Quiz 3 Flashcards
Transcription | Vowel analysis | Consonants 1
What factors determine how much of the recorder audio gets transcribed
- The amount of time at hand
- Amount of data to transcribe
- Amount of $ to hire transcribers
- Quality of audio
T or F : MS-Word aligns the transcription text with the audio
F - it is a simple word processor but it doesn’t mean that it aligns
Types of software : ELAN - what are the pros
- Transcribing video data
- Allows many tiers
Types of software : PRAAT - what are the pros
- One of easiest ways to transcribe audio data
- Transcription is time-aligned with AUDIO
- Allows many tiers for diff levels to annotate
What are the 3 symbol systems for transcription
- IPA
- SAMPA
- ARPAbet
Symbol systems - what does SAMPA use and WHY pick it?
ASCII characters to rep. phonetic symbols
It simplifies IPA symbols for easy use in computer systems
Earlier software systems had limited ability to handle IPA symbols
Symbol systems - what does ARPAbet use and WHEN is it used? WHY pick it?
- ARPAbet uses a certain set of ASCII characters per phoneme - 2-3 LETTER CODE
- Used in text-to-speech systems
- Simplifies English phonemes for the use of speech tech
- Makes an easier integration of speech synthesis
When deciding how much of a recorded interview to transcribe, which of the following processes should be considered to ensure efficient and meaningful data usage?
A. The quality of the interviewer’s questions and how much they contribute to the conversation.
B. The type of symbol system chosen (e.g., SAMPA vs. ARPAbet).
C. The balance between data transcription depth and the project’s purpose, time
constraints, and budget.
D. The need for code-switching analysis within the transcript.
C. The balance between data transcription depth and the project’s purpose, time
constraints, and budget.
What are resonance frequencies?
Literally frequencies that are more resonant, they turn into what we call FORMANTS
T or F : our vocal folds generate a LOT of frequencies, but only some of them get through
True - some of them amplified and others get dampened
What are amplified frequencies?
Formants
F0 = ? = ?
F0 = fundamental frequency = harmonic 1
What is the difference in peaks between a broad and narrow band spectrum
Broad - peak of general formant
Narrow - peak of each harmonic (component)
Ella wants to know how many component frequencies are available under 1000 Hz. What type of spectrum would you suggest to try?
A. narrowband
B. wideband
C. either will be good
A. narrowband
James has generated a narrowband spectrum of an [u] vowel. Which of the following is true?
A. he has no way to estimate the frequency of the third formant
B. he has enough info to find the frequency of the F0
B. he has enough info to find the frequency of the F0
What is vowel normalization? How do we achieve it?
Techniques that account for the variation in our vocal tracts
- using Z-SCORES
What are two reasonings from speaker variability?
- Physiology
- Dialects
Goals of vowel normalization (4 main ones)
- Preserve sociolinguistic diffs in vowel quality
- Preserve phon. distinctions among vowels
- Eliminate physiological variation caused by mouth size
- Model cog processes in vowel normalization
What are the 4 types of vowel normalization techniques? think 2 V & 2S
- Vowel intrinsic method
- Vowel extrinsic method
- Speaker-Intrinsic
- Speaker-Extrinsic Method
Vowel intrinsic method + example
- Vowel intrinsic method - use info from a single vowel TOKEN (can be many speakers)
- Formant Frequency Adjustments : adjusting F values based on AVERAGE values for sim. speakers
Vowel extrinsic method + e.g.
use info of F values across diff vowels of the same person
- Grand Mean Norm : normali. vowels using a GRAND mean calculated from ALL V’s by the person
Speaker intrinsic + e.g.
normalize based on data from a SINGLE person’s V’s
- Individual Speaker Baseline : making a baseline for F freqs specific to an individual speaker
Speaker extrinsic + e.g.
Speaker-Extrinsic Methods: taking data from MANY speakers
- Mean Adjustment Method : normalize using mean values taken from MANY speakers
Stops : Hawaiian is know for what KIND of stops?
voiceless
What are the (3-4) main acoustic parts of stops?
- Time of closure
- VOT - combined by prevoicing and aspiration
- Formant transition
T or F : during the making of a stop, at a complete closure, the velar port is OPEN
F - has to be closed
What happens to the pressure buildup behind the constriction spot when then is a release?
It gets equalized
The _____ vibrations and _____ open __________ are gestures that can help create DIFF kinds of stops
What kinds?
- Vocal fold
- wide; vocal folds
Voiced and voiceless
Bella needs to transcribe some audio files where she wants to annotate the sounds with multiple layers of information/coding. She also needs to make sure in her annotation that the beginning and ending of all the phonetic segments are marked with clear boundaries showing timing information. Which of the following option will not work for her purpose? (One correct answer.)
A. MS Word
B. Praat TextGrid
C. ELAN
D. None of the above will work, in fact
A. MS Word
How is the SAMPA transcription system different from the ARPAbet convention? (One correct answer.)
A. SAMPA is primarily used for American English transcription, while ARPAbet is used for British English.
B. SAMPA uses only ASCII characters, while ARPAbet uses both ASCII and special symbols.
C. ARPAbet represents phonemes with two- or three-letter codes based on American English, while SAMPA closely follows IPA conventions with simpler ASCII characters.
D. ARPAbet provides a wider range of phonetic symbols than SAMPA.
B. SAMPA uses only ASCII characters, while ARPAbet uses both ASCII and special symbols.
John is planning to use DARLA (Dartmouth Linguistic Automation) to annotate 10 audio files. What things would John be able to perform with DARLA? (Two correct answers.)
A. John needs to upload a transcription TextGrid having only a single tier.
B. John will be able to upload all his 10 files at a single step instead of one file at a time, but it will take more time for DARLA to return the aligned TextGrids.
C. DARLA will return a TextGrid having exactly two different tiers.
D. DARLA will return two separate TextGrids with time-aligned annotations for individual words as well as phonemes.
A. John needs to upload a transcription TextGrid having only a single tier.
C. DARLA will return a TextGrid having exactly two different tiers.
Mary is studying some vowels and she is curious what kind of information a broadband sprectrum can provide? Which of the following can Mary do with a broadband sprectrum? (Two correct answers.)
A. Identify the frequencies of all the harmonics below 5000 Hz
B. Identify the frequencies of all the formants below 6000 Hz
C. Find our how loud (i.e., amplitude) a component frequency was in the sound signal
D. Identify the frequency of the fundamental frequency (~pitch) for the speaker
B. Identify the frequencies of all the formants below 6000 Hz
John is curious how he can generate a narrowband spectrum in Praat. (One correct answer.)
He needs to specify…
A. the Window length as 0.005 or less under Formant settings
C. the Window length as 0.005 or less under Spectrogram settings
D. the Window length as 0.030 or more under Spectrogram settings
E. the Window length as 0.030 or more under Formant settings
D. the Window length as 0.030 or more under Spectrogram settings
Which of the following are true? (Two correct answers.)
A. Even though spectra and spectrograms look dramatically different, they are both based on the same underlying information.
B. Compared to spectrograms, spectra make it easier to see the dynamic changes throughout the vowel
C. Spectra show frequency information at a single point in time, while spectrograms display frequency changes over time.
D. Unlike spectra, spectrograms do not show information about the amplification of frequencies.
A. Even though spectra and spectrograms look dramatically different, they are both based on the same underlying information.
C. Spectra show frequency information at a single point in time, while spectrograms display frequency changes over time.
