Keep only f0 frames that fall inside chosen TextGrid intervals

Subsets a long-format f0 data frame — with landmark columns already attached by attach_landmarks() — to the rows whose time falls inside selected intervals of one tier: automatically detected vowel intervals, the rhyme (first vowel interval to the end of the token; monosyllabic data only), or an explicit set of labels.

Usage

filter_interval_rows(
  df,
  set,
  mode = c("vowel", "rhyme", "labels"),
  labels = NULL,
  token = "token"
)

Arguments

df

Long-format data frame with token plus the tier's landmark columns (<set>, <set>_i, ...) from attach_landmarks().

set

Landmark-set base name, i.e. the sanitised tier name used as the label column (e.g. "vowel", "segment").

mode

One of "vowel", "rhyme", or "labels". See Details.

labels

Character vector of interval labels to keep (only for mode = "labels").

token

Name of the token-ID column (used by mode = "rhyme"). Default "token".

Value

The subset of df, row order preserved.

Details

The three modes:

• "vowel": keep rows whose interval label passes ipa_vowel_label() — all vowel intervals, whatever the syllable count.

• "rhyme": keep rows in labelled intervals from the first vowel interval of each token to the token's end (vowel + coda). This assumes each token is one syllable; for multisyllabic tokens it would span from the first vowel across every following syllable. Tokens with no vowel-labelled interval are dropped entirely.

• "labels": keep rows whose (whitespace-trimmed) label equals one of labels — for tiers where the region of interest is marked explicitly (e.g. a rhyme tier, or non-IPA label schemes).

Rows with an NA label — tokens with no matching TextGrid, or frames outside the tier's span — are always excluded, as are frames in empty-labelled (silence) intervals. Callers should tell users how many tokens were lost that way rather than let them vanish silently.

See also

attach_landmarks() to add the landmark columns; ipa_vowel_label() for the vowel test.