Entity Disambiguation

1
Entity Disambiguation

• By
• Angela Maduko
• Directed by
• Amit Sheth

2
Entity Disambiguation Problem

• Emerges mainly while merging information from
  different sources
• Two major levels
• 1. Schema/Ontology level Determining the
  similarity of attributes/concepts/classes from
  the different schema/ontology to be merged
• 2. Instance level Which instances of
  concepts/classes (/tuples in relational databases
  ) refer to the same entity

3
Current approaches for both levels

• Feature-based Similarity Approach (FSA)
• Set-Theory Similarity Approach (STA)
• Information-Theory Similarity Approach (ITA)
• Hybrid Approach (HA)
• Relationship-based Similarity Approach (RSA)
• Hybrid Similarity Approach (HSA)

4
ITA

• In 1, Dekang presents a measure for the
  similarity between two concepts based on both
  their commonalities and differences
• Intuition 1 The similarity between A and B is
  related to their commonality. The more
  commonality they share, the more similar they
  are.
• Intuition 2 The similarity between A and B is
  related to the differences between them. The more
  differences they have, the less similar they are.
• Intuition 3 The maximum similarity between A and
  B is reached when A and B are identical, no
  matter how much commonality they share.

5
ITA

• Consider the concept Fruit
• A is an Apple
• B is an Orange
• Commonality of A and B?
• Common(A, B) Fruit(A) and Fruit(B)
• Measures the commonality between A and B
  I(common(A, B)) by the amount of information
  contained in common(A, B)
• Where the information content of S I(S) -logP(S)

6
ITA

• Differences is measured by I(description(A, B))
  I(common(A, B))
• Decription(A, B) is a proposition which describes
  what A and B are
• Can be applied at both levels 1 2
• Intuitively, sim(A, B)
• 1 when A and B are exactly alike
• 0 when they share no commonalities
• Proposes sim(A, B)

7
ITA

• In 2, Resnik measures the similarity between
  two concepts in an is-a taxonomy based on the
  information content of their most specific common
  super-concept
• Define P(c) as the probability of encountering an
  instance of a concept c in the taxonomy
• For any two concepts c1 and c2, define S(c1, c2)
  as the set of concepts that subsume both c1 and
  c2
• Proposes sim(c1, c2)

8
ITA

• 100 instances of concept X
• 4 instances of concept Y
• 200 instances of concept Z
• 2000 instances of all concepts
• sim(A, B)
• Sim(C, D)
• sim(A, D)
• sim(A, E)
• sim(C, D) gt sim(A, B). Should this be so?

Y
D
C
Z
F
E
X
A
B
9
ITA

• Define s(w) as the set of concepts that are word
  senses of word w. Proposes a measure for word
  similarity as follows
• Sim(w1, w2)
• Can be applied at level 1 only
• Doctor (medical and PhD)
• Nurse (medical and nanny)
• Sim(Doctor, Nurse)

10
STA

• 3 introduces a set theoretical notion of a
  matching function F based on the following
  assumptions for classes a, b, c with description
  sets A, B, C respectively
• Matching s(a, b) F(A ? B, A - B, B - A)
• Monotonicity s(a, b) s(a, c) whenever A ? B ?
  A ? C, A - B ? A - C, B - A ? C - A

11
STA

• Proposes two models
• Contrast model Similarity is defined as
• An increasing function of common features
• A decreasing function of distinctive features
  (features that apply to one object but not the
  other)
• S(a, b) ?f(A ? B) - ?f(A -B) - ?f(B - A) (?,?,?
  0)
• Function f measures the salience of set of
  features
• f depends on intensity and context factors
• Intensity physical salience (eg physical
  features)
• Context salience of features varies with
  context

12
STA

• Ratio Model
• S(a, b)
• ?,?,? 0
• Can be applied at both levels 1 2

13
STA

• 4 determines the similarity between two
  entities by the distance between them
• Defines Pij as the probability that entities ai
  and bj have the same value for their kth
  attribute, for all k
• Assigns costs to mismatching errors (ie not
  matching where should and matching where should
  not) and then calculates a cost function based on
  Pij to be maximized.
• Shows that the expected distance between ai and
  bj dij 1 - Pij, substitutes this in the cost
  function
• Calculates the distance between two entities as a
  linear combination (weighted average) of the
  distances between their common attributes

14
STA

• Relationships amongst common attributes such as
  key attributes functional dependencies are
  exploited
• Obtains attribute weights from user
• Can be applied at level 2 only

15
CA

• In 5, the authors present a model-based k-means
  clustering algorithm for name disambiguation in
  citations
• Randomly assigns citations to N clusters
• Estimates prior probabilities of each cluster
• Computes probability that a cluster produces a
  given citation c
• Assigns c to the cluster with the highest
  probability of producing it
• Applied at level 2 only

16
CA

• My comments on this paper
• The drawback of this approach is in the
  estimation of the model parameters, the data
  necessary for this may not be readily available.
• Even if a user supplies estimates of these
  parameters, these estimates may not be unbiased
  and consistent

17
HA

• 7 combines clustering and information content
  approaches for entity disambiguation (Scalable
  Information Bottleneck (LIMBO) method)
• Attempts to cluster entities in such a way that
  the clusters are informative about the entities
  within them
• Model A set T of n entities (relational tuples),
  defined on m attributes (A1, A2, , Am) .Domain
  of attribute Ai is the set Vi Vi,1, Vi,2, ,
  Vi, di
• Let T and V be two discrete random variables that
  can take values from T and V respectively
• Initially, assigns each entity to a cluster ie
  clusters entities. Let Cq denote this initial
  clustering, then the mutual information of Cq and
  T, I(Cq, T) the mutual information of V and T,
  I(V, T)

18
HA

• Assumes number of distinct entities k is known
• Seeks a clustering Ck of V such that I(Ck, T)
  remains as large as possible or the information
  loss I(V, T) - I(Ck, T) is minimal

19
HSA

• In 8, Kashyap and Sheth introduce the concept
  of semantic proximity (semPro) between entities
  to capture their similarity
• In addition to context, employs relationships and
  features of entities in determining their
  similarity
• semPro(O1,O2) ltContext, Abstraction, (D1, D2),
  (S1, S2)gt
• Context ? context in which objects O1 and O2 are
  being compared
• Abstraction ? abstraction/mappings relating
  domains of the objects
• (D1, D2) ? domain definitions of the objects
• (S1, S2) ? states of the objects

20
HSA

• Abstractions
• Total 1-1 value mapping
• Partial many-one mapping.
• Generalization/specialization.
• Aggregation.
• Functional dependencies.
• ANY
• NONE

21
HSA

• Semantic Taxonomy
• Defines 5 degrees of similarity between objects
• Semantic Equivalence
• Semantic Relationship
• Semantic Relevance
• Semantic Resemblance
• Semantic Incompatibility

22
HSA

• Semantic Equivalence strongest measure of
  semantic proximity
• Two objects are said to be semantically
  equivalent when they represent the same real
  world entity ie
• semPro(O1,O2) ltALL, total 1-1 value mapping,
  (D1, D2), - gt (domain Semantic Equivalence)
• semPro(O1,O2) ltALL, M, (D1, D2), (S1, S2)gt
  where M a total 1-1 value mappings between (D1,
  S1) and (D2, S2) (state Semantic Equivalence)

23
HSA

• Semantic Relationship weaker than semantic
  equivalence.
• semPro(O1,O2) ltALL, M, (D1 ,D2) , _)gt where M
  a partial many-one value mapping,
  generalization or aggregation
• Requirement of a 1-1 mapping is relaxed such
  that, given an instance O1, we can identify an
  instance of O2, but not vice versa.

24
HSA

• Semantic Relevance
• Two objects are semantically relevant if there
  exists any mapping between their domains in some
  context
• semPro(O1,O2) ltSOME, ANY, (D1 ,D2) , _)gt

25
HSA

• Semantic Resemblance weakest measure of semantic
  proximity.
• There does not exists any mapping between their
  domains in any context
• Have same roles in some contexts with coherent
  definition contexts

26
HSA

• Semantic Incompatibility
• Asserts semantic dissimilarity.
• Asserts that there is no context and no
  abstraction in which the domains of the two
  objects are related.
• semPro(O1,O2) ltNONE, NONE, (D1,D2), _gt

27
HSA

• 4 encompasses both the feature-based and
  relationship-based approaches
• Represents entity classes using 3 components,
  assesses similarity of two classes using 3
  different similarity measures wrt to the 3
  components viz
• Synonym set (to address polysemy and synonymy)
• Set of distinguishing features or differentiae
  (functions, parts and attributes)
• Set of semantic inter-relations amongst entity
  classes (mainly hyponymy and meronymy)

28
HSA

• Applied at level 1 only
• Defines the semantic neighbourhood of a class a
  with radius r as follows N(a, r) ci ? ?i
  d(a, ci) r where d(a, c) is the shortest path
  connecting the two classes in the ontology

29
HSA

• Proposes S(a, b) ?wSw(a, b) ?uSu(a, b)
  ?nSn(a, b)
• Sw, Su and Sn are the respective similarities
  between synonym sets, features and semantic
  neighbourhoods of classes a and b
• ?w, ?u and ?n 0 are the respective weights of
  the similarity of each component
• S(a, b) based on a normalization of Tverskys3
  model, with 0?1

30
HSA

• Assumes similarity is asymmetric, with more
  similarity from a class to its super-class than
  vice-versa
• The function ? evaluates this asymmetry, defined
  thus ?(a, b)
• where depth(a) returns the shortest path from
  class a to an imaginary root connecting the roots
  of the two ontologies
• Applies word matching in the synonym sets of a
  and b for Sw

31
HSA

• For Su, applies matching over corresponding
  differentiae (functions Sf, parts Sp, attributes
  Sa with corresponding weights ?f, ?p and ?f 0)
  such that Su(a,b) ?fSf(a,b) ?pSp(a,b)
  ?fSf(a,b) where ?f ?p ?f 1
• For Sn, compares entity classes in semantic
  neighbourhoods based on synonym sets or
  differentiae of these classes.

32
HSA

• In 5 Cho et al propose a model derived from the
  edge-based approach, employing information
  content of the node based approach based on these
  facts
• There exists a correlation between similarity and
  of shared parent concepts in a hierarchy
• Link type (hyponymy, meronymy etc) ? semantic
  relationship

33
HSA

• Conceptual similarity between a node and its
  adjacent child node may not be equal
• As depth increases in the hierarchy, conceptual
  similarity b/w a node and its adjacent child node
  decreases
• Population of nodes is not uniform over entire
  ontological structure (links in a dense part of
  hierarchy ? less distance than that in a less
  dense part )

34
HSA

• Proposes S(ci, cj) D(Lj ?i)?0kn
  W(tk)d(ck1?k)f(d) ( maxH(c) ), where
• f(d) is a function that returns a depth factor
  (topological location in hierarchy)
• d(ck1?k) is a density function
• D(Lj ?i) is a function that returns a distance
  factor between ci and cj (shortest path from one
  node to the other)
• W(tk) is a weight function that assigns weights
  to each link type (W(tk) 1 for is-a link)
• H(c) is information content of super-concepts of
  ci and cj
• For level 1 only

35
References

• Dekang Lin, An Information-Theoretic Definition
  of Similarity, Proceedings ofthe Fifteenth
  International Conference on Machine Learning,
  p.296-304, 1998
• Philip Resnik, Using Information Content to
  Evaluate Semantic Similarity in a Taxonomy,
  IJCAI, 1995.
• Tversky Amos, Features of Similarity,
  Psychological Review 84(4), 1977, pp 327 - 352.
• Debabrata Dey, A Distance-Based Approach to
  Entity Reconciliation in Heterogeneous Databases,
  IEEE Transactions on Knowledge and Data
  Engineeing, 14 (3), May/June 2002.
• Hui Han, Hongyuan Zha and C. Lee Giles, A
  Model-based K-means Algorithm for Name
  Disambiguation in Proceedings of the Second
  International Semantic Web Conference (ISWC-03)
  Workshop on Semantic Web Technologies for
  Searching and Retrieving Scientific Data. 2003
• M. Andrea Rodriguez and Max J. Egenhofer,
  Determining Semantic Similarity Among Entity
  Classes from Different Ontologies, IEEE
  Transactions on Knowledge and Data Engineering ,
  15 (2) 442-456, 2003
• Periklis Andritsos, Renee J. Miller and
  Panayiotis Tsaparas, Information-Theoretic Tools
  for Mining Database Structure from Large Data
  Sets, SIGMOD Conference 2004 731-742
• Vipul Kashyap, Amit Sheth, Semantic and schematic
  similarities between database objects a
  context-based approach, VLDB Journal 5, no. 4
  (1996) 276--304. 367
• Miyoung Cho, Junho Choi and Pankoo Kim, An
  Efficient computational Method for Measuring
  Similarity between Two Conceptual Entities, WAIM
  2003 381-388