本文選題：知識(shí)圖譜　切入點(diǎn)：嵌入模型　出處：《中山大學(xué)》2017年博士論文　論文類型：學(xué)位論文

【摘要】：知識(shí)圖譜是三元組的集合,其中三元組的形式是(主語,謂詞,賓語),主語和賓語是實(shí)體,謂詞是關(guān)系。每個(gè)三元組(例如(奧巴馬,出生地,檀香山))表示一個(gè)事實(shí)。當(dāng)被應(yīng)用于問答系統(tǒng)中時(shí),只有當(dāng)一個(gè)知識(shí)圖譜覆蓋了問答所對(duì)應(yīng)的事實(shí),它才能夠提供所需要的答案。盡管已經(jīng)有多個(gè)大規(guī)模、開放領(lǐng)域的知識(shí)圖譜問世,它們距離完備仍然有很遠(yuǎn)的距離,例如Freebase中有30%的人物實(shí)體缺少記錄他們父母親信息的三元組。知識(shí)圖譜補(bǔ)全就是向一個(gè)已有的知識(shí)圖譜中增加新的三元組,且加入的三元組必須是客觀事實(shí)。主要有兩個(gè)渠道的信息可以用于補(bǔ)全知識(shí)圖譜:1.從一個(gè)知識(shí)圖譜已有的三元組來推理新的三元組。2.從文本中抽取新的實(shí)體和新的三元組。為了利用第一個(gè)渠道的信息,近年涌現(xiàn)了大量知識(shí)圖譜嵌入方面的工作,它們?yōu)槊總�(gè)實(shí)體學(xué)習(xí)一個(gè)稠密的向量表示,同時(shí)基于實(shí)體的向量表示計(jì)算每個(gè)三元組的可信度。這些嵌入模型能被用于推理信息抽取模型從文本中抽取得到的三元組。由于上述兩個(gè)渠道是互補(bǔ)的,所以合并嵌入模型與信息抽取模型能夠表現(xiàn)出較之單一模型更好的性能。我們將現(xiàn)有知識(shí)圖譜嵌入模型存在的弱點(diǎn)以及將其與信息抽取模型合并所存在的挑戰(zhàn)總結(jié)如下:1.業(yè)界領(lǐng)先的知識(shí)圖譜嵌入模型—TransE不能妥善地處理具有自反或者一對(duì)多/多對(duì)一/多對(duì)多性質(zhì)的關(guān)系。2.在訓(xùn)練一個(gè)知識(shí)圖譜嵌入模型時(shí),現(xiàn)有的負(fù)采樣算法有可能產(chǎn)生假陰性樣本。3.對(duì)于從文本中抽取的三元組,其主語和賓語是詞。如果這個(gè)三元組的主語或賓語無法鏈接至所考慮的知識(shí)圖譜中的某個(gè)實(shí)體,現(xiàn)有的嵌入模型因?yàn)槿鄙賹?shí)體的向量表示進(jìn)行計(jì)算,無法對(duì)其進(jìn)行推理。在本文中,我們提出一系列技術(shù)去解決上述問題。本文的主要貢獻(xiàn)包括:1.我們表明了上述首個(gè)問題源自于Trans E將每種關(guān)系建模成對(duì)于實(shí)體向量的平移操作。于是,我們提出一個(gè)新的知識(shí)圖譜嵌入模型Trans H。該模型通過在進(jìn)行平移操作之前首先將實(shí)體向量投影至為每種關(guān)系定義的超平面,解決了Trans E存在的上述弱點(diǎn)。同時(shí),Trans H避免了增加過多模型復(fù)雜度。2.我們提出了一個(gè)數(shù)據(jù)驅(qū)動(dòng)的、每種關(guān)系獨(dú)有的分布,用于采樣負(fù)例來訓(xùn)練知識(shí)圖譜嵌入模型。該分布能減少抽樣到假陰性樣本的機(jī)會(huì)。同時(shí),該分布的參數(shù)可以由每種關(guān)系的基本統(tǒng)計(jì)量確定。3.我們首先表明,在詞嵌入模型—Word2Vec中,詞之間的隱式關(guān)系可以被解釋成對(duì)于詞向量的平移操作,類似于Trans E對(duì)于知識(shí)圖譜中關(guān)系的建模�；诖�,我們提出了一個(gè)聯(lián)合嵌入模型,去為每個(gè)實(shí)體和每個(gè)詞都學(xué)習(xí)一個(gè)稠密的向量表示。我們的聯(lián)合嵌入模型能夠?yàn)橥瑫r(shí)涉及詞和實(shí)體的三元組計(jì)算可信度。據(jù)我們所知,我們的聯(lián)合嵌入模型是能夠處理此類三元組的首個(gè)方法。4.我們提出三個(gè)分別基于實(shí)體鏈接,實(shí)體名稱,實(shí)體描述的對(duì)齊模型。用于訓(xùn)練這些模型的監(jiān)督信息都易于獲取且是規(guī)模大的。經(jīng)驗(yàn)性評(píng)估顯示,這些模型能有效將詞被嵌入的向量空間與實(shí)體被嵌入的向量空間所對(duì)齊。我們做了大量的實(shí)驗(yàn)去比較提出的模型與基準(zhǔn)方法。實(shí)驗(yàn)結(jié)果表明,我們的方法在性能上優(yōu)于業(yè)界領(lǐng)先的方法,而且更為細(xì)致的實(shí)驗(yàn)結(jié)果分析肯定了我們提出模型的動(dòng)機(jī)。
[Abstract]:Knowledge map is a collection of three tuples, which is in the form of three tuple (subject, predicate, object, subject and object) is the entity relationship. Each predicate is three tuples (e.g. (Obama, born in Honolulu, said a fact)). When applied to QA system, only when a the knowledge map covering answers corresponding to the fact that it can provide the required answer. Although there have been a number of large-scale, the advent of knowledge map in open field, they still have a long distance from complete, such as Freebase in figure 30% the lack of solid three tuple record their parents information. Knowledge map is complete to the knowledge map of an existing three increase in new tuples, and adding the three tuple must be objective facts. There are two main channels of information can be used to complement the knowledge map: 1. from the three tuple a knowledge map to an existing The new inference three tuple.2. extraction from text in the new entity and three new tuples. In order to use the first channel information in recent years, the emergence of a large number of knowledge embedded in the work, they are learning for each entity a dense vector, and the vector based on the calculation of each entity credibility. These three tuple the model can be used to embed three tuple reasoning model for information extraction from text extraction. Due to the above two channels are complementary, so with embedded model and information extraction model can show the performance of the single model is better. We will present the knowledge map embedded model and its weakness and the existing information summary the challenges are as follows: extraction model with TransE not knowledge map embedding model - 1. industry-leading deal with reflexive or one to many / many to one / more of Many properties in relation to.2. in training a knowledge map embedding model, the existing negative sampling algorithm may produce false negative samples.3. for three tuple extraction from the text, the subject and the object is the word. If an entity knowledge maps of the three tuple cannot be linked to the subject or object in the considered the existing models of embedded, because lack of vector representation of the entity is calculated to reasoning on it. In this paper, we propose a series of techniques to solve the above problems. The main contributions of this thesis include: 1. we show that the first question comes from Trans E of each entity relationship modeling for vector translation the operation. Then, we propose a new model of Trans H. embedded knowledge map by the model before the first translation operation entity is the vector projection in every kind of relationship defined by hyperplanes, solved The weakness of Trans E. At the same time, Trans H avoided the excessive increase in the complexity of the model.2., we propose a data-driven, unique distribution of every kind of relationship, for example sampling negative training knowledge map embedding model. The distribution can reduce the sampling to false negative samples. At the same time parameters of the machine, the distribution can be determined by the basic statistics of every kind of relationship.3. we first show that, in the word embedding model - Word2Vec, implicit relations between words can be interpreted as the word vector translation operation, similar to the Trans E for modeling the relationship between knowledge map. Based on this, we propose a joint embedding model for each entity, to learn every word and a dense vector. We can model the joint embedding into three tuple words and also relates to the credibility of computational entity. As far as we know, our combined embedded mode .4. is the first method can deal with such three tuples we propose three respectively based on the physical link, entity name, entity description alignment model. To monitor the information in the training of these models are easy to obtain and large scale. Empirical evaluation shows that the model can effectively be word vector space and entity embedding by embedding the vector space alignment. We have done a lot of experiments to compare the proposed model with the reference method. Experimental results show that our method outperforms the method of leading industry, and a more detailed analysis of the experimental results that our proposed model of motivation.

【學(xué)位授予單位】：中山大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2017
【分類號(hào)】：TP391.1

【相似文獻(xiàn)】

相關(guān)博士學(xué)位論文 前2條

1 王楨;基于嵌入模型的知識(shí)圖譜補(bǔ)全[D];中山大學(xué);2017年

2 陳曦;面向大規(guī)模知識(shí)圖譜的彈性語義推理方法研究及應(yīng)用[D];浙江大學(xué);2017年

，

本文編號(hào)：1649765