Skip to content

Commit

Permalink
Better phonology -- ispras/lingvodoc-react#1117 (#1502)
Browse files Browse the repository at this point in the history 
* Separated functionality

* debug

* test_phonology fixes

* more optimized

---------

Co-authored-by: Ivan Beloborodov <[email protected]>
  • Loading branch information
@vmonakhov @myrix
vmonakhov and myrix authored Apr 2, 2024
1 parent eb2d5b7 commit c1676cd
Show file tree
Hide file tree
Showing 3 changed files with 95 additions and 78 deletions.
1 change: 1 addition & 0 deletions lingvodoc/schema/gql_cognate.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1945,6 +1945,7 @@ def acoustic_data(
markup_entity_id,
markup_url,
storage,
False,
__debug_flag__))

if textgrid_result_list is None:
Expand Down
1 change: 1 addition & 0 deletions lingvodoc/schema/query.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -5130,6 +5130,7 @@ def perform_phonological_statistical_distance(
markup_id,
markup_url,
storage,
vowel_selection,
__debug_flag__))

row_count += 1
Expand Down
171 changes: 93 additions & 78 deletions lingvodoc/views/v2/phonology.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2273,7 +2273,7 @@ def before_after_text(index, interval_list, join_set = None):
return before_text, after_text


def process_sound(tier_data_list, sound):
def process_sound(tier_data_list, sound, vowel_selection = None):
"""
Analyzes sound intervals corresponding to vowel-containing markup.
"""
Expand All @@ -2300,6 +2300,17 @@ def process_sound(tier_data_list, sound):
if len(interval_list) <= 0:
continue

# Computing average sound interval length.

total_interval_length = (

sum(
end - begin
for raw_index, (begin, end, text) in raw_interval_list))

mean_interval_length = (
total_interval_length / len(raw_interval_list))

# Looking in particular at longest interval and interval with highest intensity, and at all
# intervals in general.

Expand All @@ -2309,26 +2320,7 @@ def process_sound(tier_data_list, sound):
max_intensity_interval = interval_list[max_intensity_index]
max_length_interval = interval_list[max_length_index]

max_intensity_i_list = (
sound.get_interval_intensity(*max_intensity_interval[:2]))

max_length_i_list = (
sound.get_interval_intensity(*max_length_interval[:2]))

max_intensity_f_list = (
sound.get_interval_formants(*max_intensity_interval[:2]))

max_length_f_list = (
sound.get_interval_formants(*max_length_interval[:2]))

intensity_list = [
sound.get_interval_intensity(begin_sec, end_sec)
for begin_sec, end_sec, text in interval_list]

formant_list = [
sound.get_interval_formants(begin_sec, end_sec)
for begin_sec, end_sec, text in interval_list]

# Pitches are calculated for the whole sound content at once.
pitch_list = sound.get_pitch()

# Getting desired intervals and mean values within them from the obtained 'pitch_list'.
Expand Down Expand Up @@ -2364,16 +2356,85 @@ def process_sound(tier_data_list, sound):
print(f"XL Mean: {xl_p_mean:.3f} Hz", file=f)
print(f"XI Mean: {xi_p_mean:.3f} Hz\n", file=f)

# Computing average sound interval length.
max_intensity_i_list = []
max_length_i_list = []
max_intensity_f_list = []
max_length_f_list = []
interval_data_list = []

total_interval_length = (
if vowel_selection is None or vowel_selection == True:
#print('Calculating max_intensity_ and max_length_ lists..')

sum(
end - begin
for raw_index, (begin, end, text) in raw_interval_list))
max_intensity_i_list = (
sound.get_interval_intensity(*max_intensity_interval[:2]))

mean_interval_length = (
total_interval_length / len(raw_interval_list))
max_length_i_list = (
sound.get_interval_intensity(*max_length_interval[:2]))

max_intensity_f_list = (
sound.get_interval_formants(*max_intensity_interval[:2]))

max_length_f_list = (
sound.get_interval_formants(*max_length_interval[:2]))

if vowel_selection is None or vowel_selection == False:
#print('Calculating lists for all intervals...')

intensity_list = [
sound.get_interval_intensity(begin_sec, end_sec)
for begin_sec, end_sec, text in interval_list]

formant_list = [
sound.get_interval_formants(begin_sec, end_sec)
for begin_sec, end_sec, text in interval_list]

# Preparing data of all other intervals.

str_list = [

'{0} {1:.3f} {2:.3f} [{3}]'.format(
text.strip(),
end_sec - begin_sec, intensity,
len(''.join(text for raw_index, (begin, end, text) in
raw_interval_list[:interval_idx_to_raw_idx[seq_index][index]])))

for index, ((intensity, _, _), (begin_sec, end_sec, text)) in
enumerate(zip(intensity_list, interval_list))]

source_index_list = [
interval_idx_to_raw_idx[(seq_index, index)]
for index in range(len(interval_list))]

# Compiling results.

interval_data_list = [

(interval_str,
(end - begin) / mean_interval_length,
f'{p_mean:.3f}',
[f'{i_min:.3f}', f'{i_max:.3f}', f'{i_max - i_min:.3f}'],
list(map('{0:.3f}'.format, f_list)),
'+' if index == max_length_index else '-',
'+' if index == max_intensity_index else '-',
source_index)

for (
index,
(interval_str,
p_mean,
(_, i_min, i_max),
f_list,
(begin, end, text),
source_index)) in

enumerate(
zip(
str_list,
pitch_means,
intensity_list,
formant_list,
interval_list,
source_index_list))]

# Preparing data of maximum length and maximum intensity intervals.

Expand All @@ -2393,53 +2454,6 @@ def process_sound(tier_data_list, sound):
len(''.join(text for index, (begin, end, text) in
raw_interval_list[:interval_idx_to_raw_idx[seq_index][max_intensity_index]])))

# Preparing data of all other intervals.

str_list = [

'{0} {1:.3f} {2:.3f} [{3}]'.format(
text.strip(),
end_sec - begin_sec, intensity,
len(''.join(text for raw_index, (begin, end, text) in
raw_interval_list[:interval_idx_to_raw_idx[seq_index][index]])))

for index, ((intensity, _, _), (begin_sec, end_sec, text)) in
enumerate(zip(intensity_list, interval_list))]

source_index_list = [
interval_idx_to_raw_idx[(seq_index, index)]
for index in range(len(interval_list))]

# Compiling results.

interval_data_list = [

(interval_str,
(end - begin) / mean_interval_length,
f'{p_mean:.3f}',
[f'{i_min:.3f}', f'{i_max:.3f}', f'{i_max - i_min:.3f}'],
list(map('{0:.3f}'.format, f_list)),
'+' if index == max_length_index else '-',
'+' if index == max_intensity_index else '-',
source_index)

for (
index,
(interval_str,
p_mean,
(_, i_min, i_max),
f_list,
(begin, end, text),
source_index)) in

enumerate(
zip(
str_list,
pitch_means,
intensity_list,
formant_list,
interval_list,
source_index_list))]

textgrid_result_list[-1][2].append(

Expand Down Expand Up @@ -2475,6 +2489,7 @@ def process_sound_markup(
markup_entity_id,
markup_url,
storage,
vowel_selection = None,
__debug_flag__ = False):
"""
Extracts phonology data from a pair of sound recording and its markup, using cache in a manner
Expand Down Expand Up @@ -2605,7 +2620,7 @@ def process_sound_markup(
# Analysing sound, showing and caching analysis results.

textgrid_result_list = process_sound(
tier_data_list, sound)
tier_data_list, sound, vowel_selection)

log.debug(
'{0}:\n{1}'.format(
Expand Down Expand Up @@ -4778,7 +4793,7 @@ def no_vowel_selected_f(tier_number, tier_name, transcription_list, selected_lis
vowel_range_list if args.interval_only else None))

textgrid_result_list = (
process_sound(tier_data_list, sound))
process_sound(tier_data_list, sound, args.vowel_selection))

caching.CACHE.set(cache_key, textgrid_result_list)

Expand Down Expand Up @@ -7161,7 +7176,7 @@ def no_vowel_selected_f(tier_number, tier_name, transcription_list, selected_lis

sound = AudioPraatLike(pydub.AudioSegment.from_wav(wav_path))

textgrid_result_list = process_sound(tier_data_list, sound)
textgrid_result_list = process_sound(tier_data_list, sound, args.vowel_selection)
result_list.append(textgrid_result_list)

result_string = '\n'.join(
Expand Down

0 comments on commit c1676cd

Please sign in to comment.