英语论文网

elated to the object and subjective world and language,it has attractedmany philosophers,logicians as well as linguistics.Many achievements have been madeboth at home and abroad.

2.1.1 Studies on English modality abroad

In regard to the classification of modality,four modes have been clarified by Wright(1951):epistemic modal,alethic modal,existential modal and deontic modal.Wright’sclassification is of great importance to later researches of modality.Palmer(1979)clarifiesEnglish modality into three types:epistemic,dynamic and deontic.Palmer holds the ideathat the degrees of modality can be classified in terms of‘necessity’and‘possibility’,andthere is a feasible fourth type,namely,‘neural’modality.Nonetheless,there is no definitedistinction between neural and other three modalities。

Most scholars reckon that most of the modal verbs are polysemous(Wright,1951;Lyons,1977;Palmer,1979;Leech 1980;Coates,1983;Quirk,1985;Sweetser,1990).Hyland(1998:105)holds the idea that:“The modals cover a wide semantic range and eachone represents an overlap of meanings and stylistic differences”.Leech and Coates(1980)study the semantic indeterminacy of modal verbs from three different angles:merger,ambiguity and gradience.They believe that the meaning of modal verbs is a mixture of polysemy and monosemy and a mixture of categorical and non-categorical contracts.Coates(1983)studies the meaning of modal auxiliaries in modern British English.Sheholds the idea that the meanings of modal verbs are fuzzy.She conducts her study on thebasis of a large corpus and the fuzzy set theory of Zadeh(1972)and takes stylisticvariation into consideration.She believes that not only root but also epistemic meaning isfuzzy.

2.2 Studies on feature selection

Feature selection,which has become a hot research topic in practice andmethodology,has been applied in many fields,for instance,image recognition,faultdiagnosis,data analysis in biology and so on.Feature selection methods can be dividedinto filter,wrapper and embedded.Embedded method and wrapper method are often usedin the feature selection process of continuous decision system,while filter method isusually used in the evaluation function of wrapper method.Feature selection plays a rolein pattern recognition and compressing the processing scale of big data,from which theoptimal features can be saved and the redundant features will be eliminated.

2.2.1 Studies on feature selection abroad

Many scholars have defined feature selection from different angles:Koller(1996) clarifies feature selection from the perspective of distribution as:select as small a featuresubset as possible under the condition that the distribution of the result class is as similaras that of the original data class as possible;The definition given by Dash(1997)isselecting the smallest feature subset that meets the two conditions of not significantlyreducing the classification accuracy and not significantly changing the class distribution.The starting points of the above definitions are different and have their own emphases,butthe goal is to find a minimum feature subset that can effectively identify the target.Basedon the Parzen,Kwak and Choi(2002)propose a new method to calculate the mutualinformation between input variables and class variables.

Chapter 3 Research Design............................12本论文由英语论文网提供整理，提供论文代写，英语论文代写，代写论文，代写英语论文，代写留学生论文，代写英文论文，留学生论文代写相关核心关键词搜索。