semantic search - a guide to web research: lecture 4
TRANSCRIPT
![Page 1: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/1.jpg)
Semantic SearchA Guide to Web Research: Lecture 4
Yury Lifshits
Steklov Institute of Mathematics at St.Petersburg
Stuttgart, Spring 2007
1 / 32
![Page 2: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/2.jpg)
The challenge of the Semantic Web, therefore, is toprovide a language that expresses both data and rules forreasoning about the data and that allows rules from anyexisting knowledge representation system to be exportedonto the Web.
T. Berners-Lee, J. Hendler, O. LassilaSemantic Web, 2001
2 / 32
![Page 3: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/3.jpg)
Outline
1 Introduction to Semantic WebConcept and History of DevelopmentArchitecture of Semantic WebConcept of Semantic Search
2 Three Algorithms for Semantic SearchMinimal AnswersConcept MatchingComputing Interconnections
3 Directions for Further Research
3 / 32
![Page 4: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/4.jpg)
Outline
1 Introduction to Semantic WebConcept and History of DevelopmentArchitecture of Semantic WebConcept of Semantic Search
2 Three Algorithms for Semantic SearchMinimal AnswersConcept MatchingComputing Interconnections
3 Directions for Further Research
3 / 32
![Page 5: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/5.jpg)
Outline
1 Introduction to Semantic WebConcept and History of DevelopmentArchitecture of Semantic WebConcept of Semantic Search
2 Three Algorithms for Semantic SearchMinimal AnswersConcept MatchingComputing Interconnections
3 Directions for Further Research
3 / 32
![Page 6: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/6.jpg)
Part ISematic Web
What is it?
What is already done?
What remains to be done?
4 / 32
![Page 7: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/7.jpg)
Motivating Scenarios
A person asking his web-agent:
Book the ticket for the movie “The Lives of Others”in the nearest cinema that shows it today evening
Find a suitable wine for every item in this menu. Ifpossible, choose French
Microwave, please, go to the website of the dishmanufacturer and download the optimal parametersfor cooking
5 / 32
![Page 8: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/8.jpg)
Motivating Scenarios
A person asking his web-agent:
Book the ticket for the movie “The Lives of Others”in the nearest cinema that shows it today evening
Find a suitable wine for every item in this menu. Ifpossible, choose French
Microwave, please, go to the website of the dishmanufacturer and download the optimal parametersfor cooking
5 / 32
![Page 9: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/9.jpg)
Motivating Scenarios
A person asking his web-agent:
Book the ticket for the movie “The Lives of Others”in the nearest cinema that shows it today evening
Find a suitable wine for every item in this menu. Ifpossible, choose French
Microwave, please, go to the website of the dishmanufacturer and download the optimal parametersfor cooking
5 / 32
![Page 10: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/10.jpg)
Timeline
1994: Foundation of W3C. They develop standards such as:HTML, URL, XML, HTTP, PNG, SVG, CSS
1998: Tim Berners-Lee published “Semantic Web Road Map”
1999: W3C launched groups for designing Sematic Webfoundations, the first version of RDF is published
2000: American defence research institution startedinvestigations for ontology descriptions (DAML+OIL project)
2001: “The Sematic Web” paper in Scientific American
2004: New version of RDF, ontology description languageOWL
2006: Candidate recommendation of SPARQL, a querylanguage for Semantic Web
6 / 32
![Page 11: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/11.jpg)
Naıve Plan
1 Develop a MEGA-language that is powerful
enough to describe all human knowledge and
is machine understandable at the same time.
2 Force all web publishers translate their
websites to this language
3 Write programs that can search in and
reason about all the information in the web
There is a more practical solution for the first step
7 / 32
![Page 12: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/12.jpg)
Naıve Plan
1 Develop a MEGA-language that is powerful
enough to describe all human knowledge and
is machine understandable at the same time.
2 Force all web publishers translate their
websites to this language
3 Write programs that can search in and
reason about all the information in the web
There is a more practical solution for the first step7 / 32
![Page 13: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/13.jpg)
RDF and OWL
Tim Berners-Lee suggested to separate development ofsyntax and semantic of this MEGA-language:
Resource Description Framework (RDF) is a syntax fordocuments of Semantic Web. It uses links to ontologies
Ontology Web Language (OWL) is a language forontology description
Ontology describes classes of objects, their propertiesand relationships in some domain, e.g. toy shops
8 / 32
![Page 14: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/14.jpg)
RDF and OWL
Tim Berners-Lee suggested to separate development ofsyntax and semantic of this MEGA-language:
Resource Description Framework (RDF) is a syntax fordocuments of Semantic Web. It uses links to ontologies
Ontology Web Language (OWL) is a language forontology description
Ontology describes classes of objects, their propertiesand relationships in some domain, e.g. toy shops
8 / 32
![Page 15: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/15.jpg)
Semantic Web Step-by-Step
1 Syntax for knowledge representation (done: RDF)
2 Ontology description language (done: OWL)
3 Web-services description language (started: OWL-S)
4 Tools for reading/publishing Semantic Webdocuments (started: Jena, Haystack, Protege)
5 Query language for data represented by RDF(started: SPARQL)
6 Logic reasoning about RDF statements (to be done)
7 Semantic search and semantic agents (to be done)
9 / 32
![Page 16: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/16.jpg)
Semantic Web Step-by-Step
1 Syntax for knowledge representation (done: RDF)
2 Ontology description language (done: OWL)
3 Web-services description language (started: OWL-S)
4 Tools for reading/publishing Semantic Webdocuments (started: Jena, Haystack, Protege)
5 Query language for data represented by RDF(started: SPARQL)
6 Logic reasoning about RDF statements (to be done)
7 Semantic search and semantic agents (to be done)
9 / 32
![Page 17: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/17.jpg)
Semantic Web Step-by-Step
1 Syntax for knowledge representation (done: RDF)
2 Ontology description language (done: OWL)
3 Web-services description language (started: OWL-S)
4 Tools for reading/publishing Semantic Webdocuments (started: Jena, Haystack, Protege)
5 Query language for data represented by RDF(started: SPARQL)
6 Logic reasoning about RDF statements (to be done)
7 Semantic search and semantic agents (to be done)
9 / 32
![Page 18: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/18.jpg)
Semantic Web Step-by-Step
1 Syntax for knowledge representation (done: RDF)
2 Ontology description language (done: OWL)
3 Web-services description language (started: OWL-S)
4 Tools for reading/publishing Semantic Webdocuments (started: Jena, Haystack, Protege)
5 Query language for data represented by RDF(started: SPARQL)
6 Logic reasoning about RDF statements (to be done)
7 Semantic search and semantic agents (to be done)
9 / 32
![Page 19: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/19.jpg)
Semantic Web Step-by-Step
1 Syntax for knowledge representation (done: RDF)
2 Ontology description language (done: OWL)
3 Web-services description language (started: OWL-S)
4 Tools for reading/publishing Semantic Webdocuments (started: Jena, Haystack, Protege)
5 Query language for data represented by RDF(started: SPARQL)
6 Logic reasoning about RDF statements (to be done)
7 Semantic search and semantic agents (to be done)
9 / 32
![Page 20: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/20.jpg)
Semantic Web Step-by-Step
1 Syntax for knowledge representation (done: RDF)
2 Ontology description language (done: OWL)
3 Web-services description language (started: OWL-S)
4 Tools for reading/publishing Semantic Webdocuments (started: Jena, Haystack, Protege)
5 Query language for data represented by RDF(started: SPARQL)
6 Logic reasoning about RDF statements (to be done)
7 Semantic search and semantic agents (to be done)
9 / 32
![Page 21: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/21.jpg)
Semantic Web Step-by-Step
1 Syntax for knowledge representation (done: RDF)
2 Ontology description language (done: OWL)
3 Web-services description language (started: OWL-S)
4 Tools for reading/publishing Semantic Webdocuments (started: Jena, Haystack, Protege)
5 Query language for data represented by RDF(started: SPARQL)
6 Logic reasoning about RDF statements (to be done)
7 Semantic search and semantic agents (to be done)9 / 32
![Page 22: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/22.jpg)
Cake of Tim Berners-Lee
10 / 32
![Page 23: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/23.jpg)
Concept of Semantic Search
What is sematic search?
Assistance to classical web search
Question answering systems
Queries that returns concepts (nodes in XMLdocuments), not documents themselves
Query is a complex concept (small XML tree),semantic search returns the most similar object
SQL-like queries to database of RDF statements
Automated logical inference for RDF statements
11 / 32
![Page 24: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/24.jpg)
Concept of Semantic Search
What is sematic search?
Assistance to classical web search
Question answering systems
Queries that returns concepts (nodes in XMLdocuments), not documents themselves
Query is a complex concept (small XML tree),semantic search returns the most similar object
SQL-like queries to database of RDF statements
Automated logical inference for RDF statements
11 / 32
![Page 25: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/25.jpg)
Concept of Semantic Search
What is sematic search?
Assistance to classical web search
Question answering systems
Queries that returns concepts (nodes in XMLdocuments), not documents themselves
Query is a complex concept (small XML tree),semantic search returns the most similar object
SQL-like queries to database of RDF statements
Automated logical inference for RDF statements
11 / 32
![Page 26: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/26.jpg)
Concept of Semantic Search
What is sematic search?
Assistance to classical web search
Question answering systems
Queries that returns concepts (nodes in XMLdocuments), not documents themselves
Query is a complex concept (small XML tree),semantic search returns the most similar object
SQL-like queries to database of RDF statements
Automated logical inference for RDF statements
11 / 32
![Page 27: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/27.jpg)
Concept of Semantic Search
What is sematic search?
Assistance to classical web search
Question answering systems
Queries that returns concepts (nodes in XMLdocuments), not documents themselves
Query is a complex concept (small XML tree),semantic search returns the most similar object
SQL-like queries to database of RDF statements
Automated logical inference for RDF statements
11 / 32
![Page 28: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/28.jpg)
Concept of Semantic Search
What is sematic search?
Assistance to classical web search
Question answering systems
Queries that returns concepts (nodes in XMLdocuments), not documents themselves
Query is a complex concept (small XML tree),semantic search returns the most similar object
SQL-like queries to database of RDF statements
Automated logical inference for RDF statements
11 / 32
![Page 29: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/29.jpg)
Concept of Semantic Search
What is sematic search?
Assistance to classical web search
Question answering systems
Queries that returns concepts (nodes in XMLdocuments), not documents themselves
Query is a complex concept (small XML tree),semantic search returns the most similar object
SQL-like queries to database of RDF statements
Automated logical inference for RDF statements11 / 32
![Page 30: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/30.jpg)
Part IIIThree Algorithms for Semantic
Search
Finding the most specific answer
Concept matching
Identifying related nodes in XML documents
12 / 32
![Page 31: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/31.jpg)
XRANK: Model
Database is a set of XML documentsThere are hyperlinks between nodesEvery node contain some textQuery is a short list of keywords
A complete answer is a node that togetherwith its descendants contain all query terms
13 / 32
![Page 32: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/32.jpg)
XRANK: Model
Database is a set of XML documentsThere are hyperlinks between nodesEvery node contain some textQuery is a short list of keywords
A complete answer is a node that togetherwith its descendants contain all query terms
13 / 32
![Page 33: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/33.jpg)
Minimal Answers
A node v is called to be a minimal answer if
∀k ∈ Q :[v contains k]
OR[∃u son of v s.t. u contains∗ kAND u is not complete answer]
Search task: find all minimal answers and rankthem accordingly to the link/containement popularity
14 / 32
![Page 34: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/34.jpg)
Minimal Answers
A node v is called to be a minimal answer if
∀k ∈ Q :[v contains k]
OR[∃u son of v s.t. u contains∗ kAND u is not complete answer]
Search task: find all minimal answers and rankthem accordingly to the link/containement popularity
14 / 32
![Page 35: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/35.jpg)
Dewey Code
Nodes in database have Dewey codes n1.n2. . . . nh
For example, Dewey code 7.2.12 denotes the 12thleft son of the 2nd left son of the root of the 7thdocument in our collection.
For every keyword Dewey inverted index store alist of Dewey codes of nodes (DIL) that directly containthis keyword
15 / 32
![Page 36: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/36.jpg)
Dewey Code
Nodes in database have Dewey codes n1.n2. . . . nh
For example, Dewey code 7.2.12 denotes the 12thleft son of the 2nd left son of the root of the 7thdocument in our collection.
For every keyword Dewey inverted index store alist of Dewey codes of nodes (DIL) that directly containthis keyword
15 / 32
![Page 37: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/37.jpg)
Illustration from XRANK paper
16 / 32
![Page 38: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/38.jpg)
Minimal Answers Problem
Given Dewey inverted lists for all query terms toreturn a list of Dewey codes of all minimal answers
17 / 32
![Page 39: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/39.jpg)
Algorithm for Minimal Answers (1/2)
Single pass: every time reada next code in union of DILs
Keep an auxiliary data structure Dewey stackfor the last scanned read node v :
for every predecessor of vkeep a set of keywordsthat are contained∗ prior-or-equal to vignoring complete nodes
18 / 32
![Page 40: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/40.jpg)
Algorithm for Minimal Answers (1/2)
Single pass: every time reada next code in union of DILs
Keep an auxiliary data structure Dewey stackfor the last scanned read node v :
for every predecessor of vkeep a set of keywordsthat are contained∗ prior-or-equal to v
ignoring complete nodes
18 / 32
![Page 41: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/41.jpg)
Algorithm for Minimal Answers (1/2)
Single pass: every time reada next code in union of DILs
Keep an auxiliary data structure Dewey stackfor the last scanned read node v :
for every predecessor of vkeep a set of keywordsthat are contained∗ prior-or-equal to vignoring complete nodes
18 / 32
![Page 42: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/42.jpg)
Algorithm for Minimal Answers (2/2)
Update for Dewey stack from v to u:
1 find a lowest common predecessor w for v and u
2 Sequentially consider ancestors of u from bottom totop, add keywords of u to their set in Dewey stack
3 Stop at root, or with identical set update or on thefirst complete node
4 In latter case output this node to the list of minimalanswers
19 / 32
![Page 43: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/43.jpg)
Conceptual Graph Matching
Query is a tree with labelled edges and nodes
Database is a family of trees
Domain information: similaritybetween edge/node labels
Task: to find a tree in DBwith maximal similarity to query tree
20 / 32
![Page 44: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/44.jpg)
Conceptual Graph Matching
Query is a tree with labelled edges and nodes
Database is a family of trees
Domain information: similaritybetween edge/node labels
Task: to find a tree in DBwith maximal similarity to query tree
20 / 32
![Page 45: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/45.jpg)
Illustration from Conceptual MatchingPaper
21 / 32
![Page 46: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/46.jpg)
Similarity Formula
TreeSim(Q, R) = NodeSim(q0, r0)+
+ maxchildren matching π
(∑i
EdgeSim(q0qi , r0rπi) · TreeSim(Q|qi
, R |rπi)
)
22 / 32
![Page 47: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/47.jpg)
Recursive Algorithm for GraphMatching
Compare query tree with every tree in DB separately:
1 Compute TreeSim for every pair of Q and R roots’children
2 Find the best matching by applying Bellman-Fordalgorithm
Complexity for l -branch trees of depth d :C (d + 1) = l2C (d) + l4 + constC (d) = O(l2d+2) = O(n2l2)
In general, time complexity is O(n4)
23 / 32
![Page 48: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/48.jpg)
Recursive Algorithm for GraphMatching
Compare query tree with every tree in DB separately:
1 Compute TreeSim for every pair of Q and R roots’children
2 Find the best matching by applying Bellman-Fordalgorithm
Complexity for l -branch trees of depth d :C (d + 1) = l2C (d) + l4 + constC (d) = O(l2d+2) = O(n2l2)
In general, time complexity is O(n4)
23 / 32
![Page 49: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/49.jpg)
Recursive Algorithm for GraphMatching
Compare query tree with every tree in DB separately:
1 Compute TreeSim for every pair of Q and R roots’children
2 Find the best matching by applying Bellman-Fordalgorithm
Complexity for l -branch trees of depth d :C (d + 1) = l2C (d) + l4 + constC (d) = O(l2d+2) = O(n2l2)
In general, time complexity is O(n4)23 / 32
![Page 50: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/50.jpg)
XSEarch Model
Database: huge XML tree with labelson internal nodes and keywords on leafs
Query terms: “label:keyword”, “label:”, “:keyword”
Answer: a set of interconnected nodesthat together satisfy all query terms
24 / 32
![Page 51: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/51.jpg)
XSEarch Model
Database: huge XML tree with labelson internal nodes and keywords on leafs
Query terms: “label:keyword”, “label:”, “:keyword”
Answer: a set of interconnected nodesthat together satisfy all query terms
24 / 32
![Page 52: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/52.jpg)
Illustration from XSEarch Paper
25 / 32
![Page 53: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/53.jpg)
Interconnection
Nodes u and v are interconnected iff on theshortest path between them only labels of u and v cancoincide
26 / 32
![Page 54: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/54.jpg)
Properties of Interconnection
For u being ancestor of v :
InCon[u, v ] = InCon[u, parent(v)]&(label(u) 6= label(parent(v))) & InCon[sonv(u), v ]&
(label(sonv(u)) 6= label(v))
Otherwise:
InCon[u, v ] = InCon[u, parent(v)]& (label(u) 6=label(parent(v))) & InCon[parent(u), v ]&
(label(parent(u)) 6= label(v))
Using these formulas we can compute InCon for all pairs in
O(|T |) for all pairs by dynamic programming
27 / 32
![Page 55: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/55.jpg)
Properties of Interconnection
For u being ancestor of v :
InCon[u, v ] = InCon[u, parent(v)]&(label(u) 6= label(parent(v))) & InCon[sonv(u), v ]&
(label(sonv(u)) 6= label(v))
Otherwise:
InCon[u, v ] = InCon[u, parent(v)]& (label(u) 6=label(parent(v))) & InCon[parent(u), v ]&
(label(parent(u)) 6= label(v))
Using these formulas we can compute InCon for all pairs in
O(|T |) for all pairs by dynamic programming
27 / 32
![Page 56: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/56.jpg)
Properties of Interconnection
For u being ancestor of v :
InCon[u, v ] = InCon[u, parent(v)]&(label(u) 6= label(parent(v))) & InCon[sonv(u), v ]&
(label(sonv(u)) 6= label(v))
Otherwise:
InCon[u, v ] = InCon[u, parent(v)]& (label(u) 6=label(parent(v))) & InCon[parent(u), v ]&
(label(parent(u)) 6= label(v))
Using these formulas we can compute InCon for all pairs in
O(|T |) for all pairs by dynamic programming27 / 32
![Page 57: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/57.jpg)
Directions for Further Research
Algorithms for online conceptual graph matching
Queries using arithmetic: “what is the most popularmovie (according to IMDB) I have not seen yet?”
Automated inference for RDF statements?Semantic search for the case when the answer is notin the DB, but can be derived from it.
28 / 32
![Page 58: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/58.jpg)
Call for participation
Know a relevant reference?Have an idea?
Find a mistake?Solved one of these problems?
Knock to my office 1.156
Write to me [email protected]
Join our informal discussions
Participate in writing a follow-up paper
29 / 32
![Page 59: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/59.jpg)
Highlights
XRANK: merging Dewey inverted lists by a singlepass
Concept matching: finding the most similar tree tothe query tree
XSEarch: computing interconnection by dynamicprogramming
Vielen Dank fur Ihre Aufmerksamkeit!Fragen?
30 / 32
![Page 60: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/60.jpg)
Highlights
XRANK: merging Dewey inverted lists by a singlepass
Concept matching: finding the most similar tree tothe query tree
XSEarch: computing interconnection by dynamicprogramming
Vielen Dank fur Ihre Aufmerksamkeit!Fragen?
30 / 32
![Page 61: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/61.jpg)
References (1/2)
Course homepage
http://logic.pdmi.ras.ru/~yura/webguide.html
L.Guo, F.Shao, C.Botev, J.Shanmugasundaram
XRANK: Ranked Keyword Search over XML Documents
http://www.cs.fiu.edu/~vagelis/classes/COP6727/publications/XRank.pdf
S.Cohen, J.Mamou, Y.Kanza, Y.Sagiv
XSEarch: A Semantic Search Engine for XML
http://wwwdb.informatik.uni-rostock.de/Archiv/vldb2003/papers/S03P02.pdf
J.Zhong, H.Zhu, J.Li, Y.Yu
Conceptual Graph Matching for Semantic Search
http://apex.sjtu.edu.cn/docs/iccs2002.pdf
31 / 32
![Page 62: Semantic Search - A Guide to Web Research: Lecture 4](https://reader036.vdocument.in/reader036/viewer/2022071602/613d6078736caf36b75c9ad7/html5/thumbnails/62.jpg)
References (2/2)
R.Guha, R.McCool, E.Miller
Semantic Search
http://learning.ncsa.uiuc.edu/lmarini/papers/p700-guha.pdf
S.Harris
SPARQL query processing with conventional relational database systems
http://eprints.ecs.soton.ac.uk/11126/01/harris-ssws05.pdf
E.Brill, S.Dumais, M.Banko
An Analysis of the AskMSR Question-Answering System
http://www.stanford.edu/class/linguist180/EMNLP2002.pdf
T.Berners-Lee, J.Hendler, O.Lassila
Semantic Web
http://wireless.ictp.trieste.it/school 2002/lectures/canessa/0501berners-lee.ps
32 / 32