The day after tomorrow, the college entrance examination will be held. Are you already nervous? Although I don't take part in the college entrance examination, I still routinely pay attention to a wave of college entrance examination composition questions every year. After all, they are all abused people. I still vaguely remember those years of "loofah vine and meat bean whiskers", "reporting my father's driving and making phone calls", "Duke Qi, Guan Zhong and uncle Bao"
Before the college entrance examination, in addition to teachers and examinees, melon eaters will also actively participate in the major project of "Chinese composition challenge".
Recently, a cat of up master @ Turing of station B got a new job, using AI to predict the 7 most likely propositions for this year's college entrance examination.
In the video, the up master shows the "core word", "prediction topic" and "model representation" of the seven propositions respectively:
Difficulties and hopes
Struggle and self-improvement
Exploration and dream
Innovation and development
Struggle, value
Fairness and moral education
Beauty and environment
How about it? Is it very tasty?
How does this prediction model work? According to the introduction of up, similar to the manual proposition, which needs to extract and screen topics from a large number of hot events and policy documents, AI also needs to mine rules from a large number of data. The more data it obtains, the more reasonable the distribution, and the higher the accuracy of prediction.
In order to obtain as much data as possible, up master specially wrote a set of crawlers to crawl the articles, periodicals and newspapers published by the people's daily, people's daily, CCTV news and other official media in the past year, as well as the important documents and speeches of the high-level meetings.
After that, through the pre training model and topic modeling algorithm in naturallanguageprocessing, this AI system can understand the news and articles in the Chinese context, so as to learn and imitate the idea of writing questions for the college entrance examination in the past few years, and refine the final topic phrases in combination with the most important events in the past year.
However, in recent years, the college entrance examination composition topics have become more and more close to the "big pattern" topics such as life values and feelings of family and country. Therefore, composition topics with strong speculative nature like the (previous) Jiangsu volume and Shanghai volume are not within the prediction set of the up master.
The up master has opened the code of this algorithm on GitHub. The portal:< https://github.com/EssayKillerBrain/EssayTopicPredict >
This video triggered a lot of discussion among netizens of station B. The postgraduate entrance examination party and the public entrance examination party have sought to predict the postgraduate entrance examination English composition questions, postgraduate entrance examination political questions, public entrance examination national examination questions and so on. These questions are indeed similar to today's college entrance examination composition propositions. Of course, many people (such as myself) who have the correct view of money want AI to predict the lottery.
The netizens who love trembling and are smart have also begun to be nervous for the proposition group. However, perhaps the proposition group has already used another set of algorithms to counter predict.
Others are making more specific predictions on the composition of the national paper based on these prediction results.
As for the accuracy of the prediction results, it will not be known until two days later. However, many candidates say that they have indeed encountered these predicted topics in the mock test, and they are within the range predicted by their teachers. But from another point of view, this prediction result seems to be a little "listen to your words, such as listening to your words". These core words and themes are the hot spots of the past year. I can bet without AI It is also common for the proposition group not to play cards according to common sense. Too many key points are drawn, which is equivalent to no key points.
Some netizens doubt whether the prediction results are too fitting. After all, some hot events have a strong sense of existence in the news, but they may not have much relevance with the proposition of Chinese composition.
In addition, we should also consider the legitimacy of the use of AI products that help people write compositions, correct homework, predict test questions and so on. As mentioned by up in the video, once similar AI models appear, they will be used by many people who smell business opportunities to make profits. When you go to the second-hand trading platform, you will find that many people are already selling various "writing artifacts" and "automatic AI article generators", and some of them are completely copied from the AI writing models developed by up, so it is difficult to protect intellectual property rights.
For most of us, it is better to use what is useful than to enjoy what is useless
Finally, we wish the national examinees a successful start and top the list.
Station B video: