机器翻译:语义还是数据?——人工智能视域下的语义问题与哲学反思

Machine Translation: Semantics or Data?——Semantic Problems and Philosophical Reflections from the Perspective of Artificial Intelligence

翻译行为可以看作维特根斯坦"语言游戏"的一种。"积木世界"的尝试,说明机器翻译同样也是一种"语言游戏",涉及哲学上语义问题的争论。随着人工智能的出现,经由统计派到神经翻译再到问答系统的发展过程,机器翻译具备足够的数据支撑来解答语义问题。数据的强大以至于计算机科学界出现了语义问题的争论已无甚意义的声音,但哲学界并不这么认为。针对语义问题的不同观点显示了人工智能视域下机器翻译语义问题的解答仍然是开放性的,实证的计算机科学为人类展现了人工智能的各种可能性,但科学提供的事实并不能完全解决专属于哲学领域的概念和意义问题。

Translation belongs to Wittgenstein s “language game”. The attempt at“world of building blocks” illustrates that machine translation is also a“l anguage game”, involving philosophical and semantic disputes. With the emergence of artificial intelligence, through the development process from statistics to neural translation to question answering system, machine translation has enough data support to answer semantic questions. Data is so powerful that there is no meaningful debate in computer science about semantics, but not in philosophy. Different views on the semantic problem show that the answer to the semantic problem of machine translation from the perspective of artificial intelligence is still open. Empirical computer science shows various possibilities of artificial intelligence for human beings, but the facts provided by science cannot completely solve the problems of concepts and meanings exclusive to the field of philosophy.