Clark Kevin B
Research and Development Service, Veterans Affairs Greater Los Angeles Healthcare System, Los Angeles, CA, USA.
NeuroLex Laboratories, Inc., Atlanta, GA, USA.
Commun Integr Biol. 2018 Apr 27;11(2):1-2. doi: 10.1080/19420889.2018.1445899. eCollection 2018.
A recent theoretical treatment by Christiansen and Chater attempts to address fundamental challenges significant to language processing and evolution with one major operational constraint called the "Now-or-Never" bottleneck. The authors' "Chunk-and-Pass" processing putatively mitigates the severe multilevel Now-or-Never bottleneck via fast linguistic coding and compression, hierarchical language representation and pattern duality, and incrementally learned item-based predictions useful for grammaticalization over wide spacetime scales. Despite being a promising explanation of language processes, structure, and development, the Chunk-and-Pass model manages the Now-or-Never constraint with seeming reliance on optimal joint source-channel coding, a set of computational attributes for natural and artificial speech based on Shannon's noisy channel theorems. Restating the Now-or-Never bottleneck with information-theoretic source-channel capacity limitations stresses tradeoffs inherent in the authors' model involving multilevel lossy code-transmission rate and security. Such attributes render evolvable associative networks capable of Chunk-and-Pass speech acquisition, recognition, generation, and adaptation, suggesting Chunk-and-Pass processing represents a special case of joint source-channel coding.
克里斯蒂安森和查特最近的一项理论研究试图通过一个名为“机不可失”瓶颈的主要操作限制来应对语言处理和进化中面临的重大基本挑战。作者提出的“组块传递”处理方式据称通过快速语言编码和压缩、层次化语言表示和模式对偶性，以及在广泛时空尺度上对语法化有用的增量式基于项目的预测，缓解了严重的多级“机不可失”瓶颈。尽管“组块传递”模型对语言过程、结构和发展给出了一个很有前景的解释，但它在管理“机不可失”约束时，似乎依赖于最优联合源信道编码，这是一组基于香农噪声信道定理的自然语音和人工语音的计算属性。用信息论的源信道容量限制来重新阐述“机不可失”瓶颈，强调了作者模型中固有的权衡，这些权衡涉及多级有损码传输速率和安全性。这些属性使可进化的联想网络能够进行“组块传递”语音获取、识别、生成和适应，这表明“组块传递”处理代表了联合源信道编码的一种特殊情况。
