Using NLP Approach for Opinion Types Classifier
Using NLP Approach for Opinion Types Classifier
复制标题
使用 NLP 方法进行意见类型分类器
DOI：
10.17706/jcp.11.5.400-410
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
A. Idrees
中科院分区：
文献类型：
--
作者：
Mahmoud Othman;Hesham A. Hassan;R. Moawad;A. Idrees
以文本形式表示的信息要么是事实，要么是观点，每当我们需要做出决定时，我们经常会寻求他人的意见，这是对我们的决定最有影响的因素之一。传统上，个人可以从朋友和家人那里获得意见，而组织则通过调查、焦点小组、民意调查和顾问来获取意见。如今，通过用户生成的内容表达的观点被认为是网络上可用的重要信息类型之一，因此，已经出现了许多用于表达观点的资源，包括社交媒体等。这种情况揭示了健壮、灵活的信息提取（IE）系统的必要性，这些系统能够将网页转换为程序友好的结构，例如关系数据库，以揭示这些观点。在本文中，我们提出了一种对考虑某个对象的文档或一组文档的意见进行分类的方法。该方法已在意见数据集上实施并应用。所提出的系统发现为一个文档或一组文档中的对象提供的意见。该系统会发现不同类型的固执己见的陈述，包括固执己见的、比较性的、最高级的和非固执己见的。该系统已应用于一组 4000 个句子，并使用标准指标评估结果，它们是真阳性、真阴性、假阳性、假阴性、精确度、召回率和 F 分数。我们还提供了所提出的工作与同一领域先前提出的工作的比较。
Information that are represented as text are either facts or opinions, whenever we need to make a decision, we often seek out the opinions of others which is one of the most influencing factors for our decisions. Traditionally, individuals can get opinions from friends and family while organizations use surveys, focus groups, opinion polls and consultants. Nowadays, opinions expressed through user generated content are considered as one of the important types of information which is available on the web, therefore, many resources have been emerged for expressing opinions including social media and others. This situation has revealed the necessity for robust, flexible Information Extraction (IE) systems, these systems have the availability to transform the web pages into program-friendly structures such as a relational database to reveal these opinions. In this paper, we propose an approach to classify the opinions of a document or a set of documents considering an object. The approach has been implemented and applied on a dataset of opinions. The proposed system discover the opinions provided for an object in a document or set of documents. The system discovers different types of opinionated statements, including the opinionated, comparative, superlative, and nonopinionated. The system has been applied on a set of 4000 sentences, and the results has been evaluated using the standard metrics, they are True positive, True negative, False positive, False negative, Precision, Recall, and F-score. We also provided a comparison of the presented work with previous work that has been presented in the same field.