本文選題：自然語言處理 + 情感分析�。� 參考：《北京郵電大學(xué)》2016年碩士論文

【摘要】：新聞分析的研究是一個(gè)交叉性的前沿學(xué)科,具有廣泛的應(yīng)用前景。新聞分析的研究會促進(jìn)自然語言處理技術(shù)的發(fā)展,同時(shí)也將在信息檢索、決策支持、文本挖掘等應(yīng)用領(lǐng)域發(fā)揮積極的作用�；ヂ�(lián)網(wǎng)媒體有著大量的新聞稿件,通過該課題的研究,可以很好的了解在專題報(bào)道中新聞稿件專題所持的態(tài)度,民眾關(guān)心該專題的熱度,不同地區(qū)的民眾對該專題的參與程度,以及與該專題所涉及到的不同角度的問題,該專題新聞之間的關(guān)聯(lián)關(guān)系等,都是及時(shí)了解及觀察民眾輿情的一個(gè)重要窗口,為決策部門制定相關(guān)政策提供一定的參考,具有十分重要的實(shí)用價(jià)值與社會效益。本論文的研究將借助于自然語言處理與數(shù)據(jù)挖掘技術(shù),進(jìn)行人物、地點(diǎn)、機(jī)構(gòu)、關(guān)鍵詞的重要內(nèi)容提取和排序;借助于情感分析技術(shù),完成專題情感分析;正負(fù)面新聞分析及專題新聞基于時(shí)間的熱度分析;稿件數(shù)量時(shí)間軸分布分析,并構(gòu)建專題新聞文本集信息可視化的概念模型和展現(xiàn)模型,對關(guān)鍵內(nèi)容進(jìn)行可視化展現(xiàn)。本論文的主要研究工作有以下四點(diǎn):第一、新聞情感分析技術(shù)的研究與實(shí)現(xiàn)。本文對情感分析的研究作用于新聞層面,針對新聞文本的情感分析可以了解媒體以及領(lǐng)域?qū)＜覍δ硞�(gè)事件或國家政策的態(tài)度。方便人民大眾做出自己的判斷。對于新聞情感分析算法的研究,本文融合了自然語言處理、數(shù)據(jù)挖掘等算法。第二、實(shí)體提取算法的研究與實(shí)現(xiàn)。在本論文中,我們主要把CRF算法用在實(shí)體提取領(lǐng)域。通過訓(xùn)練模型自動(dòng)識別出人名、地名和組織機(jī)構(gòu)名三種實(shí)體。通過模板匹配的方式識別時(shí)間和日期實(shí)體。第三、關(guān)鍵詞抽取算法的研究與實(shí)現(xiàn)。我們將LDA算法作用于專題新聞的關(guān)鍵詞抽取技術(shù)。同時(shí)考慮中文的語義表達(dá)提出了組合詞生成算法,主要解決現(xiàn)有分詞系統(tǒng)的局限性。最后,本文的研究依托實(shí)體提取算法、關(guān)鍵詞提取算法和情感分析算法,將多維數(shù)據(jù)展示技術(shù)相結(jié)合,設(shè)計(jì)并實(shí)現(xiàn)了專題新聞分析系統(tǒng)。此系統(tǒng)可以展示某一專題下出現(xiàn)次數(shù)最多的實(shí)體以及和實(shí)體相關(guān)的新聞文章;可以展示某一專題下最能表現(xiàn)主題思想的關(guān)鍵詞;可以將某一專題新聞的情感分析結(jié)果以柱狀圖的形式展現(xiàn)出來;可以展現(xiàn)以時(shí)間為線索的新聞熱點(diǎn)關(guān)注。本論文的研究有以下創(chuàng)新之處:在情感分析算法中,對于新聞標(biāo)題和新聞?wù)牟扇×瞬煌乃惴ā?biāo)題情感分析中加入了中性識別算法。正文情感分析加入了主觀句識別和主體詞識別算法。
[Abstract]:The research of news analysis is an intersecting frontier subject, which has a wide application prospect. The research of news analysis will promote the development of natural language processing technology, and will also play an active role in information retrieval, decision support, text mining and other applications. Internet media have a large number of news articles. Through the research on this topic, we can have a good understanding of the attitude of news articles in special reports, the public's concern about the heat of the topic, and the degree of participation of people from different regions in this topic. As well as the problems from different angles involved in the topic, the relationship between the news of the topic, and so on, it is an important window for timely understanding and observation of public opinion, which provides a certain reference for the policy-making departments to formulate relevant policies. It has very important practical value and social benefit. The research of this paper will use natural language processing and data mining technology to extract and sort the important contents of people, places, institutions and keywords, and complete the thematic emotional analysis with the help of affective analysis technology. The analysis of positive and negative news and the time-based heat analysis of special news, the analysis of time axis distribution of contribution quantity, and the construction of conceptual model and presentation model of information visualization of feature news text set, to visualize the key content. The main research work of this thesis is as follows: first, the research and realization of news emotion analysis technology. In this paper, the research of emotional analysis plays an important role in news. The emotional analysis of news texts can understand the attitude of media and field experts to a certain event or national policy. It is convenient for the masses to make their own judgments. For the research of news emotion analysis algorithm, this paper combines natural language processing, data mining and other algorithms. Second, the research and implementation of entity extraction algorithm. In this paper, we mainly use CRF algorithm in the field of entity extraction. Three kinds of entities are identified automatically by training model, such as name of person, place name and organization name. Identify time and date entities by template matching. Third, the research and implementation of keyword extraction algorithm. We apply LDA algorithm to keyword extraction technology of special news. At the same time, considering the semantic expression of Chinese, a combinatorial word generation algorithm is proposed, which mainly solves the limitations of existing word segmentation systems. Finally, based on entity extraction algorithm, keyword extraction algorithm and emotion analysis algorithm, a special news analysis system is designed and implemented by combining multidimensional data display technology. The system can display the most frequently appeared entities under a particular topic and the news articles related to entities, and the key words that can best express the theme thought under a given topic. The emotional analysis results of a particular news can be shown in the form of a histogram, and the news hot spots with time as a clue can be displayed. The research of this thesis has the following innovations: in the affective analysis algorithm, different algorithms are adopted for news headlines and news text. The neutral recognition algorithm is added to the title emotional analysis. Subjective sentence recognition and subject word recognition are added to the text affective analysis.
【學(xué)位授予單位】：北京郵電大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2016
【分類號】：TP391.1

【參考文獻(xiàn)】

相關(guān)期刊論文 前5條

1 聞彬;何婷婷;羅樂;宋樂;王倩;;基于語義理解的文本情感分類方法研究[J];計(jì)算機(jī)科學(xué);2010年06期

2 葉強(qiáng);張紫瓊;羅振雄;;面向互聯(lián)網(wǎng)評論情感分析的中文主觀性自動(dòng)判別方法研究[J];信息系統(tǒng)學(xué)報(bào);2007年01期

3 鄭家恒,盧嬌麗;關(guān)鍵詞抽取方法的研究[J];計(jì)算機(jī)工程;2005年18期

4 鄭家恒,張輝;基于HMM的中國組織機(jī)構(gòu)名自動(dòng)識別[J];計(jì)算機(jī)應(yīng)用;2002年11期

5 李建華,王曉龍;中文人名自動(dòng)識別的一種有效方法[J];高技術(shù)通訊;2000年02期

相關(guān)博士學(xué)位論文 前2條

1 張瑩;在線新聞評論的情感分析研究[D];南開大學(xué);2013年

2 萬源;基于語義統(tǒng)計(jì)分析的網(wǎng)絡(luò)輿情挖掘技術(shù)研究[D];武漢理工大學(xué);2012年

相關(guān)碩士學(xué)位論文 前2條

1 于墨;基于情感分析的新聞瀏覽平臺關(guān)鍵技術(shù)研究[D];哈爾濱工業(yè)大學(xué);2011年

2 史海峰;基于CRF的中文命名實(shí)體識別研究[D];蘇州大學(xué);2010年

，

本文編號：2069700