translated by "her" (and not "it"). This means that, for this dictionary to be used for automatic translation, we also need a multiple inheritance hierarchy of words. Catherine: French/English named woman: woman: human being: mammalian: vertebrate: animal: living being: being (I'm simplifying.). But also [French/English named woman: French speaking person: human being] or [French/English named woman: English speaking person: human being] etc. For each subclass of nouns, we need to make the often long list of verbs and nouns it can be agent, object or complement of, except those which are already in the bigger class. For instance, a mammalian can do and undergo all things a vertebrate can do and undergo, plus sucking and suckling. We also have sometimes to subtract or replace verbs of the super-class in one of the 2 languages, or both. For instance, a woman, though a mammalian, doesn't "suckle", she "breast-feeds", but the meaning is the same. This work has also to be done with verbs, adjectives, adverbs, though it might be more difficult. to breast-feed: to suckle: to feed: to bring up: to act.
Another thing to write is the words' etymological relations. For instance, if a text uses both "a drive" and "a driver", the automatic translator should try to translate them by two related words if possible, because this may be meaningful. These relations are both language-dependent and inter-linguistic. The latter characteristic may be important in case of a 2-language text, such as a text in a language quoting a text, or just a word, of an other language, as I've just done here above. So we need to record expressions relationships in a table of expressions, differing from defined meanings relationships to be recorded in a table of defined meanings.
If and only if we're able to describe languages with this level of precision, not in a loose human way as in Wiktionary, but with precise fields usable by a machine, then we can hope for a high quality automated translation thanks to OmegaWiki.
This is an adaptation of another post of mine, about another project, there: Grammar for automatic translation. --Fiable.biz 12:59, 26 August 2009 (UTC)

In the long term I agree. WordNet is a good example of what we could achieve with respect to multiple inheritance hierarchy. --Kipcool 07:43, 3 September 2009 (UTC)
I hope we'll do better that that. I couldn't find there any list of verbs a noun can be subject of, for instance. --Fiable.biz 02:37, 6 September 2009 (UTC)

Human reader's needs[edit]

Picture(s)[edit]

Compared to Wiktionary, this dictionary is not very beautiful, and not very popular. The possibility to upload pictures to defined meanings would be welcome. Moreover, it will ease the comprehension much. I'd need many words to define a дээл (defined meaning), the Mongolian traditional gown, as different from any other gown. But if I put a photo, you'll understand very quickly. The same is true for "horse", presently defined as "A large animal with four legs which people ride on or use for carrying things or pulling vehicles.". This definition could also fit for a camel (I took myself a photo of a camel pulling a chart in Sainshand, capital of the Oriental Gobi.). --Fiable.biz 23:40, 26 August 2009 (UTC)

In fact what we need is the possibility to have access to the images of Wikimedia Commons. There is a project that does just that, but I cannot find it anymore. Becoming an official project of the WMF would also be a solution... --Kipcool 07:47, 3 September 2009 (UTC)
This is it: http://www.mediawiki.org/wiki/InstantCommons --Tosca 14:26, 3 September 2009 (UTC)
I guess getting access to Commons' pictures is not very difficult and would improve much the appearance of this dry dictionary. --Fiable.biz 01:33, 6 September 2009 (UTC)

Examples[edit]

Examples are often easier to understand than definitions and would also be an important element to make OmegaWiki more popular. I think a few examples should be allowed in English, and translated into the other languages. If the define meaning is not an English concept (for instance a дээл, Mongolian traditional gown), then the examples should be in the source language, and translated into the other ones. --Fiable.biz 00:26, 27 August 2009 (UTC)

It is already possible to add examples. They are just too well hidden in my opinion. --Kipcool 07:50, 3 September 2009 (UTC)
This is why I didn't find them at first. But these examples are not translated. Moreover, I think that 2 kinds of examples are needed: examples of the expression (signifier)'s use in a sentence, as seems to be proposed now, and examples of signifié. For instance, it's very difficult to define what "Strepsirrhini" is exactly, and an example of this word's use taken from real world would be from a scientific paper and won't help much an ordinary reader. But saying that a lemur is an example of that sub-order would help. Giving an example of a sentence including the word "antonym" would not help much, but giving as example that "obscure" is the antonym of "clear" would help. Some lexical items will need both kinds of examples. --Fiable.biz 01:45, 6 September 2009 (UTC)

Need for a policy to produce DMs[edit]

Words are about cutting and dividing the reality. A word's domain stops at another word's domain. All languages don't cut the reality exactly the same way. The cutting borders are not exactly identical. Even inside one language, the field, the context, the groups of people may lead to different domain borders for a given word. The more educated people are, the more they can sub-cut and distinguish several "meanings" for a word, though, in every day life, the word is usually thought as having one meaning. For instance, the word "child" seems quite clear and should have exact synonyms in all languages because this reality exists everywhere: a young human being. In fact, what is not clear is the limits. The UN Convention on the Rights of the Child, states, art. 1: "For the purposes of the present Convention, a child means every human being below the age of eighteen years unless under the law applicable to the child, majority is attained earlier." so including foetus and embryos (the preamble says: "the child, by reason of his physical and mental immaturity, needs special safeguards and care, including appropriate legal protection, before as well as after birth"). But even this technical definition explicitly depends on... the countries legislation. The legal age of majority goes from 9 years old (for Iranian females) up to 21 years old (in several countries including Madagascar or Mississippi). In other contexts, languages, cultures, the same word used to translate "child" in this UN Convention would be opposed to "foetus" or, more often, to "teenager" (a word itself depending strongly on the English language's numbers in "-teen"), or to "adolescent". In Mongolian, the same word is also used to refer to students, by contrast with the university's teachers and staff. In many languages, including English, the same word is also used to refer to filiation: one's children can be 40 years old. Moreover, the UN definition is based on the notion of "age", counted from birth in most countries, but from conception in China and sometimes in Mongolia. In French, the same word can also be feminine : "une enfant", which, according to traditional French grammar, is to be regarded as a different word. So, how to choose and redact DMs? In other words, how should Omegawiki cut the reality?

  1. A solution is to cut the word's domain in sub-domains which will be the intersection of all meanings and will be our DMs, so that these DMs don't overlap but form a partition (in the mathematical meaning of the term: not intersecting subsets whose union is the full set) of the whole domain of the word. In this case, DMs for "child" should be "female embryo", "male embryo", "female foetus", "male foetus", "female born human being having reach neither 13 years after birth neither legal age of majority", "Iranian female between age of majority and 13 years after birth", "male born human being having not reach 13 years after birth", "female human being born at least 13 years ago but less than 18 years ago and not having reach age of majority" etc. In this case, the actual domain of one language's word for "child" would be a grouping of some of these very narrow sub-domains.
  2. Another way would be to say that the DM has to describe the full actual domain of the word in a given language. So, in English, the unique DM of "child" could be "Human being regarded as young or as born from another one". Such a wide definition would get exact translations in many languages. But it's very imprecise and hides the complexness of the word usages. We miss the context precision. A given person would be designated as "child" by law, but not at school. The definition doesn't say this. Neither does it says that a 22 years old student may be a "хүүхэд" ("child", in Mongolian) but in no circumstance a "child", except as someone's child. This proposal has the advantage of keeping the general idea behind the word, but hides the fact that the implementation of this idea differs according to languages and fields.
  3. Another solution can be to define DMs as the actual domain of the word in the combination of one language, one field and one grammatical word (same word class, same gender, same inflexion group etc.). So, in law context, a DM could be UN Convention on the Rights of the Child's article 1. But the convention says "For the purposes of the present Convention", and it's probable that, even in the domain of law, there are other definitions incompatible with this one, so that the "law" domain should be subdivided into "international law"→ one DM, "criminal law" → another DM, civil law → another DM etc.. The same is probably true in medicine: I bet the psychiatrist doesn't have the same definition of "child" than the cardiac surgeon. And the geneticist is above all interested in the biological filiation meaning.
  4. Yet another proposal is solutions 2 and 3 together: including in OmegaWiki both the widest DM, giving the general idea of a word, and the precise DM, and a relationship saying that these precise ones are implementation of the widest one.

There are certainly other solutions. I think that the 4th solution above is the more fruitful. Although it leads to many DMs, there will be less than with the 1st solution. All the 4 solutions hides hyperonymy and other approximations. In fact, the Mongolian word "хүүхэд" has a wider usage than the English "child", so that there are of course contexts where translating one by the other will produce things difficult to understand. For instance, if a Mongolian says in Mongolian : "I've been a хүүхэд for 22 years, but I'll graduate in June.", the sentence is difficult to translate into English because it's related to the meaning of "хүүхэд" in Mongolian, which, in solution 1 and 3, are split into several different "meanings". And solution 2 fails to say that the general idea of "young human being", which lies under this sentence, has not the same concrete application in Mongolian and in English. However, we need a policy to produce DMs, i.e. a policy to decide how narrow the definitions should be, how to split a word's meaning into different meanings.--Fiable.biz 18:24, 11 February 2010 (UTC)

I agree that 4) is a good solution. We need a general DM and special DMs. Every legal system probably needs its own definition of things like "child", "murder" etc. Maybe this could be realized as DMs with SubDMs.
I'm also concerned about words "losing" their true or exact meaning because they only appear as translations to DMs that were created with a different language in mind. Maybe in the long run we could also have monolingual dictionaries within OmegaWiki to make sure that all words are properly defined and then sort of mapped to each other.
I find it hard to establish a policy right now, because Omegawiki is still so new and also small. I would prefer the "learn as you go" approach for now. If a specific problem arises, we can discuss it. --Tosca 15:55, 13 February 2010 (UTC)

There are several topics here and need to be treated separately. I will start by writing down some thought, hopefully we can crystallize some general principles from this discussion
  • We have to accept the fact that certain words are purposefully ambiguous. An advertisement that says "We organize birthday parties for children" uses the word child vaguely and a dictionary should provide a definition that covers this vagueness. On the other hand there are technical texts that carry a very precise meaning when they use a word and there is a need for their definitions as well.
  • We have to distinguish between lexical and non-lexical definitions. The age at which a person is no longer a child is a matter of legal definition and can be avoided in a lexical definition. You can simply define child as "a young person who has not reached a the age of majority" and leave it at that. The phrase "age of majority" masks all the ambiguity about children in Iran or Mississipi. Definitions should be general enough to cover all variations of a concept.
  • Some languages don't have words for separate concepts, so a definition may have many "or"s in it. Consider the word aunt which may be defined as "a sister of a mother or sister of a father, or the wife of the brother of the mother or the wife of the brother of the father." Turkish has separate words for each of these relatives, so aunt really should have a separate definition for each of these relation terms. In general, if a definition refers to specific entities separated by the word "or" we should consider splitting it.
  • However, the last rule I suggested made me think of the definition of finger. You can think of defining finger as "the thumb or the index finger or the middle finger or ...". Should the definition of finger be split into five: 1) thumb, 2) index finger, 3) middle finger, etc? No, I think this is a different case. Finger can sometimes be used ambiguously on purpose as in "he lost two fingers in the explosion" and sometimes specifically as in "after the vows the groom put the ring on the bride's finger". So, this is a similar situation to the first point I made above about "child" used ambiguously.
  • The example about aunt I gave above also made me think about an alternative definition. You may think of defining aunt as "a female member of one's extended family in the same generation as one's parents". This definition suddenly simplifies the messy definition with three "or"s but requires an abstraction. Interestingly, there is no Turkish word that corresponds to this broader concept. So, should the definition of aunt be split into the definitions of the four types of family members it refers to, or be kept as the single more abstract definition? I guess the answer should be both.
  • Just as Wiktionary usesw the motto "all word in all languages", perhaps OmegaWiki should have the motto of "all definitions in all languages".
--InfoCan 14:10, 12 August 2012 (CEST)

InfoCan just wrote: "should the definition of aunt be split into the definitions of the four types of family members it refers to, or be kept as the single more abstract definition? I guess the answer should be both.". This is a possible answer, whose consequence is that any expression of any language thought, in that language, to have one meaning should be linked with one definedMeaning, even if, in other languages, it is considered to be several meanings or, at the opposite, a specialised meaning. For instance, in Mongolian, there are special words for "a foal of less than 1 year old", "a foal between 1 and 2 years old", "a foal between 2 and 3 years old", "a calf of less than 1 year old" etc.. At the opposite (as in many languages) there is a word meaning "younger sibling" and words meaning "relative from my mother side", "relative from my father side". In Malagasy, there is also a word meaning "older sibling". All these ways to cut the reality should be definedMeanings. I have to think a bit more, but I think I can agree with your idea, though this will imply quite many definedMeanings. --Fiable.biz 05:50, 21 August 2012 (CEST)

Don't verbs and corresponding nouns of action express a same meaning?[edit]

I wonder whether "to act" and "action", "to delete" and "deletion" etc. are not a same definedMeaning? The difference is syntactical, not semantic, is it? Is there any semantic difference between "I regret my action." and "I regret having acting this way?", between "I saw his jump." and "I saw him jumping"? In many cases, the use of a noun of action or of a verb is language-dependent or, as we've just seen, style-dependent. For instance, French language uses extensively nouns of actions, English language, less, and Mongolian language, even less. Infinitive and participles can be used as nouns, and, in Mongolian, they are very developed (accepting many more that the 2 tenses they can have in English, voices, aspects etc.). Moreover, quite often one of the 2 (noun of action, and verb) just doesn't exist. For instance, in contemporary French, there is no noun of action meaning "the action of following" in the literal meaning. In the figurative meaning we can say, for instance "La suite du Christ est difficile." ("Following Christ is difficult."), but we would not say "La suite d'un camion est difficile." ("Following a lorry is difficult."). But such a word exists in Malagasy. At the opposite, in English there is the noun "crisis" and no verb "to crise" (One cannot say "Wall street crised in 2008."), but such a word exists in Mongolian. Merging verbs and noun of action would reduce the number of definedMeanings, and make the "part of speech" ("word class" would be better) a crucial piece of information.--Fiable.biz 06:26, 21 August 2012 (CEST)

I think this question goes into the difference between interpretation and translation. If you want to convey what is meant (interpret) by "I regret my action" in a natural way, you may choose to say, or even, may have no choice but to say, the equivalent of "I regret having acted this way". However, speakers of a language that distinguishes between these two forms of expression, would disagree with you that they are equivalent. For example "action speaks louder than words" doesn't quite say the same thing as "to act speaks louder than words", because here action is the subject of the verb to speak. I think we have to acknowledge that certain concepts don't have expressions in every language. Certain DMs will just not have SynTranses in every language. We'll have the DM for "a foal between 2 and 3 years old" and it will have a corresponding expression in Mongolian but no SynTranses into English, French, etc.
It seems we have to bring back the approximate meaning field back, and just make sure that it is used properly. --InfoCan 18:09, 21 August 2012 (CEST)
A native speaker would find a difference between "It is better to give than to receive." and "Giving is better than receiving.", at least because the former is a proverb (so a lexical item) and the latter is not and because of the order of the terms (A gerund is place at the beginning of the sentence.), or between "It's no use crying over spilt milk." and "It's no use to cry over spilt milk", the former being a proverb too. But does this imply a so big semantic meaning that we would need 2 definedMeanings for "giving" and "to give", "to cry" and "crying"? No. The semantic differences are, to my mind, those of the gerund and the infinitive in English. The same way, the small semantic differences you point out seem to me included in the general semantic difference between a noun and a verb, so that noting the word class would provide for the whole information. In many languages, like in French, you can only use one form to express things like "It is better to give than to receive.", you don't have the choice between gerund and infinitive. In Omegawiki, we already accept some small semantic differences inside a definedMeaning, such as those in the "usage" field ("colloquial", "unofficial" etc.). And the slight differences you point out seem to me no bigger than those ones. I'm wondering if we are not just injecting our grammatical categories into the semantic field, and pretend that nouns and verbs are different meanings just because this is a word class difference, but gerund and infinitive have the same definedMeaning just because it's an inflectional difference. But the very notion of word class itself is a partly artificial. "following" is classified as a noun because it behaves as many other words that can't be regarded as inflections, like "table" or "chair". But this lexicalisation in "-ing" is so frequent in English that we could nearly regard the noun "following" as an inflected form ("substantivised form") of the verb "to follow", and accept that this form doesn't exist for all verbs (just as the 1st person of singular doesn't exist for some verbs: there is no form "I rain.") and accept there are (many) verbs with irregular substantivised form, like "to deleted" whose irregular "substantivised form" would be "deletion" instead of "deleting". Unfortunately I don't know any language with no noun of action, but I bet there are. And, in very old languages, such as proto-indo-european, and even proto-Greek, things now expressed by verb inflection (specially the aspect) would be included in the verb root itself. Even in contemporary French, some old verbs have 2 or even 3 roots (for instance "aller": "all-", "v-" and "ir-") which are artificially considered as forming one verb's inflection forms, just to follow the way the other verbs inflect, but these were different words. What is the difference between an inflection and a lexicalisation? Mainly the systematicity of the process. If the process is nearly systematic, then the brain just remembers the inflection rule, and exceptions, like for the English preterite "-ed", and a quite long list of irregular verbs (like "came" instead of "comed"). When the exceptions are too many, the brain cannot work like this and has to remember all the words, one by one, such as the actor ending "-er": we say "killer", "teacher" or "worker", but "president" (not "presider") or "prostitute" (not "prostituer"). A proto-european man would possibly not regard "eating" and "have eaten" as the same definedMeaning because, in his language (that I must admit I don't know) they might have been expressed by 2 different words. But since we can now express this difference in a systematic way, I think we can regard this as one definedMeaning, and say the proto-indo-european man: we will regard your 2 verbs as one irregular verb with 2 roots. To take another example, in Malagasy, there is a circumstantial voice. In French, we have the form "ouvrant" ("opening"): active voice, "ouvert" ("opened"): passive voice, and "ouvré" ("when is opened"), in "jour ouvré": "working day". We regard "ouvré" as an adjective, so a different word, because there is generally no "circumstantial voice" in French (except for this verb). But we could (and I would say we should) consider, following Malagasy language, more developed as far as this is concerned, that "ouvré" is the circumstantial participle of the verbe "ouvrir". Yet another example: in English, "to go" and "to send" are 2 different words, as well as "to appear" and "to see". But in Mongolian, whose word building is far more logical, "to send" is just an inflection of "to go": its causative voice ("to make go"), and "to appear" is also the same verb as "to see", at the passive voice ("to be seen"), though the passive of "to look at" is more used. So I would say that since there is a systematic way of producing the causative meaning, there is no need of a definedMeaning "send" different from "go". We can regard "to go" and "to send" as one verb, "send" being its irregular causative voice. (For "to appear", I can see a small difference from "to be seen", so I would let 2 definedMeanings). The same is true for "dog" and "bitch", "king" and "reining queen" etc.: I would say that, even if in English these differences are lexical, this doesn't create 2 definedMeanings, but one, whose difference be noted by the "gender" field. The same way, I would say to the modern indo-european man, whose substantivising systems are quite disordered and archaic, that his substantivised forms, expression actions, of verbs don't form new definedMeanings, but will be regarded are more or less irregular forms of the verb. And, of course, we have to record in OmegaWiki the irregular forms, as in good paper dictionaries, because they are lexical items. Of course, we can decide, at the opposite, that anything regarded as a word in any language should have its definedMeaning even if, in other languages, the difference is just inflectional. But this means many definedMeanings. Moreover what about the irregular "inflections", such as "came": will they need their proper definedMeaning too? --Fiable.biz 05:39, 22 August 2012 (CEST)
You make interesting arguments. Our current community consensus is to not include into OmegaWiki inflectional forms of verbs, and simple inflections of nouns (like plurals). But there are many other inflections in various languages that can generate nouns from nouns, verbs from nouns, nouns from verbs, verbs from verbs, etc. Such inflections are not universal: in Turkish the word for bed, yatak is derived from the verb yat (to lie) and the suffix -ak, which is used in a number of other words also to mean "place where". But I think most speakers of other languages would object to make the noun "bed" an inflectional form of the verb "to lie". This may be an extreme example but my point is that it is not obvious where to draw the line to decide what constitutes an inflectional form. My inclination is to go with your last suggestion, that is "anything regarded as a word in any language should have its definedMeaning even if, in other languages, the difference is just inflectional."
I am also inclined to think that part of speech is an important aspect of a DM, and for that reason all SynTranses that are labeled identical_meaning for a DM should have the same part of speech annotation. If you need to translate "giving" into a language that doesn't have gerunds, then "to give" can be given as an non-identical meaning. As I have said before, we have to admit that not every language has a word for every concept that exists in other languages.
These are my initial thoughts and I may be dissuaded later. It would be good to have more participation in this discussion. --InfoCan 17:51, 23 August 2012 (CEST)
Maybe the difference between derivational affix and inflectional affix there could help you. We could use it, but it is again a language-dependent criterion, so I think it is not relevant for definedMeaning.--Fiable.biz 11:39, 24 August 2012 (CEST)

Policy to create separate definedMeanings or not[edit]

We have to answer clearly the following questions and put the answer in the help page. Do we need to create separate definedMeanings in each of the following cases:

Different words in any language to express noun gender difference[edit]

Example: "dog" and "bitch", or in Mongolian тэмээ (male camel) and ингэ (female camel). The "gender" field would be enough.

  • The practise seems to be 2 definedMeanings.
  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)
    • Oppose bitch is not just a female dog - has other meanings as well just like the Dutch word teef an impolite way to address a woman.  Klaas|Z4␟V:  09:53, 31 July 2013 (CEST)
OK. What I meant is one DM for both "male dog" and "female dog". --Fiable.biz (talk) 11:47, 31 July 2013 (CEST)
With the idea of 'all dogs are equal ? Au contraire monsieur/madame/mademoiselle - there are thousands different kind of dogs not just what gender is concerned. Perhaps the word horse in Dutch (paard) is a nice example (scores of different types here as well). A male we call 'hengst' female 'merrie' (different meanings here too - both though the latter more in combination with 'nacht' [i.e. 'night']) so I insist: bad idea IMNSHO I leave the example 'cat' (also Gaelic! for an eventual "adult" section ;-) Kind regards from Arezzo province,  Klaas|Z4␟V:  10:23, 3 August 2013 (CEST)
My idea is not one DM for all the different kinds of dogs, but not to create 2 DMs only based on sex. I had forgotten the neutral. In the "paard | hengst | merrie" example, there would be, according to my proposal, 1 DM. The rationale is that this difference is so systematic that it can be expressed by a simple "sex" choice (neutral | male | female), without having to repeat all definition translations (with just the difference "male"/"femelle"), word translations etc.. Moreover, in many languages, the difference is expressed quite systematically too, nearly like an inflexion, or by adding the words "male" or "female". In French, nearly all the professions take a feminine form, like "professeur" → "professeure". The same in Spanish ("profesor"→"profesora"). If I understand you correctly, you would like 3 DMs in such cases: "teacher", "male teacher" and "female teacher". And what would you do for "ant": create one DM, waiting to see if there exist a language distinguishing male and female ants, or creating in all cases 3 DMs too? Please try to answer to following questions too. --Fiable.biz (talk) 13:40, 3 August 2013 (CEST)
Looked at it from a monolingual standpoint of using OmegaWiki, one would likely expect three different definitons for a triple of words like the Dutch "paard | hengst | merrie" since one would likly want to be informed precisely enough so as to make an optimal choice between them when e.g. translating or writing or speaking. --Purodha Blissenbach (talk) 17:00, 10 October 2013 (CEST)

Irregular inflexion in any language to express noun gender difference[edit]

Example in French: âne (male donkey) and ânesse (femelle donkey).

  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)
Now I understand finally ;-) a bit what you mean. Simplification. What about e.g. mule a cross between a horse and a donkey (in two variations they might be called differently in some languages)  Klaas|Z4␟V:  17:55, 4 August 2013 (CEST)
The French âne / ânesse is a clear cut case of two definitions necessary. Two different words, two different meanings. --Purodha Blissenbach (talk) 10:24, 3 November 2013 (CET)

Different words in any language to express adjective gender difference[edit]

Example: "beautiful" (for a woman) and "handsome" (for a man).

  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)
Agree In many languages one uses the same word 'nl:mooi' or almost 'it:bell{o|a}'  Klaas|Z4␟V:  17:59, 4 August 2013 (CEST)
  • 1 DM is imho complete nonsense here, since the two words have almost nothing in common, meaingwise, and their use does not relate to gender either. --Purodha Blissenbach (talk) 19:23, 5 September 2013 (CEST)

Irregular inflexion in any language to express adjective gender difference[edit]

Example in French: "beau" (beautiful, for a masculine noun) and "belle" (beautiful, for a feminine noun).

  • The practise is: one definedMeaning.
  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)
    • See previous - positive  Klaas|Z4␟V:  18:02, 4 August 2013 (CEST)
  • Impossible to get along with one DM since from one either you can infer gender, if used on their own, thus meanings differ clearly. --Purodha Blissenbach (talk) 19:24, 5 September 2013 (CEST)

Different words in any language to express verb voice difference[edit]

Example: "to see" and "to show" (="to cause to see", causative voice, inflexion of "to see" in Mongolian), "to learn" and "to teach" (="to cause to learn", inflexion of "to learn" in Malagasy), or "to cook" and "cooker" (="the cooking agent", agent voice, inflexion of "to cook" in Mongolian).

  • The practise is 2 definedMeanings when these the voice is not regarded as such in the traditional grammar of that language (in which case the speakers regard the 2 words as different) but 1 definedMeaning when the voice is regarded as such the traditional grammar of that language (in which case the speakers regard the 2 words as just different forms of one verb).
  • My suggestion: 1 definedMeaning for the active (for instance "make"), the passive ("be made"), the agent ("maker"),
but separate DMs for every kind of causative provided that, in any language, it is regarded as a word different from the word expressing the active. For instance "teach" (something) should be a different DM than "learn" because, in English, it is a different word (though in Malagasy, it is not.) and "teach" (as an activity) would be yet a different word because, in Mongolian, it is a different word ("багшилах"), though in English it is not.
I propose also different DMs for the circumstantial voices provided that, in any language, it is regarded as a word different from the word expressing the active. For instance French "ouvré" (corresponding the English adjective "working", as in "working day") would be a different DM from "to work" because in French it is regarded as a different word.
The rationale is that there are very many kinds of possible causative and circumstantial voices (theoretically 768 possible kinds of causatives, following Dixon's classification), that no language treat all of them inflectionally and that the number of possible causative or circumstantial voices seem to depend much of the action. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)
  • Whether of not such verb form are regarded separate verbs (lexemes) or not is imho rather unimportant. By the way, there are not so few even major languages where various classes of scholars support either view, and we cannot and must make not decisions on their behalf. We do't have to. I'ts merely a matter of how data are entered than of anything else. For instance, if you have word pairs such as the Russian покупать and купить, they can be possibly automagically entered and annoted and a meaning stored from one another, possibly not. Meanings of grammatical forms or inflected forms may be seen under varying formalsims depending on the language we are using to describe them. Another example is the word group of English: little house, German: kleines Haus or Häuschen, Turkish: evjik. All of them mean the same, but the first two forms happen to be periphrasistic and thus can be analyzed further, the other two are derivational, and inflectional plus agglutinative as well, in that order, and can be further analyzed only on the morpheme level, but not as separate words. --Purodha Blissenbach (talk) 20:32, 5 September 2013 (CEST)

Irregular inflexion in any language to express verb voice difference[edit]

Examples: "breaking" and "broken", "withdrawing" and "withdrawn", in French "prenant" (taking) and "pris" (taken).

  • The practise is: 1 definedMeaning.
  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)

Different words in any language to express verb mood difference[edit]

Example: "I am" and "I would be" ("am" and "be" clearly don't have any common etymology).

  • The practise is: 1 definedMeaning.
  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 13:59, 3 August 2013 (CEST)

Irregular inflexion in any language to express verb mood difference[edit]

Example in French: "je suis" ("I am", indicative) and "que je sois" ("I be", subjunctive).

  • The practise is: 1 definedMeaning.
  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 13:59, 3 August 2013 (CEST)

Different words in any language to express verb tense or person difference[edit]

Examples: "I am" and "you are" ("am" and "are" clearly don't have any common etymology), "I am" and "I was" ("am" and "was" clearly don't have any common etymology), in French "je vais" ("I go") and "j'allais" ("I was going"). This also includes some Latin irregular verbs and Greek thematic aorists. Since, in modern languages, normal verbs are supposed to have all tenses and persons, these words of different origin are regarded as different forms of one verb.

  • The practise in to use one definedMeaning.
  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)

Irregular inflexion in any language to express verb tense or person difference[edit]

Examples: "come" and "came", "break" and "broke", in French "prends" (take) and "pris" (took).

  • The practise is: 1 definedMeaning.
  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)

Different words in any language to express noun connotation difference[edit]

Example: "animal" and "beast", in French "résistant" ("Resistance fighter") and "terroriste" ("terrorist").

  • Kip said here above: 2 definedMeanings.
  • My suggestion: 2 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)

Different words in any language according to the level of language[edit]

Examples: "to get worse", "to worsen" and "to deteriorate" or in French "réanimation" (intensive-care medicine) and "réa" (colloquial word).

  • The practise is to use one definedMeaning, and the "usage" rolling list.
  • I entirely agree with one definedMeaning: this is a language difference, not a semantic one. The "usage" rolling list should be more complete, including "formal" and "literary".--Fiable.biz (talk) 12:42, 28 July 2013 (CEST)

Different words in any language to express similar things according to the subject[edit]

Examples: "to suckle" and "to breast-feed", "paw" and "leg", in Mongolian "гар" (hand) and "мутар" (respectful word for God's hand or a king's hand).

  • Kip said here above: 2 definedMeanings.
  • My suggestion: 2 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)

Affix in any language to express similar things according to the subject[edit]

Examples: "hand" and "handy", in French "main" and "menotte" (child's hand). These examples are not very good, because the difference is not only of the subject but convey a connotation. I'll think of better ones. --Fiable.biz 10:53, 24 August 2012 (CEST)

  • My suggestion: 2 definedMeanings. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)

Different words in any language to express word class difference[edit]

Example: "to ask" and "question", in French "entendre" ("to hear") and "audition" ("hearing"). The "part of the speech" field would be enough.

  • The practise: 2 definedMeanings.
  • InfoCan wrote here above: 2 definedMeanings.
  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)

Affix in any language to express word class difference[edit]

Examples: "to delete" and "deletion", "to preside" and "presidence". The "part of the speech" field would be enough.

  • The practise: 2 definedMeanings.
  • InfoCan wrote here above: 2 definedMeanings.
  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)

One word considered as two in any language because it covers 2 word classes[edit]

Examples: "following" (gerund of the verb "to follow", and noun), "discouraged" (part participle of "to discouraged", and adjective). Note that, even in one language, this depends much of the grammar. For instance "very" can be seen as having one word class: "adverb", or 2: ad-adjective as in "very tall" and ad-adverb as in "very well".

  • My suggestion: 1 definedMeaning. --Fiable.biz (talk) 00:29, 21 March 2013 (CET)
    • If meanings are indeed identical, which happens not to be the case in some of those instances in some languages, e.g. German, then we can maybe get along with one definition. Yet gammatical annotations have to of cource be employed to make them two or more distinct words. In the same context, see English, where appending "-ly" creates adverbs from most adjectives. But there are notable exceptions. Neither *smally nor *litt(e)ly nor *bigly exist in English, while largely and widely are widely used in English. Which means that, when we want to make the English adjective→adverb rule automatic, we have to ask for user confirmation when it is applied - unless the exceptions are in a known closed word class or doubtlessly enumerable list, which I do not know at the moment. --Purodha Blissenbach (talk) 20:51, 5 September 2013 (CEST)

Warning: Omegawiki is not only Indo-European[edit]

Although many people in this planet (in Europe, Russia, India, Central Asia, Middle-East, North and South America, Black Africa) and in Omegawiki speak Indo-European languages, if OmegaWiki wants to be for all languages, our approach of these subjects shouldn't be too much Indo-European-centred. Let's not forget there are also quite many Chinese, Japanese, Indonesians etc.. More precisely,

  • our word building process and the very limited number of our voices make us think some words as having different meanings while other languages have a much more systematic way to produce words and voices, so that the semantic difference is sometimes for them just an inflectional difference. This is even truer for English, a language with an hybrid origin (old German and Norman languages) with a huge vocabulary but lacking logic, specially in word structure. On the other hand, the absence of logic makes clearer subtle differences which can be hidden to speakers of more systematic languages. For instance "to send" is not exactly "to cause to go" as Mongolian language pretends it: it is to be the initial cause of the moving, not a continuous cause as for "to propel".
  • A word class is a class of words which can play a same set of roles (grammatical functions) in a sentence. For instance, a verb is a word that can easily be predicate. But verbs gerund or infinitive can be subject or object too, like nouns and this fact doesn't, as we understand it, make a participle or an infinitive enter the "noun" class.
            squeezing in: I disagree for many languages --Purodha Blissenbach (talk) 21:02, 5 September 2013 (CEST)
    Our word classes meet reasonably the need of our languages, but not all needs of all languages. Moreover, other classification are possible. For instance, Mongolian traditional grammar has a "place and time word" class because these words can modify the verb, like an adverb ("He will come tomorrow.") or be subject or object, like a noun ("Tomorrow is another day"). The same fact is interpreted by Oxford dictionary by 2 words "tomorrow": an adverb and a noun, because there is no "place and time word" class in traditional English grammar. At the opposite, there is no "adverb" class in traditional Mongolian grammar, because most of what we call "adverbs" are just adjectives. This fact also exists in English for quite many words, such as "a fast horse" and "He runs fast." but English grammarians say these are 2 words "fast" while Mongolians would say it is the same used in 2 functions. And there is, in Mongolian, a small "decorative word" class for the few words which can modify some words meaning but never a noun, such as "very" or "much". So basing the definedMeaning rules on Western grammar seems to me quite arbitrary and inappropriate. But I must admit it is easier to understand by the majority of OmegaWiki's present users.

I'd prefer some strict logic in our policy, even if it's at first not so intuitive for Indo-European speakers. Logic rules, once understood, are quite easy to remember, and thus to apply. To my mind, definedMeaning rules should be based on meaning, not on particular grammars and languages, and expression rules should be based on language. --Fiable.biz 10:46, 24 August 2012 (CEST)


What may be logical to one person may appear bizarre to someone of a different linguistic background. My knowledge of linguistics is limited so I am not sure whether we can come up with rules that would be logical to every person of the world. I think our policy should be to avoid surprise for the end user, whether their language is Indo-European or Ural-Altaic. This would mean going for the common principles and, otherwise, be as inclusive as possible. For this reason, we may end up using classifications that may seem inappropriate or redundant to speakers of another language group. This may well be unavoidable when creating a universal dictionary. --InfoCan 17:06, 24 August 2012 (CEST)

To my mind, dictionaries without surprise already exist, for instance Wiktionaries. OmegaWiki is a good surprise. --Fiable.biz 09:01, 27 August 2012 (CEST)

Thinking more on the subject, I have to agree with most of what you say. If a language does not have all for of speeches that another language has, of course SynTranses cannot be constrained to have the same form of speech. An infinitive in one language may well be the translation of a gerund in another language.

To logic behind certain forms of inflections may be opaque to speakers of other languages. As you pointed out, it is not obvious that send is the causative form of to go (to make go). I could argue that to take away or to chase also have the same meaning. So, the fact that in Mongolian to send is a derived form of to go should not interfere with how speakers of other languages create DMs for either of these verbs.

On a related topic, because inflectional forms vary from language to language, I think they have to be indicated as annotations at the language (SynTrans) level and not be used to create new DMs. For example, not every language has the genitive case, so the genitive case of a noun should not be a separate DM. However the spelling of this inflected form should be searchable in the database, and be linked to the lemma (root word) through the annotations. --InfoCan 17:51, 26 August 2012 (CEST)


Causatives[edit]

In fact I understood that "causative" is a large notion. See Wikipedia: there are several kind of causatives, even if the languages usually don't acknowledge this fact, so that they often specialized there causative voice (as in Mongolian the causative of "to go" means "to send", not "to propulse" or "to drive") if they have one, use the active to express some causative ("Caesar pontem fecit.": "Caesar made a bridge."), use specialised verbs ("to kill" = "to directly cause to die") and/or use several modals to express the other causatives. Examples: "John killed the dog.", "John let the dog die.", "John caused the dog to die.", "John had the dog killed.". There exist useful classifications but even inside them, according to the action, there may be sometimes several ways to cause the action. So I now think we might better go on considering causative forms as special definedMeaning if they are different words in any language. --Fiable.biz (talk) 23:38, 20 March 2013 (CET)

OmegaWiki is better in semantics where the wiktionaries are better in grammar. Ideally we create a best of both worlds by merging those two powers. For the dozens of languages without wiktionary OW can be leading. On Meta is a vivid discussion with a new group calling themselves 'LangWiki'. They seem to try to re-invent the wheel. Is it so difficult to be more co-operative?  Klaas|Z4␟V:  10:05, 31 July 2013 (CEST)