Scribd
Upload
52 views11 pages

5.Topic-Sensitive PageRank S5

The document discusses topic-sensitive PageRank, an improvement to PageRank that weights pages more heavily based on their topic. It describes how topic-sensitive PageRank can be applied to search engines by choosing topics, computing topic vectors, and using the vectors for relevant topics. The document also discusses classifying documents by topic using characteristic words.

Uploaded by

Houssam Fouki
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
52 views11 pages

5.Topic-Sensitive PageRank S5

The document discusses topic-sensitive PageRank, an improvement to PageRank that weights pages more heavily based on their topic. It describes how topic-sensitive PageRank can be applied to search engines by choosing topics, computing topic vectors, and using the vectors for relevant topics. The document also discusses classifying documents by topic using characteristic words.

Uploaded by

Houssam Fouki
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

BIG DATA ANALYTICS

Topic-Sensitive PageRank
Topic-Sensitive PageRank

• Improvement to PageRank
• Weights certain pages more heavily because of their topic
• Alter the way random surfers behave:
• prefer to land on a page that is known to cover the chosen
topic
• Can also be applied to negate the effects of ”link spam”
Motivation

• Distinct interests may be expressed using the same term:


• E.g., ”jaguar”: the animal, the automobile, a version of the
MAC operating system ...
• If we can deduce that the user is interested in automobiles
=⇒ we can do a better job
• A private PageRank vector for each user?
General Idea

• The topic-sensitive PageRank:


• one vector for each topic
• bias the PageRank to favor pages of that topic
• classifies users according to their interest in each of the topics

• We lose some accuracy...


• But we store only a short vector for each user
Using Topic-Sensitive PageRank (1)

• How to integrate topic-sensitive PageRank into a search


engine?
1. Choose the topics and create specialized PageRank vectors
2. Pick a teleport set for each of these topics
3. Use it to compute the topic-sensitive PageRank vector
4. Determine the set of topics that are most relevant for a
particular search query
5. Use the PageRank vectors for that topic
Using Topic-Sensitive PageRank (2)
Selecting the topic set:
• use the top-level topics of the Open Directory (human
classification)
To determine the set of topics that are most relevant for a
particular search query:

• Allow the user to select a topic from a menu


• Infer the topic(s):
• the words from the recent Web pages or recent queries
• the information about the user:
• bookmarks

• stated interests on Facebook


• ...
Classifying documents by topic: main idea

• Topics are characterized by words that appear


surprisingly often in documents on that topic
• e.g. neither ”fullback” nor ”measles” appear very often in
documents on the Web
• ”Fullback”: pages about sports

• ”Measles”: pages about medicine.


Classifying web pages by topic (1)

• A large, random sample of the Web: the background


frequency of each word
• A large sample of pages known to be about sports:
• Identify the words that appear significantly more frequently in
the sports sample than in the background
• To avoid misspelling: put a floor on the number of times a
word appears
Classifying web pages by topic (2)

• S1 , S2 , ..., Sk : the sets of words that have been determined to


be characteristic of each of the topics
• P : be the set of words that appear in a given page P
• Compute the Jaccard similarity between P and each of the
Si ’s.
• Classify the page: topic with the highest Jaccard similarity
• All Jaccard similarities may be very low
• =⇒ pick reasonably large sets Si to cover all aspects of the
topic
Back to Topic Sensitive PageRank

• Classify the pages the user has most recently retrieved


• Blend the topic-sensitive PageRank vectors
• Same procedure on the bookmarked pages or combine both
References

• J. Leskovec, A. Rajaraman and J. D. Ullman Mining of


Massive Datasets (2014), Chapter 5

• O. Klopp, M. Panov, S. Sigalla and A. Tsybakov. Assigning


Topics to Documents by Successive Projections. (2021)
https://arxiv.org/pdf/2107.03684.pdf

You might also like