Text is an especially malleable medium for human and machine creativity. When guided by the appropriate symbolic and/or statistical models, even a small and seemingly superficial change at the formal level can result in a predictable yet profound change at the semantic and pragmatic level. Text is also a virtually unlimited resource on the web, which offers abundant, free-flowing channels of topical texts for almost every genre and register. In this paper we consider diverse approaches to transforming these input channels into new and creative streams of machine-generated outputs. We focus on the specific kind of linguistic creativity associated with metaphor, yet also demonstrate that divergent approaches to metaphor generation can, in turn, enable divergent uses and applications for machine creativity.

中文翻译：

---

不断变化的渠道：文本创意流的不同方法

文本是人类和机器创造力的一种特别具有延展性的媒介。在适当的符号和/或统计模型的指导下，即使是形式层面的微小且看似表面的变化，也可能导致语义和语用层面的可预测但深刻的变化。文本也是网络上几乎无限的资源，它为几乎所有流派和注册提供了丰富、自由流动的主题文本渠道。在本文中，我们考虑了将这些输入通道转换为机器生成输出的新的和创造性的流的多种方法。我们专注于与隐喻相关的特定类型的语言创造力，同时也证明了隐喻生成的不同方法可以反过来使机器创造力的不同用途和应用成为可能。