基底神经节和规则支配性语言在血管性和变性疾病中的作用

The Declarative/Procedural Model of Pinker, Ullman and colleagues claims that the basal ganglia are part of a fronto-striatal procedural memory system which applies grammatical rules to combine morphemes (the smallest meaningful units in language) into complex words (e.g. talk-ed, talk-ing). We tested this claim b y investigating whether striatal damage or loss of its dopaminergic innervation is reliably associated with selective regular past tense deficits in patients wi th subcortical cerebrovascular damage, Parkinson’s disease or Huntington’s dis ease.We focused on past tense morphology since this allows us to contrast the re gular past tense (jump-jumped), which is rulebased,with the irregular past tens e (sleep-slept), which is not We used elicitation and priming tasks to test pat ients’ability to comprehend and produce inflected forms. We found no evidence o f a consistent association between striatal dysfunction and selective impairment of regular past tensemorphology, suggesting that the basal ganglia are not esse ntial for processing the regular past tense as a sequence of morphemes, either i n comprehension or production, in contrast to the claims of the Declarative/Proc edural Model. All patient groups showed normal activation of semantic and morpho logical representations in comprehension, despite difficulties suppressing seman tically appropriate alternatives when trying to inflect novel verbs. This is con sistent with previous reports that striatal dysfunction spares automatic activat ion of linguistic information, but disrupts later language processes that requir e inhibition of competing alternatives.

Lexical Tones and Word Learning in Mandarin-Speaking Children at Three Years of Age(Invited paper)

Unlike Indo-European languages,Mandarin relies heavily on lexical tones to distinguish word identity. Using the intermodal preferential looking paradigm, this study examined 3-year-old Mandarinspeakers＇ ability to use Mandarin lexical tones in learning new words. Results showed that when children were presented with Tone 2（rising） and Tone 4（falling）pairs, children successfully learned both words.However, when children were presented with Tone 2and Tone 3（dipping） pairs, they learned the Tone 2word but not the Tone 3 one. Children were then divided into two groups based on their learning performance on the Tone 3 word. Successful learning of Tone 3 words was observed in the high performers but not in the low performers, who consistently misused Tone 3 as Tone 2. This study showed that Mandarinspeaking 3-year-olds could use lexical tones to learn words under experimental conditions, and that the difficulty of Tone 3 acquisition may be related to its lower level of perceptual distinctiveness compared with other tones.

Term-Dependent Confidence Normalisation for Out-of-Vocabulary Spoken Term Detection

An important component of a spoken term detection （STD） system involves estimating confidence measures of hypothesised detections.A potential problem of the widely used lattice-based confidence estimation,however,is that the confidence scores are treated uniformly for all search terms,regardless of how much they may differ in terms of phonetic or linguistic properties.This problem is particularly evident for out-of-vocabulary （OOV） terms which tend to exhibit high intra-term diversity.To address the impact of term diversity on confidence measures,we propose in this work a term-dependent normalisation technique which compensates for term diversity in confidence estimation.We first derive an evaluation-metric-oriented normalisation that optimises the evaluation metric by compensating for the diverse occurrence rates among terms,and then propose a linear bias compensation and a discriminative compensation to deal with the bias problem that is inherent in lattice-based confidence measurement and from which the Term Specific Threshold （TST） approach suffers.We tested the proposed technique on speech data from the multi-party meeting domain with two state-ofthe-art STD systems based on phonemes and words respectively.The experimental results demonstrate that the confidence normalisation approach leads to a significant performance improvement in STD,particularly for OOV terms with phonemebased systems.