Am Mo., 6. März 2023 um 14:56 Uhr schrieb Samuel Okepefi <[email protected]>:
> Hello, > > Apologies for posting here. I just found out about this threads and I want > to address MF-Warburg concerns. > Thanks a lot for your reply! > > As far as I can tell, wikinews articles are mostly news summary from > respected newspapers as sources (at most 2 sources per article). This > policy we have adhered to. Most of the news articles on enwikinews are > stub-like if compared to enwiki articles see examples here: > > > https://en.wikinews.org/wiki/More_than_100_killed_in_mosque_bombing_in_Pakistan > > https://en.wikinews.org/wiki/Earthquake_strikes_Aksu_Prefecture,_Xinjiang_Region,_China;_no_casualties_reported > > https://en.wikinews.org/wiki/US_singer_Madonna%27s_elder_brother_Anthony_Ciccone_dies_at_age_66 > > https://en.wikinews.org/wiki/Glasshouse_in_to_Tier_1_women%27s_cricket_final_on_Australia%27s_Sunshine_Coast > > When we started this project, we cite two to three sources but decided to > stick to one source since ours is summarising a news article and then > translating it. This is a community decision. You can see examples here. > > > https://incubator.wikimedia.org/wiki/Wn/guw/Folasade_Ogunsola_l%E1%BA%B9zun_y%E1%BB%8Dnnu_tintan_heyin_m%E1%BA%B9b%E1%BB%8Ddo-azinpon%E1%BB%8D_UNILAG_t%E1%BB%8Dn_go > > > https://incubator.wikimedia.org/wiki/Wn/guw/Naijilia_na_klan_eyin_Bola_Tinubu_l%E1%BA%B9zun_t%C3%B2g%C3%A1n_w%E1%BA%B9_Obaseki_d%E1%BB%8D > > As to posting short edits, I correct the sentences in most cases and that > gives a lot of issues to me. So I encouraged writing text in bits and > making sure the tonal markers are right. This reduces the level of my > corrections on the body of texts. I also come in to add the proper > formatting (once we get approval importing and then using the templates is > usually easy and this will reduce this work) and audio to the news. You can > confirm this trends on guwwiki or guwwiktionary. It's not a tactic as you > concluded. Gungbe is tonal and a word if not marked properly can have > different meanings e.g owhé(house), owhe(year), owhè(sun). > So if I understand it correctly, several users are contributing the content, while you see to it that the formatting is done properly. This leads me to wonder: does this project rely entirely on you being there? I don't mean this in a negative way, I appreciate what you do a lot, but if one person has to "fix" every article, that would be a problem if that one person once happens to be unavailable. > > Concerning using Google translate to translate our article and Google > suggesting it as Yoruba. A proper test I think you should have done is > reversing the Google translated text in English back to Yoruba and you will > be amazed to see that no single word or expression will be the same as the > previous text in gungbe. But thanks for this observations, it makes me to > be aware of the possibility of gungbe being added to Google translate > languages soon. My explanation for this is that there are large body of > texts online for Gungbe and Google's algorithm is based on how close a > written text is to an existing language. I found this interesting paper > https://metatext.io/datasets/jw300. And did the same test you did on this: > > https://www.jw.org/guw/Wes%E1%BA%B9dot%E1%BA%B9n/linlinwe-l%E1%BA%B9/at%C3%B2%CC%A3h%C3%B2%CC%A3-nuhih%C3%B3%CC%A3-l%E1%BB%8D-t%E1%BB%8Dn-s%E1%BB%8Dha1-2019-jan-f%C3%A9vr/naw%E1%BA%B9-jiwhey%E1%BA%B9whe-n%E1%BB%8D-yin/ > > > https://www.jw.org/guw/nuhe-biblu-pl%E1%BB%8Dn-m%C3%AD/kanbi%E1%BB%8D-l%E1%BA%B9/pipeh%E1%BA%B9-z%C3%B2%CC%A3n-j%E1%BA%B9magb%E1%BB%8D/ > The texts are in gungbe but Google gives an intelligent translation and > detects it as Yoruba. But when done in the reverse. It doesn't match the > previous text. (You can try it too) > > I'll suggest when you do such test next time, try to do the reverse and > see if it matches the previous text. > Well yes - but a Google Translate "re-translation" never gives back the original text, regardless of the languages involved. I was merely surprised to see this translation being understandable, so I wondered about whether the pages are not actually written in Yoruba. But if I understand it correctly, it is just a case of mutual intelligibility? > > In addition to the above, you can see that most of our sports news article > are longer than most enwikinews articles. You can see here: > > > https://incubator.wikimedia.org/wiki/Wn/guw/Kane_do_ag%C3%B3_200t%E1%BB%8D_et%E1%BB%8Dn_to_Premier_League_m%E1%BA%B9_dile_Tottenham_duto_Man_City_ji > > > https://incubator.wikimedia.org/wiki/Wn/guw/Man_City_h%C3%B2_Chelsea,_nado_s%E1%BA%B9p%E1%BB%8D_m%E1%BA%B9he_to_aga_to_todohukanji_EPL_t%E1%BB%8Dn_m%E1%BA%B9 > > Or a full list in the sports category: > > > https://incubator.wikimedia.org/wiki/Category:Wn/guw/Aihundida_lanm%E1%BA%B9yiya_t%E1%BB%8Dn > > Let me know if further clarifications are needed from me. > Once again, apologies for posting here as I assume it is not open to the > public. > No worries, thanks for addressing the questions so we can move this forward! > > Best > Samatics > _______________________________________________ > Langcom mailing list -- [email protected] > To unsubscribe send an email to [email protected] >
_______________________________________________ Langcom mailing list -- [email protected] To unsubscribe send an email to [email protected]
