Meaning And Computing

As a core task in NLP (Natural Language Processing) and Al (Artificial Intelligence), semantic computing (or meaning computing) can be used widely in a lot of fields. For example, text minning, information retrieval, machine translation, human-computer interaction. It takes meaning computing as the understanding of natural language for computer in the computational ism way. On one hand, philosophy of language, semantics, etc. provide some theory of meaning with operability; on the other hand, computer science needs to process the meaning theory formally, so as to manage the meaning computing. Above mentioned, it results in an intricate contradiction, i.e. meaning and computing.In computability theory, all computable functions can carry out their computing procedures via a changing of symbols. It is requirement of the meaning theory uses meaning representation languages to represent the meaning of a natural language, and provides a meaning in formal way as well. Logical semantics (or Truth-conditional semantics) may be a qualified candidate. Because of inexistence of the contradiction, if meaning computing is possible, it will make a range of philosophical presuppositions, which including truth-conditional semantics, the principle of meaning compositionality, and the formal representation of the world (or context). Truth-conditional semantics will automatically generate the meaning of every sentence for a natural language; the principle of meaning principle will explain learnability and understandablity of a language; the formal representation of the world can assign semantic values to natural language expressions. However, the meaning of natural languages is more complicated than truth conditions, and the three presuppositions are prima facie reasonable, or even ad hoc hypothesis.A Sentence’truth-condition is part of its meaning. However, there are so many sentences without truth-conditions and having meanings. The sort of sentence are performative, or illocutionary acts. They don’t have truth-conditions, because they perform acts, not describe facts. It is necessary for computers to understand these performative expressions, provided that they can understand a natural language. Illocutioanry acts are controlled by constitutive rules or regulative rules, and their meanings are associated with speaker’s intentions and contexts very closely, even which depend on these factors. These factors mentioned which can be represented formally are processed by computers, which is supported by an epistemological presupposition that all knowledge would be expressed in the form of propositions, even formal propositions. In fact, this presupposition doesn’t determinate the only meaning of a sentence. That we construct a model which is same as, or very familiar our actual world is impossible, and the key reason is rules, context, or world functions in a holism way, whereas computers need discrete, isolated data.J.R.Searle argued against the possibility of computers understanding a natural language in terms of’Chinese Room’thought experiment. He thought that understanding is not a syntactical calculation procedure, which is guaranteed by intentions. Some philosophers explain intentions in computationism way, considering than selective representations and modeling will simulate intentional states in minds. As the same reason, these two do not form a system’s intention, and organisms’deep background and local cultural practice which we only have all result in intentions in a holism way.In principle, this dissertation will not make a conclusion that meaning computing is impossible in a hurry. We just demonstrate some intricate difficulties in front of meaning computing. Only do we have to say that processing our language understanding not only doesn’t simply extend mircoworld theories, but also adopt ideal experiments in physics.