Web Characteristics CE-324: Modern Information Retrieval Sharif - - PowerPoint PPT Presentation
Web Characteristics CE-324: Modern Information Retrieval Sharif University of Technology M. Soleymani Spring 2020 Most slides have been adapted from: Profs. Manning, Nayak & Raghavan (CS-276, Stanford) Some slides have been adapted from:
Sharif University of Technology
Most slides have been adapted from: Profs. Manning, Nayak & Raghavan (CS-276, Stanford) Some slides have been adapted from: Profs. Leskovec, Rajaraman, and Ullman (Mining
} No design/co-ordination } Distributed content creation, linking,
} Content includes truth, lies, obsolete information,
} Unstructured (text, html, …), semi-structured (XML,
} Scale much larger than previous text collections …
} Growth
} Content can be dynamically generated
2
Web
Results 1 - 10 of about 7,310,000 for miele. (0.12 seconds) Miele, Inc -- Anything else is a compromise At the heart of your home, Appliances by Miele. ... USA. to miele.com. Residential Appliances. Vacuum Cleaners. Dishwashers. Cooking Appliances. Steam Oven. Coffee System ... www.miele.com/ - 20k - Cached - Similar pages Miele Welcome to Miele, the home of the very best appliances and kitchens in the world. www.miele.co.uk/ - 3k - Cached - Similar pages Miele - Deutscher Hersteller von Einbaugeräten, Hausgeräten ... - [ Translate this page ] Das Portal zum Thema Essen & Geniessen online unter www.zu-tisch.de. Miele weltweit ...ein Leben lang. ... Wählen Sie die Miele Vertretung Ihres Landes. www.miele.de/ - 10k - Cached - Similar pages Herzlich willkommen bei Miele Österreich - [ Translate this page ] Herzlich willkommen bei Miele Österreich Wenn Sie nicht automatisch weitergeleitet werden, klicken Sie bitte hier! HAUSHALTSGERÄTE ... www.miele.at/ - 3k - Cached - Similar pages Sponsored Links CG Appliance Express Discount Appliances (650) 756-3931 Same Day Certified Installation www.cgappliance.com San Francisco-Oakland-San Jose, CA Miele Vacuum Cleaners Miele Vacuums- Complete Selection Free Shipping! www.vacuums.com Miele Vacuum Cleaners Miele-Free Air shipping! All models. Helpful advice. www.best-vacuum.comSearch
3
} Can be modeled as a directed graph
4
} Relevance is not enough } Other desirable qualities (non IR!!)
} Content:Trustworthy, diverse, non-duplicated, well maintained } Web readability: display correctly & fast } No annoyances: pop-ups, etc.
} On the web, recall seldom matters
} What matters
} Precision at 1? Precision above the fold? } Comprehensiveness – must be able to deal with obscure queries
¨ Recall matters when the number of matches is very small
5
} Need [Brod02, RL04]
} Informational – want to learn about something (~40% / 65%) } Navigational – want to go to that page (~25% / 15%) } Transactional – want to do something (web-mediated) (~35% / 20%)
} Access a service } Downloads } Shop
} Gray areas
} Find a good hub } Exploratory search “see what’s there”
6
7
} “Tuning” your web page to rank highly in the algorithmic search
} Alternative to paying for placement } Thus, intrinsically a marketing function
8
} Commercial, political, religious, lobbies } Promotion funded by advertising budget
} Contractors (Search Engine Optimizers) for lobbies, companies } Web masters } Hosting services
} E.g.,Web master world ( www.webmasterworld.com )
} Search engine specific tricks } Discussions about academic papers J
9
} The top-ranked pages for the query maui resort were the ones
} e.g., maui resort maui resort maui resort } Often, the repetitions would be in the same color as the
} Repeated terms got indexed by crawlers } But not visible to humans on browsers
10
11
Is this a Search Engine spider? N Y SPAM Real Doc
12
} Pages optimized for a single keyword that re-direct to the
} Mutual admiration societies, hidden links, awards } Domain flooding: numerous domains that point or re-direct to
} Fake query stream – rank checking programs
} “Curve-fit” ranking programs of search engines
} Millions of submissions via Add-Url
13
} Votes from authors (linkage signals) } Votes from users (usage signals)
} Anti robot test
} Ignore
} Use
} Training set based on known spam
} Linguistic
} For images: flesh tone detectors,
} Blacklists } Top queries audited } Complaints addressed } Suspect pattern detection
14
} http://help.yahoo.com/help/us/ysearch/index.html } http://www.google.com/intl/en/webmasters/
15
16
} Not as common
} E.g., last-modified date the only difference between two copies
17
¨ E.g., Similarity > 80% => Docs are “near duplicates” ¨ Not transitive though sometimes used transitively
18
} Segments of a doc (natural or artificial breakpoints) } Shingles (Word N-Grams)
} Jaccard coefficient:
!∩# !∪#
19
20
21
/(!)∪/(#) based on short sketches of Doc A and B
22
} Example: the set of all k shingles.
} 𝐷𝑗𝑘 = 1 indicates presence of shingle 𝑗 in set 𝑘
23
24
} 𝑜::: # of rows where both columns are one (# of the items that
<)
} 𝑜:=: # of rows where 𝐷; contains 1 but 𝐷
< contains 0 (# of the items
<)
} and so on
< =
25
Massive Datasets, http://www.mmds.org
26
27
} ℎB(𝐷𝑗) = index of first row with 1 in column 𝐷; (after the permutation 𝜌
} The signature matrix – whose columns represent the sets and the rows
28
< until we see a 1.
<) .
<)
<
<)
} Look down columns 𝐷;, 𝐷
< until first row that at least one of 𝐷; or 𝐷 <
<.
} If both 𝐷; and 𝐷
< are non-zero in this row (ℎB 𝐷; = ℎB 𝐷 < ), the
<.
} Thus, in each permutation we indeed select a random sample from 𝐷; ∪
< and check that if it exist also in 𝐷; ∩ 𝐷 <.
} Therefore,
<
<)
29
30
} Thinking of signatures as columns of integers, the similarity of
} And the longer the signatures, the smaller will be the expected
31
} Let 𝑔 map all shingles to 0. . . 2𝑛 − 1 (e.g., 𝑔= fingerprinting).
} For i=1 to size of sketch vector
} Let 𝜌𝑗 be a random permutation } Pick 𝑇𝑙𝑓𝑢𝑑ℎP 𝑗
} If the size of sketch vector is 𝑁, we have 𝐾𝑏𝑑𝑑𝑏𝑠𝑒(𝐵, 𝐶) ≈
∑ V(/WXYZ[\ ; ]/WXYZ[^ ; )
_ `ab
c
32
64
64
64
64
33
64
64
64
64
64
64
64
64
A B
34
64
64
64
64
64
64
64
64
B A
35
} Let 𝑡𝑗𝑛[𝑡𝑙𝑓𝑢𝑑ℎd, 𝑡𝑙𝑓𝑢𝑑ℎdh] = fraction of identical elements in
} fraction of permutations where MinHash values agree
} Observe 𝑡𝑗𝑛[𝑡𝑙𝑓𝑢𝑑ℎd, 𝑡𝑙𝑓𝑢𝑑ℎdh] ≈ 𝐾𝑏𝑑𝑑𝑏𝑠𝑒(𝐵, 𝐶)
36
37
38
} Pick, say, 100 hash functions.
} ℎ;(.) gives order of rows for i-th permutation.
39
40
} Hash columns of signature matrix M to many buckets } Each
41
Massive Datasets, http://www.mmds.org
} We expect documents x and y to have the same (Jaccard)
Massive Datasets, http://www.mmds.org 42
Massive Datasets, http://www.mmds.org 43
Massive Datasets, http://www.mmds.org 44
} Make k as large as possible
45
Massive Datasets, http://www.mmds.org
Massive Datasets, http://www.mmds.org 46
47
Massive Datasets, http://www.mmds.org
} Signatures of 100 integers (rows) } Therefore, signatures take 40Mb
Massive Datasets, http://www.mmds.org 48
} Since sim(C1, C2) ³ s, we want C1, C2 to be a candidate pair: We want
} i.e.,
} We would find 99.965% pairs of truly similar documents
49
Massive Datasets, http://www.mmds.org
} Since
} In other words, approximately 4.74% pairs of docs with similarity 0.3%
} They are false positives since we will have to examine them (they are
Massive Datasets, http://www.mmds.org 50
} The number of Min-Hashes (rows of M) } The number of bands b, and } The number of rows r per band
51
Massive Datasets, http://www.mmds.org
2
1
52
Massive Datasets, http://www.mmds.org
Massive Datasets, http://www.mmds.org 53
2
1
} Prob. that all rows in band equal = sr } Prob. that some row in band unequal = 1 - sr
54
Massive Datasets, http://www.mmds.org
r
b
r / 1
55
Massive Datasets, http://www.mmds.org
2
1
Massive Datasets, http://www.mmds.org 56
} 50 hash-functions (r=5, b=10)
Massive Datasets, http://www.mmds.org 57
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
58
Massive Datasets, http://www.mmds.org
} We used hashing to assign each shingle an ID
} We used similarity preserving hashing to generate signatures
} We used hashing to get around generating random permutations
} We used hashing to find candidate pairs of similarity ³ s
Massive Datasets, http://www.mmds.org 59
60