Biber的多维度分析=数据挖掘&文本聚类分析？
- 主题发起人 守卫猪猪
- 时间
回复: Biber的多维度分析=数据挖掘&文本聚类分析？
MD analysis examines a large number of observable linguistic features for a small number of unobserval, underlying constructs or factors. The factors are called dimensions becasue each represent a continuous scale. Just as, in physical check-ups, different scales measure different aspects of the patients' phyical condition , for example, his blood pressure, pulse, weight, body temperature, etc., the dimesions here measure differenct aspects of language use in texts. Each dimension usually has two complementary sets of linguistic features that co-occur in texts. The presence ( in varying degrees) of the positive loading features means the absence ( in varying degrees) of the negative loading features or the other way around.
Central to MD analysis is the statistical procedure of factor analysis. Ideally, the dimensions extracted should explain a large portion of the total shared variance. However, neither Biber himself nor his followers have been successful in this regard. In fact, few MD studies have yielded dimesions that can account for more than 50% of the variance. For example, Kanoksilapatham (2007)'s 7 dimensions account for only 33.5% of the total varience.
I have problem with MD anaysis chiefly because I don't see how the results can be put to pratical use. I'm not satisfied with those implifications discussed in general terms. It seems to me that many MD studies were carried out just for fun.
MD analysis examines a large number of observable linguistic features for a small number of unobserval, underlying constructs or factors. The factors are called dimensions becasue each represent a continuous scale. Just as, in physical check-ups, different scales measure different aspects of the patients' phyical condition , for example, his blood pressure, pulse, weight, body temperature, etc., the dimesions here measure differenct aspects of language use in texts. Each dimension usually has two complementary sets of linguistic features that co-occur in texts. The presence ( in varying degrees) of the positive loading features means the absence ( in varying degrees) of the negative loading features or the other way around.
Central to MD analysis is the statistical procedure of factor analysis. Ideally, the dimensions extracted should explain a large portion of the total shared variance. However, neither Biber himself nor his followers have been successful in this regard. In fact, few MD studies have yielded dimesions that can account for more than 50% of the variance. For example, Kanoksilapatham (2007)'s 7 dimensions account for only 33.5% of the total varience.
I have problem with MD anaysis chiefly because I don't see how the results can be put to pratical use. I'm not satisfied with those implifications discussed in general terms. It seems to me that many MD studies were carried out just for fun.
回复: Biber的多维度分析=数据挖掘&文本聚类分析？
楼上这位朋友说的有道理，目前我阅读了部分文献，部分的作者仅仅是抱着机械地采用这个多维度方法来分析一个语域或多个语域的变异，得出的结论也只能说是理论层面的，缺少实用的意义。并且在他们的文章中，不同的方法运用让我晕头转向，他们并没有解释为何实用此方法，会对结果造成什么样的影响，这样不知道是否会对自己的文章可靠性产生一些影响呢？
但在我看来，多数作者运用了MD方法对多种语域进行了分析，得出的理论还是有实际的价值的，有的可以用于指导翻译，有的可以用于指导英语教学，这些都可以算是实用意义了吧。
您提到的Biber解释因子分析的结果，也就是语言共现的功能的诠释，我在此方面也不甚理解，这个解释到底是基于什么语言学理论来判断语言共现特征所代表的意义呢？是功能语言学吗？
楼上这位朋友说的有道理，目前我阅读了部分文献，部分的作者仅仅是抱着机械地采用这个多维度方法来分析一个语域或多个语域的变异，得出的结论也只能说是理论层面的，缺少实用的意义。并且在他们的文章中，不同的方法运用让我晕头转向，他们并没有解释为何实用此方法，会对结果造成什么样的影响，这样不知道是否会对自己的文章可靠性产生一些影响呢？
但在我看来，多数作者运用了MD方法对多种语域进行了分析，得出的理论还是有实际的价值的，有的可以用于指导翻译，有的可以用于指导英语教学，这些都可以算是实用意义了吧。
您提到的Biber解释因子分析的结果，也就是语言共现的功能的诠释，我在此方面也不甚理解，这个解释到底是基于什么语言学理论来判断语言共现特征所代表的意义呢？是功能语言学吗？