ndx-hierarchical-behavioral-data Extension for NWB

Installation

pip install ndx-hierarchical-behavioral-data

Usage

Use pre-made hierarchical transcription tables:

from ndx_hierarchical_behavioral_data.definitions.transcription import TIPhonemes, HBTSyllables, HBTWords, HBTSentences

# Phonemes level
phonemes = TIPhonemes()
phonemes.add_column('max_pitch', 'maximum pitch for this phoneme. NaN for unvoiced')
for i, p in enumerate('abcdefghijkl'):
    phonemes.add_interval(label=p, start_time=float(i), stop_time=float(i+1), max_pitch=i**2)

# Syllables level
syllables = HBTSyllables(lower_tier_table=phonemes)
syllables.add_interval(label='abc', next_tier=[0, 1, 2])
syllables.add_interval(label='def', next_tier=[3, 4, 5])
syllables.add_interval(label='ghi', next_tier=[6, 7, 8])
syllables.add_interval(label='jkl', next_tier=[9, 10, 11])

# Words level
words = HBTWords(lower_tier_table=syllables)
words.add_column('emphasis', 'boolean indicating whether this word was emphasized')
words.add_interval(label='A-F', next_tier=[0, 1], emphasis=False)
words.add_interval(label='G-L', next_tier=[2, 3], emphasis=True)

# Sentences level
sentences = HBTSentences(lower_tier_table=words)
sentences.add_interval(label='A-L', next_tier=[0, 1])

View individual tiers:

sentences.to_dataframe()

words.to_dataframe()

syllables.to_dataframe()

phonemes.to_dataframe()

Hierarchical dataframe:

sentences.to_hierarchical_dataframe()

Hierachical columns, flattened rows:

from hdmf.common.hierarchicaltable import flatten_column_index

flatten_column_index(sentences.to_hierarchical_dataframe(), 1)

Denormalized dataframe:

from hdmf.common.hierarchicaltable import to_hierarchical_dataframe

to_hierarchical_dataframe(sentences).reset_index()

Denormalized dataframe with flattened columns:

from hdmf.common.hierarchicaltable import flatten_column_index

flatten_column_index(sentences.to_hierarchical_dataframe(), 1).reset_index()

This extension was created using ndx-template.