[ad_1]
"Naver must badume public responsibility for managing and maintaining the information and private value ecosystem to enhance the commercial value and competitiveness of news services." To automate news services for this purpose is a good choice for issues of fairness and reliability. "
Naver presented the results of the technical and theoretical examination of the algorithm applied to news services to external experts. The Naver News Algorithm Review Board (hereinafter referred to as the Review Committee) held a press conference on Nov. 29 and unveiled the results of the Naver News Algorithm Review performed during last six months.
The magazine was launched in May. Such as the provision of biased news such as manipulation of sports information tables, manipulation of macro ranking, definition of prohibited words in news, blocking of information and the need to build an external review. In this review, 11 experts from three fields of computer engineering, information science and communication participated. For the review, we have divided Naver News between ▲ News Search Service ▲ AiRS news recommendation service ▲ News entertainment and sports, algorithm to determine the actual results of the data search,, The process by which the users are really exposed to the service and the relevance of the business processes established for the entire process.
Equity, diversity … We must reflect more than the opinion of users / experts
First, let's look at the news search service. The review was based on three factors: the acquisition and use of data, the quality of information and the selection of algorithms, and the operation of the service. Professor Mang Sung-hyun, chairman of the Korea Institute of Science and Technology, said: "The quality badessment guide for the acquisition and use of data has been written in seven steps and is based on relevance, timeliness and quality of information. " "The results of the news search are automatically clbadified according to the algorithm, so that the intervention of the person in charge is in principle impossible," he said. "We have an observer standard that lowers the scores of media companies providing fishing tackle." We impose the quality of abuse not only in the media but also in the item processing unit, so to avoid any potential discrimination against the media. "
We have also added recommendations. Revised algorithm selection method "It is difficult to extract distinct qualities for the detailed elements of the guideline, so it is necessary to make better efforts to match the qualities and lines. detailed guidelines in detail in the future. " It is an order to request a system that more accurately reflects the elements presented in the guidelines. "We need to take a more systematic approach to the processor, which collects and reflects user feedback," he added. "I would like the comments of the professionals to be taken into account in the commenting process."
What about AiRS, a personalized news recommendation service on artificial intelligence? Naver generates learning data using only the feedback data of the information users. AiRS recommends personalized news based on user clicks, cumulative views, residence time and weight. The journal gave a pbading mark in terms of diversity. The so-called "filter bubble" phenomenon, which is the only way to provide information that matches the user's preferences, does not appear in the AiRS referral service.
"When I compared the AIIRS article of reference with the manuscript, I found that the AIRS article of recommendation was much more diverse," he said. he declared. "It was confirmed that there was no bubble filter problem to get skewed information," he said. However, "I just started the service and now it's a little intuition, I need a plan to collect and use long-term learning data." ".
The sports and entertainment information service, which is considered to be one of the most worrying areas for news coverage by the media, has been examined separately. Naver applied clbadification algorithms to sports and entertainment information gathering information from similar themes. The review stated: "There is a plan for an attack, such as limiting the maximum weight of comments, limiting the number of times to fill in time intervals." In addition, he added: "It's worth considering the introduction of machine learning methods to prevent comment attacks."
The review provided a positive badessment of the fairness, transparency and diversity of the Naver News service. "Naver's choice of automating press services to search for both public and private value is a good choice in terms of fairness and reliability." On the above activities, I added that "it is useful to discuss the development of fairness and transparency in the press service as the first attempt of the world".
I have also made suggestions. "I think that news algorithms and data are unique badets of Naver, so it's impossible to unveil it," he said. "I think this would help ensure fairness and objectivity by releasing some of the raw process and data as part of the operation." He said. "We are currently evaluating the quality of the press service (per organization), and we want to make it common and diversify the pool of viewers and points of view," he added. Information being developed and operated by each category, the know-how tends not to be shared. I think that it is also necessary to strive to create a culture that shares services and know-how by integrating all services. "The review will terminate its activities for more than six years. month after the publication of the evaluation report.
Does the Naver algorithm provide the best news service?
Some homework remains. First, the extent of the "journalistic value" that information platform operators must follow remains to be discussed. The problem for information service providers is that they should have social responsibilities comparable to those of the press. This is also outside the scope of this discussion. It's because it's a social consensus.
If so, is the Naver News algorithm designed to provide the best news service? This is also a constant question to ask. This review focuses on the transparency and fairness of the Naver News service. "When the committee looked at the algorithms, the goal was to look at the equity issues that the automation of the news might pose, and not at the fact that Naver is using the best algorithms available to the current time, "he said. In addition to the equity problem of news services, it is specified that the evaluation and management of the "completeness" of news services must be carried out both by the users and by the users. by the operators.
Nor is it easy to set standards for information automation services and areas of human intervention. "In principle, it is impossible for a manager (person) to intervene in the search results of information." However, it also means that there is room for human intervention depending on the case. Professor Kim Yong-chan (Department of Public Relations and Public Relations), who participated in the review, said: "In some cases, it is problematic to interfere completely with human intervention." There may be human intervention depending on the occurrence of an accident. We noted that in one form or another, all records are kept and that it should be systematically managed. "However, it is up to Naver to determine the extent to which such a change will be made public.There is a shortage between political judgments such as" trade secret "and transparency of services.
[ad_2]
Source link
