Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

26
  • date post

    19-Dec-2015
  • Category

    Documents

  • view

    214
  • download

    1

Transcript of Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

Page 1: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.
Page 2: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

Contents

Introduction

Procedure

The Results

Conclusion4

1

2

3

2

Page 3: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

Contents

Introduction

Procedure

The Results

Conclusion4

1

2

3

3

Page 4: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

4

Background

• The rapid growth of the e.Publications.• Most web search engines provide knowledge from

different areas.– e.g. Google.com

• Some databases are focusing on specialized fields.– e.g. PubMed

• Each search engine is ranking the results depending on pre-defined factors.

Page 5: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

5

The Problem

• The variety of ranking and classifications might hinder the researchers’ work.

• Researchers and doctors spend long time to read all articles to find strong evidence to support their work.

• Beginner researchers’ decisions might be affected by relying on articles have low – quality with high – ranking or low-quality with large number of citations.

• The need to study and evaluate the popular ranking systems to focus on the most important publications.

Page 6: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

6

Project Aim

• Test the rank of 2,075 medical publications in four web ranking systems and a specialized citation index, which are:– Google PageRank.– Yahoo WebRank.– Bing ranking systems.– Alexa ranking systems.– PubMed citation index.

• Compare the results with the Strength of Recommendation Taxonomy ( SORT ) methodology.

Page 7: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

7

The Strength of Recommendation Taxonomy (SORT)

• The SORT classification has been designed to determine the strength of evidences in the medical publications.

• SORT uses three criteria to classify the publications:– The strength of recommendation for the body of evidence.

• A consistent and high quality patient-oriented evidence.• B inconsistent or limited quality patient-oriented evidence. • C opinions, disease-oriented, consensus and usual practice as well

as case studies for treatment, diagnosis, screening or prevention.

– The quality of the individual studies. ( 1, 2 , 3 ).

– The consistency of the publication. ( Consistent or inconsistent )

Page 8: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

Contents

Introduction

Procedure

The Results

Conclusion4

1

2

3

8

Page 9: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

9

Development Steps

Page 10: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

10

Step 1: Generating the PubMed Publications’ Links

• PubMed Publications’ ID have been given.• Automatic links generating.• The link = http://www.ncbi.nlm.nih.gov/pubmed/ + Publication ID

The Given Data The Generated links

Page 11: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

11

Steps 2-5: Testing Web Ranking Systems

• Automatic links testing framework have been developed.

• Pre-developed script has been used.• AJAX technologies have been used to improve the

test speed.

Ranking System Approximate number of links per test

Google PageRank 200

Yahoo WebRank 30

Bing 50

Alexa 50

Page 12: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

12

Step 6: Testing PubMed Database

• PubMed citation index does not provide web service to implement the test.

• The PubMed publication pages have been studied to develop an automated method to fetch the number of citations.

Page 13: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

13

Step 6: Testing PubMed Database

Example of PubMed Page

Page 14: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

14

Step 6: Testing PubMed Database

Example of PubMed Citation Format

• The pages have been copied and analysed automatically.

• The system is able to process around 20 links per test.

• Some articles do not have citations.

Page 15: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

15

Step 6: Testing PubMed Database

Example of PubMed test results

Page 16: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

Contents

Introduction

Procedure

The Results

Conclusion4

1

2

3

16

Page 17: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

17

Google PageRank Results

0 1 2 3 4 50

50

100

150

200

250

300

350

400

450

ABC

Google PageRank Results

14

130

65

229

206

145

372

404

182

17

176

224

32

122

3 20Publication

rank

Number of Publications

Page 18: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

18

Google PageRank Results

• The correlation between the SORT classification and Google PageRank has been calculated using Pearson, Spearman and Kendall Correlation Coefficients as the following: – Pearson Correlation = 0.10793– Spearman Rank Correlation = 0.09971– Kendall Correlation = 0.08813

Page 19: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

19

Yahoo, Bing and Alexa Results

• Yahoo assigned all results to zero.

• Bing assigned all results to zero.

• Alexa assigned all results as same as the root page; which is: http://www.nih.gov/ . e.g. In 2 / November / 2011, the rank of http://www.nih.gov/

was 355.

all publications’ ranks = 355

Page 20: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

20

PubMed Results

0 10 20 30 40 50 60 70 80 90 1000

100

200

300

400

500

600

ABC

Google PageRank Results

432

7 5101417

577

130

44406

116

393

4

83

31

131

Number ofCitations

Number of Publications

6 2371011 2

269

44

8

81

2222

Page 21: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

21

PubMed Results

• The correlation between the SORT classification and PubMed citation index has been calculated among Pearson, Spearman and Kendall Correlation Coefficients as the following: – Pearson Correlation = 0.07157– Spearman Rank Correlation = 0.05207– Kendall Correlation = 0.04267

Page 22: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

22

Results comparison

Correlation Google PageRank PubMed

Pearson 0.10793 0.07157

Spearman 0.09971 0.05207

Kendall 0.08813 0.04267

Page 23: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

Contents

Introduction

Procedure

The Results

Conclusion4

1

2

3

23

Page 24: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

24

Conclusion

• This study attempted to test whether the web ranking systems and citation indexes can help to determine the strength of 2,075 clinical publications according to their SORT classifications.

• Three generic measures have been used to calculate the correlations.

• Neither publication’s rank in the four popular systems nor the number of citations in PubMed database are correlated to the Strength Of Recommendation Taxonomy.

Page 25: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

25

Recommendations

• Studying other citation indexes such as Google Scholar, ScienceDirect, and Scopus.

• Track the positions of the authors who rank the publications and putting the publication date into the consideration

Page 26: Contents Introduction Procedure The Results Conclusion 4 1 2 3 2.

Thank You!