東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Text Mining for Sentiment Analysis.

Zhang, Zhe.

FindBook

Google Book

Amazon

博客來

Text Mining for Sentiment Analysis.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Text Mining for Sentiment Analysis./
作者:	Zhang, Zhe.
面頁冊數:	111 p.
附註:	Source: Dissertation Abstracts International, Volume: 76-07(E), Section: B.
Contained By:	Dissertation Abstracts International76-07B(E).
標題:	Computer science. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3690404
ISBN:	9781321584585

Text Mining for Sentiment Analysis.
Zhang, Zhe.

Text Mining for Sentiment Analysis. - 111 p.

Source: Dissertation Abstracts International, Volume: 76-07(E), Section: B.

Thesis (Ph.D.)--North Carolina State University, 2014.

Over the past few years, with the development of web services and the emergence of userdriven social media, more and more people express their sentiments publicly, generating a large amount of opinionated data. Sentiment analysis is such a research field that aims at developing automated approaches to accurately extract sentiments from opinionated data. Researchers have devoted considerable attention to this field. However, due to the complexity and diversity of linguistic expressions, we are still far from a satisfying solution.

ISBN: 9781321584585Subjects--Topical Terms:

523869
Computer science.

Text Mining for Sentiment Analysis.
LDR:03774nmm a2200313 4500 001 2065821
005 20151212141526.5
008 170521s2014 ||||||||||||||||| ||eng d
020 $a 9781321584585
035 $a (MiAaPQ)AAI3690404
035 $a AAI3690404
040 $a MiAaPQ $c MiAaPQ
100 1 $a Zhang, Zhe. $3 1684351
245 1 0 $a Text Mining for Sentiment Analysis.
300 $a 111 p.
500 $a Source: Dissertation Abstracts International, Volume: 76-07(E), Section: B.
500 $a Adviser: Munindar P. Singh.
502 $a Thesis (Ph.D.)--North Carolina State University, 2014.
520 $a Over the past few years, with the development of web services and the emergence of userdriven social media, more and more people express their sentiments publicly, generating a large amount of opinionated data. Sentiment analysis is such a research field that aims at developing automated approaches to accurately extract sentiments from opinionated data. Researchers have devoted considerable attention to this field. However, due to the complexity and diversity of linguistic expressions, we are still far from a satisfying solution.
520 $a In this dissertation, we have identified four challenges that may hinder current research progress: basic sentiment expressing unit, paucity of labeled data, domain dependence, and author modeling. Accordingly, we propose two approaches, ReNew and Arch, to address these challenges.
520 $a ReNew, a semi-supervised sentiment analysis framework, can leverage unlabeled opinionated data to automatically generate a domain-specific sentiment lexicon and trains a sentiment classifier. The domain-specific sentiment lexicon uses dependency relation pairs as its basic elements to capture the contextual sentiment of words. The sentiment classifier leverages relationships between consecutive sentences, clauses, and phrases to infer sentiments. We evaluate the effectiveness of ReNew using a hotel review dataset. Empirical results show that ReNew greatly reduces the human effort for building a domain-specific sentiment lexicon with high quality. Specifically, in our evaluation, working with just 20 manually labeled reviews, it generates a domain-specific sentiment lexicon that yields weighted average F-Measure gains of 3%. The sentiment classifier achieves approximately 1% greater accuracy than a state-of-the-art approach based on elementary discourse units.
520 $a Arch, a probabilistic model for unsupervised sentiment analysis, can discover sentimentaspect pairs from unlabeled opinionated data. By incorporating authors explicitly as a factor, Arch can capture the association of sentiments and aspects with authors. The generated interpretable author profiles can be used for (1) summarizing authors' preferences in terms of sentiments and aspects and (2) measuring similarities among authors. To assess the generalizability of Arch, we use four datasets in two domains for evaluation. Results show that Arch successfully discovers sentiment-aspect pairs with higher semantic coherence than those generated by state-of-the-art approaches. The author profiles are well correlated with ground truth. To exhibit the prospects for potential applications, we demonstrate the effectiveness of Arch for authorship attribution and document-level sentiment clasifification.
520 $a In both ReNew and Arch, we use segments as basic sentiment expressing units to capture fine-grained sentiments.We also present a rule-based segmentation algorithm based on discourse relations.
590 $a School code: 0155.
650 4 $a Computer science. $3 523869
690 $a 0984
710 2 $a North Carolina State University. $b Computer Science. $3 2099755
773 0 $t Dissertation Abstracts International $g 76-07B(E).
790 $a 0155
791 $a Ph.D.
792 $a 2014
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3690404