第12行: |
第12行: |
| 情感分析的目的和挑战可以通过一些简单的例子来说明。 | | 情感分析的目的和挑战可以通过一些简单的例子来说明。 |
| | | |
− | = = = Simple cases简单案例 = = | + | === Simple cases简单案例 === |
| | | |
| * Coronet has the best lines of all day cruisers. | | * Coronet has the best lines of all day cruisers. |
第25行: |
第25行: |
| * 我不喜欢旧的游艇。 | | * 我不喜欢旧的游艇。 |
| | | |
− | = = More challenging examples更具挑战性的例子 = = | + | === More challenging examples更具挑战性的例子 === |
| | | |
| * I do not dislike cabin cruisers. ([[Negation]] handling) | | * I do not dislike cabin cruisers. ([[Negation]] handling) |
第350行: |
第350行: |
| 它指的是确定对实体的不同特征或方面表达的意见或感情,例如,手机、数码相机或银行。功能或方面是一个实体的属性或组成部分,例如,手机的屏幕,餐厅的服务,或照相机的图像质量。基于特征的情感分析的优势在于可以捕捉感兴趣对象的细微差别。不同的特征可以产生不同的情绪反应,例如,酒店可以有一个方便的地点,但平庸的食物。这个问题涉及几个子问题,例如,识别相关实体,提取它们的特征/方面,以及确定对每个特征/方面表达的意见是积极的、消极的还是中性的。特征的自动识别可以通过句法方法、主题建模或者深度学习来实现。关于这一层次的情感分析的更详细的讨论可以在刘的作品中找到。 | | 它指的是确定对实体的不同特征或方面表达的意见或感情,例如,手机、数码相机或银行。功能或方面是一个实体的属性或组成部分,例如,手机的屏幕,餐厅的服务,或照相机的图像质量。基于特征的情感分析的优势在于可以捕捉感兴趣对象的细微差别。不同的特征可以产生不同的情绪反应,例如,酒店可以有一个方便的地点,但平庸的食物。这个问题涉及几个子问题,例如,识别相关实体,提取它们的特征/方面,以及确定对每个特征/方面表达的意见是积极的、消极的还是中性的。特征的自动识别可以通过句法方法、主题建模或者深度学习来实现。关于这一层次的情感分析的更详细的讨论可以在刘的作品中找到。 |
| | | |
− | = Methods and features方法和特征 = | + | == Methods and features方法和特征 == |
| | | |
− | Existing approaches to sentiment analysis can be grouped into three main categories: knowledge-based techniques, statistical methods, and hybrid approaches.<ref name ="“Cambria"> | + | Existing approaches to sentiment analysis can be grouped into three main categories: knowledge-based techniques, statistical methods, and hybrid approaches.<ref name="“Cambria"> |
| {{cite journal | | {{cite journal |
| | first1 = E | | | first1 = E |
第388行: |
第388行: |
| |archive-date = 2015-11-23 | | |archive-date = 2015-11-23 |
| }} | | }} |
− | </ref> Some knowledge bases not only list obvious affect words, but also assign arbitrary words a probable "affinity" to particular emotions.<ref name ="Stevenson"> | + | </ref> Some knowledge bases not only list obvious affect words, but also assign arbitrary words a probable "affinity" to particular emotions.<ref name="Stevenson"> |
| {{cite journal | | {{cite journal |
| | first1 = Ryan | | | first1 = Ryan |
第408行: |
第408行: |
| | doi-access = free | | | doi-access = free |
| }} | | }} |
− | </ref> Statistical methods leverage elements from [[machine learning]] such as [[latent semantic analysis]], [[support vector machines]], "[[bag of words]]", "[[Pointwise Mutual Information]]" for Semantic Orientation,<ref name = "Turney02"> | + | </ref> Statistical methods leverage elements from [[machine learning]] such as [[latent semantic analysis]], [[support vector machines]], "[[bag of words]]", "[[Pointwise Mutual Information]]" for Semantic Orientation,<ref name="Turney02"> |
| {{cite conference | | {{cite conference |
| | first = Peter | last = Turney | | | first = Peter | last = Turney |
第452行: |
第452行: |
| }} | | }} |
| </ref> | | </ref> |
| + | |
| | | |
| | | |
第486行: |
第487行: |
| 在情感分析中需要一个人工分析组件,因为自动化系统不能分析个人评论者或平台的历史趋势,而且在他们表达的情感中常常被错误地分类。自动化影响了大约23% 被人类正确分类的评论。然而,人们往往不同意,并认为人际协议提供了一个上限,自动情绪分类器最终可以达到。 | | 在情感分析中需要一个人工分析组件,因为自动化系统不能分析个人评论者或平台的历史趋势,而且在他们表达的情感中常常被错误地分类。自动化影响了大约23% 被人类正确分类的评论。然而,人们往往不同意,并认为人际协议提供了一个上限,自动情绪分类器最终可以达到。 |
| | | |
− | = Evaluation 评估 = | + | == Evaluation 评估 == |
| | | |
| The accuracy of a sentiment analysis system is, in principle, how well it agrees with human judgments. This is usually measured by variant measures based on [[precision and recall]] over the two target categories of negative and positive texts. However, according to research human raters typically only agree about 80%<ref> | | The accuracy of a sentiment analysis system is, in principle, how well it agrees with human judgments. This is usually measured by variant measures based on [[precision and recall]] over the two target categories of negative and positive texts. However, according to research human raters typically only agree about 80%<ref> |
第502行: |
第503行: |
| }} | | }} |
| </ref> | | </ref> |
| + | |
| | | |
| | | |