Information Retrieval & Data Mining Winter Semester 2015/16 Saarland University, Saarbrücken https://www.mpi-inf.mpg.de/de/departments/databases-and-information-systems/ teaching /winter-semester-201516/information-retrieval-and-data-mining/ Jilles Vreeken Gerhard Weikum [email protected][email protected]Teaching Asssistants: Abdalghani Joanna Robin Sreyasi Mohamed Adam Abujabal Biega Burghartz Chowdury Gad-Elrab Grycner Dhruv Yusra Saskia Panagiotis Natalia Erdal Amy Gupta Ibrahim Metzler Mandros Prytkova Kuzey Siu Coordinators IRDM 2015 1-1
50
Embed
Information Retrieval & Data Miningresources.mpi-inf.mpg.de/departments/d5/teaching/ws15_16/irdm/... · Information Retrieval & Data Mining ... African singers who covered Dylan songs?
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
build and analyzeWeb graph,index all tokensor word stems
server farm with 100 000‘s of computers,distributed/replicated data in high-performance file system,massive parallelism for query processing
fast top-k queries,query logging,auto-completion
scoring functionover many dataand context criteria
GUI, user guidance,personalization
IRDM 2015 1-9
Content Gathering and Indexing
Documents
Internet crisis:users still love
search engines
and have trust
in the Internet
Internet
crisis
users
...
Extractionof relevantwords
Internet
crisis
user
...
Linguisticmethods:stemming
Internet
Web
crisis
user
love
search
engine
trust
faith...
Statisticallyweightedfeatures(terms)
Index
(B+-tree)
crisis love ...URLs
Indexing
Thesaurus(Ontology)
Synonyms,Sub-/Super-Concepts
......
.....
......
.....
Crawling
Bag-of-Words representations
IRDM 2015 1-10
Ranking bydescendingrelevance
Vector Space Model for Content Relevance Ranking
Search engine
Query(set of weightedfeatures)
||]1,0[ Fid Documents are feature vectors
(bags of words)
||]1,0[ Fq
||
1
2||
1
2
||
1:),(
F
j
j
F
j
ij
F
j
jij
i
qd
qd
qdsim
Similarity metric:
IRDM 2015 1-11
Vector Space Model for Content Relevance Ranking
Search engine
Query(Set of weightedfeatures)
||]1,0[ Fid Documents are feature vectors
(bags of words)
||]1,0[ Fq
||
1
2||
1
2
||
1:),(
F
j
j
F
j
ij
F
j
jij
i
qd
qd
qdsim
Similarity metric:Ranking bydescendingrelevance
e.g., using: k ikijij wwd
2/:
jikk
ij
ijfwithdocs
docs
dffreq
dffreqw
#
#log
),(max
),(1log:
tf*idf
formula
IRDM 2015 1-12
Link Analysis for Authority Ranking
Search engine
Query(Set of weighted features)
||]1,0[ Fq
Ranking by descendingrelevance & authority
+ Consider in-degree and out-degree of Web nodes:
Authority Rank (di) :=
Stationary visit probability [di]
in random walk on the Web
Reconciliation of relevance and authority (and …) by weighted sum
IRDM 2015 1-13
Google‘s PageRank [Brin & Page 1998]
random walk: uniformly random choice of links + random jumps
PR( q ) j(q ) (1 ) p IN ( q )
PR( p ) t( p,q )
Authority (page q) = stationary prob. of visiting q
Idea: links are endorsements & increase page authority,
authority higher if links come from high-authority pages
with
Nqj /1)(
p)outdegree(qpt /1),(
and
Social Ranking
Extensions with
• weighted links and jumps
• trust/spam scores
• personalized preferences
• graph derived from
queries & clicks
IRDM 2015 1-14
Indexing with Inverted Lists
crisis
B+ tree on terms
17: 0.344: 0.4
...
Internet... trust...
52: 0.153: 0.855: 0.6
12: 0.514: 0.4
...
28: 0.144: 0.251: 0.652: 0.3
17: 0.128: 0.7
...
17: 0.317: 0.144: 0.4
44: 0.2
11: 0.6index lists with postings(DocId, score)sorted by DocId
Google etc.:> 10 Mio. terms> 100 Bio. docs> 50 TB index
q: Internetcrisistrust
Vector space model suggests term-document matrix,but data is sparse and queries are even very sparse better use inverted index lists with terms as keys for B+ tree
terms can be full words, word stems, word pairs, substrings, N-grams, etc.(whatever „dictionary terms“ we prefer for the application)
• index-list entries in DocId order for fast Boolean operations
• many techniques for excellent compression of index lists
• additional position index needed for phrases, proximity, etc.(or other precomputed data structures)
IRDM 2015 1-15
Search Result Quality: Evalution Measures
Capability to return only relevant documents (no false positives):
Precision (Präzision) = r
rtopamongdocsrelevant#
Recall (Ausbeute) = docsrelevant#
rtopamongdocsrelevant#
Capability to return all relevant documents (no false negatives):
00,20,40,60,8
1
0 0,2 0,4 0,6 0,8
Recall
Pre
cisi
on
Typical quality
00,20,40,60,8
1
0 0,2 0,4 0,6 0,8
Recall
Pre
cisi
on
Ideal quality
typically for r = 10, 100, 1000
typically for r = corpus size
ideal measure is user satisfactionheuristically approximated by benchmarking measures(on test corpora with query suite and relevance assessment by experts)