Gyara Fassarar Na’urori (Machine Translation Post-Editing (MTPE))

An ciro daga littafin:

BIƊA DA TANADI A DUNIYAR INTANET
(Sana’o’i, Kasuwanci, Ƙwadago)

Na

Abu-Ubaida SANI
Abdulrahman ALIYU

Manazartar Littafi: Sani, A-U. & Aliyu, A. (2026). Biɗa Da Tanadi a Duniyar Intanet (Sana'o'i, Kasuwanci, Ƙwadago). Amsoshi Academic Publishers. ISBN: 978-978-62480-1-1. DOI: https://doi.org/10.5281/zenodo.21361768

Tuntuɓa Domin Samun Littafi: 

Dr. Abu-Ubaida Sani: Imel - abuubaidasani5@gmail.com | WhatsApp - 08133529736

Dr. Abdulrahman Aliyu: WhatsApp - 08036954354

Gyara Fassarar Na’urori (Machine Translation Post-Editing (MTPE))

Gyaran fassarar na’urori (MTPE) yana nufin gyara fassarorin da na’urorin fassara suka samar. Ɗaya daga cikin manyan sauye-sauye da aka samu a fannin kimiyyar harsuna shi ne na’urorin fassara. Kamfanoni da dama sun samar da manhajojin fassara. Yanzu da ilimin ƙirƙirarriyar basira (AI) yake ƙara bunƙasa, dabarun samar da manhajojin fassara suna ƙara yawa a duniya. Hausa tana ɗaya daga cikin harsunan da suke samun wannan tagomashi. Tana cikin jerin harsunan da suke kan manhajojin fassara da dama. Daga cikinsu akwai:

1. Bing Translator (https://www.bing.com/translator)

2. Google Translate (https://translate.google.com)

3. Lanfrica (HausaMT) (https://lanfrica.com)

4. Masakhane (NLP Project) (https://www.masakhane.io)

5. MyMemory Translation (https://mymemory.translated.net)

6. Translated.net (https://www.translated.net)

Kasuwar ayyukan MTPE tana ƙara bunƙasa. Akwai manyan dalilai biyu da suke haifar da hakan. Na farko shi ne kasancewar ana ƙara samun kamfanonin da suke bayar da horaswa ga manhajojinsu na fassara. Yayin aiwatar da hakan, dole ne su fitar da samfurin fassarorin da injunan suka yi domin a yi musu gyare-gyare. Da waɗannan gyare-gyaren ne ake ƙara inganta basirar na’urorin.

A ɓangare ɗaya kuwa, ana ci gaba da samun ƙaruwar injunan fassara na Hausa masu nagarta. Wannan ya sa waɗansu masu buƙatar fassara suke amfani da su. Da yake har yanzu injunan ba su kai kason nagarta da ake buƙata ba (ana kan inganta su), dole ne a bayar da aikin MTPE domin yin gyararraki ga ire-iren fassarorin da aka yi ta hanyar amfani da injunan. Hakan ya haifar da ƙaruwar ayyukan MTPE na Hausa.[1]

Binciken Mykhalevych (2024) ya nuna yadda harsunan da ake yawan amfani da su a kan intanet suke ƙara samun ingantuwar fassara a kan injin ɗin fassara na Google (Google Translate). A shekarar 2022, cibiyar bincike ta Preply ta nazarci mizanin ingancin fassarar GT. Sakamakon da ta samu ya kasance kamar haka:

Fassarar Ingilishi zuwa Sifaniyanci ya kai matakin 96.58% na daidaituwa. Ingilishi zuwa Jamusanci kuwa ya kai 94.56%. Shi kuwa Ingilishi zuwa Italiyanci ya kai 94.12%. Yayin da aka fassara rubutun harunan nan zuwa Inglishi kuwa, makin daidaiton ya kai 97.18%. (Mykhalevych, 2024 SL. 13).

Binciken Schäferhoff (2023 SL. 46) ya nuna cewa, harsunan da aka fi amfani da su a kan intanet ne suka fi samun daidaitacciyar fassara a kan injunan fassara.[2] Binciken Taira et al. (2021) ya tabbatar da cewa har asibitoci sukan dogara da fassarorin da injunan fassara suke samarwa a yau (p. 3361). Duk da haka, fassarar GT ba ta kai a dogogara da ita ba tukunna (p. 3364). A duba misalin aikin gyaran fassarar na’urori cikin harshen Hausa da aka turo a cikin hoton da yake ƙasa:

Hoto Na 4.4: Aikin gyaran fassarar na’urori cikin harshen Hausa

Madogara: Saƙon imel daga kamfanin fassara na Transparent

---

[1] Ya zuwa shekaru masu zuwa, haƙa na iya cimma ruwa inda injunan fassara na Hausa za su iya kai wajajen mizanin samar da daidaitacciyar fassara da ta kai kashi casa’in cikin ɗari (90%). Idan haka ta faru, lallai ayyukan fassara za su yi matuƙar raguwa. A maimakon haka, ayyukan MTPE ne kawai za su yawaita.

[2] Taira et al. (2021 sh. 3364) ma sun jaddada yadda ingancin fassara ya danganta daga harshe zuwa harshe. Wannan ya sa ya zama babban ƙalubale ga Hausawa musamman manazarta da su duƙufa wajen ganin Hausa ta ci gaba da shiga cikin duniyar intanet.