INGLIZ VA O‘ZBEK YOSHLAR SLENGI MA’NOLARINI AVTOMATIK FARQLASH: DIAGNOSTIK TAJRIBA

Annotatsiya: Maqolada ingliz va o‘zbek yoshlar slengining ma’nolarini avtomatik farqlashga oid dastlabki diagnostik tajriba bayon qilinadi. Tadqiqot maqsadi so‘z shakli, leksik birikuvchanlik va oldingi replikaning ma’no tanlashga qo‘shgan hissasini tekshirishdan iborat. Material manbasi ko‘rsatilgan o‘n oltita parchadan tuzildi: har bir tilda sakkiztadan misol tanlandi. Inglizcha cooked va salty, o‘zbekcha beton va sharoit birliklarining odatiy hamda sleng talqinlari solishtirildi. Python dasturida kontekstsiz tanlov, aynan mos so‘zlarni hisoblash va asos bo‘yicha qidiruvni oldingi replika bilan birlashtirish usullari bajarildi. Dastlabki usul sakkizta, ikkinchisi to‘qqizta, uchinchisi o‘n to‘rtta misolda tayanch talqinga mos javob qaytardi. Uchinchi usul ikkita misolda ma’no tanlamadi. Natijalar qo‘shimchali shakllarni tanish hamda suhbat kontekstini saqlash zarurligini ko‘rsatdi. Inkor, leksik ma’no va so‘zlovchi bahosi alohida izohlandi. Tanlanma oldindan ko‘rilgan parchalarni qamraydi; ko‘rsatkichlar mustaqil sinov aniqligi sifatida talqin qilinmaydi. Mualliflar yoshi tekshirilmaganligi va belgilashning ekspertlar tomonidan tasdiqlanmaganligi xulosalar doirasini chegaralaydi. Tajriba hisoblash qarorlarini kuzatish hamda korpus annotatsiyasida noaniq holatlarni ajratish uchun dastlabki mezonlarni aniqlashtiradi.

Abstract: The article reports a preliminary diagnostic experiment on automatic sense disambiguation in English and Uzbek youth slang. It examines how word forms, lexical cooccurrence and preceding turns affect sense selection. Sixteen documented excerpts were assembled, with eight examples per language. Ordinary and slang readings of English cooked and salty and Uzbek beton and sharoit were compared. Three Python procedures were implemented: contextless selection, exact lexical matching, and prefix matching combined with preceding turns. The procedures returned reference-consistent answers for eight, nine and fourteen examples, respectively. The third procedure abstained on two examples. The findings indicate the importance of recognising inflected forms and retaining conversational context. Negation, lexical meaning and evaluative stance were analysed separately. The sample contains previously inspected passages, so the scores do not estimate independent test accuracy. Unverified speaker ages and the absence of expert validation restrict the conclusions. The experiment clarifies criteria for corpus annotation and evaluation.