Recently, the 6th Natural Language Processing and Chinese Computing Conference (NLPCC 2017) organized by the China Computer Federation (CCF) was successfully held in Dalian.
As the first internationally oriented conference in the field of domestic NLP, NLPCC has shown a vibrant atmosphere in terms of the format of the conference, the number of participants, and the quality of the reports, and is expected to lead China's NLP to the international market.
In order to gain a deeper understanding of the current development status and prospects of NLP in China, CCF's efforts in the direction of NLP, and the development of the NLPCC conference, Lei Feng.com asked Zhou Ming, Director of the CCF Chinese Information Technology Committee and Vice President of Microsoft Research Asia (who is also the
The executive director of the Chinese Information Society of China (CIPS) and the president-elect of the International Association for Computational Linguistics (ACL) and the Secretary-General of the CCF Chinese Information Technology Committee and Professor Zhao Dongyan of Peking University (Leifeng.com will report later) conducted an exclusive interview.
The main content of this article is Dr. Zhou Ming's in-depth introduction to the progress of NLP research and the development status and prospects of NLP in China from the perspective of the CCF Chinese Information Technology Committee.
His opening remarks were as follows: Currently, governments of various countries (including the United States, Germany, Japan, China, etc.) are formulating some plans for artificial intelligence, but China has the clearest plan for artificial intelligence.
Combining the relevant contents of the State Council's "China Artificial Intelligence Development Plan" (July 2017) and the General Secretary's "Report to the Nineteenth National Congress" (October 2017), it can be seen that China plans the development of artificial intelligence into two stages.
, the first stage is to enter the world's advanced level in 2020, and the second stage is to reach the top level in 2030.
Our domestic natural language processing is basically in sync with the country’s plan for artificial intelligence.
In other words, we will enter the world's advanced level by 2020 and look forward to reaching the world's top level by 2030.
What is the big difference between the advanced level and the top level?
The advanced level means that you follow the most developed countries in the world, and you also master all the key technologies, but you are not the proposer of the key technologies, that is, you are not the leader; the top level is actually you leading the way, and you tell the world which direction to go.
Go in the right direction, you put forward the key theoretical model, and others are following you.
That's the only difference.
In the field of NLP, we in China are now very good followers. Once any technology appears in the world (mainly the United States), we will immediately learn to master it and apply it quickly, and our application is no worse than that of the United States.
The only difference now is that we were not the first to propose this technology and method.
Therefore, our CCF Chinese Information Technology Committee believes that we can now be said to be basically at the world's advanced level, and will fully reach the world's advanced level in three years, that is, in 2020.
On this basis, we look forward to reaching the world's top level in 2030.
This is our vision.
The following is Dr. Zhou Ming’s in-depth explanation. Lei Feng.com has simplified and edited it based on the interview content without changing the original meaning for the benefit of readers.
1. NLP is the core of cognitive intelligence Lei Feng.com: What is the position of NLP in the entire AI field?
Zhou Ming: In recent years, artificial intelligence has entered a period of rapid development due to the four major elements of big computing, big data, algorithm models (represented by deep learning) and implementation scenarios.
Its main development direction: perceptual intelligence and cognitive intelligence.
The so-called perceptual intelligence refers to the perceptual abilities of vision (image), hearing (speech), etc.
Everyone knows that perceptual intelligence has advanced by leaps and bounds, such as the ImageNet evaluation of image recognition, the Switchboard evaluation of speech recognition, etc. They have reached or even exceeded the human level in this test set.
Research progress in this area has also promoted the development of many applications, such as security, face recognition, object detection, and the application of speech recognition on mobile phones, smart homes and other devices.
Cognitive intelligence, in layman’s terms, means “the ability to understand and think.”
Cognitive intelligence has many things, and its core includes language intelligence, knowledge graphs, user portraits, etc.
On this basis, it supports several applications, such as intelligent writing, chat dialogue, poetry creation, text generation, gaming, etc.
Some are doing very well, such as the game system represented by AlphaGo; but some are still unsatisfactory.
At present, compared to perceptual intelligence, cognitive intelligence is generally half behind in introducing deep learning, but it is currently catching up.
For example, the quality of neural machine translation is getting better and better, and so are chat systems and human-computer dialogue.
Natural language understanding is at the core of cognitive intelligence.
Its progress will lead to the progress of knowledge graphs, enhance user understanding, and further promote the entire reasoning ability.
On this basis, chatting, problem solving, translation, dialogue, etc. will also be improved.
Once cognitive intelligence advances, coupled with the advancement of perceptual intelligence, the overall artificial intelligence will further develop.
Bill Gates once said, "Language understanding is the crown jewel of artificial intelligence." Dr. Shun Xiangyang also said, "Those who understand language will conquer the world." They all emphasized the importance of NLP.