CS4706 - Spring 2010
Homework 4 - "Intonation Variation and Annotation (100 pts)"

Submit in Courseworks

IMPORTANT: Submit all your files in one compressed file named "hw4-YourUni.zip" Example: "hw4-fb2175.zip".


I. (35 pts)   Record yourself saying "Minnie Maloney made her mother many millions." using the following ToBI contours:

    a. H* L- L% (declarative)

    b. H* L- H% (continuation rise)

    c. L* H- H% (yes-no question contour)

    d. H* H- H% (high-rise question contour)

    e. H* !H* L- L% (downstepped contour)

    f. L*+H L- H% (fall-rise contour)


    For each recording, create a Praat TextGrid file with the following tiers:  tones (including HiF0), breaks, orthography (words), and miscellaneous (to use for any comments).  Fill in each tier     with the appropriate labels.  Submit the recordings and TextGrids.  All files should be named with your uni and an indicator of the contour type.

Note:  If you have trouble producing a contour you can manipulate the pitch contour in Praat.  If you do this, submit the original recording, the manipulation file, and the resynthesized version of the wav file.  The TextGrid in these cases should be aligned with the manipulation.


II. (40 pts)     Collect 2 tokens of Barack Obama's speech from the web, from different genres, e.g. a stump speech, a one-on-one-interview, a debate, an infomercial -- you choose.  Save each token to a wav file indicating the genre and your uni (e.g. barack-fb2175-debate.wav)

Files should meet the following constraints:


a.                         Each token should be at least 5sec long, sampled at 16khz.


b.                         The speech should be as clean as possible so that you can get a good pitch track when Obama is speaking, so try to avoid overlapping applause and other speakers.


In a README.<uni>.hw4 file, you should annotate the following for each utterance:

a.      Filename

b.   Genre

c.   Where and when was the utterance made?

d.      What is *your own* native language (i.e., the first language you learned as a baby)?


You should provide a full ToBI labeling for each utterance (orthographic tier, tonal tier (including HiF0), break index tier, and miscellaneous tier), including, on the miscellaneous tier, time marks for all disfluences in the speech, if any.  Also indicate coughs or laughs or other non-speech noise. (Submit the 2 twav files with their aligned TextGrids files, one for each wav file: e.g, barack-fb2175-debate.TextGrid)


III. (25 pts) Write an essay of 250-500 words on the ToBI system (name the file tobi-<uni>.{txt,doc,pdf} answering the following questions:


    a.  What did you find hardest to label?  Easiest?

    b.  What additional information might be included in the ToBI labeling scheme and why?  What purpose do you think it might serve?


To summarize, your submission should include 6 wavfiles and aligned TextGrids for Part I; 2 wavfiles and aligned TextGrids plus a README for Part II; one text/doc/pdf file for Part III.  Please make sure you have named the files appropriately and included all the files in your submission.  Create a zip file with all files names as noted above.  Name the file hw4-<uni>.zip.  This file should expand to a directory containing all your files.  Submit the zip file in Courseworks to the HW4 folder.