import polars as pl
from aligned_textgrid import AlignedTextGrid, Word, Phone,\
to_df
tg = AlignedTextGrid(
textgrid_path="../resources/josef-fruehwald_speaker.TextGrid",
entry_classes=[Word, Phone]
)Outputting and saving as a DataFrame
You can ouput and save any given aligned_textgrid as a polars dataframe with the to_df() function.
Outputting as a data frame
Single Intervals
Bottom of the hierarchy
If you pass a single interval from the bottom of the sequence hierarchy, you’ll get back a fairly minimal dataframe with the start and end times, the label, and an ID for the interval.
one_interval = tg[0].Phone[1]
one_interval_df = to_df(one_interval)
one_interval_df
shape: (1, 5)
| Phone_id | Phone_tier_index | Phone_label | Phone_start | Phone_end |
|---|---|---|---|---|
| str | i64 | str | f64 | f64 |
| "0-0-1-0" | 1 | "HH" | 0.11 | 1.97 |
Top of the hierarchy
If you pass to_df() an interval from higher up in the hierarchy, by default it will output its data, as well as the data for every interval below it in the hierarchy, concatenated horizontally.
word_interval = tg[0].Word[1]
word_interval_df = to_df(word_interval)
word_interval_df
shape: (4, 10)
| Word_id | Word_tier_index | Word_label | Word_start | Word_end | Phone_id | Phone_tier_index | Phone_label | Phone_start | Phone_end |
|---|---|---|---|---|---|---|---|---|---|
| str | i64 | str | f64 | f64 | str | i64 | str | f64 | f64 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-0" | 1 | "HH" | 0.11 | 1.97 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-1" | 2 | "W" | 1.97 | 2.09 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-2" | 3 | "EH1" | 2.09 | 2.13 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-3" | 4 | "N" | 2.13 | 2.2 |
However, if you want just a simplified, single row output for an interval, regardless of its location within the hierarchy, pass to_df(..., with_subset = False).
word_interval_df2 = to_df(word_interval, with_subset=False)
word_interval_df2
shape: (1, 6)
| id | tier_index | label | start | end | entry_class |
|---|---|---|---|---|---|
| str | i64 | str | f64 | f64 | str |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "Word" |
Tiers
If you pass a tier to to_df(), it will output a dataframe for ever interval in the tier concatenated vertically. By default, this means intervals high in the hierarchy will have their rows repeated for every interval they contain, but if you want one row per interval in the output, you can pass to_df(..., with_subset = False).
tier_df1 = to_df(tg[0].Word)
tier_df1.shape(1191, 10)tier_df1.head(10)
shape: (10, 10)
| Word_id | Word_tier_index | Word_label | Word_start | Word_end | Phone_id | Phone_tier_index | Phone_label | Phone_start | Phone_end |
|---|---|---|---|---|---|---|---|---|---|
| str | i64 | str | f64 | f64 | str | i64 | str | f64 | f64 |
| "0-0-0" | 0 | "" | 0.0 | 0.11 | "0-0-0-0" | 0 | "" | 0.0 | 0.11 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-0" | 1 | "HH" | 0.11 | 1.97 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-1" | 2 | "W" | 1.97 | 2.09 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-2" | 3 | "EH1" | 2.09 | 2.13 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-3" | 4 | "N" | 2.13 | 2.2 |
| "0-0-2" | 2 | "the" | 2.2 | 2.26 | "0-0-2-0" | 5 | "DH" | 2.2 | 2.22 |
| "0-0-2" | 2 | "the" | 2.2 | 2.26 | "0-0-2-1" | 6 | "AH0" | 2.22 | 2.26 |
| "0-0-3" | 3 | "sunlight" | 2.26 | 2.72 | "0-0-3-0" | 7 | "S" | 2.26 | 2.39 |
| "0-0-3" | 3 | "sunlight" | 2.26 | 2.72 | "0-0-3-1" | 8 | "AH1" | 2.39 | 2.44 |
| "0-0-3" | 3 | "sunlight" | 2.26 | 2.72 | "0-0-3-2" | 9 | "N" | 2.44 | 2.52 |
# 1 row per interval
tier_df2 = to_df(tg[0].Word, with_subset=False)
tier_df2.shape(377, 6)tier_df2.head(10)
shape: (10, 6)
| id | tier_index | label | start | end | entry_class |
|---|---|---|---|---|---|
| str | i64 | str | f64 | f64 | str |
| "0-0-0" | 0 | "" | 0.0 | 0.11 | "Word" |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "Word" |
| "0-0-2" | 2 | "the" | 2.2 | 2.26 | "Word" |
| "0-0-3" | 3 | "sunlight" | 2.26 | 2.72 | "Word" |
| "0-0-4" | 4 | "strikes" | 2.72 | 3.22 | "Word" |
| "0-0-5" | 5 | "raindrops" | 3.22 | 3.79 | "Word" |
| "0-0-6" | 6 | "in" | 3.79 | 3.89 | "Word" |
| "0-0-7" | 7 | "the" | 3.89 | 4.02 | "Word" |
| "0-0-8" | 8 | "air" | 4.02 | 4.45 | "Word" |
| "0-0-9" | 9 | "" | 4.45 | 4.61 | "Word" |
TierGroups and TextGrids
The behavior for TierGroups and TextGrids are similar. By default, the to_df() function will either return a dataframe representing the entire hierarchy structure, or will return one row for each interval in the TextGrid.
full_df1 = to_df(tg)
full_df1.shape(1191, 10)full_df1.head(10)
shape: (10, 10)
| Word_id | Word_tier_index | Word_label | Word_start | Word_end | Phone_id | Phone_tier_index | Phone_label | Phone_start | Phone_end |
|---|---|---|---|---|---|---|---|---|---|
| str | i64 | str | f64 | f64 | str | i64 | str | f64 | f64 |
| "0-0-0" | 0 | "" | 0.0 | 0.11 | "0-0-0-0" | 0 | "" | 0.0 | 0.11 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-0" | 1 | "HH" | 0.11 | 1.97 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-1" | 2 | "W" | 1.97 | 2.09 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-2" | 3 | "EH1" | 2.09 | 2.13 |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "0-0-1-3" | 4 | "N" | 2.13 | 2.2 |
| "0-0-2" | 2 | "the" | 2.2 | 2.26 | "0-0-2-0" | 5 | "DH" | 2.2 | 2.22 |
| "0-0-2" | 2 | "the" | 2.2 | 2.26 | "0-0-2-1" | 6 | "AH0" | 2.22 | 2.26 |
| "0-0-3" | 3 | "sunlight" | 2.26 | 2.72 | "0-0-3-0" | 7 | "S" | 2.26 | 2.39 |
| "0-0-3" | 3 | "sunlight" | 2.26 | 2.72 | "0-0-3-1" | 8 | "AH1" | 2.39 | 2.44 |
| "0-0-3" | 3 | "sunlight" | 2.26 | 2.72 | "0-0-3-2" | 9 | "N" | 2.44 | 2.52 |
# 1 row per interval
full_df2 = to_df(tg, with_subset=False)
full_df2.shape(1568, 6)full_df2.head(5)
shape: (5, 6)
| id | tier_index | label | start | end | entry_class |
|---|---|---|---|---|---|
| str | i64 | str | f64 | f64 | str |
| "0-0-0" | 0 | "" | 0.0 | 0.11 | "Word" |
| "0-0-1" | 1 | "when" | 0.11 | 2.2 | "Word" |
| "0-0-2" | 2 | "the" | 2.2 | 2.26 | "Word" |
| "0-0-3" | 3 | "sunlight" | 2.26 | 2.72 | "Word" |
| "0-0-4" | 4 | "strikes" | 2.72 | 3.22 | "Word" |
full_df2.tail(5)
shape: (5, 6)
| id | tier_index | label | start | end | entry_class |
|---|---|---|---|---|---|
| str | i64 | str | f64 | f64 | str |
| "0-0-374-1" | 1186 | "R" | 111.83 | 111.92 | "Phone" |
| "0-0-375-0" | 1187 | "B" | 111.92 | 112.02 | "Phone" |
| "0-0-375-1" | 1188 | "L" | 112.02 | 112.08 | "Phone" |
| "0-0-375-2" | 1189 | "UW1" | 112.08 | 112.31 | "Phone" |
| "0-0-376-0" | 1190 | "" | 112.31 | 115.065034 | "Phone" |
Saving a DataFrame
To save one of these dataframes, use one of the methods from polars, like DataFrame.write_csv()
full_df1.write_csv("test.csv")Reuse
GPLv3