Page 153 - JOURNAL OF LIBRARY SCIENCE IN CHINA 2018 Vol. 42

152 Journal of Library Science in China, Vol. 8, 2016

Continued
Rule Example
目前关于文本分类算法的研究很多，概括起来主要分为以下几类：1)基于统计的方法，
如朴素贝叶斯，KNN、类中心向量、支持向量机、最大熵等方法；2)基于连接的方法，
如人工神经网络；3)基于规则的方法，如决策树等
方法+总结|包括|分为+ Currently, there are many studies on text classification algorithms which can be generalized
以下|下面|如下+N+类| into following several types: 1) Statistics-based method, e.g. Naïve-Bayes, KNN, Centroid
种|类型： vector, Support Vector machine, Maximum Entropy. 2) Link based methods, e.g. artificial neural
*** methods can be network. 3) Rule-based method, e.g. decision tree
generalized into *** 关于传统的网页结构相似性度量方法，可以总结为两种，一种是完全基于树结构的计
types. 算，一种是使用统计方法挖掘不同网页之间的共同特征
Methods that measure the similarity of webpage structure can be generalized into two types:One
is the tree-structure based method, and the other uses the statistical method to mine common
features of different webpages
传统的竞争情报分析方法，如SWOT分析方法、定标比超分析方法、关键成功因素分析
方法、核心竞争力分析方法等
Traditional competitive intelligence analysis, e.g. SWOT analysis, benchmarking analysis, key
方法+如M1,M2,M3和 success analysis and core competing capability
（以及）Mn 自从最大频繁项集的概念被提出之后，研究者们提出了许多挖掘MFI的高效算法，如
Method, e.t. MaxMiner、MAFIA、GenMax、Pincer Search、基于Diffset的方法以及HBMFI等
Ever since the proposition of maximum frequency itemset, researchers have proposed many
effective methods for mining MFI, e.g. MaxMiner, MAFIA、GenMax, Pincer Search, Diffset
based method and HBMFI

…统称为|并列为…的 基于案例推理、基于规则推理、基于模型推理并列为“知识推理”的三大推理方法
几+类|大+方法 Case-based reasoning, rule-based reasoning and model-based reasoning are regarded as the three
*** are regarded as ** reasoning methods for “knowledge reasoning”

3.2.2 Dynamic relations between methods
Dynamic relations between methods include improvement, inheritance, evolvement and
substitution. These relations are described using words like “propose” and “improve”, etc.
Innovatively proposing a method means that there is a new problem, new data or new needs and
thus the author proposes a new solution or method. Rules of proposing method rules are simple.
Most of these descriptions use feature words like “propose”, and they also use sentence structures
like “aiming at ... we designed...”. Sometimes there is noise, for example, “we proposed an
improvement on ...”. Here although there is the word “propose”, it is in fact just an improvement.
Borrowed or improved methods are the introduction of methods from other fields to a certain
field. Usually applicability analysis is conducted, and then the borrowed methods are improved
or methods in the current field are improved. Table 3 shows dynamic relation description rules
and examples.

3.3 Method feature description

From a judgement point of view, method feature description can be categorized into advantage
description, disadvantage description and neutral characteristic description. From a description