A subset of the LDP transcripts have received morphosyntactic coding – an additional level of annotation that describes the morphosyntactic properties and relations among the word tokens in an utterance.
These transcripts have a mor column (for both parent and child) containing morphological coding and a syn column containing syntactic coding for each utterace.
The tables below provide a key to the part-of-speech tags and grammatical-relation labels used in these two columns.
Here’s an example of an annotated utterance:
I spilled it . # utt
pro|I v|spill-PAST pro|it . # mor
1|2|SUBJ 2|0|ROOT 3|2|OBJ 4|2|PUNCT # syn
The first line shows the utterance as it would be transcribed in the utt column, the second line shows the morphology annotation that would be contained in the corresponding mor column, and the third line shows the syntactic annotation that would be contained in the correspoding syn column.
We can represent the above example in a slightly different format, combining the annotation of the two columns, with one word token per line:
# WORD LEMMA SUFFIX POS DEP REL
1 I I pro 2 SUBJ
2 spilled spill PAST v 0 ROOT
3 it it pro 2 OBJ
4 . . . 2 PUNCT
The morphology annotation indicates that the first word (“I”) is tagged as a pronoun (pro) and the second word (“spilled”) is tagged as a verb (v) and contains a past-tense suffix (PAST).
The syntactic annotation indicates that the first word (“I”) is dependent on the second word (“spilled”), the verb of the utterance, and is playing the role of grammatical subject in relation to the verb.
Note how it represents the grammatical dependency relations:
┌── PUNCT ──┐
I spilled it .
└─ SUBJ ─┘ └─ OBJ ──┘
Note
For more information about the various steps involved in adding morphosyntactic coding see our Morphosyntax Annotation Guide.