四元组PLSA：法人识别中的合并实体及其评级资源-CSDN文库

Opinion

mining;

40 浏览量 2021-03-18 06:20:11 上传评论收藏 330KB PDF 举报

资源推荐

资源详情

资源评论

Quad-tuple PLSA: Incorporating Entity

and Its Rating in Aspect Identiﬁcation

Wenjuan Luo

1,2

, Fuzhen Zhuang

, Qing He

, and Zhongzhi Shi

The Key Laboratory of Intelligent Information Processing, Institute of Computing

Technology, Chinese Academy of Sciences, Beijing 100190, China

Graduate University of Chinese Academy of Sciences, Beijing 100039, China

{luowj,zhuangfz,heq,shizz}@ics.ict.ac.cn

Abstract. With the opinion explosion on Web, there are growing re-

search interests in opinion mining. In this study we focus on an important

problem in opinion mining — Aspect Identiﬁcation (AI), which aims to

extract aspect terms in entity reviews. Previous PLSA based AI methods

exploit the 2-tuples (e.g. the co-occurrence of head and modiﬁer), where

each latent topic corresponds to an aspect. Here, we notice that each

review is also accompanied by an entity and its overall rating, resulting

in quad-tuples joined with the previously mentioned 2-tuples. Believ-

ing that the quad-tuples contain more co-occurrence information and

thus provide more ability in diﬀerentiating topics, we propose a model

of Quad-tuple PLSA, which incorporates two more items — entity and

its rating, into topic modeling for more accurate aspect identiﬁcation.

The experiments on diﬀerent numbers of hotel and restaurant reviews

show the consistent and signiﬁcant improvements of the proposed model

compared to the 2-tuple PLSA based methods.

Keywords: Quad-tuple PLSA, Aspect Identiﬁcation, Opinion Mining.

1 Introduction

With the Web 2.0 technology encouraging more and more people to participate

in online comments, recent years have witnessed the opinion explosion on Web.

As large scale of user comments accumulate, it challenges both the merchants

and customers to analyze the opinions or make further decisions. As a result,

opinion mining which aims at determining the sentiments of opinions has become

a hot research topic.

Additionally, besides the simple overall evaluation and summary, both cus-

tomers and merchants are becoming increasingly concerned in certain aspects

of the entities. Take a set of restaurant reviews as example. Common restau-

rant aspects include “food”, “service”, “value” and so on. Some guests may be

interested in the “food” aspect, while some may think highly of the “value” or

“service” aspect. To meet these personalized demands, we need to decompose

the opinions into diﬀerent aspects for better understanding or comparison.

On the other hand, it also brings out perplexity for merchants to digest all

the customer reviews in case that they want to know in which aspect they

P.-N. Tan et al. (Eds.): PAKDD 2012, Part I, LNAI 7301, pp. 392–404, 2012.

 Springer-Verlag Berlin Heidelberg 2012

Quad-tuple PLSA: Incorporating Entity and Its Rating 393

lack behind their competitors. As pointed out in [12], the task of aspect-based

summarization consists of two subtasks: the ﬁrst is Aspect Identiﬁcation (AI),

and the second is sentiment classiﬁcation and summarization. The study in this

paper mainly focuses on the ﬁrst task, which aims to accurately identify the

aspect terms in the reviews for certain type of entities.

Hotel: Quality Inn & Suites Downtown Rating:

ƾƾƾƾƾ

Review1: If you are looking for the most elegant hotel, this is not it. If you are looking for

the cheapest, this is not it. If you are looking for the best combination of price, location,

rooms, and staff , the Quality Inn and Suites is a no brainer. A few blocks from the French

Quarter, clean nice rooms, great price, and the staff was awesome. ----by Sabanized

Hotel: L.A. Motel Rating:

ƾƾƾƾ

Review2: Good motel location and good quality! The front desk was helpful, by the way, the

beds could be larger. ----by Jim Porter

Hotel: Hotel Elysee Rating:

Review3: The manager was impatient. Beds were small and dirty. Hot water was not

running and the room was smelly. Anyway, it was cheap. ----by Kate Jeniffer

Fig. 1. Sample Reviews

AsshowninFigure1,thereare3reviews on diﬀerent hotels, where the de-

scription for the same aspect is stained in the same color. One of a recent works

in this area argues that it is more sensible to extract aspects from the phrase

level rather than the sentence level since a single sentence may cover diﬀerent

aspects of an entity (as shown in Figure 1, a sentence may contain diﬀerent col-

ored terms) [5]. Thus, Lu et al. decompose reviews into phrases in the form of

(head, modiﬁer) pairs. A head term usually indicates the aspect while a modiﬁer

term reﬂects the sentiment towards the aspect. Take the phrase “excellent staﬀ”

for example. The head “staﬀ” belongs to the “staﬀ/front desk” aspect, while the

modiﬁer “excellent” shows a positive attitude to it. Utilizing the (head, modiﬁer)

pairs, they explore the latent topics embedded in it with aspect priors. In other

words, they take the these 2-tuples as input, and output the latent topics as the

identiﬁed aspects.

In this study, we observe that besides the (head, modiﬁer) pairs each review

is often tied with an entity and its overall rating. As shown in Figure 1, a hotel

name and an overall rating are given for each review. Thus, we can construct

the quad-tuples of

(head, modiﬁer, rating, entity),

which indicates that a phrase of the head and modiﬁer appears in the review

for this entity with the rating. For example, the reviews in Figure 1 include the

following quad-tuples,

剩余12页未读，继续阅读

评论收藏

内容反馈

weixin_38692707

粉丝: 8
资源: 901

四元组PLSA：法人识别中的合并实体及其评级

四元组 absoluteOrientation

python 实现四数之和

第四章四元组求log2n1

c# 小型语言的词法分析器 输入代码 生成四元组

统计特殊四元组1

基于四元组和面积比的点模式匹配 (2013年)

tyf.rar_四元组_语义分析_语义分析程序

基于C#实现词法分析关键字识别与变色【100012472】

3D_Math_Quaternion

LeetCode 四数之和

天线下倾角计算工具及型号对照表

pythagorean_tuples:to生成勾股勾股三元组，四元组和n元组的直接方法

unity 面试题目以及答案

Fully-Developed-Compiler-based-on-LL1-Parsing-Technique:这是一个完全开发的编译器，可以进行词法分析、LL1 解析、符号表生成和四元组生成

BS模式系统权限控制模型研究

深度强化学习+DQN+Pytorch+CartPole-v0

leetcode2sumc-four-sum:四和

Qt 5实现串口调试助手 （源工程文件、0积分下载）

【SystemVerilog】路科验证V2学习笔记（全600页）.pdf

AutoSAR标准协议4.2.2

光伏-储能并网系统仿真.rar

XCP协议的规范文档

GD32替换STM32注意事项.pdf

NPPJSONViewer.zip

蓝牙BLE协议中文版.pdf

CANoe通过CAPL脚本实现自动测试

AD20官方中文教程.pdf

最新资源

c# 小型语言的词法分析器输入代码生成四元组

Qt 5实现串口调试助手（源工程文件、0积分下载）