search

bendichter / ndx-speech

Other
0 stars 0 forks source link

readme

ndx-speech Extension for NWB:N

This in a NeuroData Extension (NDX) for Neurodata Without Borders: Neurophysiology (NWB:N) for transcriptions of natural speech. It consists of a single neurodata_type, Transcription, which is an NWBDataInterface that contains TimeIntervals objects labelled "sentences", "words", "syllabels", "phonemes", and "phoneme features". Each of these TimeIntervals objects is optional, and may be included or omitted based on the resolution of the transcription. schematic

Installation

pip install ndx-speech

Usage

python

from ndx_speech import Transcription
from pynwb.epoch import TimeIntervals

# build it yourself:
sentences=TimeIntervals()
sentences.add_row(start_time=0.1, stop_time=0.5, label='I never said she stole my money')
sentences.add_row(start_time=1.2, stop_time=1.3, label='Little pigs are bad carpenters')

# or use pandas DataFrames:
import pandas as pd
sentences = TimeIntervals.from_dataframe(pd.read_csv('some_file.csv'))

# similar for words, syllables, phonemes, and phoneme features

transcription = Transcription(sentences=sentences)

MATLAB

documentation under construction