结构化文本和非结构化文本
在数据挖掘方面，结构化文本和非结构化文本有什么区别？选择/开发数据挖掘方法来分析这些不同文本时的主要考虑因素是什么？
With respect to the data mining, what are the differences between structured text and unstructured text? What are the major considerations when choosing/developing data mining approaches for analyzing these different texts?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
我首先要说的是，在回答这些类型的问题时，您所处理的特定领域非常重要。在您的问题中添加一些背景信息将有助于得到更有用的答复。
一般情况下，结构化文本和非结构化文本之间的主要区别在于一个简单的事实：结构化文本具有易于消化的形式，而非结构化文本则不然。对于某些文本挖掘，这可能像词袋模型一样简单（每个单词出现多少次？），一直到极其复杂的 NLP 方法，试图提取更深层次的语言结构，例如词性或实体检测/解析。结构化数据的日常示例可能是 Twitter 上帖子的元数据（用户名/时间戳/转发信息等），其中相关的非结构化数据将是帖子本身的文本。
在不确切知道您感兴趣的内容的情况下，一个重要的考虑因素是一个简单的事实：结构化文本通常是简单机器学习模型的便捷形式，而非结构化文本很少如此，因为它不能轻易被视为一堆二进制/真实数据- 有价值的特征并放入您最喜欢的统计模型中。
希望这对高水平有所帮助——如果我的回答过于宽泛，请随时更新原始帖子的详细信息 =)
I'll preface this by saying that the specific domain you are dealing with matters a great deal when answering these types of questions. Adding some context to your question will allow much more helpful responses.
The central difference between structured and unstructured text, in the general case, is the simple fact that structured text has an easily digested form and unstructured text does not. For some text mining, this may be as simple as a bag-of-words model (how many times does each word occur?), all the way up to extremely complicated NLP approaches that attempt to pull out deeper language structures like parts of speech or entity detection/resolution. An every-day example of structured data could be the metadata of a post on Twitter (username/time stamp/retweet info/etc.) where the related unstructured data would be the text of the post itself.
Without knowing exactly what you are interested in, a large consideration is the simple fact that structured text is often in a convenient form for simple machine learning models, while unstructured text rarely is, since it cannot be easily treated as a bunch of binary/real-valued features and thrown into your favorite statistical model.
Hope this helps on a high level -- feel free to update the original post with details if I'm being too broad with my response =)