Тест на ИИ в переводе на язык жестов

Ответы и пояснения появляются по мере прохождения теста.

1. Why does translating a signed language require more than a gesture-to-English-word lookup?
   • Every sign follows the same word order as spoken English
   • Signed languages have their own grammatical and spatial structures
   • Handshape recognition alone captures every linguistic feature
   • Each sign always corresponds to exactly one English word
2. Что такое «неручные маркеры» на жестовых языках?
   • Специальные перчатки, которые носят подписывающие лица
   • К видео добавлены письменные субтитры
   • Перед подписанием используется дезинфицирующее средство для рук
   • Выражения лица и движения тела, имеющие грамматическое значение.
3. In a pose-based sign-language pipeline, what does keypoint extraction produce from video frames?
   • Coordinates describing selected hand, face, and body landmarks
   • A guaranteed complete translation with no sequence model
   • A rule that the original video has already been deleted
   • Only the recording's file size and filename
4. Why does continuous sign-language recognition need to address temporal alignment?
   • Every sign is a static image with no temporal information
   • Signed languages contain no grammatical structure
   • Signs and transitions can coarticulate without an obvious pause at every boundary
   • A video's frame rate reveals the target sentence without interpretation
5. Основными узкими местами в обучении этих систем являются:
   • Слишком много высококачественных размеченных наборов данных.
   • Камеры запрещены
   • Языки жестов, не имеющие грамматики
   • Нехватка больших аннотированных наборов данных на языке жестов.
6. What does involving Deaf communities and language experts add when designing sign-language AI tools?
   • A guarantee that all signed languages share the same grammar
   • Knowledge of real needs and linguistic and cultural context for design and evaluation
   • Elimination of the need to evaluate translations
   • A reason to assume one laboratory benchmark covers every signer
7. What does a gloss annotation represent in the sign-language research pipeline described here?
   • A universal spoken-language sentence that every signed language uses
   • Only a camera's color profile
   • A textual label for a sign, rather than a complete grammatical spoken-language translation
   • A guarantee that all facial grammar has been fully preserved
8. Which language and domain characterize the PHOENIX-2014T benchmark introduced in the cited 2018 work?
   • German Sign Language weather broadcasts paired with German text
   • Every signed language recorded in all clinical settings
   • Only English speech recordings with no signing video
   • Only static photographs of handwritten alphabets