Artificial intelligence Facebook can understand children’s stories published by the Facebook data of Sina Technology News Beijing time on February 19th afternoon news, Facebook today announced for the training of artificial intelligence software data. Artificial intelligence software Facebook can understand children’s stories, and to predict the missing words in sentences. The size of this dataset is more than 1.6GB, which is an annex to a previous published academic paper. Facebook CEO Mark · Zuckerberg wrote on Facebook on Thursday, and introduced Facebook’s research. He said: "for computers, language is the most difficult to understand a kind of information. For humans, guess how to complete a sentence may be very simple, but the machine is not so. In the past, the computer can predict simple words, such as’ on ‘or’ at ‘and’ run ‘and’ eat ‘the verb, but the computer can predict similar’ ball ‘,’ table ‘, and the names of noun." "In this study, our team taught computers to analyze the context of the statement, in order to more accurately predict some difficult words, such as nouns and names. This is usually the most important part of a sentence. By analyzing the appropriate context, the computer can make the most accurate judgments. This context can’t be too much or too much. We will call this "Goldilocks principle"." This data set from the "Gutenberg project" in the book, future researchers will be provided to the academic and research institutions and other companies. This will help optimize the application of the language understanding system. Similar to other large Internet Co, Facebook has previously opened source code for some AI software. In addition, the company also shared the design of artificial intelligence server. Through public data sets, Facebook provides more tools for the advancement of artificial intelligence research. YAHOO recently published 13TB data sets to assist machine learning research. However, these data are only available to academic staff. (Li Li)

Facebook的人工智能已经可以理解儿童故事 Facebook公布的部分数据   新浪科技讯 北京时间2月19日下午消息,Facebook今天公布了用于训练人工智能软件的数据。Facebook的人工智能软件能理解儿童故事,并预测语句中缺漏的单词。   这一数据集的大小超过1.6GB,是此前发表的一篇相关学术论文的附件。Facebook CEO马克·扎克伯格周四在Facebook上撰文,介绍了Facebook的这项研究。   他表示:“对计算机来说,语言是最难理解的一类信息。对人类来说,猜测如何完成一个整句可能很简单,但对机器而言并非如此。以往,计算机可以预测简单的单词,例如‘on’或‘at’,以及‘run’和‘eat’等动词,但计算机无法预测类似‘ball’、‘table’,以及人名等名词。”   “在这项研究中,我们的团队教会计算机去分析语句的上下文,从而更准确地预测某些困难的单词,例如名词和人名。这通常是语句中最重要的部分。通过分析适当的上下文,计算机可以做出最准确的判断。这样的上下文不能太多也不能太少。我们将这称为‘金凤花原理’。”   这一数据集来自“古登堡项目”中的图书,未来将被提供给学术研究机构以及其他公司的研究者。这将有助于优化应用的语言理解系统。   与其他大型互联网公司类似,Facebook此前也曾开源部分人工智能软件的源代码。此外,该公司还分享了人工智能服务器的设计。通过公开数据集,Facebook又提供了更多工具,用于推进人工智能研究。   雅虎近期公布了13TB的数据集,协助机器学习的研究。不过,这些数据只被提供给学术机构人员。(李丽)相关的主题文章:

Similar Posts