Annotation 2: Segments

[previous] [next]

[1. Basis of the reconstruction] [2. Symbols and features]

1. Basis of the reconstruction

The smallest unit of tokenization in the OGR is the Segment (seg). The Segments are not intended to provide a phonetic transcription of the text but instead a consistent and phonemically-informed interpretation of the spelling, which in these early texts is broadly phonemic. The segmental annotation respects the following principles:

1. Each Segment is represented by at least one character in the text, i.e. there are no unwritten Segments.
2. There are no unpronounced characters, with the exception of those within digraphs and Latin initial h.
3. Each Segment is maximally underspecified. In all contexts in which a particular phonological feature is non-distinctive (e.g. obstruent voicing in word-final position), it is left underspecified in the annotation. This also applies where the precise phonetic value of a Segment is uncertain (e.g. in Latinate loanwords or proper names).
4. This notwithstanding, where phonological features are distinctive, this is reflected in the Segment annotation even if it is not shown in the spelling. E.g. the letter e in tonic syllables may be transcribed /e/, /ɛ/, or /æ/.
5. Although the phonemic symbols have a standard value (see below) the Segment annotation is not standardized across the corpus. The same word may be transcribed differently in different texts to reflect graphical and phonological differences.

More extensive documentation of the segmental annotation will be found in published scientific presentations of the corpus.

2. Symbols and features

Segment-level annotation is contained in four core tags:

• seg_plus and seg_minus: two comma-separated lists of alphabetically-ordered phonological features, which are specified as “+” or “-” respectively for the Segment. The more features included on these lists, the more specified the segment.
• seg_phoneme: a character denoting a particular matrix of phonological features. IPA characters are generally used with their standard value. Underspecified segments are represented by non-IPA characters or capitals.
• seg_matches: a string of all the characters that match the matrix of phonological features. For example, the palatal nasal stop “ɲ” also matches the underspecified characters “N” (nasal consonant) and “C” (consonant). The seg_matches string is therefore CNɲ. In practical terms, this means that the query seg_matches=/.*N.*/ will match all nasal consonants.

When searching for classes of sounds, use the either seg_plus and seg_minus or seg_matches tags, e.g.:

• Find all vowels and glides:
  • seg_minus=/.*cons.*/
  • OR seg_matches=/.*V.*/
• Find all coda nasals:
  • onc="C" _i_ seg_plus=/.*nas.*/
  • OR onc="C" _i_ seg_matches=/.*N.*/
• Find all word-final obstruents:
  • word _r_ seg_minus=/.*son.*/
  • OR word _r_ seg_matches=/.*Q.*/
• Find all word-final coronal stops:
  • seg_plus=/.*CORONAL.*/ _=_ seg_minus=/.*cont,.*son,.*strident.*/ & word _r_ #1
  • OR word _r_ seg_matches=/.*T.*/

In the TXM version of the corpus, Segment-level annotation is available through the word-level phon tag, which concatenates all the seg_phonemes in the word as a single string. The seg_plus, seg_minus and seg_matches tags are not available.

2.1 Vowels and glides

2.2 Consonants

[previous] [next]