AUTHOR NAME DISAMBIGUATION AND CROSS SOURCE DOCUMENT COREFERENCE

  • : Ms Word, Ms Word Format
  • : 70 Pages
  • : ₦3,000 | $25 | ₵60 | Ksh 2720
  • : 1-5 Chapters
  •  
  • Click to DOWNLOAD Materials

AUTHOR NAME DISAMBIGUATION AND CROSS SOURCE DOCUMENT COREFERENCE

ABSTRACT

This thesis deals with two research problems: author name disambiguation in digital library and cross-source document coreference.

The first problem comes from the digital library, which is an important technological tool to maintain the information used by users. However, due to the problem of ambiguous author names, users can not distinguish the exact authors of the articles in the digital library. This ambiguity mainly comes from two problems: polyseme, an author name shared by multiple authors, and synonym, an author with multiple name variant. Successfully addressing this ambiguous author name problem can improve the search quality of the digital library when one intends to search a specific author, which happens quite frequently in the digital library. In addition, when one attempts to compute statistics such as the reputation of an author based on his publications, disambiguating the author name enhances the accuracy. In this thesis, we present a comprehensive and synthetically summarization of the author name disambiguation algorithms. We also survey the evaluation datasets and metrics used in the papers. In addition, based on the survey, we suggest several possible directions and interesting ideas in the future.

 

For the cross-source document coreference problem, it is a new and important research direction. Cross-source document coreference deals with the problem of disambiguating the entities in documents of one source to their corresponding identities, if exists, in another source. For example, one source, which is called general source, can be World Wide Web and another source can be the Wikipedia, which is called canonical source. The success of cross-source document coreference can benefit many scenarios. We can automatically construct and enrich the entity information in one source according to its information in another source. Search results of entities in one source can also be grouped by their identities in the other sources. Furthermore, we can compute the reputation of one entity discussed in one source from the information of other sources. In this thesis, we utilize one information extraction tools, OpenCalais, and develop a large number of features, 88 features, to help cross-source document coreference. We also make use of a state-of-the-art machine learning algorithm, random forests, to this new area. In the experiment, we compare the random forests model with three traditional models: Decision Tree, Naïve Bayes and Bayes Network. The experiment results demonstrate that random forests model significantly outperforms Naïve Bayes and Bayes Network by 10.67% and 12.65%. Random forests algorithm also outperforms the Decision Tree by 2.88%.

Keywords: author name disambiguation, digital library, cross-source document coreference, random forests

TABLE OF CONTENTS

List of Figures  vii

List of Tables ……………………………………………………………………………………………………………. iix

Acknowledgement      x

1  Introduction 1

1.1  Motivation and Problem Definition         1

1.2  Contribution of the Thesis            2

1.3  Outline of the Thesis        4

2  Author Name Disambiguation in Digital Library   5

2.1  Introduction to Author Name Disambiguation in Digital Library           5

2.2  Author Name Disambiguation Algorithms          7

2.2.1  Algorithms for Polyseme Problem        8

2.2.2  Algorithm for Synonym Problem          13

2.2.3  Algorithms for Both Problems  15

2.2.4  Summary           45

2.3  Evaluation Datasets and Metrics  51

2.3.1  Dataset  51

2.3.2  Evaluation Metrics        57

2.3.3  Summary           66

2.4  Summary on Author Name Disambiguation in Digital Library  67

3  Cross-source Document Coreference Using Random Forests        69

3.1  Introduction to Cross-source Document Coreference      69

3.2  Related Work       72

3.3  Methodology        74

3.4  Experiments          81

3.5  Summary on Cross-source Document Coreference Using Random Forests       84

4  Conclusion and Future Work          85

References      88

 

 

 

List of Figures

FIGURE 1: CROSS-SOURCE DOCUMENT COREFERENCE EXAMPLE. GENERAL SOURCE IS ON THE LEFT AND CANONICAL SOURCE IS ON THE

RIGHT …………………………………………………………………………………………………………………………………………… 2

FIGURE 2: CITATION LABELING ALGORITHM, CITED FROM (LEE ET AL., 2005) ………………………………………………………………… 9

FIGURE 3: BLOCK-BASED ALGORITHM, CITED FROM (LEE ET AL., 2005) …………………………………………………………………….. 13

FIGURE 4: AGGLOMERATIVE CLUSTERING ALGORITHM, CITED FROM (SONG ET AL., 2007) ……………………………………………….. 19

FIGURE 5: HEURISTIC-BASED HIERARCHICAL CLUSTERING ALGORITHM, CITED FROM (COTA ET AL., 2007) ……………………………… 21

FIGURE 6: STEP 2 OF COAUTHOR-BASED ALGORITHM, CITED FROM (KANG ET AL., 2009) ………………………………………………… 24

FIGURE 7: PROCESS OF WEB-BASED ALGORITHM, CITED FROM (PEREIRA ET AL., 2009) …………………………………………………… 25

FIGURE 8: CLUSTERING FUNCTION OF WEB-BASED AUTHOR DISAMBIGUATION, CITED FROM (PEREIRA ET AL., 2009) ………………… 27

FIGURE 9: ONLINE SVM AND DBSCAN CLUSTERING AUTHOR DISAMBIGUATION, CITED FROM (HUANG ET AL., 2006) ……………… 33

FIGURE 10: AN AUTHOR NAME WITH THREE DIFFERENT AFFILIATIONS, CITED FROM (CULOTTA ET AL., 2007) …………………………. 36

FIGURE 11: TRAINING PHASE OF THE ERROR-DRIVEN LEARNING ALGORITHM, CITED FROM (CULOTTA ET AL., 2007)………………….. 37

FIGURE 13: HISTOGRAM OF NAME COUNTS IN MEDLINE, CITED FROM (TORVIK AND SMALHEISER, 2009) …………………………… 53

FIGURE 14: A DBLP WEBPAGE FOR A SPECIFIC AUTHOR ……………………………………………………………………………………….. 54

FIGURE 15: SEARCH “SUPPORT VECTOR MACHINE” IN CITESEERX …………………………………………………………………………….. 55

FIGURE 16: PERCENTAGE/RANK RATIO FOR TWO ALGORITHMS, CITED FROM (LEE ET AL., 2005) ………………………………………… 58

FIGURE 17: EXAMPLE OF THE RESULTS BASED ON ACCURACY FOR TOP-K CANDIDATES, CITED FROM (LEE ET AL., 2005) ………………. 60

FIGURE 18: EXAMPLE OF RESULTS USING “ACCURACY BASED ON CONFUSION MATRIX” (CITED FROM (HAN ET AL., 2005)) ………….. 61

FIGURE 19: EXAMPLE OF THE RESULTS BASED ON F1P AND F1C THE NUMBERS IN X-AXIS INDICATING DIFFERENT AUTHOR NAMES (CITED

FROM (SONG ET AL., 2007)) ……………………………………………………………………………………………………………… 63

FIGURE 20: RELATIONSHIP BETWEEN CROSS DOCUMENT COREFERENCE AND CROSS-SOURCE DOCUMENT COREFERENCE …………….. 73

FIGURE 21: CONSTRUCT A TREE IN RANDOM FORESTS ALGORITHM (TREERATPITUK AND GILES, 2009) …………………………………. 75

 

List of Tables

TABLE 1: PARTIALLY CITATION CLUSTERS OF THREE DISAMBIGUATED AUTHORS OF THE SAME NAME LABEL “J. E. SMITH” , CITED FROM

(HAN ET AL., 2005) ………………………………………………………………………………………………………………………… 16

TABLE 2: SUMMARY OF THE ALGORITHMS ……………………………………………………………………………………………………….. 47

TABLE 3: SUMMARY OF DATASETS USED IN THE ALGORITHMS DISCUSSED IN CHAPTER 2 ……………………………………………………. 51

TABLE 4: EXAMPLE OF METADATA IN THREE MEDLINE PAPERS, CITED FROM (TREERATPITUK AND GILES, 2009) ……………………… 52

TABLE 5: SUMMARY OF MEDLINE, DBLP AND CITESEERX …………………………………………………………………………………… 56

TABLE 6: EXAMPLE OF CONFUSION MATRIX ………………………………………………………………………………………………………. 60

TABLE 7: MATCHING MATRIX, CITED FROM (KANG ET AL., 2009) …………………………………………………………………………….. 61

TABLE 8: SUMMARY OF THE EVALUATION METRICS …………………………………………………………………………………………….. 64

TABLE 9: ENTITIES EXTRACTED FROM OPENCALAIS AND USED IN THIS CHAPTER …………………………………………………………….. 76

TABLE 10: CONFUSION MATRIX FOR DIFFERENT MODELS ………………………………………………………………………………………. 83

TABLE 11: DISAMBIGUATION PRECISION FOR DIFFERENT MODELS …………………………………………………………………………….. 83

 

Acknowledgement

 

This research thesis would not have been possible without the support of many people. I wish to express my gratitude to my supervisor, Dr. Giles who was abundantly helpful and offered invaluable assistance, support and guidance.

 

Deepest gratitude is also due to the members of the thesis committee, Dr. Yen and Dr. Wu for their unselfish and unfailing support as my thesis committees.

 

Special thanks also to all my graduate friends, especially group members: Jian Huang, Treeratpituk Puck, and Wenying Xiong for sharing the literature and invaluable assistance. Not forgetting to my bestfriends who always been there.

 

I would also like to convey thanks to the Ministry and Faculty for providing the financial means and laboratory facilities.

 

I wishes to express my love and gratitude to my beloved families, for their understanding and endless love, through the duration of my studies.

 

Ke Dang

 

1 Introduction

1.1 Motivation and Problem Definition

In the area of information science and technology, which is the interdisciplinary science dealing with information, technology and people, digital library is an important technological tool to maintain the information used by people. For example, the digital library MEDLINE

(Torvik and Smalheiser, 2009) contains 15.3 million records of articles in the 2006 versions. With so much information stored in the digital library, it is very important to ensure that the records are correct. However, due to various problems such as the data-entry errors, ambiguous format and imperfect citation gathering (On et al., 2005), it is challenge to achieve this goal. Ambiguous author name is just one of these challenge problems and it is a very common problem in the digital libraries (Lavender et al., 2008). Author name disambiguation in digital library is the problem of identifying the true authors from the ambiguous author names in the papers recorded in the digital library. It is the first problem discussed in this thesis.

 

Author name disambiguation in digital library deals with the problem of disambiguating entities within one source. The second problem in this thesis is cross-source document coreference. Cross-source document coreference (CSDC) deals with the problem of disambiguate the entities in documents of one source to their corresponding identities, if exists, in another source. In this thesis, we will use the individuals with the same name as the example of ambiguous entities. For example, when we search one person name, “Paul Collins” in the search engine, as the left graph in below, the results retrieved by the search engine is one source. We have another canonical source like Wikipedia. There are 12 different “Paul Collins” in the Wikipedia as in the right graph. Cross-source document coreference algorithms disambiguate the person names in the search results to their matching Wikipedia pages.

 

 

Figure 1: Cross-source document coreference example. General source is on the left and canonical source

is on the right

1.2 Contribution of the Thesis

As discussed above, this thesis mainly deals with two research problems. The first one is author name disambiguation in digital library and the second one is cross-source document coreference.

 

For author name disambiguation in digital library, there are two challenges in this research area: (1) How to develop an algorithm to disambiguate the author names. (2) How to evaluate the performance of the algorithms. For the first challenge, we will synthetically survey the author name disambiguation algorithms. We will analyze them according to the problems they solve: algorithms for polyseme problem, algorithms for synonym problem and algorithms for both problems. The second research challenge comes from the evaluations of the performance of the author name disambiguation algorithms. First, various evaluation metrics are used to measure the performance of author disambiguation algorithms. It is partly because of the different expressions of this research problem. Although all algorithms attempt to address the author name disambiguation problem as discussed above, the forms of the expressions over this problem are different. Second, various datasets are used in the papers. Thus, now the algorithms are evaluated by different evaluation metrics and over different datasets. In this thesis, we will review these evaluation metrics and datasets.

 

Cross-source document coreference is a new research direction. In order to conduct cross-source document coreference, there are two steps: constructing the canonical source and disambiguating entities in the documents of other source to the canonical source. The first step is done by the creators of Wikipedia, DBPedia and etc. This paper focuses on the second step: based upon an existing canonical source, designing algorithm to disambiguate the entities in documents of other sources to their corresponding identities in the canonical source. The contribution of this thesis comes from two perspectives. First, this thesis leverage one information extraction tools, OpenCalais, and developed a large number of features, 88, to help cross-source document coreference. Second, it utilizes a state-of-the-art machine learning algorithm, random forests, to this new research area. The experiment results demonstrate that random forests outperform all the traditional models in the cross-source document coreference problem.

1.3 Outline of the Thesis

This thesis is structured as following: Chapter 2 synthetically surveys the author name disambiguation algorithms and studies the evaluation datasets and metrics for author name disambiguation algorithms. Chapter 3 provides our algorithm for the cross-source document coreference problem. Finally, we summarize this paper and discuss the future work in the chapter 4.

AUTHOR NAME DISAMBIGUATION AND CROSS SOURCE DOCUMENT COREFERENCE

Sharing is caring!

Leave a Reply