Largest-chunking and group formation: Two basic strategies for a cognitive model of linguistic processing

The present study aims at shedding further light on how Agreement Groups (AG) processing (e.g. Drienkó 2020a) and Largest Chunk (LCh) segmentation (e.g. Drienkó 2018a) can be combined to model the emergence of language. The AG model is based on groups of similar utterances which enable combinatorial mapping of novel utterances. LCh segmentation is concerned with cognitive text segmentation, i.e. with detecting word boundaries in a sequence of linguistic symbols. Previous cross-linguistic research on French, English, and Hungarian texts (Drienkó 2020b) demonstrated that LCh segmentation is not efficient when words are the basic segmentation units and utterances are the target sequences. However, almost all utterance boundaries were identified at the expense of inserting relatively many extra boundaries. These extra boundaries delineated reoccurring fragments for building longer utterances. The present analysis of English mother-child data confirms previous findings that in spite of the relatively low efficiency of word-based LCh segmentation with respect to utterance boundaries, LCh segments can still prove to be useful word combinations for AG processing. Furthermore, compared with the previous experiments, the data suggest higher boundary precision (42%) and higher coverage (85%). These findings, on the one hand, support the claim that LCh fragments can be useful in linguistic processing (with AGs), and, on the other hand, are in line with a view that mother-child language facilitates processing more than other speech contexts.