Dictionary Structure

I can’t help thinking that there are some serious problems with the dictionary structure.

This idea of mapping one Czech word to one English word is a real problem even if the alternatives feature works.

Each Czech word can have many meanings as can an English word.

A specific Czech word only translates to a specific English word within a specific context.

Note also that any given word may also have synonyms. So for example if I translate ‘velký’ as ‘big’ or ‘large’ it makes no difference. But if I translate ‘jestli’ as ‘whether’ or ‘if’ it does matter as these words have different meanings (they are not synonyms).

Compare http://dictionary.reference.com/browse/whether to http://dictionary.reference.com/browse/if


• Jestli chcete. → If you please/like/want to. • Jestli je vám to nevhod ... → If it doesn't suit you ... • Jestli se nepletu, ... → If I'm not mistaken, ... • Podívej se jestli to drží. → Check if it holds. • Říkal jsem si, jestli byste ... → I was wondering if you ... • Zajímalo by mne, jestli ... → I would like to know/I wonder if... • Jsem zvědavý, jestli to pošle. → I wonder if he'll send it. • Ať se propadnu, jestli ... → I'll be damned if ... • jestli ano, je-li tomu tak → if so • Jestli si to někdo zaslouží, pak jsi to ty. → If anyone deserves it, it's you. • Jestli to máme zvládnout ... → If we are to manage it ... • jestli sis nevšiml → in case you didn't notice • Jestli bylo nějaké zklamání, bylo jen dočasné. → If there was any disappointment it was temporary. • Jestli máte chuť → If you are so inclined • Jestli ti to nevyhovuje/se ti to nehodí ... → If it is inconvenient for you ... • Nepoznáš, jestli si dělá legraci. → You can't tell whether he's joking. • jestli je to dobrý nebo špatný nápad → whether it's a good or bad idea

What would help would be if the dictionary allowed different meanings to be entered. Synonyms should be treated as correct answers. But often a memrise course is testing the learner knows a certain meaning of the word.

For example in my course “Staré pověsti české a moravské” I use the word “připomínat” with the meaning ‘to remind’ however the dictionary forces me to take the meaning as ‘to resemble’

In fact “připomínat” has five main meanings:

  1. remind sb of/to do sth
  2. nag sb to do sth
  3. remind sb of sb
  4. resemble sb
  5. suggest sth

Any ideas?

Posted by predrup 6/1/12, last update 6/4/12 (2 years ago)
  • This is a good point. The "primary definition" shouldn't be viewed as being, or as trying to be the only translation of a word. This is just the first one that you learn. The alternative translations should all also be added, and are visible during learning. We have built a tool for adding in sample sentences in a format that will allow them to be tested from as well. This will mean that you can create sample sentences that show the meaning of the work in different contexts, and use those to test those other, richer uses.

    That way, rather than learn all the different ways that a Czech word is used by learning multiple English translations, you will learn the different ways that the Czech word is used by seeing it in multiple contexts and understanding how it is used. So the first meaning that you learn is just a mental bridge that you start to build to understanding the actual way that the Czech word is used in Czech. Does that make sense?

    Synonyms should indeed be allowed as correct answers - they should be added to the "alternative Czech" or "alternative english" fields. You can do that by going to the "word detail page" and either adding the alternative directly (if you are a curator) of by leaving a comment suggesting it, in which case it will be added by a curator.

    I hope that all makes sense,

    Best wishes,


    Posted by BenWhately Staff   6/1/12 (2 years ago)
  • @predrup, I really feel your frustration on this one. I wish ben was more right when he says the alternatives are visible during learning. That's not really true though, they are only visible when first presented with the word, and then ONLY if you press the more button..

    You make a good point with your sample sentences, and it really goes to show how we really need to be able to choose which of the many meanings of a word get shown as the main meaning in the wordlists we individually create. I have changed připomínat to have the main definition of "to remind", but at the end of the day, the way things currently are, there is no way to please everyone, and someone else may come along and want the main definition to be "resemble".

    @Ben, this is especially frustrating for wordlists that revolve around a particular piece of text, or song, where really you want the main definition to be the meaning used within the text..

    I am also a little confused about memrises long term goals concerning alternative meanings in the target language. To me it makes perfect sense to list multiple english meanings of a czech word. This is rather like a dictionary.. each target language word has multiple definitions listed as entries in the dictionary. But NO dictionary attempts to list multiple meanings in the target language too.. That's sort of like trying to cross a thesaurus with a 2way language dictionary, and personally I just can't see how that is doable..

    as an example take predrups example of the word připomínat, that has multiple english translations, to remind, to commemorate, to resemble, to nag, to suggest etc.. all these words can be correctly translated as připomínat depending on context.. so then what can be added as czech alternatives? If the main definition of the word is listed as to resemble, someone might add vypadat as an alternative definition, but the moment you think of the meaning of to remind, the alternative czech would be wrong..

    So maybe the alternative czech should only be used for words that are 100% identical with the czech entry? like for adjectives just the alternative nominative endings, and for nouns, just the diminutive endings.. it does seem that the alternative czech fields should not really be used much beyond that because few words outside of the odd noun or 2 have truly identical meanings.

    Posted by mangodurian 6/2/12 (2 years ago)
  • @mango, you are absolutely right about the difficulty of creating a database with multiple two-way translations for each word. It is very hard. But as long as the testing direction is kept English -> Czech, it is not so difficult. In that case the Czech field should be used to put any Czech word that can be a translation for the given English primary meaning. This is how the learning flow currently works:

    • you are presented with a Czech word. You see a primary English translation.
    • You take a look at the other ways the word can be translated into english to get a richer sense of its meaning. Read through any sample sentences as well.
    • ideally you make a mem that helps to combine these meanings in your head, but highlights the primary one. It might just focus on the primary one though.

    • Test: you are presented with the English word, produce the Czech word. You could also answer with another Czech word that is the correct translation of that English word.

    • next stage: testing from sample sentences, allowing you to test different meanings of the Czech word, using context. So you will have a sentence with the english word for one of the secondary means in it, and will have to fill in the blank in the Czech sentence.

    This current situation doesn't allow you, as you say, to pick the primary meaning that you want to highlight in a particular wordset. There are a couple of ways to approach a fix for that:

    • we could allow you to fix the primary definitions on a per wordset basis. This has complications in which mems are shown etc, but it would be possible.

    • we could allow you to fix which sample sentences are used for testing in each wordset, meaning that you would have the same initial learning experience for each word, learning the primary definition with all the mems etc that are associated with that, and then start being tested on the particular meaning that is relevant to that wordset. This might be a more elegant solution. It would also allow you to use, eg, sentences from the story/videosong as the sample sentence being tested from.

    What do you think?

    Best wishes


    Posted by BenWhately Staff   6/3/12 (2 years ago)
  • Hi Ben,

    I have to agree with Mango this is confusing.

    Making an absolute link between a course and a dictionary entry would only work if there was only one course allowed.

    People are rightly creating courses where the context of the course is part of the learning experience. Mango has been creating courses that assist in learning a specific song I have been creating courses that assist people to read a specific book. This shows the value of memrise.

    But forcing the concept of what you describe as a ‘primary definition’ across every course makes no sense. There is no such thing as a primary definition nor does learning a primary definition universally help someone to learn a language.

    The assumption is being made that through the sample sentences a learner can deduce how the language works. Languages have complex ideas that cannot be expressed through simply studying samples. I see the power of memrise as a tool for the student to reinforce whatever language course and learning style they are following.

    I would also like to be able to create courses where each side is Czech. This is so I can create courses that practice import concepts in the Czech language, examples would be:

    • Matching verb aspect pairs
    • Matching adjectives with their corresponding adverb forms
    • Changing the tense of a verb
    • Identifying the preposition/case that goes with a verb usage
    • Creating the comparative/superlative form for an adjective/adverb

    Speaking a language fluently is less about translating and more about working in the language itself.


    Posted by predrup 6/3/12 (2 years ago)
  • Sorry to be confusing! it was not my intention to set down a correct way, but to ask your opinion on how to do it, and to discuss that.

    I definitely agree that pure translation is not the best way to be learning once you reach a high standard. much better to work solely in the target language. I think that each topic should have a separate wiki that is just in the target language, so that advanced learners can use that.

    But that is also the reason that I don't particularly like the idea of having multiple different translations of the same word being taught in different places. That seems to be to be learning about translating, not about using the language.

    The sample sentences don't require you to deduce how the language works - they are not grammar tests. Definitely not. These tests are only about seeing the context that the word is used in. You will have the english translation of the sentence, so you don't need to know the grammar involved. You just need to work out what the right word to use in that context it. That way you would be able to set the tests so that the usage that was tested was the one that was relevant to that course.

    The idea of a "primary" definition is just that it helps with the very first step of creating a bridge for your memory to engage with the new word. By keeping that to a single meaning, we focus the creative efforts of everyone who is learning onto one item and therefore improve the material for that one meaning as much as possible. Then we can move on to focusing on the other meanings, and elucidating those. This is all of course a bit speculative, because that testing tool doesn't yet exist. I appreciate that that is annoying. But it is how we have to think since we are developing tools that don't yet exist!

    We have limited resources to build new features, and when we do so we want to be sure that we use them in the right way. I am not yet convinced that building extra machinery to test different primary definitions in different courses is the right way to do it - though it might be. It seems to me that that would be a reinforcement of the translation method, rather than a move beyond that to a richer, more engaging learning experience. Which is why is currently seems to me that that might be the best way to proceed.

    So I think that at the moment I would be in favour of focusing efforts on getting the word seeded in your memory by learning one meaning first, taking advantage of all the great mems that other people have made to make that first experience as engaging and pleasurable as possible, and then being able to set a focus for later tests on a per-course basis if that were needed, using a selection of sample sentences and testing from those. Or you could just sequentially elucidate all the other shades of meaning of the word if you were learning a course without particular context.

    So in summary - and apologies for rambling on - learning the "primary meaning" is just about building the first bridge to getting beginning an understanding of the word. After that the really deep work of expounding and enriching your understanding of the word and the language starts. I would prefer to focus our limited resources on improving first the quality and ease of building that initial memory bridge, and then on the expounding and enriching stage; rather than on improving the choices that you have over different initial "bridges" to start understanding a word.

    Does that make sense?

    Best wishes and thank you again for all of your feedback,


    Posted by BenWhately Staff   6/3/12 (2 years ago)
  • Hi Ben,

    The argument we are trying to make is that words have multiple meanings. This is a fact and translating these meanings into other languages is another matter. Trying to arbitrarily assign a word one primary meaning and then attempting to hang all the other meanings off it as alternatives doesn’t ultimately help.

    In the case of Czech we are trying to use alternatives to catch aspect pairs, synonyms and different meanings and the list will go on.

    Conventional dictionaries explicitly list the different meanings and explain them for very good reason.

    When we step into translating we have to allow for the fact that Czech makes a distinction between verb aspects and in the case of verbs of motion determinate/indeterminate and if you are walking or using some form of transport.

    To give an example say I am presented with the word “to leave”, memrise tells me it is a verb to limit the options, but I still have any of the following as correct answers: nechávat, nechat, odejit, odcházet, odjet, odjíždět, opustit, opouštět, odcestovat (this is not an exhaustive list and we don't have all these forms in the dictionary yet).

    With the current database structure we could add ‘to leave’ multiple times with the context described in brackets, but I can see that becoming a mess.

    What is the primary meaning of to leave? There are at least five meanings http://dictionary.reference.com/browse/leave.


    Posted by predrup 6/3/12 (2 years ago)
  • Thank you for this Peter, and sorry to be slow in picking up the core of your argument.

    Just to be sure that I am understanding you correctly:

    • further context is needed beyond the single English definition, to give the verb aspect, whether the verb is determinate or indeterminate, and direction. How would you prefer to do this, if not by adding this information in to the definition? Would adding more meanings to the main definition field solve this issue?

    • I m not totally sure I agree that it doesn't help to learn a single translation first, and then to learn all of the other translations after. I actually find that to be incredibly helpful myself, don't you? trying to create a link between a single word in a foreign language and multiple english words all at once is a painful process. I appreciate that at the moment the issue of multiple correct Czech translations for "to travel" is an issue, but the plan that I outlined would mean that after the first few tests, you would always be being tested on words used in context, so that would reduced the number of correct alternatives dramatically. Each sample sentence would also have a comments section explaining the use of the word in that context, and how it could be translated.

    So when you learn a word for "to travel", you will be tested by seeing "to travel" a few times. Yes, you could answer any of the words for "to travel"; but there will likely be only one that you have just planted. So you will know which one it is looking for. Once the word reaches a threshold for further meanings to be introduced (after a few days) then you will being deepening your understanding of how that word is used.

    • what is the alternative plan? I am not sure I understand it properly. Is it just to show all of the english translations at once on every test? We could to that for the first tests, and then start introducing the sample sentences in the same way as outlined above. Would that work, do you think?

    In that case, the only difference between the two plans would be that in the second one, all the meanings would be displayed during testing for the first few tests, before the sentence testing kicked in.

    My worry about this is really focused on new learners, for whom seeing long lists of definitions is a real turn off (we see this in the figures for when people drop out of learning). But perhaps we need to have settings that can be changed to be tailored to the needs of more advanced learners. So one "definition lite" version for beginners, and then the option to see all of the definitions right from the start for more advanced learners. Then ultimately the option to see the definition only in the target language or really advanced learners.

    Does that sound like a better solution?

    THank you again for you patience with giving all this feedback!

    Best wishes


    Posted by BenWhately Staff   6/4/12 (2 years ago)
  • Hi Ben, I've spent most of my life working as a developer. So my instinct is to draw a diagram, would that help? I guess I can't make attachments? Would an ERD work for you?

    Posted by predrup 6/4/12 (2 years ago)
  • That would be perfect, thank you!

    Best wishes


    Posted by BenWhately Staff   6/4/12 (2 years ago)
  • @ben,

    I'm a little slow in responding, but you wrote something that up until now I was unaware of.. about how the "Czech alternative field should be used to put any Czech word that can be a translation for the given English primary meaning".. This is not altogether clear, and I think mostly it's being used to add words that are synonyms of the czech listed word.. how you describe it's usage is really quite different, and I have to add that to me it doesn't make complete sense.

    having the czech alternatives as alternative czech words corresponding to the listed english primary word, will mean that if you ever make it possible to have an alternative meaning listed primarily in a given wordset, then the alternative czech meanings will most likely be wrong. you would have to have several lists of alternate czech meanings, one for each of the listed english alternatives. I suppose ultimately that's doable, but surely it would make more sense to restrict the czech alternatives to true synonyms of the czech word and not tie it to any one of the often multiple english alternatives??

    The way I see it working best, for czech at least, is that when we create a wordlist, we get to choose which of the english alternatives gets listed primarily, and when we write a mem, we get asked whether the mem is specific to a particular wordlist (useful for lists revolving around a particular piece of text/song lyrics), or a particular english alternative (to be shown only with lists that have a particular alternative listed as the main definition) - maybe not make the mems entirely inaccessible in other word lists, but hidden until a show other mems button is pressed, and if we find the mem useful in another list, make it possible to add it to that list too (maybe on a per user basis).

    Although this is clearly more complicated than how things now stand, I don't personally view it as being over complicated.


    I think the different verb issue (aspect/determinate/indeterminate) can be easily solved, by just adding more to the part of speech. (ie imperfective determinate verb).. I think we've already agreed on this previously, although there are clearly word entries that still need to be added to reflect this in the PoS.. I haven't gotten around to all of them yet, but I've done some. If you agree that this addresses this problem, then please edit whatever PoS you know of that need changing.. (to be honest, I have still not learned all the differences, so am unsure of them myself, and would appreciate you taking care of them).. If you have changed some and they are waiting to be fixed while patrolling, then apologies that they are not fixed up as the patrol button (and moderate) are still not working for czech (any ideas when they will be ben?).

    peter, love your idea for creating courses where each side is czech.. I think that they would be great contributions, and I'm sure we could fix up the conventions to allow such courses.. for example we could add a part of speech labeled "aspective match", have the english side saying , počkat (perfective) and the czech side saying čekat (imperfective).. If you think such word pairs should not be listed in the dictionary, we could always reject them while moderating, which would mean they would stay as they were in the wordlist, but would not be listed while searching for them..

    do you think it would be possible to post a link to the ERD, I'd love to see what idea you've come up with too (alternatively send me a copy via email maybe?).


    Posted by mangodurian 6/4/12 (2 years ago)
  • Hi Mango, I will email you what I come up with. But I need some time to think about it. Peter

    Posted by predrup 6/4/12 (2 years ago)

Recent threads

This forum doesn't have any recent activity