Artificial Intelligence The development of language recognition in China

Since 70,000 years ago, humans have been thinking and expressing in ways that have never been seen before, and have developed a unique human phonetic system. Perhaps it is precisely because of the ability of human beings to master language that they can stand out in the competition between species.

The use of language is really the most natural and fastest way for human beings to communicate. At present, we have entered the era of "artificial intelligence", and the Internet of Everything, the intelligence of all things, and human-computer interaction are becoming more and more frequent. Therefore, it is particularly important to teach the machine to "hear" to "understand", and this technology is "speech recognition" technology.

For human beings, speech recognition is like an instinct, and we don't even need to teach it, but we will learn to distinguish the voices of different people and the movements of different creatures by listening to the sounds around us autonomously. However, for machines, this is not an easy task.

First of all, the machine has to input the language, then calculate, then recognize and understand, and then convert it into text or commands, which can be said to be a difficult and complex process. In order for machines to work better with humans, learning to understand human language seems to be the only option.

Only by creating such an auditory system for machines, so that machines can naturally convert language into executed commands, just like humans, can artificial intelligence play a real role.

In 2012, at the "21st Century Computing Conference" held in Tianjin, simultaneous interpretation was completed for the first time by machines, so that machines can understand human language and develop more possibilities, such as translation, whether it is interpretation or translation, these jobs will undoubtedly be replaced by machines.

It can be said that "speech recognition" is an obstacle that must be overcome in the development journey of artificial intelligence, and it is also a difficulty that must be overcome.

Artificial Intelligence Language.

In the process of research and development of artificial intelligence, the problem of artificial intelligence language has been noticed from the very beginning. In the early days of AI development, AI language was researched and developed. In fact, more than 100 AI languages have appeared in more than 40 years, but many have been eliminated.

There are about three of them. The first is the study of the theory of computability by computer scientists. For example, the LISP language is designed to deal with the large number of symbolic programming problems in artificial intelligence, and its theoretical basis is the theory of recursive functions on the set of symbols.

It has been proven that any computable function on a symbol set can be programmed with lisp. The prolog language is designed to deal with logical reasoning problems that also occur in large numbers in artificial intelligence (first and foremost to solve the problem of natural language understanding). Its theoretical basis is the proof of the dissolution of the first-order predicate calculus (first of all, the calculus of its subset horn clause), and its computational power is equivalent to the problem it faces, which is also logical reasoning.

However, prolog is backward reasoning, and ops5 is forward reasoning. The theoretical basis of OPS5 is the generative system of POST, whose computing power is also equivalent to that of Lisp. The second is the research results of cognitive science.

A variety of cognitive models have been developed, and knowledge representation languages have been designed for these models. For example, generative representations, framework representations, semantic network representations, etc., actually have their cognitive models as backgrounds. As mentioned above, OPS5 is a language for generative representation, SRL, FRL, FEST, etc. are framework languages, and concept maps and SNETTI are both semantic web representation languages.

Object-oriented programming is based on the convergence of two ideas in Simula's class program and Minsky's framework representation (it applies to all areas of computer software, not just artificial intelligence).

Due to the characteristics of AI research problems and the particularity of the methods to solve them, in order to build AI systems conveniently and effectively, it is necessary to develop specialized AI languages. What are the characteristics of an AI language, that is, what are the characteristics of an AI language?

In general, an AI language should have the following characteristics:

1.Symbolic processing capabilities (i.e., non-numerical processing capabilities) are required;

2.It is suitable for structured programming and is easy to program; (The ability to break down a system into small units that are easy to understand and process, so that one part of the system can be changed more easily without destroying the whole system.) ）

3.There should be recursive and backtracking functions;

4.Human-computer interaction capabilities are required;

5.suitable for reasoning;

6.It is necessary to have the ability to mix processes with explanatory data structures, as well as a pattern matching mechanism to discern data and determine control.

AI language is a kind of computer programming language adapted to the field of artificial intelligence and knowledge engineering, with symbol processing and logical reasoning capabilities. It can be used to write programs to solve various complex problems with intelligence, such as non-numerical calculation, knowledge processing, reasoning, planning, and decision-making.

One might ask, what is the difference between using AI language to solve problems and traditional methods?

The traditional method usually expresses all the knowledge of the problem in a fixed program in various models, and the solution of the problem is completely guided by the program and executed step by step (article by article) according to the pre-arranged steps. Problem solving ideas with Feng. The Neumannian computer architecture coincides.

At present, the large-scale database method, the mathematical model method, and the statistical method are all strictly structured methods.

For the problems to be solved by AI technology, it is often not possible to embody all the knowledge in a fixed program. It is usually necessary to build a knowledge base (containing facts and reasoning rules), and the program decides its own actions according to the environment and the input information given and the problem to be solved, so it is a reasoning process guided by the environmental model. This approach is flexible, conversational, self-explanatory, and learning.

This method is better than traditional methods for solving unstructured problems with unclear or incomplete conditions and objectives (i.e., not well formalized and difficult to describe), and it usually uses heuristics and heuristics to solve the problem.

The difference between an AI program and a traditional program.

When dealing with some simple problems, there is no difference between the general traditional method and the method used by artificial intelligence. But when it comes to solving complex problems, AI approaches differ from traditional approaches. Artificial Intelligence Approach:

The problem that artificial intelligence wants to solve cannot embody all knowledge in a fixed program. It builds a knowledge base (containing facts and reasoning rules), and the program decides its own actions according to the environment and the input information given and the problem to be solved, so it is a reasoning process guided by the environmental model. This approach is flexible, conversational, self-explanatory, and learning.

This method is better than traditional methods for solving some ill structured problems. Weak structure means that the "x" and "y" are not well defined or complete, that is, they cannot be well formalized and described. "- Use the method of temptation.

AI has not yet evolved to fully solve all of these problems. These kinds of problems are the ones that AI research aims to solve. Subsequently, it is also hoped that the computer hardware structure will also come to a revolution and break through Feng.

Neumann architecture.

All computer programming languages artificial intelligence involves artificial intelligence is a big concept.

Artificial intelligence is mainly involved in these modules:

machine learning,deep learning.

recommendation system.

nature language system.

practical computer vision.

Programming languages, on the other hand, are just tools.