【摘要】：隨著搜索引擎的不斷發(fā)展,搜索引擎的個(gè)性化逐漸成為目前商業(yè)搜索引擎發(fā)展的目標(biāo),以便針對(duì)不同用戶提供不同的搜索服務(wù)。本文主要圍繞個(gè)性化搜索逐步展開(kāi)研究。首先,在深入剖析基于本體的用戶興趣模型的建立和用戶興趣的變化的基礎(chǔ)上,提出了衰減-激勵(lì)的用戶興趣模型更新策略,以多向激活擴(kuò)散的興趣模型更新算法,實(shí)現(xiàn)了用戶興趣隨著時(shí)間不斷變化。然后,通過(guò)分析他人在個(gè)性化排序方面工作的基礎(chǔ)上,提出了基于用戶興趣與初始搜索結(jié)果序列的重新排序算法。所做的具體工作如下:(1)本文給出了用戶興趣模型的形式化定義,針對(duì)于用戶興趣的變化,分別從興趣衰減和興趣增強(qiáng)兩個(gè)方面給予了充分地討論,并提出了用戶興趣的多向激活擴(kuò)散算法。首先概述了用戶興趣模型的相關(guān)應(yīng)用以及興趣建模技術(shù),接下來(lái),結(jié)合本體的相關(guān)知識(shí),給出了基于本體的用戶興趣模型一個(gè)形式化定義,基于本體的用戶興趣模型是指本體建模元語(yǔ)表示的用戶興趣知識(shí)的集合,是一個(gè)關(guān)于某領(lǐng)域的概念類層次結(jié)構(gòu)。相對(duì)于其他表示方法,本體可以更加深入細(xì)致地反映用戶興趣概念的性質(zhì)和關(guān)系,更加準(zhǔn)確地描述用戶的興趣知識(shí)。然后,深入地討論了基于本體的用戶興趣模型的構(gòu)建方法以及用戶興趣和領(lǐng)域本體維護(hù)的相關(guān)技術(shù),最后,根據(jù)人的遺忘特性提出了遺忘-激勵(lì)策略,以多向激活擴(kuò)散的興趣模型更新算法,實(shí)現(xiàn)了用戶興趣隨著時(shí)間的不斷變化。(2)本文提出了基于用戶興趣模型的查詢分析算法。首先概述了用戶的搜索行為以及用戶意圖的相關(guān)知識(shí),總結(jié)分析了擴(kuò)展查詢的相關(guān)技術(shù),比較了各種技術(shù)的優(yōu)缺點(diǎn)。最后,針對(duì)于基于用戶興趣模型的查詢分析算法涉及的相關(guān)技術(shù)給予了詳細(xì)的討論,并給出了算法的描述。(3)本文提出了基于用戶興趣與初始搜索結(jié)果序列的重新排序算法。首先概述了搜索引擎的相關(guān)排序算法,為了實(shí)現(xiàn)用戶感興趣的頁(yè)面盡量排在前面,在計(jì)算用戶興趣與搜索結(jié)果之間語(yǔ)義相似度的基礎(chǔ)上,提出了基于用戶興趣與初始搜索結(jié)果序列的重新排序算法。
[Abstract]:With the continuous development of search engine, the individualization of search engine has gradually become the goal of the development of commercial search engine, in order to provide different search services for different users. This paper mainly focuses on personalized search step by step. Firstly, based on the in-depth analysis of the establishment of ontology-based user interest model and the change of user interest, an attenuated and motivated user interest model updating strategy is proposed, and the multi-direction activation diffusion interest model updating algorithm is used to realize the continuous change of user interest over time. Then, based on the analysis of the work of others in personalized sorting, a reordering algorithm based on user interest and initial search result sequence is proposed. The specific work done is as follows: (1) in this paper, the formal definition of user interest model is given, and the change of user interest is fully discussed from two aspects: interest attenuation and interest enhancement, and a multidirectional activation diffusion algorithm for user interest is proposed. Firstly, the related applications of user interest model and interest modeling technology are summarized. Then, combined with the related knowledge of ontology, a formal definition of ontology-based user interest model is given. Ontology-based user interest model refers to the collection of user interest knowledge represented by ontology modeling meta-language, and is a conceptual class hierarchy about a certain domain. Compared with other representation methods, ontology can reflect the nature and relationship of the concept of user interest more deeply and carefully, and describe the interest knowledge of users more accurately. Then, the construction method of ontology-based user interest model and the related technologies of user interest and domain ontology maintenance are discussed in depth. Finally, according to the forgetting characteristics of human beings, a forgetting-incentive strategy is proposed to update the interest model with multi-direction activation diffusion to realize the continuous change of user interest with time. (2) in this paper, a query analysis algorithm based on user interest model is proposed. Firstly, the related knowledge of user search behavior and user intention is summarized, the related technologies of extended query are summarized and analyzed, and the advantages and disadvantages of each technology are compared. Finally, the related technologies of query analysis algorithm based on user interest model are discussed in detail, and the description of the algorithm is given. (3) in this paper, a reordering algorithm based on user interest and initial search result sequence is proposed. Firstly, the related sorting algorithm of search engine is summarized. In order to realize the page of user interest as far as possible, on the basis of calculating the semantic similarity between user interest and search result, a reordering algorithm based on user interest and initial search result sequence is proposed.
【學(xué)位授予單位】：國(guó)防科學(xué)技術(shù)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2013
【分類號(hào)】：TP391.3
，

本文編號(hào)：2507496