A TextGrid file contains data about intervals, segments, times etc. of the corresponding signal file (audio in wav, mp3, aif…). Because grids are in plain-text – they can be analysed / checked / extracted automatically, or parsed.
In case you are a linguist/phonetician you might be using Praat, a small, but very powerful, programme for phonetic analysis. Chances are have a lot of speakers and recordings. You will probably segment signals in Praat, and save the segmentation in TextGrids.
Thanks to Margaret Mitchell and Steven Bird, who contributed the parser for Praat TextGrid to Natural Language Toolkit, automated analysis is now much easier.
I am grateful to the authors, because they saved me a lot of time during segmentation checks. All that was needed was a Python script that uses the above code to load TextGrid content, and then write a set of checks for each file/speaker.
Checking file 03-speaker-im.TextGrid
Checking proper tier names...
Checking if tiers contain 32 items...
Checking if all tiers have valid text...
Checking if the diphthongs have pairs...
Checking if all words are present...
Checking if the words and diphthongs match...
Mismatch: "ay_l" not allowed in "dice", at position 24.
It should say "ay_s".
Here, for example, my script warned me that I have a wrong label for a diphthong in the file number 3. To spot that “manually” it would require a lot of time and attention.
It’s not always possible to find a good searchable phonetic dictionary. That is why I created a free and open source program that searches phonetically transcribed words and filters the results against some basic rules. It uses BEEP and Moby Hyphenator II sources.
For several month I have been working on my M.A. in experimental phonetics. One of the prerequisites is an acceptable corpus. My work is about the English diphthongs. However, diphthongs have to be pronounced after voiced plosives and before voiced/unvoiced plosives, and the words containing diphthongs should preferably be monosyllabic.
Making a corpus is not an easy task and it involves a painstaking search for suitable material. I had no searchable phonetic dictionary of any sort (a version of Macmillan Advanced Dictionary refused to work). It was a pure luck, then, to come across a paper where the bibliography listed one interesting source: University of Cambridge public FTP server. That is where I found BEEP and MH2 and decided to compile my own searchable dictionary, hopefully usable for the making the corpus.
What FONRYE is, and what it is not
FONRYE (named after fonetski rječnik in Serbian) is a very simple program (or script, if you like) written in Python 2.6. It is a specific piece of software I created for personal use: to search for diphthongs in a phonetic context. It does not have any fancy search rules or regular expression syntax. The plan was to use regexp, but it was very slow to run – I guess it can be improved if needed. So, please bear in mind that it was not planned for releasing: the code may contain strange comments, bad spelling etc.
Its settings are contained in the script itself, in 4 lines of code, which will be explained later. Here’s an example:
before = ('m', 'n', 'r', 'l' ),
after = sounds['voiceless'] + sounds['voiced'],
diphthongs = sounds['diphthongs'],
syllable = 0
The user enters desired search conditions, executes the program, which then saves the results in a folder, accompanied by a short info.
How to use FONRYE
On Windows/Mac: Download Python, but a version lower than 3.0. The version 2.6 is preferred. On Linux: You already have Python installed, but make sure you have an “old” version as well (again, prior to the version 3).
Download FONRYE files, and unpack them. Please make sure you do not delete ‘results’ folder or the program will not work. On Windows go to Start menu and find IDLE inside Python folder. On Linux: Use any plain text editor which supports code editing, such as gedit. Or, install Python IDLE from your OS repository. Edit file run.py, enter your settings and save the file. Finally, run the program (double click run.py or press F5 in IDLE).
The program will start the search, and after it finishes the results will be in results/fonyre_results_n, where n is the search counter.
Settings and results format
In the step 3 above you opened run.py file. Here is how to enter the “settings”. First, locate these lines:
before = (),
after = (),
diphthongs = ()
syllable = 0
Do not modify anything except content inside the brackets and syllable number (that is, unless you are familiar with programming). By the way, syllable = 0 means words with 1 syllable, syllable = 1 with 2 syllables etc. Enter your phonemes in the brackets. For example, the settings:
before = ('b', 'd'),
after = ('p', 't'),
diphthongs = ('ay',)
syllable = 0
…will search for all words containing diphthong ay (IPA: aɪ) if the diphthong is between b/d and p/t. After the search is done, go to ‘results/fonrye_search’ folder and locate search_info.txt (here is a vowel search info, a sample) – that is info about your search, including unique mark (ID) placed in all result files to keep track of the searches/results. The folder ‘files’ is where your searches are placed. For the provided sample search the program produced the following file/results:
BIGHT b ay t
BITE b ay t
BLIGHT b l ay t
BRIGHT b r ay t
BRIGHTS b r ay t s
BY-PASS b ay p aa s
DIGHT d ay t
I could create and use this little project, and place it on the Net, thanks to two people who provided the core of the project: a phonetic dictionary and a hyphenation dictionary. The phonetic dictionary was compiled by Toby Robinson from Cambridge University Engineering Department; Moby Hyphenation dictionary was created by Grady Ward. Both the projects were placed into the public domain in 1996. See Bibliography page for FTP addresses.
My credits are for some fast-writing not-so-good-looking slow Python code, which you are free to improve and share.