LING 15 Lecture Notes - Lecture 30: Logogram, Syllabary, Speech Synthesis
Document Summary
Top down: holistic approach; recognizing a second of sound rather than a sliver of sound. Software knows what to do w/input (audio data, sentence you speak) and database looks up info it needs. Bottom up parser: figure out sound-by-sou(cid:374)d what"s said a(cid:374)d pie(cid:272)e together i(cid:374)to a word software will start with sou(cid:374)d (cid:894)for(cid:373)a(cid:374)ts, frequencies) and look up database what matches these measurements. Speech recognition deals with sounds not written text. Use example on short answers; short answers 5pts each. Co(cid:373)(cid:271)i(cid:374)atorialis(cid:373): do(cid:374)"t see i(cid:374) a(cid:374)i(cid:373)al (cid:272)o(cid:373)(cid:373)u(cid:374)i(cid:272)atio(cid:374) e(cid:454)(cid:272)ept (cid:373)a(cid:455)(cid:271)e in bird songs. Co(cid:373)(cid:271)i(cid:374)e letters i(cid:374)to words; (cid:272)a(cid:374)"t reall(cid:455) do that with e(cid:373)ojis (cid:894)use o(cid:374)l(cid:455) emojis in sentences to convey a message- not always effective) Morphemes are chunks used to form words. :) mean smiley face but when apart, : and ) mean diff things. Emojis indicate affect- smiling face = happiness, very iconic; represents a(cid:374) e(cid:373)otio(cid:374) (cid:374)ot a word would(cid:374)"t sa(cid:455) i a(cid:373) so :(cid:895) that i wo(cid:374) this.
