Python implementation of the SFC intonation model.

The Superposition of Functional Contours (SFC) model is a prosody model that is based on the decomposition of prosodic contours into functionally relevant elementary contours [1]. It proposes a generative mechanism for encoding socio-communicative functions, such as syntactic structure and attitudes, through the use of prosody. The SFC has been successfully used to model different linguistic levels, including: attitudes, dependency relations of word groups, word focus, tones in Mandarin, etc. It has been used for a number of languages including: French, Galician, German and Chinese. Recently, the SFC model has been extended into the visual prosody domain through modelling facial expressions and head and gaze motion.

The SFC model is based on neural network contour generators (NNCGs) each responsible for encoding one linguistic function on a given scope. The prosody contour is then obtained by overlapping and adding these elementary contours. NNCG training is done using an analysis-by-synthesis loop that distributes the error and usual backpropagation training at each iteration. Four syllable position ramps are used by the NNCGs to generate pitch and duration coefficients for each syllable.

[1] Bailly, Gérard, and Bleicke Holm. "SFC: a trainable prosodic model." Speech communication 46, no. 3 (2005): 348-364.

PySFC is a Python implementation of the SFC model that was created with two goals: i) to make the SFC more accessible to the scientific community, and ii) to serve as a foundation for future improvements of the prosody model. The PySFC also implements a minimum set of tools necessary to make the system self-contained and fully functional.

Python was chosen as an implementation language because of the powerful scientific computing environment that is completely based on free software. It is based on NumPy within the SciPy ecosystem. The neural networks and their training have been facilitated through the use of the Multi Layer Perceptron (MLP) regressor in the powerful scikit-learn module. Great attention was put on code readability, which is also one of the features of good Python, augmented with detailed functions docstrings, and comments. The code is segmented in Spyder cells for rapid prototyping. Finally, the whole implementation has been licensed as free software with a GNU General Public License v3.

PySFC comprises the following modules:

• sfc.py - main module that controls the application of the SFC model to a chosen dataset.
• sfc_params.py - parameter setting module that includes:
  • data related parameters - type of input data, phrase level functional contours, local functional contours,
  • SFC hyperparameters - number of points to be sampled from the pitch, number of iterations for analysis-by-synthesis, as well as NNCG parameters,
  • SFC execution parameters - use of preprocessed data, use of trained models, plotting.
• sfc_corpus.py - holds all the functions that are used to consolidate and work with the corpus of data that is directly fed and output from the SFC model. The corpus is a Pandas data frame object, which allows easy data access and analysis.
• sfc_data.py - comprises functions that read the input data files and calculate the f_0 and duration coefficients,
• sfc_learn.py - holds the SFC training function analysis_by_synthesis() and the function for NNCG initialisation,
• sfc_dsp.py - holds DSP functions for smoothing the pitch contour based on SciPy,
• sfc_plot.py - holds the plotting functions based on matplotlib and seaborn.

Currently, PySFC supports the proprietary SFC fpro file format as well as standard Praat TextGrid annotations. Pitch is calculated based on Praat PointProcess pitch mark files, but integration of state-of-the-art pitch extractors is planned for the future. PySFC also brings added value, by adding the possibility to adjust the number of samples to be taken from the pitch contour at each rhythmical unit vowel nucleus, and with its extended plotting capabilities for data and performance analysis.

Here are a few example plots with PySFC just to show case what it can do. The plotted files are included as examples in the examples/ directory.

Figure 1. Example PySFC intonation decomposition for the French utterance: Son bagou pourrait faciliter la communauté. into constituent functional contours: declaration (DC), dependency to the left/right (DG/DD), and cliticisation (XX, DV).

Figure 2. Example PySFC intonation decomposition for the Chinese utterance: Tā men céng zài jī cāng nèi géi lǔ kè diǎn gē hè shēng rì, céng ná zhē shuí guǒ nái fěn qù tàn wàng yóu tā men zhuǎn sòng qù yī yuàn de lǔ kè chǎn fù. into constituent functional contours: declaration (DC), tones (C0-4), word boundaries (WB), and independence (ID).

Figure 3. Example PySFC expansion in left and right context for the dependency to the right (DD) functional contour, numbers next to the plots show the number of occurences of that scope in the data.

Figure 4. Example PySFC plots of f_0 reconstruction losses for all NNCGs for attitude DC per iteration for French.