Narrowing the distance to language

Abstract

Infants' earliest attempts at word segmentation appear to be guided by a single source of information (e.g., English-learners initially rely on the predominant stress pattern of words). This initial strategy successfully identifies many potential words in the input, but mis-segments others. However, simply breaking the input into smaller chunks helps learners to identify other possible cues to the location of word boundaries in utterances. Because no one source of information is completely reliable, listeners must eventually rely on multiple cues to segment words. The development of such skills is not critical for developing a native language vocabulary, but also for acquiring the grammatical organization of utterances. Tracking familiar sound patterns, such as function words and grammatical morphemes, may help in learning about syntactic organization. One factor that facilitates learning about the distribution of such elements is sensitivity to boundaries of prosodic phrases. Access to such linguistically-relevant chunks also helps in tracking the distribution of words in the input.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 32 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.