Shulou(Shulou.com)11/24 Report--

For a special administrative region government, how to achieve effective exchange of information and services between simplified Chinese, traditional Chinese and English through government websites and citizens with different needs, so as to improve the experience and efficiency of government services and reduce the cost of government services? This is a very realistic but extremely troublesome practical problem-the emergence of the general model of Xiao I Huazang, which solves this problem very well.

At the end of October, Xiao I Robot released the Huazang large model ecosystem, officially launched the achievement landing and commercial realization of Xiao I Huazang general large model. The senior director of Xiao I Robot shared an interesting commercial landing story of the big model with the media at the press conference: based on the base of Xiao I Huazang General Model, Xiao I Robot completed an important case of intelligent government services with the government of a special administrative region. It has helped the SAR government save 70% of its labor costs in the extraction, provision and maintenance of real-time information on government affairs-how is this done?

It is reported that in the above-mentioned case of commercial co-creation of the Huazang model, the SAR government has the following typical needs: to publish all kinds of government information and other useful information to relevant government departments' websites in a timely and accurate manner, and then need to extract, classify and extract the stock information and data on these government websites and external websites, so as to provide precise service support to citizens in need.

Before addressing the above needs, there is a thorny problem that needs to be solved: because the citizen structure of the Special Administrative region is highly internationalized, it is necessary to seamlessly switch between traditional Chinese, simplified Chinese and English in order to meet the different service needs of different citizens, but before that: there is no other large model that can solve traditional Chinese problems very smoothly-- large models including ChatGPT. It can better solve the contextual needs of English and simplified Chinese, but the effect is not good for traditional Chinese.

According to Wen Chen Cheng's conclusion, before the intervention of the general model of China and Tibet, the SAR Government had three obvious pain points in the above needs:

The SAR government does not update and maintain the information of the government website in a timely manner, and cannot carry out the update and maintenance, because the content of the website is too much, and after the policy has been released, it has to rely on manpower to complete the background update and replace it, and the immediacy is not high enough.

The manual workload to complete the information extraction and matching is too large, because the content information on the website is more complex, involving all the departments under the government of the special administrative region. If you have to rely on manpower, it will take a long time to find these contents alone.

When citizens log on to the website for consultation, they often cannot find the entrance, and find the accurate information they need in the complicated website information, such as looking for a needle in a haystack, there is no entrance, and there is no unified entrance so that people can find all the information.

The language style needs to cover Chinese simple, complex and English, not only the same piece of information needs to be maintained three times on the website, but also needs to meet the needs of different styles when providing information services to the public.

In the aspect of cross-stylistic ability, after the intervention of Xiao I Hua Zang general model, using the industry first-class NLP (natural language processing) ability and strong deep learning ability accumulated by Xiao I robot for more than 20 years, we have quickly and efficiently completed the learning and training of traditional Chinese, and achieved very good results.

At the same time, in terms of data extraction, analysis, extraction and intelligent matching of service information, according to Chen Cheng, the small I Huazang big model grabs all the information on the website through the big model technology. After obtaining the data, then organize the content into a unified private library, as long as you connect to this private library, as long as one entry can answer all the users' questions.

"in the past, when citizens entered various websites, they needed to query information through different entry points, but now they do not need it. Now they only need one entry, because we have got all the information into the private library, which is equivalent to classifying its fragmented data into the private library. This is the formation of the private library. On the other hand, we also obtain the external information outside the SAR government in real time through the large model to help them do the real-time update of the government website and the real-time update of the data, which is the service efficiency, which greatly improves the efficiency of their government services. We calculated at that time that they used to do it manually, but now through the robot customized based on the Huazang model, the efficiency has been increased by 70%, that is, 70% of the manpower has been saved. " Wen Chen Cheng said.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.