| Summary: | 碩士 === 靜宜大學 === 資訊傳播工程學系 === 107 === Since 2014, numerous well-known News media organizations used the “Robot Journalist” to report stories. To date, few journalisms have used robotic journalists to generate news, including Associated Press, The New York Times, Los Angeles Times, and so on. “Automated Journalism” is suitable for regular and high repeatability News reports. The procedure of Automated Journalism are: data-collecting, data-analyzing, writing-description, release the stories. The data-collecting and data-analyzing belong to the field of “Web Mining”. Written description and release belong to the field of “Natural Language Generation”. The advantage of automatic journalism is to release news quickly and save more human resources. Moreover, it can have error minimization when they are quoting data. The disadvantages is the high repetitive articles, boring and stiff. Hence, the writing style can not be accepted by most readers.
This paper will take the “Central Weather Bureau, MOTC” - meteorological open data platform for instance. There are many documents on the data platform in XML format which can collect the meteorological data and analyze it quickly. Then, according to the structured data generate weather press release automatically. Based on experiment, we try to investigate whether Automatic Journalism have more vividly description and application for these data, and solve the problem of the immutable article style.
The study used the machine learning approach to forecast future weather. We assume that there is a regression relationship between the weather data. We can apply the historical weather data and supervised-learning to train a weather forecast model. This forecast model can employ the past data to analyze and forecast the data of future weather. Furthermore, researcher examine a way that people can less depend on meteorological open data platform by using the forecast model. The final experimental results are prove that both the root-mean-square error (RMSE) of machine learning and the accuracy of predicted temperature six hours later can achieve good results. The automated news of weather forecasting in the future can save a lot of routine works and make human reporters concentrate on the richness and in-depth analyzing of the news.
|
|---|
