Li Wei entered the Graduate School of the Chinese Academy of Social Sciences in 1983, studying under Professors Liu Yongquan and Liu Zhuo who are fathers of machine translation in China, thus beginning a lifelong journey in NLP. After graduation, he continued MT research at the Institute of Linguistics (CASS), then pursued doctoral work in the United Kingdom and Canada, earning a PhD in Computational Linguistics from Simon Fraser University. Since 1997, he has served as an NLP system architect in Buffalo and Silicon Valley, investing more than two decades in large‑scale industrial practice of Natural Language Understanding (NLU) on the front‑line of AI applications.
Guo Jin received his PhD in Computer Science from the National University of Singapore in 1994 with a focus on Chinese tokenization and statistical language modelling, work published in Computational Linguistics and related venues. Moving to the United States in 1998, he held research posts at Motorola, Amazon, and the JD Silicon Valley Research Center, exploring applications that fuse machine learning, NLP, and human–computer interaction across internet and IoT scenarios.
From the 1980s onward, the AI community has witnessed a “two‑track contest” between rationalism and empiricism in NLP. The ascendancy of machine learning has gradually eclipsed the grammar school, and computational grammar risks a generational break.
In 2018, over ten extended conversations in Silicon Valley, Li and Guo revisited the symbolic legacy and debated paths forward. Those dialogues became the backbone of the present volume, calling for a rationalist renaissance to dismantle the cognitive citadels that still impede AI.
零 缘起
自20世纪80年代起, 人工智能领域见证了理性主义 (rationalism) 与经验主义(empiricism) 的“两条路线斗争”。其中, 自然语言学界的“斗争”结果是, 文法学派(grammar school) 与统计学派 (statistical school) 此消彼长, 机器学习渐 成主流, 计算文法 (computational grammar)则有断代之虞。
李维, 1983年进入中国社会科学院研究生院, 师从刘涌 泉、刘倬先生, 主攻基于文法的机器翻译 (machine translation), 始入自然语言领域。毕业后在中国社会科学院语言研究所从事机器翻译研究, 继而留学英国、加拿大, 获Simon Fraser University (SFU) 计算语言学 (Computational Linguistics) 博士。1997年起, 在美国水牛城、硅谷, 从事自然语言理解 (Natural Language Understanding, NLU) 工业实 践20余载, 为人工智能(Artificial Intelligence, AI) 应用第一 线的系统架构师。
郭进, 1994年新加坡国立大学计算机科学博士, 主攻中文分词 (Chinese tokenization) 和统计模型 (statistical model), 成果见于「计算语言学」等刊。1998年赴美, 先后在摩托罗拉、亚马逊、京东硅谷研究院等从事人工智能研究, 探索将机器学习 (machine learning)、自然语言处理 (Natural Language Processing, NLP) 等人机交互技术应用于互联网与物联网的解决方案。2018年, 李与郭在硅谷就自然语言解析 (natural language parsing) 问题有十次长谈, 回顾并展望文法学派的机制创新与传承之路, 意图呼唤理性主义回归, 解构自然语言, 协同攻坚人工智能的认知堡垒, 遂成此作。