m-markowska/Shupamem-tones

# Finite-state-tonal-representations-in-Shupamem-possessives # Shupamem Tones This repository contains Python code for modeling **tonal patterns in Shupamem nouns and possessive constructions** using finite-state transducers (FSTs) with Pynini. The script `shupamem_tone_pipeline.py`: 1. Builds a small **lexical corpus** of Shupamem nouns and possessive pronouns. 2. Computes **underlying tonal representations (URs)** for several noun classes. 3. Applies **tone rules** (e.g., H-spreading, L-insertion, multiple linking) using Pynini FSTs. 4. Generates a TSV file with the predicted **surface tones in possessive phrases**. The code is meant as an executable companion to my research on **finite-state modeling of Shupamem tone** and tonal morphophonology more generally. --- ## Project Structure - `shupamem_tone_pipeline.py` Main script. Creates the corpus, derives URs, applies tone rules, and writes the final output file. - `shupamem_nouns_corpus.tsv` TSV corpus file generated by `shupamem_tone_pipeline.py` (lexical items, tones, POS tags, and English glosses). - `shupamem_poss_output.tsv` TSV output file generated by `shupamem_tone_pipeline.py` with predicted surface tones for noun + possessive pronoun combinations. > Note: `shupamem_nouns_corpus.tsv` and `shupamem_poss_output.tsv` are created/overwritten each time you run `shupamem_tone_pipeline.py`. --- ## Dependencies - Python 3.8+ (tested with 3.9/3.10) - Pynini - Standard library: - `csv` - `string` ### Installing Pynini Pynini can be installed via `pip` (prebuilt wheels) or from source. A typical installation on Linux / macOS might look like: ```bash pip install pynini