本文選題：Android系統(tǒng) + 自然語言處理��； 參考：《北京郵電大學(xué)》2013年碩士論文

【摘要】：隨著電子商務(wù)的高速飛躍式的發(fā)展,電子營銷這種銷售方式受到了廣泛的關(guān)注。郵件憑借其龐大的用戶量受到了電子營銷業(yè)的追捧。然而在生活節(jié)奏快速的今天,人們雖然可以接受基于郵件的電子營銷方式,但是由于時(shí)間有限很難分出精力去關(guān)注這些郵件打折促銷信息。但是值得關(guān)注的是在信息飛速發(fā)展的今天,各種打折消息往往都會通過郵件傳播。于是為了不讓人們錯(cuò)過這些對自己有用的消息,本課題對于郵件中的打折信息進(jìn)行有效消息的過濾,提取出人們更關(guān)注的消息,從而節(jié)省人們寶貴的時(shí)間。又由于移動互聯(lián)網(wǎng)的飛速發(fā)展,人們利用移動設(shè)備連接網(wǎng)絡(luò)的時(shí)間在飛速的增長,在電腦旁的上網(wǎng)時(shí)間在不斷下降,加之Android系統(tǒng)的51.7%的市場占有率,穩(wěn)穩(wěn)的排在智能操作系統(tǒng)的首位,于是本論文選擇Android平臺作為軟件最終的使用平臺,力圖為人們提供符合人們使用習(xí)慣的基于Android系統(tǒng)的郵件打折信息排行軟件。 本系統(tǒng)通過分析大量的郵件打折信息發(fā)現(xiàn),自然語言處理詞性標(biāo)注之后的打折信息有諸多的特征,可以幫助我們更準(zhǔn)確快速的提取出郵件中的打折信息和商店信息。為了加強(qiáng)自然語言處理中分詞步驟的準(zhǔn)確性我們提出了針對于打折郵件的CRF分詞算法。另外為了加強(qiáng)自然語言處理中詞性標(biāo)注的準(zhǔn)確程度本次研究分析了基于自然語言處理的隱馬爾科夫模型,提出了基于增強(qiáng)隱馬爾科夫模型的完整的自然語言處理詞性標(biāo)注解決方案。 主要包括以下內(nèi)容： 1)分析了電子商務(wù)的快速發(fā)展以及郵件的廣泛普及,介紹了Android平臺的廣泛應(yīng)用前景以及基于Android的通信協(xié)議,分析了Android平臺的主要應(yīng)用軟件。 2)提出了處理廣告郵件的迫切性,分析了SMTP、POP3、IMAP等郵件收發(fā)協(xié)議,找到了可以屏蔽郵件的不同協(xié)議類型,用統(tǒng)一的入口去獲取郵件內(nèi)容的接口。 3)分析了自然語言處理技術(shù)的廣泛應(yīng)用范圍,介紹了自然語言處理領(lǐng)域的現(xiàn)存問題以及自然語言處理的三種模型——隱馬爾科夫模型和最大熵模型以及條件隨機(jī)場模型。 4)針對于要處理的郵件信息的特點(diǎn)提出了基于條件隨機(jī)場的郵件分詞算法并驗(yàn)證了其正確性。這部分解決了郵件中的交集型及組合型切分歧義問題。 5)提出了改進(jìn)隱馬爾科夫模型的完整詞性標(biāo)注方案。對于增強(qiáng)型隱馬爾科夫模型做了驗(yàn)證,實(shí)驗(yàn)證明增強(qiáng)型隱馬爾科夫模型在詞性標(biāo)注上的顯著優(yōu)勢。 6)實(shí)現(xiàn)了基于自然語言處理技術(shù)的Android廣告郵件排行軟件,介紹了本軟件的類構(gòu)成,軟件設(shè)計(jì)流程,軟件模塊構(gòu)成及操作方式。
[Abstract]:With the rapid development of e-commerce, e-marketing has received wide attention. Email, with its huge number of users, has been sought after by the e-marketing industry. However, in today's fast-paced life, although people can accept email based on electronic marketing, but due to time constraints, it is difficult to focus on these discounted email promotional information. But it is worth noting that in the rapid development of information today, all kinds of discounted messages are often spread through email. Therefore, in order not to let people miss these useful messages, this paper filters the discount information in email, extracts the information that people pay more attention to, and saves people valuable time. Because of the rapid development of the mobile Internet, the time people use mobile devices to connect to the network is growing rapidly, and the time of using the Internet next to the computer is declining, and the market share of Android system is 51.7%. This paper chooses Android platform as the final use platform of software, and tries to provide people with email discount information ranking software based on Android system, which accords with people's usage habits. By analyzing a large number of discounted email information, the system finds that the discounted information after dealing with part of speech labels has many features, which can help us to extract discount information and store information more accurately and quickly. In order to enhance the accuracy of word segmentation in natural language processing, we propose a CRF segmentation algorithm for discounted mail. In addition, in order to enhance the accuracy of part of speech tagging in natural language processing, this study analyzes the hidden Markov model based on natural language processing. A complete natural language processing solution for part of speech tagging based on enhanced Hidden Markov Model is proposed. It mainly includes the following: 1) the rapid development of electronic commerce and the popularization of mail are analyzed. The application prospect of Android platform and the communication protocol based on Android are introduced. The main application software of Android platform is analyzed. 2) this paper puts forward the urgency of dealing with the advertisement mail, analyzes the email sending and receiving protocols such as SMTP / POP3 / IMAP, finds out the different types of protocols which can shield the mail, and uses the unified entry to obtain the mail content interface. 3) the extensive application of natural language processing technology is analyzed. The existing problems in the field of natural language processing and the three models of natural language processing, hidden Markov model, maximum entropy model and conditional random field model, are introduced. 4) according to the characteristics of the message information to be processed, the algorithm of message segmentation based on conditional random field is proposed and its correctness is verified. This part solves the problem of intersection and combination segmentation ambiguity in mail. 5) an improved hidden Markov model is proposed to complete part of speech tagging. The enhanced Hidden Markov Model is verified and the experimental results show that the enhanced Hidden Markov Model has a significant advantage in part of speech tagging. 6) the Android advertisement mail ranking software based on natural language processing technology is implemented. The class structure, software design flow, software module structure and operation mode of this software are introduced.
【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】：2013
【分類號】：TP316;TP311.52

【參考文獻(xiàn)】

相關(guān)期刊論文 前5條

1 趙巖;王曉龍;劉秉權(quán);關(guān)毅;;融合聚類觸發(fā)對特征的最大熵詞性標(biāo)注模型[J];計(jì)算機(jī)研究與發(fā)展;2006年02期

2 張曉艷;王挺;陳火旺;;命名實(shí)體識別研究[J];計(jì)算機(jī)科學(xué);2005年04期

3 趙海;揭春雨;;基于有效子串標(biāo)注的中文分詞[J];中文信息學(xué)報(bào);2007年05期

4 宋彥;蔡東風(fēng);張桂平;趙海;;一種基于字詞聯(lián)合解碼的中文分詞方法[J];軟件學(xué)報(bào);2009年09期

5 于江德;谷川;葛文英;樊孝忠;;一種基于字和子串聯(lián)合標(biāo)注的漢語分詞方法[J];山西大學(xué)學(xué)報(bào)(自然科學(xué)版);2011年03期

，

本文編號：1840268