Target Child Productions

Concrete, imageable object words spoken by the infant who was the research subject were also coded in a similar manner as nouns spoken by other adults and children in the recordings.

The key differences for child productions (applying to target child only, not siblings or other recorded children in the file) are as follows:

1. For child productions, annotators would write the intended object word, not the noun as it was spoken in the file. For adults and other children, the noun was annotated exactly as spoken. As an example, if the target child said, "buh-buh," and their mother responded, "yes, that's your belly button," an annotator would code the noun "belly+button" with speaker "CHI" (target child) and "belly+button" with speaker "MOT". However, if a parent refers to an object in baby-talk form, e.g. the child's mother says, "Here's your baba," meaning bottle, the annotator would code "baba" with speaker "MOT". Note: If the parent refers to an object by a baby-talk nickname, and then the target child for that file subsequently uses the same form in that recording, then the baby-talk form was used for the child as well as the parent. This happened most frequently with forms of bottle (e.g. "baba," "ba," "bob").

2. Child productions were phonetically transcribed by specially trained coders. These phonetic transcriptions use a "translation" of the IPA (International Phonetic Association) alphabet into ASCII symbols called "UNIBET" (MacWhinney, 1992). The phonetic transcriptions appear in a column of SEEDLingS Nouns called "transcription". 10% of each file's child productions were phonetically transcribed by a second specially trained coder to assess reliability between phonetic transcribers.

Note: other kinds of codes were assigned to the target child in the SEEDLingS dataset, but do not appear in SEEDLingS Nouns. These include tags for multi word utterances spoken by the child, the first appearance of pointing by the target child in videos, the first appearance of walking by the target child in videos, and the child's first five words spoken in the recordings (including words that were not object words and therefore wouldn't normally have been coded).