發(fā)布時(shí)間：2018-05-19 21:10

  本文選題：語(yǔ)義標(biāo)注 + 網(wǎng)頁(yè)廣告加載模型　； 參考：《武漢大學(xué)》2010年博士論文

【摘要】：隨著互聯(lián)網(wǎng)的迅速發(fā)展,網(wǎng)頁(yè)廣告也得到了快速的發(fā)展。與傳統(tǒng)媒體廣告相比,網(wǎng)頁(yè)廣告能通過(guò)多媒體方式進(jìn)行全天候、全球性的展示,具有成本低、互動(dòng)性強(qiáng)等特點(diǎn)。網(wǎng)頁(yè)廣告越來(lái)越受到廣告商的青睞,并成為許多學(xué)者研究的對(duì)象。 近幾年,網(wǎng)頁(yè)廣告不論在商業(yè)應(yīng)用還是研究領(lǐng)域都取得了極大的發(fā)展。然而隨著網(wǎng)頁(yè)廣告爆炸式的投放,橫幅廣告點(diǎn)擊率不斷下降,網(wǎng)頁(yè)廣告的發(fā)展前景受到了諸多質(zhì)疑。為了提高網(wǎng)頁(yè)廣告的服務(wù)效率,在特定目標(biāo)網(wǎng)頁(yè)的上下文中尋求當(dāng)前用戶和網(wǎng)頁(yè)廣告之間的“最佳匹配”是當(dāng)前業(yè)界與學(xué)術(shù)界的一個(gè)研究焦點(diǎn)。 本文將“最佳匹配”界定為：一方面目標(biāo)網(wǎng)頁(yè)和網(wǎng)頁(yè)廣告相關(guān),另一方面網(wǎng)頁(yè)廣告在一定程度上符合用戶的興趣。在此界定下,針對(duì)網(wǎng)頁(yè)廣告加載過(guò)程中目標(biāo)網(wǎng)頁(yè)、用戶興趣和網(wǎng)頁(yè)廣告相互間不匹配的現(xiàn)狀,本文從網(wǎng)頁(yè)廣告排序過(guò)程的形式化分析入手,提出了基于語(yǔ)義標(biāo)注的方法來(lái)解決三個(gè)廣告主體間的相關(guān)性匹配問(wèn)題,并對(duì)整個(gè)網(wǎng)頁(yè)廣告加載過(guò)程進(jìn)行了系統(tǒng)建模,進(jìn)而對(duì)其中的關(guān)鍵算法進(jìn)行了系統(tǒng)地分析與研究。研究的關(guān)鍵算法包括基于網(wǎng)頁(yè)內(nèi)容的網(wǎng)頁(yè)廣告排序算法和基于用戶興趣的網(wǎng)頁(yè)廣告重排算法。 論文的主要工作和創(chuàng)新點(diǎn)主要體現(xiàn)在如下幾個(gè)方面。 首先,提出基于語(yǔ)義標(biāo)注的方法解決網(wǎng)頁(yè)廣告加載過(guò)程中的廣告主體間的相關(guān)性匹配問(wèn)題。具體過(guò)程為：首先,對(duì)目標(biāo)網(wǎng)頁(yè)、用戶興趣和網(wǎng)頁(yè)廣告分別進(jìn)行語(yǔ)義標(biāo)注,并提取它們的語(yǔ)義特征；然后,提取目標(biāo)網(wǎng)頁(yè)和網(wǎng)頁(yè)廣告的相關(guān)性特征,按兩者的相關(guān)度進(jìn)行排序,得到首輪網(wǎng)頁(yè)廣告排序結(jié)果；最后,提取用戶興趣和網(wǎng)頁(yè)廣告的相關(guān)性特征,基于兩者的相關(guān)度對(duì)首輪網(wǎng)頁(yè)廣告排序結(jié)果進(jìn)行重排,得到最終的網(wǎng)頁(yè)廣告排序結(jié)果�；谡Z(yǔ)義標(biāo)注的方法將廣告主體間的相關(guān)性匹配問(wèn)題轉(zhuǎn)換成文本語(yǔ)義相關(guān)性排序問(wèn)題,進(jìn)而可以使用成熟的文本處理技術(shù)對(duì)問(wèn)題進(jìn)行分析和求解。 其次,在網(wǎng)頁(yè)廣告排序過(guò)程形式化模型基礎(chǔ)上,提出了基于網(wǎng)頁(yè)內(nèi)容的網(wǎng)頁(yè)廣告排序算法。在此階段,主要是對(duì)目標(biāo)網(wǎng)頁(yè)和網(wǎng)頁(yè)廣告的相關(guān)性匹配問(wèn)題進(jìn)行研究。分兩個(gè)步驟進(jìn)行：首先,提取目標(biāo)網(wǎng)頁(yè)和網(wǎng)頁(yè)廣告的匹配特征。和前人方法不同的是,本文提出的方法除了使用傳統(tǒng)的向量空間模型匹配特征和語(yǔ)義關(guān)聯(lián)匹配特征外,還使用了新引入的統(tǒng)計(jì)匹配特征和潛在主題匹配特征；然后,基于多種匹配特征對(duì)網(wǎng)頁(yè)廣告進(jìn)行排序。本文使用RSVM排序模型對(duì)基于網(wǎng)頁(yè)內(nèi)容的網(wǎng)頁(yè)廣告排序算法進(jìn)行了學(xué)習(xí),該排序模型能有效的融合各種匹配特征,并提高網(wǎng)頁(yè)廣告排序的性能。 再次,提出了基于用戶興趣的網(wǎng)頁(yè)廣告重排算法,將網(wǎng)頁(yè)廣告的個(gè)性化推薦過(guò)程形式化為基于用戶興趣的網(wǎng)頁(yè)廣告重排問(wèn)題。具體步驟為：首先,分析用戶瀏覽行為和用戶興趣的關(guān)系,建立用戶興趣模型,本文使用簇興趣模型表示用戶興趣,使用質(zhì)心模型和高斯模型量化用戶興趣；然后,基于用戶興趣模型,通過(guò)Web日志挖掘?qū)τ脩襞d趣進(jìn)行提��；最后,提取用戶興趣和網(wǎng)頁(yè)廣告的相關(guān)性特征,使用兩者的相關(guān)度對(duì)基于網(wǎng)頁(yè)內(nèi)容的網(wǎng)頁(yè)廣告排序結(jié)果進(jìn)行重排,得到最終的網(wǎng)頁(yè)廣告排序結(jié)果。使用重排技術(shù)可以有效地平衡目標(biāo)網(wǎng)頁(yè)內(nèi)容、用戶興趣和網(wǎng)頁(yè)廣告之間的相關(guān)性匹配,為解決網(wǎng)頁(yè)廣告與其加載的上下文之間的相關(guān)性匹配問(wèn)題找到了切實(shí)可行的解決方法。 最后,依據(jù)本文提出的廣告主體語(yǔ)義標(biāo)注和網(wǎng)頁(yè)廣告排序的一整套方法,本文設(shè)計(jì)實(shí)現(xiàn)了一個(gè)基于語(yǔ)義的網(wǎng)頁(yè)廣告加載原型系統(tǒng)。該系統(tǒng)采用層次化的思路進(jìn)行設(shè)計(jì),并能有效和其他系統(tǒng)進(jìn)行集成。 綜上所述,本文使用基于語(yǔ)義標(biāo)注的方法對(duì)網(wǎng)頁(yè)加載模型相關(guān)問(wèn)題進(jìn)行研究是自然語(yǔ)言處理技術(shù)在計(jì)算廣告學(xué)中的一個(gè)有益嘗試,有助于在深層次上實(shí)現(xiàn)目標(biāo)網(wǎng)頁(yè)、用戶興趣和網(wǎng)頁(yè)廣告的一致性,對(duì)于搜索引擎優(yōu)化及互聯(lián)網(wǎng)信息獲取也有一定的意義。
[Abstract]:With the rapid development of Internet , the advertisement of web page has been developed rapidly . Compared with traditional media advertisement , the advertisement of web page can carry on all - weather , global display through multi - media mode . It has the characteristics of low cost , strong interactivity and so on . Web advertisement is more and more popular with advertisers and becomes the object of many scholars .

In recent years , web - based advertising has been greatly developed in both commercial and research fields . However , with the launch of the advertisement of web page and the decline of banner ad spot rate , the development prospect of web advertisement has been questioned . In order to improve the service efficiency of the web advertisement , " best match " between the current user and the web advertisement in the context of the specific target web page is a research focus of the current industry and the academic circle .

This paper defines " best match " as : on the one hand , the target web page and the web page advertisement are related , on the other hand , the web page advertisement fits the user ' s interest to a certain extent . In this definition , the paper starts with the formal analysis of the web page advertisement ordering process , proposes a method based on the semantic annotation to solve the problem of correlation matching among the three advertisement bodies , and then systematically analyzes and studies the key algorithms in the web page advertisement loading process . The key algorithms of the research include the web page advertisement sorting algorithm based on the web content and the webpage advertisement rearrangement algorithm based on the user interest .

The main work and innovation points of the thesis are mainly embodied in the following aspects .

Firstly , the semantic tagging method is proposed to solve the problem of the matching between the advertisement bodies in the web advertisement loading process . The specific process is as follows : Firstly , the semantic annotation of the target web page , user interest and web page advertisement are carried out respectively , and their semantic features are extracted ;
secondly , extracting relevance characteristics of the target webpage and the webpage advertisement , and sequencing according to the correlation degree of the two , so as to obtain the first round webpage advertisement ranking result ;
finally , the relevance characteristic of the user interest and the webpage advertisement is extracted , the advertisement ranking result of the first round webpage is rearranged based on the correlation degree of the two , so as to obtain the final webpage advertisement ranking result .

Secondly , based on the formal model of web page advertisement ordering process , a web page advertisement ranking algorithm based on web page content is proposed . In this stage , the matching features of target web page and web page advertisement are studied .
Then , the webpage advertisements are sorted based on a plurality of matching characteristics , and the webpage advertisement sorting algorithm based on the webpage content is studied by using RSVM sequencing model , and the sorting model can effectively fuse various matching features and improve the performance of the webpage advertisement ranking .

The method comprises the following steps : firstly , analyzing the relationship between user browsing behavior and user interest , establishing a user interest model , expressing user interest using a cluster interest model , and quantifying user interest using a centroid model and a Gaussian model ;
Then , based on the user interest model , the user interest is extracted through the Web log mining ;
finally , the relevance characteristic of the user interest and the webpage advertisement is extracted , and the webpage advertisement ranking result based on the webpage content is rearranged by using the correlation degree of the two , so as to obtain the final webpage advertisement ranking result .

Finally , according to the whole set of methods of advertising principal semantic tagging and web page advertisement sequencing , this paper designs a semantic - based web advertisement loading prototype system . The system is designed with hierarchical thinking and can be integrated with other systems .

In conclusion , using semantic tagging method to study the problem of web page loading model is a useful attempt in computational advertising , which helps to achieve the consistency of target web page , user interest and web page advertisement in the deep level , and has some meaning for search engine optimization and Internet information acquisition .
【學(xué)位授予單位】：武漢大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2010
【分類號(hào)】：TP393.092

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 唐培麗;解飛;陳志雨;;基于概念檢索的中文搜索引擎研究[J];長(zhǎng)春大學(xué)學(xué)報(bào);2006年04期

2 肖德榮;;廣告的戰(zhàn)略轉(zhuǎn)移與廣告教育的應(yīng)對(duì)[J];長(zhǎng)沙鐵道學(xué)院學(xué)報(bào)(社會(huì)科學(xué)版);2006年02期

3 李信利;呂月娥;;基于概念的論文相似性檢索[J];計(jì)算機(jī)工程與應(yīng)用;2007年21期

4 程豪;黃磊;劉金剛;;基于筆畫(huà)提取和顏色模型的視頻文字分割算法[J];計(jì)算機(jī)工程;2009年04期

5 唐成;;基于概念的搜索引擎[J];科技信息(學(xué)術(shù)研究);2007年09期

6 馮翱,劉斌,盧增祥,路海明,王普,李衍達(dá);Open Bookmark——基于Agent的信息過(guò)濾系統(tǒng)[J];清華大學(xué)學(xué)報(bào)(自然科學(xué)版);2001年03期

7 王建,周源華;一種基于紋理能量的JPEG圖像文本定位算法[J];上海交通大學(xué)學(xué)報(bào);2004年09期

8 朱征宇,裴仰軍,陳華月,付關(guān)友;個(gè)性化服務(wù)中用戶近期興趣視圖的生成[J];計(jì)算機(jī)工程與設(shè)計(jì);2005年04期

9 張選平;蔣宇;袁明軒;馬琮;梁平;;一種基于概念的信息檢索查詢擴(kuò)展[J];微電子學(xué)與計(jì)算機(jī);2006年04期

10 ;Automatic character detection and segmentation in natural scene images[J];Journal of Zhejiang University Science A(Science in Engineer;2007年01期

相關(guān)博士學(xué)位論文 前2條

1 應(yīng)曉敏;面向Internet個(gè)性化服務(wù)的用戶建模技術(shù)研究[D];中國(guó)人民解放軍國(guó)防科學(xué)技術(shù)大學(xué);2003年

2 潘建國(guó);基于語(yǔ)義的用戶建模技術(shù)與應(yīng)用研究[D];上海大學(xué);2009年

，

本文編號(hào)：1911690