[09:48:12] 5681 [11:14:21] I agree with the idea of creating lexemes on-the-fly. Furthermore, I think it would be better to expand the scope of lexemes not only to proper nouns but also to all lexical categories. [11:14:22] For example, adjectives that represent colors are combined with most nouns, but the lexemes that consist of such combinations and are listed in dictionaries are often limited to those that can be used frequently (for example, silver screen, red carpet, etc.). [11:15:20] I myself have been registering Japanese lexemes since 2020, but I feel that there are still many parts that are insufficient in terms of quality and quantity to be able to mechanically translate existing Wikipedia articles. Therefore, in order to enrich the content of AW, it may be necessary for the time being not only to improve wikifunctions by looking at the [11:15:20] rendering results [11:15:22] of abstract content, but also to register lacking lexemes. [11:15:23] For this purpose, I would like to propose the following two points: [11:15:25] 1) With English as the translation hub, priority should be given to registering English lexemes so that they can be rendered in English at least. [11:15:26] 2) For lexemes that are not registered in each language, English should be rendered as fallback, and if people who see them register lexemes of each language, they can be replaced with sentences of each language. [12:45:38] I think it is probably unwise to think of English as a translation hub, but it’s not clear where you think it might be such a hub. It is a good idea to give people a call-to-action link as part of a fallback representation, but I don’t see why this link would not be in the target language, based on the item label. If the label exists in the target language, there is still [12:45:39] a p [12:45:40] roblem with determining whether each of its lexical constituents is a form in a lexeme for that language, but that shortcoming should be addressed by *T419561* (which is not currently an urgent priority, for some reason). (re @higa4: I myself have been registering Japanese lexemes since 2020, but I feel that there are still many parts that are insufficient in ...) [13:28:50] I was thinking that if one can't render an AW article in one's language, instead of providing a link, just render that part in English, for example, and one will know what kind of lexeme to register in that language. Maybe we don't have to fix it in English. [13:28:50] People who create AW contents are usually speakers of that language. So, if a Japanese speaker creates AW contents, for example, the missing lexeme will be found easily because s/he can't create using missing lexeme. [13:28:52] However, for example, when a Japanese speaker try to render an AW content created by an Arabic speaker, if the missing lexeme is empty or has an error, I think it will be difficult to find the missing lexeme. [13:28:53] In that case, I suppose that wikifunctions based on the new type that may be prototyped soon, will try to create on-the-fly lexemes by tracing the knowledge graph using registered values such as P5137 (item for this sense) and P9970 (predicate for) as clues. However, at this moment, the registration status of lexemes is not enough. So, I feel that there is too little [13:28:53] information [13:28:55] to create on-the-fly. [13:28:56] Anyway, it is just an idea depending on my humble understanding, so it may be something that should be considered after actually trying a new type functions. (re @Al: I think it is probably unwise to think of English as a translation hub, but it’s not clear where you think it might be such a hu...) [14:03:23] In the end, I took the decision to move the Corner to June 8, due to internal availability of the team. Stay tuned! (re @Sannita: ) [14:19:55] Maybe. If an item has linked lexemes in a small number of languages, any of those lexemes could be “translated”, but that highlights another shortcoming: we can only find lexemes for a particular language. We could extend that to a limited set of languages that have plenty of lexemes, but “linked lexeme references for all available languages” would require a new built-in [14:19:55] [14:19:56] function and might be considered too resource-intensive. When checking lexemes for item-label constituents, on the other hand, the language and form-representation are both givens. (re @higa4: I was thinking that if one can't render an AW article in one's language, instead of providing a link, just render that part in E...) [15:32:37] 87