Hypertextual Web Search Engine

The major problem that the paper attempts to 

address is the retrieval of high precision 

results from a search engine. To address this issue, 

a new novel algorithm Page Rank is introduced. 

Page Rank algorithm is an attempt by Google to deal 

with the challenges in information retrieval. 

The paper presents various attributes of Page Rank 

in relation to hyperlinks such as anchor text, font 

size etc that help Google produce quality results 

over their competitors. Later in the paper essential 

issues on scalability and search quality in Google 

that have been significantly improved are presented. 

The issue of improved search quality and scalability

 are central issue on which the paper revolves. 

Lastly the paper deals with the idea that Google 

can be used as a research academic search engine.

 

The central issue of harnessing the attributes of

 hyperlinks to improve search quality results 

(Page Rank) is a very novel and a unique method. 

None of the search engines including Yahoo in 

1992-1997 used this methodology like Google did. 

This alone is a major strength of the paper of

 looking at information retrieval in a different

 light. Other strengths of the paper are that the

 authors acknowledge the distinction between the web 

and the controlled collection of documents, 

limitations in computer hardware and their reasons 

for why the web crawler is a challenging task. 

Taking these issues into considerations, the authors 

present their algorithm. The paper has succinctly 

covered the central issues and the critical 

components of the authors' novel search retrieval 

from a very high perspective. This allows the reader 

to appreciate the algorithm without getting bogged 

down into low level details. The authors present 

storage statistics on page 14/20 that shows how 

Google will scale effectively as the size of the 

web grows by using storage efficiently through a 

repository. With these statistics it is clear that 

in 1997 the authors did grasp the notion of the 

explosion of the internet and a strength of the paper.

 

The paper fails to address any short comings of 

the Page Rank algorithm. Potentially spammers 

can create websites with derogatory anchor texts

 all pointing to a victim site. This increases 

the ranking of the victim site in relation to that 

derogatory anchor text. The paper also does not 

present special cases of web pages that have no 

outgoing links. How is page rank calculated for 

such kind of pages? The authors assume a constant 

of .85 damping factor ('d') in their algorithm. 

This is hard to fathom because sites vary in their 

content from plain text to lots of images and 

animated text which allows for a wider distribution

 of the damping factor then just keeping it a 

constant .85. An extension to this idea can be 

incorporating the Page Rank Algorithm with Latent 

Semantic Analysis and whether this can yield 

better results?