U.S. patent application number 14/639176 was filed with the patent office on 2016-03-17 for extraction of inference rules from heterogeneous graphs.
The applicant listed for this patent is International Business Machines Corporation. Invention is credited to Apoorv Agarwal, Kenneth J. Barker, Jennifer Chu-Carroll, Aditya A. Kalyanpur, Christopher A. Welty, Wlodek W. Zadrozny.
Application Number | 20160078344 14/639176 |
Document ID | / |
Family ID | 55455065 |
Filed Date | 2016-03-17 |
United States Patent
Application |
20160078344 |
Kind Code |
A1 |
Agarwal; Apoorv ; et
al. |
March 17, 2016 |
EXTRACTION OF INFERENCE RULES FROM HETEROGENEOUS GRAPHS
Abstract
According to an aspect, a heterogeneous graph in a data store is
accessed. The heterogeneous graph includes a plurality of nodes
having a plurality of node types. The nodes are connected by edges
having a plurality of relation types. One or more intermediary
graphs are created based on the heterogeneous graph. The
intermediary graphs include intermediary nodes that are the
relation types of the edges of the heterogeneous graph and include
intermediary links between the intermediary nodes based on shared
instances of the nodes between relation types in the heterogeneous
graph. The intermediary graphs are traversed to find sets of
relations based on intermediary links according to a template. An
inference rule is extracted from the heterogeneous graph based on
finding sets of relations in the intermediary graphs. The inference
rule defines an inferred relation type between at least two of the
nodes of the heterogeneous graph.
Inventors: |
Agarwal; Apoorv; (New York,
NY) ; Barker; Kenneth J.; (Mahopac, NY) ;
Chu-Carroll; Jennifer; (Dobbs Ferry, NY) ; Kalyanpur;
Aditya A.; (Westwood, NJ) ; Welty; Christopher
A.; (Flushing, NY) ; Zadrozny; Wlodek W.;
(Charlotte, NC) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
International Business Machines Corporation |
Armonk |
NY |
US |
|
|
Family ID: |
55455065 |
Appl. No.: |
14/639176 |
Filed: |
March 5, 2015 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
14485942 |
Sep 15, 2014 |
|
|
|
14639176 |
|
|
|
|
Current U.S.
Class: |
706/14 |
Current CPC
Class: |
G06F 16/9024 20190101;
G16H 50/70 20180101; G06F 19/00 20130101; G06N 5/025 20130101; G06N
20/00 20190101 |
International
Class: |
G06N 5/02 20060101
G06N005/02 |
Claims
1. A method comprising: accessing a heterogeneous graph in a data
store, the heterogeneous graph comprising a plurality of nodes
having a plurality of node types, the nodes connected by edges
having a plurality of relation types; creating one or more
intermediary graphs based on the heterogeneous graph, the one or
more intermediary graphs comprising intermediary nodes that are the
relation types of the edges of the heterogeneous graph and further
comprising intermediary links between the intermediary nodes based
on shared instances of the nodes between the relation types in the
heterogeneous graph; traversing the one or more intermediary graphs
to find sets of relations based on the intermediary links according
to a template; and extracting inference rules from the
heterogeneous graph based on finding the sets of relations in the
one or more intermediary graphs, each of the inference rules
defining an inferred relation type between at least two of the
nodes of the heterogeneous graph.
2. The method of claim 1, wherein creating one or more intermediary
graphs further comprises: creating a source intermediary graph
comprising the intermediary nodes connected with undirected links
as the intermediary links, the undirected links based on the
relation types of the intermediary nodes sharing a common source
node in the heterogeneous graph.
3. The method of claim 2, wherein creating one or more intermediary
graphs further comprises: creating a target intermediary graph
comprising the intermediary nodes connected with undirected links
as the intermediary links, the undirected links based on the
relation types of the intermediary nodes sharing a common target
node in the heterogeneous graph.
4. The method of claim 3, wherein creating one or more intermediary
graphs further comprises: creating a target-source intermediary
graph comprising the intermediary nodes connected with directed
links as the intermediary links, the directed links based on the
relation types of the intermediary nodes having a source node that
is a target node of another relation type in the heterogeneous
graph.
5. The method of claim 4, further comprising for each intermediary
link between a first relation type and a second relation type in
the target-source intermediary graph: finding a first set of
relations in the source intermediary graph associated with the
first relation type; finding a second set of relations in the
target intermediary graph associated with the second relation type;
and determining an intersection between the first set of relations
and the second set of relations.
6. The method of claim 1, further comprising: traversing the
heterogeneous graph to extract all of the inference rules that are
inferable from the heterogeneous graph according to the
template.
7. The method of claim 6, wherein the template defines a rule
pattern as three node types having three relation types between the
three node types.
8. The method of claim 7, wherein a computational complexity to
extract all of the inference rules from the heterogeneous graph
according to the template comprising the rule pattern is of order
(n.sup.2+T.sup.3) complexity, where n is the number of nodes in the
heterogeneous graph and T is the number of relation types in the
heterogeneous graph.
Description
DOMESTIC PRIORITY
[0001] This application is a continuation of U.S. application Ser.
No. 14/485,942 filed Sep. 15, 2014, the disclosure of which is
incorporated by reference herein in its entirety.
BACKGROUND
[0002] The present disclosure relates generally to inference rule
extraction, and more specifically, to extraction of inference rules
from heterogeneous graphs.
[0003] Information extracted from literature can be summarized
either manually or automatically in networks or graphs that define
relations between nodes representing various elements. A
heterogeneous graph may include several node types and many
relation types defined between nodes of the heterogeneous graph.
Human users may examine the contents of a heterogeneous graph and
attempt to extract knowledge by looking for patterns in
relationships between various node types and relation types.
However, looking at a heterogeneous graph in a visual interface to
infer rules from the heterogeneous graph can be challenging where
semantic meaning of relations is not available. Additionally, in a
very large graph that includes millions of nodes and edges that
define relations between the nodes, it is impractical for a human
to extract all inferable rules from the graph.
SUMMARY
[0004] Embodiments include a method for inference rule extraction
from a heterogeneous graph. The method includes accessing a
heterogeneous graph in a data store. The heterogeneous graph
includes a plurality of nodes having a plurality of node types. The
nodes are connected by edges having a plurality of relation types.
One or more intermediary graphs are created based on the
heterogeneous graph. The one or more intermediary graphs include
intermediary nodes that are the relation types of the edges of the
heterogeneous graph and further include intermediary links between
the intermediary nodes based on shared instances of the nodes
between the relation types in the heterogeneous graph. The one or
more intermediary graphs are traversed to find sets of relations
based on the intermediary links according to a template. An
inference rule is extracted from the heterogeneous graph based on
finding the sets of relations in the one or more intermediary
graphs. The inference rule defines an inferred relation type
between at least two of the nodes of the heterogeneous graph.
[0005] Additional features and advantages are realized through the
techniques of the present disclosure. Other embodiments and aspects
of the disclosure are described in detail herein. For a better
understanding of the disclosure with the advantages and the
features, refer to the description and to the drawings.
BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS
[0006] The subject matter which is regarded as the invention is
particularly pointed out and distinctly claimed in the claims at
the conclusion of the specification. The forgoing and other
features, and advantages of the invention are apparent from the
following detailed description taken in conjunction with the
accompanying drawings in which:
[0007] FIG. 1 depicts a block diagram of a system for inference
rule extraction in accordance with an embodiment;
[0008] FIG. 2 depicts an example of a heterogeneous graph in
accordance with an embodiment;
[0009] FIG. 3A depicts a source intermediary graph as an
intermediary graph in accordance with an embodiment;
[0010] FIG. 3B depicts a target intermediary graph as an
intermediary graph in accordance with an embodiment;
[0011] FIG. 3C depicts a target-source intermediary graph as an
intermediary graph in accordance with an embodiment;
[0012] FIG. 4 depicts a process flow for inference rule extraction
from a heterogeneous graph in accordance with an embodiment;
[0013] FIG. 5 depicts a high-level block diagram of a
question-answer (QA) framework where embodiments of inference rule
extraction can be implemented in accordance with an embodiment;
and
[0014] FIG. 6 depicts a processing system in accordance with an
embodiment.
DETAILED DESCRIPTION
[0015] Embodiments disclosed herein relate to inference rule
extraction from a heterogeneous graph. As used herein, the term
"semantic relations" refers to relationships between concepts or
meanings. Examples related to the medical field are described
herein; however, embodiments are not limited to applications in the
medical field. Embodiments can be utilized by any application that
uses a heterogeneous graph from which inference rules can be
extracted to support data analysis and knowledge extraction,
including, but not limited to: troubleshooting and repair (e.g., to
facilitate diagnostic analysis of a system or component) and a
general question-answer (QA) system.
[0016] As one example, in the medical domain, a vast number of
knowledge sources and ontologies exist. Such information is also
growing and changing extremely quickly, making the information
difficult for people to read, process, and remember. The
combination of recent developments in information extraction and
the availability of unparalleled medical resources thus offer an
opportunity to develop new techniques to help healthcare
professionals overcome the cognitive challenges they may face in
clinical decision making. The medical domain has a vast amount of
literature found in textbooks, encyclopedias, guidelines,
electronic medical records, and many other sources. The amount of
data is also growing at an extremely high speed. Substantial
understanding of the medical domain has already been included in
the Unified Medical Language System.RTM. (UMLS) knowledge base
(KB), which includes medical concepts, relations, and definitions.
The UMLS KB is a compendium of many controlled vocabularies in the
biomedical sciences and may be viewed as a comprehensive thesaurus
and ontology of biomedical concepts. It provides a mapping
structure among these vocabularies and thus allows translation
among the various terminology systems. The 2013 version of the UMLS
KB contains information about more than 3 million concepts from
over 160 source vocabularies.
[0017] FIG. 1 depicts a block diagram of a system 100 for inference
rule extraction in accordance with an embodiment. One or more
instance of a heterogeneous graph 102 can be constructed based on
literature 104 by an automated or manual process and stored in a
data store 103. The data store 103 can be a memory device or
subsystem, such as a computer memory system, and may be distributed
between multiple physical locations or stored at a single location.
The literature 104 can include a body of documents, journals,
manuals, studies, and the like which describe information. Natural
language processing and semantic relation extraction can be used to
convert information in the literature 104 into the heterogeneous
graph 102. Annotation 106 can be performed on the heterogeneous
graph 102 to add or modify semantic relations. Alternatively,
Annotation 106 can be used to manually create a heterogeneous graph
without relying on automatic techniques. The UMLS KB is one example
of a manually constructed heterogeneous graph that includes several
million nodes, such as diseases, treatments, and symptoms, as well
as hundreds of semantic relation types defined between nodes. As
the size of the heterogeneous graph 102 grows, it may be too
unwieldy to manually inspect thousands or millions of concepts
captured in the heterogeneous graph 102 to discover and infer rules
captured therein.
[0018] As one example, if a brute force approach is taken manually
or by a computer implemented process to discover all rules
pertaining to one node, where a total number of n nodes exist in a
graph and the out-degree of each node is T (i.e., number of
relation types), overall complexity of the inspection process for
each node would be O(n.sup.2*T.sup.3) in big-O notation, i.e.,
order of the growth rate of the function. This is because for each
node, rules can be mined that pertain to two of (n-1) other nodes.
Selection can be done in (n-1)C(2) ways, where C is a node about
which rules are sought. To traverse all the directional rules
between three nodes, complexity is O(T.sup.3), and hence the
complexity for all n nodes is O(n.sup.3*T.sup.3). Exemplary
embodiments improve computing system functionality by reducing
computational complexity to O(n.sup.2+T.sup.3) to infer all rules.
Larger graph sizes would see larger degrees of computational
improvement, thus improving computer system functionality by
reducing required time to extract all rules that can be inferred
from a heterogeneous graph and increasing processing resource
availability for other tasks.
[0019] In the example of FIG. 1, inference rule extraction 108
accesses one or more templates 110 to discover a rule pattern for a
rule to be inferred and extracted from the heterogeneous graph 102.
An example of a rule pattern in the templates 110 is: "If A
relation_x B and B relation_y C then A relation_z C", where A, B,
and C are node types and relation_x, relation_y, and relation_z are
relation types. An inference rule can define an inferred relation
type between at least two of the nodes of the heterogeneous graph
102. The inference rule extraction 108 can access the heterogeneous
graph 102 in the data store 103. To infer such a rule for
particular instances of nodes and relations within the
heterogeneous graph 102, the inference rule extraction 108 creates
one or more intermediary graphs 112. The intermediary graphs 112
include intermediary nodes that are relation types of the edges of
the heterogeneous graph 102. The intermediary graphs 112 also
include intermediary links between the intermediary nodes based on
shared instances of nodes between the relation types in the
heterogeneous graph 102. The intermediary graphs 112 may be
traversed to find sets of relations based on the intermediary links
according to a rule pattern of the templates 110. An inference rule
can be extracted from the heterogeneous graph 102 and stored in
extracted inference rules 114 based on finding the sets of
relations in the intermediary graphs 112.
[0020] As one example, the intermediary graphs 112 can include a
source intermediary graph having intermediary nodes connected with
undirected links as intermediary links. The undirected links may be
based on the relation types of the intermediary nodes sharing a
common source node in the heterogeneous graph 102. The intermediary
graphs 112 can also include a target intermediary graph having the
intermediary nodes connected with undirected links as the
intermediary links, where the undirected links are based on the
relation types of the intermediary nodes sharing a common target
node in the heterogeneous graph 102. The intermediary graphs 112
may also include a target-source intermediary graph having the
intermediary nodes connected with directed links as the
intermediary links. The directed links can be based on the relation
types of the intermediary nodes having a source node that is a
target node of another relation type in the heterogeneous graph
102.
[0021] FIG. 2 depicts an example of a heterogeneous graph 200 in
accordance with an embodiment. The heterogeneous graph 200 is an
example of a portion of the heterogeneous graph 102 of FIG. 1. The
heterogeneous graph 200 includes multiple groups 202A, 202B that
have common relations, as well as other groups (not depicted).
Group 202A includes a medicine node 204A that has a value of
"magnesium sulphate", a symptom node 206A that has a value of
"pain", and a disease node 208A that has a value of "neuralgia",
where the medicine node 204A, symptom node 206A, and disease node
208A are examples of different node types. Group 202A also includes
a number of relations defined between the node types. In the
example of FIG. 2, from medicine node 204A to symptom node 206A, a
may_prevent relation 210 is defined as an edge. From symptom node
206A to disease node 208A, a definitional_manifestion_of relation
212 is defined as an edge. A may_treat relation 214 is defined as
an edge between the medicine node 204A and disease node 208A. The
may_prevent relation 210, definitional_manifestion_of relation 212,
and may_treat relation 214 are examples of different relation types
that are edges in the heterogeneous graph 200.
[0022] The group 202B includes a medicine node 204B that has a
value of "capsaicin", a symptom node 206B that has a value of
"seizures", and a disease node 208B that has a value of
"eclampsia", where the medicine node 204B, symptom node 206B, and
disease node 208B are examples of different node types. Group 202B
also includes a number of relations defined as edges between the
node types. In the example of FIG. 2, from medicine node 204B to
symptom node 206B, a may_prevent relation 210 relation is defined
as an edge. From symptom node 206B to disease node 208B, a
definitional_manifestion_of relation 212 is defined as an edge. In
an exemplary embodiment, a relation between medicine node 204B and
disease node 208B may not be defined but can be inferred as a
may_treat relation 214 as further described herein.
[0023] Upon accessing the heterogeneous graph 200 of FIG. 2, the
inference rule extraction 108 of FIG. 1 can create the intermediary
graphs 112 of FIG. 1 including a source intermediary graph 300A of
FIG. 3A, a target intermediary graph 300B of FIG. 3B, and a
target-source intermediary graph 300C of FIG. 3C.
[0024] The source intermediary graph 300A of FIG. 3A includes an
intermediary node 302A that has a value of
"definitional_manifestation_of" as a relation type of the edge:
definitional_manifestion_of relation 212 of FIG. 2. The source
intermediary graph 300A also includes an intermediary node 304A
that has a value of "may_prevent" as a relation type of the edge:
may_prevent relation 210 of FIG. 2. The source intermediary graph
300A further includes an intermediary node 306A that has a value of
"may_treat" as a relation type of the edge: may_treat relation 214
of FIG. 2. The source intermediary graph 300A can connect
intermediary nodes 304A and 306A with an undirected link 308 as an
intermediary link. The intermediary nodes 304A and 306A are
relation types that share a common source node in the heterogeneous
graph 200 of FIG. 2. For example, medicine node 204A is a common
source node of the may_prevent relation 210 and the may_treat
relation 214 of FIG. 2.
[0025] In general terms, the source intermediary graph 300A is
defined as follows: an intermediary link exists between two
intermediary nodes (e.g., relation_x and relation_y) if the source
nodes for those relations are sufficiently similar. As an example,
a set of source nodes (S) of relation_x and a set of source nodes
(S) of relation_y can be considered sufficiently similar if the
Jaccard value (J) between S(relation_x) and S(relation_y) is
non-zero, i.e., J(S(relation_x), S(relation_y))>0. A Jaccard
value measures the similarity between two sets and is defined as
the size of the intersection of the sets divided by the size of the
union of the sets.
[0026] FIG. 3B depicts the target intermediary graph 300B as one of
the intermediary graphs 112 of FIG. 1 in accordance with an
embodiment. The target intermediary graph 300B includes an
intermediary node 302B that has a value of
"definitional_manifestation_of" as a relation type of the edge:
definitional_manifestion_of relation 212 of FIG. 2. The target
intermediary graph 300B also includes an intermediary node 304B
that has a value of "may_prevent" as a relation type of the edge:
may_prevent relation 210 of FIG. 2. The target intermediary graph
300B further includes an intermediary node 306B that has a value of
"may_treat" as a relation type of the edge: may_treat relation 214
of FIG. 2. The target intermediary graph 300B can connect
intermediary nodes 302B and 306B with an undirected link 310 as an
intermediary link. The intermediary nodes 302B and 306B are
relation types that share a common target node in the heterogeneous
graph 200 of FIG. 2. For example, disease node 208A is a common
target node of the definitional_manifestion_of relation 212 and the
may_treat relation 214 of FIG. 2.
[0027] In general terms, the target intermediary graph 300B is
defined as follows: an intermediary link exists between two
intermediary nodes (e.g., relation_x and relation_y) if the sets of
target nodes for the two relations are sufficiently similar. Again,
a set of target nodes (T) of relation_x and a set of target nodes
(T) of relation_y can be considered sufficiently similar if the
Jaccard value (J) between T(relation_x) and T(relation_y) is
non-zero, i.e., J(T(relation_x), T(relation_y))>0.
[0028] FIG. 3C depicts the target-source intermediary graph 300C as
one of the intermediary graphs 112 of FIG. 1 in accordance with an
embodiment. The target-source intermediary graph 300C includes an
intermediary node 302C that has a value of
"definitional_manifestation_of" as a relation type of the edge:
definitional_manifestion_of relation 212 of FIG. 2. The
target-source intermediary graph 300C also includes an intermediary
node 304C that has a value of "may_prevent" as a relation type of
the edge: may_prevent relation 210 of FIG. 2. The target-source
intermediary graph 300C further includes an intermediary node 306C
that has a value of "may_treat" as a relation type of the edge:
may_treat relation 214 of FIG. 2. The target-source intermediary
graph 300C can connect intermediary nodes 302C and 304C with a
directed link 312 as an intermediary link from intermediary node
302C to intermediary node 304C. The intermediary nodes 302C and
304C are relation types such that the source node of one is the
target node of the other in the heterogeneous graph 200 of FIG. 2.
For example, symptom node 206A is a source node of the
definitional_manifestion_of relation 212 and is a target node with
respect to the may_prevent relation 210 of FIG. 2.
[0029] In general terms, the target-source intermediary graph 300C
is defined as follows: an intermediary link exists between two
intermediary nodes (e.g., relation_x and relation_y) if the set of
target nodes (T) of relation_x and the set of source nodes(S) of
relation_y are sufficiently similar. The Jaccard value (J) between
T(relation_x) and S(relation_y) can be used to measure similarity.
In this graph, unlike the source intermediary graph 300A and the
target intermediary graph 300B of FIGS. 3A and 3B, edges are
directional, pointing from relation_x to relation_y.
[0030] Using the combination of the source intermediary graph 300A,
the target intermediary graph 300B, and the target-source
intermediary graph 300C, one or more of the extracted inference
rules 114 can be extracted. As one example, for each edge,
(r.sub.--1, r.sub.--2), in the target-source intermediary graph
300C, a set of relations R.sub.--3.sub.--1 can be found such that
(r.sub.--1, r.sub.--3.sub.--1) exists in the source intermediary
graph 300A and a set of relations R.sub.--3.sub.--2 can be found
such that (r.sub.--2, r.sub.--3.sub.--2) exists in the target
intermediary graph 300B. A set of relations R.sub.--3 equals
R.sub.--3.sub.--1 intersected with R.sub.--3.sub.--2. This results
in mining inference rules matching a rule pattern "if A_r.sub.--1 B
and B_r.sub.--2 C then A_r.sub.--3 C" from the templates 110 of
FIG. 1.
[0031] As a generalized example, consider an inference rule
(a--r.sub.--1--b, b--r.sub.--2--c, a--r.sub.--3--c). Set S denotes
a set of source nodes for a relation type, and set T denotes a set
of target nodes for a relation type. Since this inference rule
exists, T(r.sub.--1).andgate.S(r.sub.--2) is non-empty (i.e., it
has element b). S(r.sub.--1).andgate.S(r.sub.--3) is non-empty
(i.e., it has element a). T(r.sub.--2).andgate.T(r.sub.--3) is
non-empty (i.e., it has element c). Thus, an inference rule can be
found.
[0032] By applying the inference rule extraction 108 of FIG. 1 to
group 202A of FIG. 2, the intermediary graphs 112 of FIG. 1 can
include the source intermediary graph 300A of FIG. 3A, the target
intermediary graph 300B of FIG. 3B, and the target-source
intermediary graph 300C of FIG. 3C. Where the templates 110 of FIG.
1 include the rule pattern "if A r.sub.--1 B and B r.sub.--2 C then
A_r.sub.--3 C", the extracted inference rules 114 of FIG. 1 for
group 202A of FIG. 2 can include "if magnesium sulphate may prevent
pain and pain is a definitional manifestation of neuralgia then
magnesium sulphate may treat neuralgia". This can be generalized to
an inferred rule that if a medicine node has a may_prevent relation
to a symptom node and the symptom node has a
definitional_manifestation_of relation to a disease node, then the
medicine node should also have a may_treat relation to the disease
node. If may_treat relation 214 is missing or not labeled for group
202B of FIG. 2, a may_treat relation 214 can be inferred between
medicine node 204B (FIG. 2) and disease node 208B (FIG. 2) based on
the inferred rule extracted from group 202A of FIG. 2. Thus, it can
be inferred that "if capsaicin may prevent seizures and seizures
are a definitional manifestation of eclampsia then capsaicin may
treat eclampsia".
[0033] FIG. 4 depicts a process flow 400 for inference rule
extraction from a heterogeneous graph in accordance with an
embodiment. The process flow 400 provides an example of a method
for inference rule extraction. For purposes of explanation, the
process flow 400 is described in terms of the examples of FIGS.
1-3C but can be implemented on various system configurations,
including heterogeneous graphs with millions of nodes resulting in
millions of extracted inference rules.
[0034] At block 402, a heterogeneous graph in a data store is
accessed, such as the heterogeneous graph 102 in data store 103 of
FIG. 1. The heterogeneous graph can include a plurality of nodes
having a plurality of node types. The nodes are connected by edges
having a plurality of relation types, as in the example of FIG.
2.
[0035] At block 404, one or more intermediary graphs are created
based on the heterogeneous graph, such as intermediary graphs 112
of FIG. 1 and intermediary graphs 300A-300C of FIGS. 3A-3C. The one
or more intermediary graphs include intermediary nodes that are the
relation types of the edges of the heterogeneous graph. The one or
more intermediary graphs also include intermediary links between
the intermediary nodes based on shared instances of the nodes
between the relation types in the heterogeneous graph. A source
intermediary graph, such as source intermediary graph 300A of FIG.
3, can be created as one of the intermediary graphs by connecting
intermediary nodes with undirected links as the intermediary links,
where the undirected links are based on the relation types of the
intermediary nodes sharing a common source node in the
heterogeneous graph. A target intermediary graph, such as target
intermediary graph 300B of FIG. 3B, can be created as one of the
intermediary graphs by connecting intermediary nodes with
undirected links as the intermediary links, where the undirected
links are based on the relation types of the intermediary nodes
sharing a common target node in the heterogeneous graph. A
target-source intermediary graph, such as target-source
intermediary graph 300C of FIG. 3, can be created as one of the
intermediary graphs by connecting intermediary nodes with directed
links as the intermediary links, where the directed links are based
on the relation types of the intermediary nodes having a source
node that is a target node of another relation type in the
heterogeneous graph.
[0036] At block 406, the one or more intermediary graphs are
traversed to find sets of relations based on the intermediary links
according to a template. As an example, for each intermediary link
between a first relation type and a second relation type in the
target-source intermediary graph, the source intermediary graph can
be examined to find a first set of relations in the source
intermediary graph associated with the first relation type. The
target intermediary graph can be examined to find a second set of
relations in the target intermediary graph associated with the
second relation type. An intersection between the first set of
relations and the second set of relations can be determined.
[0037] At block 408, an inference rule is extracted from the
heterogeneous graph based on finding the sets of relations in the
one or more intermediary graphs. The inference rule defines an
inferred relation type between at least two of the nodes of the
heterogeneous graph. Inference rules can be stored in the extracted
inference rules 114 of FIG. 1 for use by other processes that may
apply generalized versions of the extracted inference rules 114 to
identify missing relations, create higher level inferences, or
perform other types of rule-based analysis. The heterogeneous graph
can be traversed to extract all inference rules that are inferable
from the heterogeneous graph according to the template, which may
be one of the templates 110 of FIG. 1. A template may define a rule
pattern, such as: three node types having three relation types
between the three node types.
[0038] Turning now to FIG. 5, a high-level block diagram of a
question-answer (QA) framework 500 where embodiments described
herein can be utilized is generally shown.
[0039] The QA framework 500 can be implemented to generate a ranked
list of answers 504 (and a confidence level associated with each
answer) to a given question 502. In an embodiment, general
principles implemented by the framework 500 to generate answers 504
to questions 502 include massive parallelism, the use of many
experts, pervasive confidence estimation, and the integration of
shallow and deep knowledge. In an embodiment, the QA framework 500
shown in FIG. 5 is implemented by the Watson.TM. product from
IBM.
[0040] The QA framework 500 shown in FIG. 5 defines various stages
of analysis in a processing pipeline. In an embodiment, each stage
admits multiple implementations that can produce alternative
results. At each stage, alternatives can be independently pursued
as part of a massively parallel computation. Embodiments of the
framework 500 don't assume that any component perfectly understands
the question 502 and can just look up the right answer 504 in a
database. Rather, many candidate answers can be proposed by
searching many different resources, on the basis of different
interpretations of the question (e.g., based on a category of the
question.) A commitment to any one answer is deferred while more
and more evidence is gathered and analyzed for each answer and each
alternative path through the system.
[0041] As shown in FIG. 5, the question and topic analysis 510 is
performed and used in question decomposition 512. Hypotheses are
generated by the hypothesis generation block 514 which uses input
from the question decomposition 512, as well as data obtained via a
primary search 516 through the answer sources 506 and candidate
answer generation 518 to generate several hypotheses. Hypothesis
and evidence scoring 526 is then performed for each hypothesis
using evidence sources 508 and can include answer scoring 520,
evidence retrieval 522 and deep evidence scoring 524.
[0042] A synthesis 528 is performed of the results of the multiple
hypothesis and evidence scorings 526. Input to the synthesis 528
can include answer scoring 520, evidence retrieval 522, and deep
evidence scoring 524. Learned models 530 can then be applied to the
results of the synthesis 528 to generate a final confidence merging
and ranking 532. A ranked list of answers 504 (and a confidence
level associated with each answer) is then output.
[0043] Relation extraction plays a key role in information
extraction in the QA framework 500 shown in FIG. 5. Embodiments of
the inference rule extraction herein can be utilized by the QA
framework 500 to improve relation extraction. Embodiments can be
utilized, for example, in candidate answer generation 518, where
extracted inference rules from the answer sources 506 can be used
for potential candidate answer generation. Also, in evidence
retrieval 522 and deep evidence scoring 524, extracted inference
rules from the evidence sources 508 can be utilized to detect
implicit relations across the question and passages.
[0044] The framework 500 shown in FIG. 5 can utilize embodiments of
the inference rule extraction described herein to create learned
models 530 by training statistical machine learning algorithms on
prior sets of questions and answers to learn how best to weight
each of the hundreds of features relative to one another. These
weights can be used at run time to balance all of the features when
combining the final scores for candidate answers to new questions
502. In addition, embodiments can be used to generate a KB based on
a corpus of data that replaces or supplements commercially
available KBs.
[0045] Referring now to FIG. 6, there is shown an embodiment of a
processing system 600 for implementing the teachings herein. In
this embodiment, the processing system 600 has one or more central
processing units (processors) 601a, 601b, 601c, etc. (collectively
or generically referred to as processor(s) 601). Processors 601,
also referred to as processing circuits, are coupled to system
memory 614 and various other components via a system bus 613. Read
only memory (ROM) 602 is coupled to system bus 613 and may include
a basic input/output system (BIOS), which controls certain basic
functions of the processing system 600. The system memory 614 can
include ROM 602 and random access memory (RAM) 610, which is
read-write memory coupled to system bus 613 for use by processors
601.
[0046] FIG. 6 further depicts an input/output (I/O) adapter 607 and
a network adapter 606 coupled to the system bus 613. I/O adapter
607 may be a small computer system interface (SCSI) adapter that
communicates with a hard disk 603 and/or tape storage drive 605 or
any other similar component. I/O adapter 607, hard disk 603, and
tape storage drive 605 are collectively referred to herein as mass
storage 604. Software 620 for execution on processing system 600
may be stored in mass storage 604. The mass storage 604 is an
example of a tangible storage medium readable by the processors
601, where the software 620 is stored as instructions for execution
by the processors 601 to perform a method, such as the process flow
400 of FIG. 4. Network adapter 606 interconnects system bus 613
with an outside network 616 enabling processing system 600 to
communicate with other such systems. A screen (e.g., a display
monitor) 615 is connected to system bus 613 by display adapter 612,
which may include a graphics controller to improve the performance
of graphics intensive applications and a video controller. In one
embodiment, adapters 607, 606, and 612 may be connected to one or
more I/O buses that are connected to system bus 613 via an
intermediate bus bridge (not shown). Suitable I/O buses for
connecting peripheral devices such as hard disk controllers,
network adapters, and graphics adapters typically include common
protocols, such as the Peripheral Component Interconnect (PCI).
Additional input/output devices are shown as connected to system
bus 613 via user interface adapter 608 and display adapter 612. A
keyboard 609, mouse 640, and speaker 611 can be interconnected to
system bus 613 via user interface adapter 608, which may include,
for example, a Super I/O chip integrating multiple device adapters
into a single integrated circuit.
[0047] Thus, as configured in FIG. 6, processing system 600
includes processing capability in the form of processors 601, and,
storage capability including system memory 614 and mass storage
604, input means such as keyboard 609 and mouse 640, and output
capability including speaker 611 and display 615. In one
embodiment, a portion of system memory 614 and mass storage 604
collectively store an operating system such as the AIX.RTM.
operating system from IBM Corporation to coordinate the functions
of the various components shown in FIG. 6.
[0048] Technical effects and benefits include inference rule
extraction from a heterogeneous graph using intermediary graphs to
increase processing efficiency and reduce latency.
[0049] The present invention may be a system, a method, and/or a
computer program product. The computer program product may include
a computer readable storage medium (or media) having computer
readable program instructions thereon for causing a processor to
carry out aspects of the present invention. The computer readable
storage medium can be a tangible device that can retain and store
instructions for use by an instruction execution device.
[0050] The computer readable storage medium may be, for example,
but is not limited to, an electronic storage device, a magnetic
storage device, an optical storage device, an electromagnetic
storage device, a semiconductor storage device, or any suitable
combination of the foregoing. A non-exhaustive list of more
specific examples of the computer readable storage medium includes
the following: a portable computer diskette, a hard disk, a random
access memory (RAM), a read-only memory (ROM), an erasable
programmable read-only memory (EPROM or Flash memory), a static
random access memory (SRAM), a portable compact disc read-only
memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a
floppy disk, a mechanically encoded device such as punch-cards or
raised structures in a groove having instructions recorded thereon,
and any suitable combination of the foregoing. A computer readable
storage medium, as used herein, is not to be construed as being
transitory signals per se, such as radio waves or other freely
propagating electromagnetic waves, electromagnetic waves
propagating through a waveguide or other transmission media (e.g.,
light pulses passing through a fiber-optic cable), or electrical
signals transmitted through a wire.
[0051] Computer readable program instructions described herein can
be downloaded to respective computing/processing devices from a
computer readable storage medium or to an external computer or
external storage device via a network, for example, the Internet, a
local area network, a wide area network and/or a wireless network.
The network may comprise copper transmission cables, optical
transmission fibers, wireless transmission, routers, firewalls,
switches, gateway computers and/or edge servers. A network adapter
card or network interface in each computing/processing device
receives computer readable program instructions from the network
and forwards the computer readable program instructions for storage
in a computer readable storage medium within the respective
computing/processing device.
[0052] Computer readable program instructions for carrying out
operations of the present invention may be assembler instructions,
instruction-set-architecture (ISA) instructions, machine
instructions, machine dependent instructions, microcode, firmware
instructions, state-setting data, or either source code or object
code written in any combination of one or more programming
languages, including an object oriented programming language such
as Smalltalk, C++ or the like, and conventional procedural
programming languages, such as the "C" programming language or
similar programming languages. The computer readable program
instructions may execute entirely on the user's computer, partly on
the user's computer, as a stand-alone software package, partly on
the user's computer and partly on a remote computer or entirely on
the remote computer or server. In the latter scenario, the remote
computer may be connected to the user's computer through any type
of network, including a local area network (LAN) or a wide area
network (WAN), or the connection may be made to an external
computer (for example, through the Internet using an Internet
Service Provider). In some embodiments, electronic circuitry
including, for example, programmable logic circuitry,
field-programmable gate arrays (FPGA), or programmable logic arrays
(PLA) may execute the computer readable program instructions by
utilizing state information of the computer readable program
instructions to personalize the electronic circuitry, in order to
perform aspects of the present invention.
[0053] Aspects of the present invention are described herein with
reference to flowchart illustrations and/or block diagrams of
methods, apparatus (systems), and computer program products
according to embodiments of the invention. It will be understood
that each block of the flowchart illustrations and/or block
diagrams, and combinations of blocks in the flowchart illustrations
and/or block diagrams, can be implemented by computer readable
program instructions.
[0054] These computer readable program instructions may be provided
to a processor of a general purpose computer, special purpose
computer, or other programmable data processing apparatus to
produce a machine, such that the instructions, which execute via
the processor of the computer or other programmable data processing
apparatus, create means for implementing the functions/acts
specified in the flowchart and/or block diagram block or blocks.
These computer readable program instructions may also be stored in
a computer readable storage medium that can direct a computer, a
programmable data processing apparatus, and/or other devices to
function in a particular manner, such that the computer readable
storage medium having instructions stored therein comprises an
article of manufacture including instructions which implement
aspects of the function/act specified in the flowchart and/or block
diagram block or blocks.
[0055] The computer readable program instructions may also be
loaded onto a computer, other programmable data processing
apparatus, or other device to cause a series of operational steps
to be performed on the computer, other programmable apparatus or
other device to produce a computer implemented process, such that
the instructions which execute on the computer, other programmable
apparatus, or other device implement the functions/acts specified
in the flowchart and/or block diagram block or blocks.
[0056] The flowchart and block diagrams in the Figures illustrate
the architecture, functionality, and operation of possible
implementations of systems, methods, and computer program products
according to various embodiments of the present invention. In this
regard, each block in the flowchart or block diagrams may represent
a module, segment, or portion of instructions, which comprises one
or more executable instructions for implementing the specified
logical function(s). In some alternative implementations, the
functions noted in the block may occur out of the order noted in
the figures. For example, two blocks shown in succession may, in
fact, be executed substantially concurrently, or the blocks may
sometimes be executed in the reverse order, depending upon the
functionality involved. It will also be noted that each block of
the block diagrams and/or flowchart illustration, and combinations
of blocks in the block diagrams and/or flowchart illustration, can
be implemented by special purpose hardware-based systems that
perform the specified functions or acts or carry out combinations
of special purpose hardware and computer instructions.
[0057] The terminology used herein is for the purpose of describing
particular embodiments only and is not intended to be limiting of
the invention. As used herein, the singular forms "a", "an" and
"the" are intended to include the plural forms as well, unless the
context clearly indicates otherwise. It will be further understood
that the terms "comprises" and/or "comprising," when used in this
specification, specify the presence of stated features, integers,
steps, operations, elements, and/or components, but do not preclude
the presence or addition of one more other features, integers,
steps, operations, element components, and/or groups thereof.
[0058] The corresponding structures, materials, acts, and
equivalents of all means or step plus function elements in the
claims below are intended to include any structure, material, or
act for performing the function in combination with other claimed
elements as specifically claimed. The description of the present
invention has been presented for purposes of illustration and
description, but is not intended to be exhaustive or limited to the
invention in the form disclosed. Many modifications and variations
will be apparent to those of ordinary skill in the art without
departing from the scope and spirit of the invention. The
embodiment was chosen and described in order to best explain the
principles of the invention and the practical application, and to
enable others of ordinary skill in the art to understand the
invention for various embodiments with various modifications as are
suited to the particular use contemplated.
* * * * *