DETAILED ACTION
This Office Action is sent in response to the Applicant’s Communication received on 11/14/2025 for application number 17/738,555. The Office hereby acknowledges receipt of the following and placed of record in file: Specification, Drawings, Abstract, Oath/Declaration, IDS, and Claims.
Claims 1, 6, 7, 12, and 13 are amended.
Claims 1-13 are pending.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
35 USC 112
Starting on page 7, Applicant argues that Applicant traverses the rejection with regard to "machine learning system" because it rests on the faulty assumption that "machine learning system" is a means-plus-function term under Section 112(±). Page 4 of the Office Action simply declares, in conclusory fashion and without any underlying rationale, that this claim term "uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier." Since the Office Action does not explain why "machine learning system" would not have been understood by a POSITA as having "sufficient definite meaning" in the relevant art of knowledge graphs, the rejection should be withdrawn. The absence of "means" raises the aforementioned rebuttable presumption, and the Patent Office has not rebutted it with its naked assertion of means-plus-function status for this term. Indeed, and contrary to this naked assertion, the state of knowledge in the relevant art for the present application is such that "machine learning system" is understood with the requisite "sufficient definite meaning" as to make improper its designation as a "means-plus-function" term. In particular, the term "machine learning system" has a sufficient definite meaning in the art of artificial intelligence. Moreover, even if it were the case that "generator" is a
means-plus-function term, the specification provides a description of its corresponding structure. Since "generator" in the context of the present application is a computer-implemented limitation, the requirement that the specification describe its corresponding structure may be satisfied by disclosing the algorithm that carries out the function of this term. Here are the following descriptions from the specification that together constitute the requisite algorithm for the claimed "generator":
[0068] The generator 204 may be configured for selecting, with a selector 226, a
number of triples 228 obtained from the vector representations of entities and
relations 218 having a higher likelihood of being a fact of the knowledge graph
100 than other triples obtained from the vector representations of entities and
relations 218.
[0070] The generator 204 may be configured for determining, with a reasoner
230, the at least one third triple 224 from the number of triples 228. The reasoner
230 is for example adapted for processing the terms of the ontology to select the
at least one third triple 224 from the number of triples.
[0072] For determining the at least one second triple 214, the generator 204 may
be configured for sampling representations of two entities of the at least one
second triple 214 from representations of entities of the knowledge graph 100
and/or a relation between the two entities from representations of relations of the
knowledge graph 100.
[0074] The generator 204 may be configured for determining a triple in the
plurality of triples that comprises the first entity, and a relation. The generator 204
may be configured for determining if the relation is of a type that is allowable
according to the constraint or not. The generator 204 may be configured for
determining that the triple violates the constraint or a combination of at least some of the constraints, if the type is not allowable. This means that the triple comprises the same first entity as the knowledge graph fact.
A POSITA would have recognized in the above blurbs a description of how the claimed
generator performs its triple determination function, namely, by using vector representations of entities and relations having a higher likelihood of being a fact of the knowledge graph than other triples ([0068]), by cooperating with a reasoner to process the terms of an ontology to select the at least one third triple ([0070]), sampling representations of two entities of at least one second triple from representations of entities of the knowledge graph and/or a relation between the two entities from representations of relations of the knowledge graph ([0072]), and by determining that a triple violates a constraint or a combination of at least some constraints ([0074]). Collectively, these blurbs from the specification would have been understood by a POSITA as presenting an algorithm for implementing the claimed generator. Accordingly, Applicant submits that "generator," even if deemed a means-plus-function term, does not render claims 7-12 indefinite.
The Examiner finds the Applicant’s argument persuasive. Specifically, Applicant’s argument is persuasive regarding the term "machine learning system" having a sufficient definite meaning in the art of artificial intelligence. Furthermore, in regard to Applicant’s argument regarding the term “generator” is persuasive, as the specification provides sufficient support for a computer-implemented limitation with sufficient algorithm in paragraphs [0068], [0070], [0072], and [0074].
However, Applicant’s amendment to claims 1, 7, and 13 necessitated a new ground of 35 USC 112(b) rejection.
35 USC 101
On page 14 of the remarks section, the Applicant argues that the specification here establishes a link between the amended claims and technological improvement in the field of knowledge graph training. Paragraph [0012] explains the link between the use of the claimed multi-iteration approach to negative triple generation and improved knowledge graph training:
[0012] In accordance with an example embodiment of the present invention, for
automatically training the knowledge graph embedding model the method may
further comprise determining the at least one triple in a first iteration, adding the
at least one triple to a set of triples for a second iteration and training the
knowledge graph embedding model in the second iteration with the set of triples
for the second iteration and/or determining in the second iteration the at least one
triple with the set of triples for the second iteration. A quality of the generated
negative samples is improved iteratively, e.g. by starting with standard random
sampling of negative samples, training the knowledge graph embedding model on
them, and then exploiting the contracting predictions by the knowledge graph
embedding model for the selection of negative samples, i.e. the at least one triple,
for the next iteration of the method. Thus, the model is improved iteratively.
Thus, the above passage from the specification links the use of an iterative triple generation approach with the benefit of higher negative sample quality that improves knowledge graph training. The claims reflect this improvement by reciting the adding of a negative sample (i.e., a generated triple that violates a constraint) to a set of triples that are used to train the claimed knowledge graph, which results in an improved knowledge graph:
wherein for automatically training the knowledge graph embedding model, the method further comprises:
determining the at least one triple in a first iteration, adding the at least one triple to the set of triples for a second iteration; and
training the knowledge graph embedding model in the second iteration with the set of triples.
Applicant further argues that this represents a concrete improvement in the field of knowledge graph training that improves knowledge graph performance through the use, in knowledge graph training, of an iterative approach to generating the negative samples (i.e., "at least one triple in the plurality of triples that violates at least one constraint of the constraints or that violates a combination of at least some of the constraint"). Therefore, in view of this discussion, the claims represent an improvement in the technology of classifiers under Step 2A, Prong Two.
The Examiner respectfully disagrees. If the specification explicitly sets forth an improvement but in a conclusory manner (i.e., a bare assertion of an improvement without the detail necessary to be apparent to a person of ordinary skill in the art), the examiner should not determine the claim improves technology (see MPEP 2106.04(d)(1)). In this regard, the Applicant cited paragraph [0012] merely recites a combination of steps that are recited with a high-level of generality and in a conclusory manner. A bare assertion that exploiting the contracting predictions by the knowledge graph embedding model for the selection of negative samples and iteratively improving the quality of generated negative samples, without explicit details, would not be construed as an improvement to a person of ordinary skill in the art.
Furthermore, in regards to the cited claim limitation:
wherein for automatically training the knowledge graph embedding model, the method further comprises,
the cited limitation was analyzed under Step 2A Prong Two as an additional element that merely provides instructions to implement the abstract idea on a computer/machine learning model, or merely uses a computer/machine learning model as a tool to perform an abstract idea. Therefore, the cited additional element does not integrate the judicial exception into a practical application.
In regards to:
determining the at least one triple in a first iteration, adding the at least one triple to the set of triples for a second iteration,
The cited limitations were analyzed under Step 2A Prong One as being abstract ideas. The actions of determining one triple in a first iteration and adding the triple to the set for a second iteration are limitations recited at a high-level generality and, under the Broadest Reasonable Interpretation (BRI), can be interpreted as a procedure of human observation, evaluation, judgement, or opinion. The claimed limitations can be performed mentally with the aid of pen and paper, and are therefore a mental process.
In regards to:
training the knowledge graph embedding model in the second iteration with the set of triples,
the cited limitation was analyzed under Step 2A Prong Two as an additional element that merely provides instructions to implement the abstract idea on a computer/machine learning model, or merely uses a computer/machine learning model as a tool to perform an abstract idea. Therefore, the limitation does not integrate the judicial exception into a practical application.
Therefore, contrary to Applicant’s argument, the above limitations in combination do not integrate the judicial exception of abstract idea into a practical application.
Additionally, regarding the claim limitation “training the knowledge graph embedding model in the second iteration with the set of triples and/or determining in the second iteration the at least one triple with the set of triples for the second iteration” (appears to have been taken from claim 6, dated 05/06/2022) the Applicant argues that The Office Action’s approach is improper in view of the relevant portions of the MPEP quoted. The Patent Office concludes that the claim does not provide an inventive concept, but fails to support this conclusion with any evidence failing into categories (A)-(D) discussed above. Therefore, in view of this discussion, withdrawal of the Section 101 rejection is requested.
The Examiner respectfully disagrees. MPEP section 2106.07(a)(III) clearly recites: “At Step 2A Prong Two or Step 2B, there is no requirement for evidence to support a finding that the exception is not integrated into a practical application or that the additional elements do not amount to significantly more than the exception unless the examiner asserts that additional limitations are well-understood, routine, conventional activities in Step 2B”. Specifically, the Examiner did not construe the said limitation to be well-understood, routine, conventional activity.
Therefore, the 35 USC 101 rejection is maintained.
35 USC 103
On pages 17 and 18, Applicant argues that that the disclosure in Cai of "replacing the head entity or the tail entity of a positive triple with a random entity" is not sufficient to meet the limitation "determining... at least one triple in the plurality of triples that violates at least one constraint of the constraints". Applicant does not concede that this mapping is apt, but even if it were, it is not the case that the head/tail entity replacement of Cai meets the limitation "wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent." None of the other applied references overcomes this deficiency in Cai. Therefore, withdrawal of all prior art rejections is requested.
Examiner respectfully disagrees. Cai indeed teaches "determining... at least one triple in the plurality of triples that violates at least one constraint of the constraints" in section 3.3, page 6, column 2, paragraph 3: “we instead generate (determining) Neg(h; r; t) by uniformly sampling of Ns entities (a small number compared to the number of all possible negatives) from
ε
to replace h or t”, and in section 3.1, page 3, column 2, paragraph 2, “The negative triple (at least one triple in the plurality of triples) is generated by replacing the head entity or the tail entity of a positive triple with a random entity (violates at least one constraint of the constraints) in the knowledge graph”. Additionally, this is further clarified in the annotated Figure 1 provided below:
PNG
media_image1.png
380
995
media_image1.png
Greyscale
Therefore, Cai’s teaching of “replacing the head entity or the tail entity of a positive triple with a random entity” sufficiently meets the cited claimed limitation as negative samples (which Cai discloses) violate a constraint.
As detailed above, Cai indeed teaches “wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints”. However, the newly added claimed limitation “a union of the at least one triple with the at least one constraint or the combination is inconsistent” is taught by newly applied reference Jian. Therefore, the Applicant’s argument regarding the newly amended claim limitation has been considered, but is moot due to new ground of rejection.
Therefore, the 35 USC 103 rejection is maintained.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. Such claim limitations are:
“a generator configured to” in claim 7.
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-13 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claims 1, 7, and 13, the amended claim limitations “wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent” are indefinite for failing to clearly distinguish the sets of alternatives. Specifically, it is unclear whether the “or” preceding the limitation “the combination is inconsistent” is alternative to “wherein the at least one triple violates the at least one constraint” or alternative to “the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint”, or whether alternative to the combined limitation “wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint”. Therefore, the claim is vague and indefinite.
Dependent claims 2-6, and 8-12 inherit the deficiencies of the independent claims.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-13 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claims 1-6 are directed towards a method. Claims 7-12 are directed towards a device. Claim 13 is directed towards a non-transitory computer-readable storage medium.
Therefore, all claims are directed towards one of the four statutory categories of patent eligible subject matter.
Claim 1
Step 2A Prong 1:
Claim 1 recites:
“providing at least one first triple, wherein the first triple is a true triple of a knowledge graph;” “providing at least one second triple;” “providing an ontology including constraints that characterize correct triples;” Providing at least one first triple, wherein the first triple is a true triple of a knowledge graph, at least one second triple, an ontology including constraints that characterize correct triples are actions that can be performed mentally with the aid of pen and paper, and are therefore mental processes.
“determining a plurality of triples with the vector representations of entities and relations;” Determining a plurality of triples with the vector representations of entities and relations is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“determining, with the ontology, at least one triple in the plurality of triples that violates at least one constraint of the constraints or that violates a combination of at least some of the constraints;” Determining, with the ontology, at least one triple in the plurality of triples that violates at least one constraint of the constraints or that violates a combination of at least some of the constraints is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“determining the at least one triple in a first iteration;” Determining the at least one triple in a first iteration is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“adding the at least one triple to the set of triples for a second iteration;” Adding the at least one triple to the set of triples for a second iteration is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“A computer implemented method for automatically generating negative samples for training a knowledge graph embedding model;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“training the knowledge graph embedding model to predict triples of the knowledge graph depending on a set of triples including the at least one first triple and the at least one second triple;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
“wherein for automatically training the knowledge graph embedding model, the method further comprises;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“training the knowledge graph embedding model in the second iteration with the set of triples;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“A computer implemented method for automatically generating negative samples for training a knowledge graph embedding model;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
“training the knowledge graph embedding model to predict triples of the knowledge graph depending on a set of triples including the at least one first triple and the at least one second triple;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
“wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
“wherein for automatically training the knowledge graph embedding model, the method further comprises;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
“training the knowledge graph embedding model in the second iteration with the set of triples;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 2
Step 2A Prong 1:
Claim 2 recites:
“selecting a number of triples in the plurality of triples having a higher likelihood of being a fact of the knowledge graph than other triples in the plurality of triples;” Selecting a number of triples in the plurality of triples having a higher likelihood of being a fact of the knowledge graph than other triples in the plurality of triples is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two and Step 2B:
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. The claim is ineligible.
Claim 3
Step 2A Prong 1:
Claim 3 recites:
“determining, [with the knowledge graph embedding model], for at least one triple in the plurality of triples its likelihood of being a fact of the knowledge graph;” Determining for at least one triple in the plurality of triples its likelihood of being a fact of the knowledge graph is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“the knowledge graph embedding model;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“the knowledge graph embedding model;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 4
Step 2A Prong 1:
Claim 4 recites:
“providing a knowledge graph fact from the knowledge graph;” Providing a knowledge graph fact from the knowledge graph is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“determining a triple in the plurality of triples that includes the first entity, and a relation;” “determining whether the relation is of a type that is allowable according to the constraint or not;” “and determining that the triple violates the constraint when the type is not allowable;” Determining a triple in the plurality of triples that includes the first entity, and a relation, whether the relation is of a type that is allowable according to the constraint or not and determining that the triple violates the constraint when the type is not allowable are actions that can be performed mentally with the aid of pen and paper, and are therefore mental processes.
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“wherein the knowledge graph fact includes a first entity, and a reference relation or a representation thereof, wherein the reference relation is of a reference type;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“wherein the knowledge graph fact includes a first entity, and a reference relation or a representation thereof, wherein the reference relation is of a reference type;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim 5
Step 2A Prong 1:
Claim 5 recites:
“determining a set of triples from the plurality of triples that includes triples that violate the constraint;” Determining a set of triples from the plurality of triples that includes triples that violate the constraint is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“selecting from the plurality of triples at least one triple that is different than the triples in the set of triples;” Selecting from the plurality of triples at least one triple that is different than the triples in the set of triples is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two and Step 2B:
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. The claim is ineligible.
Claim 6
Step 2A Prong 1:
Claim 6 recites:
“determining in the second iteration the at least one triple with the set of triples for the second iteration;” Determining in the second iteration the at least one triple with the set of triples for the second iteration is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two and Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible.
Claim 7
Step 2A Prong 1:
Claim 7 recites:
“[a storage configured to] provide a knowledge graph and/or an ontology including constraints that characterize correct triples;” “[a machine learning system configured to] provide at least one first triple, wherein the first triple is a true triple of the knowledge graph;” “provide at least one second triple;” Providing a knowledge graph and/or an ontology, at least one first triple, wherein the first triple is a true triple of the knowledge graph, and at least one second triple, are actions that can be performed mentally with the aid of pen and paper, and are therefore mental processes.
“[a generator configured to] determine a plurality of triples with the vector representations of entities and relations;” Determining a plurality of triples with the vector representations of entities and relations is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“determine, with the ontology, at least one triple in the plurality of triples that violates at least one constraint of the constraints or that violates a combination of at least some of the constraints;” Determining, with the ontology, at least one triple in the plurality of triples that violates at least one constraint of the constraints or that violates a combination of at least some of the constraints is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“[wherein the machine learning system is further configured to] determine the at least one triple in a first iteration, add the at least one triple to the set of triples for a second iteration;”
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“A device for automatically generating negative samples for training a knowledge graph embedding model;” “wherein the device is configured to automatically train the knowledge graph embedding model;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“a storage configured to;” “a machine learning system;” “a generator configured to;” “wherein the machine learning system is further configured to;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“train the knowledge graph embedding model to predict triples of the knowledge graph depending on a set of triples including the at least one first triple and the at least one second triple;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
“wherein for automatically training the knowledge graph embedding model, the method further comprises;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“to train the knowledge graph embedding model in the second iteration with the set of triples for the second iteration;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“A device for automatically generating negative samples for training a knowledge graph embedding model;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
“a storage configured to;” “a machine learning system;” “a generator configured to;” “wherein the machine learning system is further configured to;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
“train the knowledge graph embedding model to predict triples of the knowledge graph depending on a set of triples including the at least one first triple and the at least one second triple;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
“wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
“wherein for automatically training the knowledge graph embedding model, the method further comprises;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
“to train the knowledge graph embedding model in the second iteration with the set of triples for the second iteration;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claims 8-12 are device claims that recite the same limitations as claims 2-6, respectively. Therefore, claims 8-12 are rejected using the same rationale as claims 2-6, respectively.
Claim 13
Step 2A Prong 1:
Claim 13 recites:
“providing at least one first triple, wherein the first triple is a true triple of a knowledge graph;” “providing at least one second triple;” “providing an ontology including constraints that characterize correct triples;” Providing at least one first triple, wherein the first triple is a true triple of a knowledge graph, at least one second triple, an ontology including constraints that characterize correct triples are actions that can be performed mentally with the aid of pen and paper, and are therefore mental processes.
“determining a plurality of triples with the vector representations of entities and relations;” Determining a plurality of triples with the vector representations of entities and relations is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“determining, with the ontology, at least one triple in the plurality of triples that violates at least one constraint of the constraints or that violates a combination of at least some of the constraints;” Determining, with the ontology, at least one triple in the plurality of triples that violates at least one constraint of the constraints or that violates a combination of at least some of the constraints is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“determining the at least one triple in a first iteration;” Determining the at least one triple in a first iteration is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
“adding the at least one triple to the set of triples for a second iteration;” Adding the at least one triple to the set of triples for a second iteration is an action that can be performed mentally with the aid of pen and paper, and is therefore a mental process.
Step 2A Prong Two
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“A non-transitory computer-readable storage medium on which is stored a computer program for automatically generating negative samples for training a knowledge graph embedding model, the computer program, when executed by a computer, causing the computer to perform the following steps;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“training the knowledge graph embedding model to predict triples of the knowledge graph depending on a set of triples including the at least one first triple and the at least one second triple;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)).
“wherein for automatically training the knowledge graph embedding model, the method further comprises;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“training the knowledge graph embedding model in the second iteration with the set of triples;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
Step 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
“A non-transitory computer-readable storage medium on which is stored a computer program for automatically generating negative samples for training a knowledge graph embedding model, the computer program, when executed by a computer, causing the computer to perform the following steps;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
“training the knowledge graph embedding model to predict triples of the knowledge graph depending on a set of triples including the at least one first triple and the at least one second triple;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
“wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent;” The limitation amounts to merely indicating a field of use or technological environment in which to apply a judicial exception. This does not amount to significantly more than the exception itself (MPEP 2106.05(h)) and cannot provide an inventive concept.
“wherein for automatically training the knowledge graph embedding model, the method further comprises;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
“training the knowledge graph embedding model in the second iteration with the set of triples;” Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)) which cannot provide an inventive concept.
Even when considered in combination, these additional elements represent mere instructions to apply an exception and therefore do not provide an inventive concept. The claim is ineligible.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-13 are rejected under 35 U.S.C. 103 as being unpatentable over Xin et al. (CN 112633478 A, see attached translation), hereinafter Xin, in view of Cai et al. (KBGAN: Adversarial Learning for Knowledge Graph Embeddings, published 2018), hereinafter Cai, and Jiang et al. (CN 111291192 A, see attached translation), hereinafter Jian.
Regarding claim 1, Xin teaches,
Training (Para 0056, iterative learning) the knowledge graph embedding model (Para 0056, graph autoencoder model) to predict triples (Para 0056, generate new triplets) of the knowledge graph depending on a set of triples (Para 0011, training triples, valid triples, and test triples) [Para 0056, We then select axioms with high scores to generate new triplets and inject them into the knowledge graph, and then return to the graph autoencoder model for iterative learning; Para 0011, Step 1: Input knowledge graph data: This step reads the knowledge graph data, including entity sets, relationship sets, training triples, valid triples, and test triples, and then initializes the read data];
determining (Para 0041, obtain) vector representations of entities (Para 0041, vector representation of the entity) and relations (Para 0041, embedding matrix representation of the relationship) with the knowledge graph embedding model [Para 0041, The graph autoencoder layer mainly consists of two parts: an entity encoder and a decoder. The relational graph convolutional neural network is used as the entity encoder to obtain the embedded vector representation of the entity in the knowledge graph. Secondly, the linear mapping-based embedding method DistMult is used as the decoder to decode the entity embedding vector to obtain the embedding matrix representation of the relationship.];
determining a plurality of triples (Para 0026, generate new triples) with the vector representations of entities (Para 0019, entity embedding E) and relations (Para 0022, relation embedding) [Para 0019, Step 5, Relation Embedding R: This step decodes the entity embedding E through DistMult; Para 0022, the axiom satisfied by the relation can be obtained by calculating the similarity between the relation embedding and the rule conclusion; Para 0026, Then, axioms with high scores can be selected to generate new triples];
providing an ontology (Para 0031, ontology semantic information) including constraints (Para 0031, rule constraints) that characterize correct triples (Para 0052, a new triple (Bob, hasFriend, Alice)) [Para 0031, IterG proposed ten OWL-like attribute axioms and formalized the axiom conditions and rule constraints to extract the ontology semantic information in the knowledge graph; Para 0052, consider the axiom SymmetricOP(hasFriend). If a knowledge graph contains the triple (Alice, hasFriend, Bob), then according to the rule form and rule conclusion of the symmetric axiom in Figures 3 and 4, a new triple (Bob, hasFriend, Alice) can be inferred];
determining the at least one triple (Para 0056, generate new triplets) in a first iteration, adding the at least one triple to the set of triples (Para 0056, inject them into the knowledge graph) for a second iteration (Para 0056, and then return to the graph autoencoder model for iterative learning) [Para 0056, We then select axioms with high scores to generate new triplets and inject them into the knowledge graph, and then return to the graph autoencoder model for iterative learning];
and training the knowledge graph embedding model (Para 0056, the graph autoencoder model) in the second iteration (Para 0056, iterative learning) with the set of triples (Para 0056, inject them into the knowledge graph, and then) [Para 0056, We then select axioms with high scores to generate new triplets and inject them into the knowledge graph, and then return to the graph autoencoder model for iterative learning].
Xin teaches a number of claim 1’s limitations including the ontology and the plurality of triples.
However, Xin does not teach a computer implemented method for automatically generating negative samples for training a knowledge graph embedding model, the method comprising the following steps: providing at least one first triple, wherein the first triple is a true triple of a knowledge graph; providing at least one second triple, determining at least one triple that violates at least one constraint of the constraints or that violates a combination of at least some of the constraints, wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent.
Cai teaches,
A computer implemented method for automatically (Sect 1, pg. 2, col 1, para 1, KBGAN) generating negative samples (Sect 1, pg. 2, col 1, para 1, negative examples) for training a knowledge graph embedding model [Sect 1, pg. 2, col 1, para 1, we propose a novel adversarial learning framework, namely, KBGAN, for generating better negative examples to train knowledge graph embedding models], the method comprising the following steps: providing at least one first triple, wherein the first triple is a true triple of a knowledge graph (Fig. 1, Box 1; Fig. 1 caption, discriminator (D) receives… the ground truth triple); providing at least one second triple (Fig. 2, Box 2; Fig. 1 caption, discriminator (D) receives the generated negative triple) [Fig. 1 caption, discriminator (D) receives the generated negative triple as well as the ground truth triple];
PNG
media_image2.png
540
1082
media_image2.png
Greyscale
determining (Sect 3.3, pg. 6, col 2, para 3, generate) at least one triple (Sect 3.3, pg. 6, col 2, para 3, Neg(h; r; t)) that violates (Sect 3.3, pg. 6, col 2, para 3, replacing the head entity or the tail entity of a positive triple with a random entity) at least one constraint of the constraints (Sect 3.3, pg. 6, col 2, para 3, head entity or the tail entity of a positive triple) or that violates a combination of at least some of the constraints [Sect 3.3, pg. 6, col 2, para 3, we instead generate Neg(h; r; t) by uniformly sampling of Ns entities (a small number compared to the number of all possible negatives) from
ε
to replace h or t; Sect 3.1, pg. 3, col 2, para 2, The negative triple is generated by replacing the head entity or the tail entity of a positive triple with a random entity in the knowledge graph].
Cai is analogous to the claimed invention as they both relate to the knowledge graph embedding models. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Xin’s teachings to incorporate the teachings of Cai and provide a method of generating negatives samples for knowledge graph embedding model including providing a plurality of triples, wherein one of the triples violates an ontological constraint [Cai, Sect 1, pg. 2, col 1, para 1] for generating better negative examples to reliably train knowledge graph embedding models.
Xin-Cai teach the above limitations of claim 1 including the at least one triple violates the at least one constraint or the combination of at least some of the constraints (Cai, Sect 3.1, pg. 3, col 2, para 2 and Sect 3.3, pg. 6, col 2, para 3) and automatically training the knowledge graph embedding model (Cai, Sect 1, pg. 2, col 1, para 1).
Xin-Cai do not teach a union of at least one triple with at least one constraint or combination is inconsistent.
Jian teaches,
a union (Para 0138, concatenated) of the at least one triple (Para 0138, triple positive sample) with the at least one constraint or the combination (Para 0138, third subject and the third relation) is inconsistent (Para 0137, Obtain a negative sample candidate set according to a query sentence formed by the third subject and the third relationship of each triple positive sample) [Para 0136, Each triplet positive sample includes the third subject, the third relation and the third object constraints. Among them, the third object constraint refers to the conditions that the third object must meet; Para 0137, Obtain a negative sample candidate set according to a query sentence formed by the third subject and the third relationship of each triple positive sample; Para 0138, After determining the triple positive samples, a negative sample candidate set is obtained based on the triple positive samples. Specifically, the third subject and the third relation contained in each triple positive sample are concatenated to obtain a query statement].
Jian is analogous to the claimed invention as they both relate to the utilization of knowledge graphs. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Xin-Cai’s teachings to incorporate the teachings of Jian and provide a union of at least one triple with at least one constraint or combination is inconsistent [Jian, para 0143] in order to assess the prediction accuracy of machine learning models.
Regarding claim 2, Xin-Cai teach all the limitations of claim 1 including the determining of the at least one triple, and the plurality of triples (as in claim 1).
Xin further teaches,
selecting a number of triples (Para 0049, scores of triplets) having a higher likelihood of being a fact (Para 0051, indicates a positive triplet) of the knowledge graph (Para 0049, knowledge graph) than other triples in the plurality of triples (Para 0051, T is the total set of all triplets) [Para 0049-0051, The cross entropy loss is then optimized to make the scores of triplets in the model higher than those in negative sampling. The embedding of entities and relationships in the knowledge graph is obtained by minimizing the cross entropy loss. The cross entropy loss function is as follows…Where T is the total set of all triplets, 1 is the activation function, and y is an indicator, where $y=1$ indicates a positive triplet and $y=0$ indicates a negative triplet.].
Regarding claim 3, Xin-Cai teach all the limitations of claims 1 and 2 including the knowledge graph embedding model (as in claim 1).
Xin further teaches,
determining (Para 0051, indicates) for at least one triple (Para 0051, a positive triplet) in the plurality of triples (Para 0051, T is the total set of all triplets) its likelihood (Para 0051, y is an indicator) of being a fact (Para 0051, $y=1$ indicates a positive triplet) of the knowledge graph (Para 0049, knowledge graph) [Para 0049-0051, The embedding of entities and relationships in the knowledge graph is obtained by minimizing the cross entropy loss. The cross entropy loss function is as follows… Where T is the total set of all triplets, l is the activation function, and y is an indicator, where $y=1$ indicates a positive triplet].
Regarding claim 4, Xin-Cai teaches all the limitations of claim 1
Cai further teaches,
providing [Fig. 1 caption, discriminator (D) receives… the ground truth triple (in the hexagonal box)] a knowledge graph fact (Fig. 1, hexagonal box; Sect 3.2, para 2, Suppose we have a ground truth triple LocatedIn(NewOrleans,Louisiana)) from the knowledge graph [Sect 1, para 1, A common representation of knowledge graph beliefs is in the form of a discrete relational triple such as LocatedIn(NewOrleans,Louisiana)], wherein the knowledge graph fact includes a first entity (Sect 3.2, para 2, NewOrleans), and a reference relation (Sect 3.2, para 2, LocatedIn) or a representation thereof, wherein the reference relation is of a reference type (Sect 3.2, para 2, geographical region) [Sect 3.2, para 2, Suppose we have a ground truth triple LocatedIn(NewOrleans,Louisiana), and corrupt it by replacing its tail entity. First, we remove the tail entity, leaving LocatedIn(NewOrleans,?). Because the relation LocatedIn constraints types of its entities, “?” must be a geographical region.];
determining a triple (Fig. 1 caption, sample one triples from the distribution) in the plurality of triples (Fig. 1 caption, a set of candidate negative triples) that includes the first entity (Fig. 1 hexagonal box, NewOrleans), and a relation (Fig. 1 hexagonal box, LocatedIn) [Fig. 1 caption, The generator (G) calculates a probability distribution over a set of candidate negative triples, then sample one triples from the distribution as the output];
PNG
media_image3.png
524
1082
media_image3.png
Greyscale
determining whether the relation is of a type that is allowable according to the constraint or not [Sect 3.2, para 2, First, we remove the tail entity, leaving LocatedIn(NewOrleans,?). Because the relation LocatedIn constraints types of its entities, “?” must be a geographical region.];
and determining that the triple violates the constraint when the type is not allowable [Sect 3.2, para 2, LocatedIn(NewOrleans,Florida) is a very useful negative triple, because it satisfies type constraints, but it cannot be proved wrong without detailed knowledge of American geography.].
Cai is analogous to the claimed invention as they both relate to knowledge graph embedding models. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Xin’s teachings to incorporate the teachings of Cai and provide the determination of a negative triple that shares an entity and a relation of a fact [Cai, sect 3.2, para 2 and sect 1, pg. 2, col 1, para 1] in order to provide useful negative triples that satisfy type constraints which generates better negative examples to train knowledge graph embedding models.
Regarding claim 5, Xin-Cai teaches all the limitations of claim 1.
Xin-Cai do not teach the limitations of claim 5 including determining a set of triples from the plurality of triples that includes triples that violate the constraint; and selecting from the plurality of triples at least one triple that is different than the triples in the set of triples.
Jian teaches,
determining a set of triples (Para 0138, negative sample candidate set) from the plurality of triples (Para 0138, triple positive samples) that includes triples that violate the constraint (Para 0138, negative sample) [Para 0138, After determining the triple positive samples, a negative sample candidate set is obtained based on the triple positive samples; Note: negatives inherently violate constraints];
and selecting (Para 0140, extract) from the plurality of triples at least one triple (Para 0140, triplet negative samples) that is different than the triples in the set of triples (Para 0140, according to the third object constraint condition) [Para 0140, Step 603: extract triplet negative samples from the negative sample candidate set according to the third object constraint condition].
Jian is analogous to the claimed invention as they both relate to the utilization of knowledge graphs. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Xin-Cai’s teachings to incorporate the teachings of Jian and provide the determination and selection of a triple that violates a constraint and is different from other triples [Cai, Sect 1, pg. 2, col 1, para 1] for generating better negative examples to train knowledge graph embedding models.
Regarding claim 6, Xin-Cai teaches all the limitations of claim 1.
Xin further teaches,
determining in the second iteration (Para 0056, iterative learning) the at least one triple (Para 0056, generate new triplets) with the set of triples (Para 0056, inject them into the knowledge graph, and then) for the second iteration [Para 0056, We then select axioms with high scores to generate new triplets and inject them into the knowledge graph, and then return to the graph autoencoder model for iterative learning].
Regarding claim 7, Xin teaches,
Train (Para 0056, iterative learning) the knowledge graph embedding model (Para 0056, graph autoencoder model) to predict triples (Para 0056, generate new triplets) of the knowledge graph depending on a set of triples (Para 0011, training triples, valid triples, and test triples) [Para 0056, We then select axioms with high scores to generate new triplets and inject them into the knowledge graph, and then return to the graph autoencoder model for iterative learning; Para 0011, Step 1: Input knowledge graph data: This step reads the knowledge graph data, including entity sets, relationship sets, training triples, valid triples, and test triples, and then initializes the read data];
determine (Para 0041, obtain) vector representations of entities (Para 0041, vector representation of the entity) and relations (Para 0041, embedding matrix representation of the relationship) with the knowledge graph embedding model [Para 0041, The graph autoencoder layer mainly consists of two parts: an entity encoder and a decoder. The relational graph convolutional neural network is used as the entity encoder to obtain the embedded vector representation of the entity in the knowledge graph. Secondly, the linear mapping-based embedding method DistMult is used as the decoder to decode the entity embedding vector to obtain the embedding matrix representation of the relationship.];
determine a plurality of triples (Para 0026, generate new triples) with the vector representations of entities (Para 0019, entity embedding E) and relations (Para 0022, relation embedding) [Para 0019, Step 5, Relation Embedding R: This step decodes the entity embedding E through DistMult; Para 0022, the axiom satisfied by the relation can be obtained by calculating the similarity between the relation embedding and the rule conclusion; Para 0026, Then, axioms with high scores can be selected to generate new triples];
provide an ontology (Para 0031, ontology semantic information) including constraints (Para 0031, rule constraints) that characterize correct triples (Para 0052, a new triple (Bob, hasFriend, Alice)) [Para 0031, IterG proposed ten OWL-like attribute axioms and formalized the axiom conditions and rule constraints to extract the ontology semantic information in the knowledge graph; Para 0052, consider the axiom SymmetricOP(hasFriend). If a knowledge graph contains the triple (Alice, hasFriend, Bob), then according to the rule form and rule conclusion of the symmetric axiom in Figures 3 and 4, a new triple (Bob, hasFriend, Alice) can be inferred];
determining the at least one triple (Para 0056, generate new triplets) in a first iteration, adding the at least one triple to the set of triples (Para 0056, inject them into the knowledge graph) for a second iteration (Para 0056, and then return to the graph autoencoder model for iterative learning) [Para 0056, We then select axioms with high scores to generate new triplets and inject them into the knowledge graph, and then return to the graph autoencoder model for iterative learning];
and training the knowledge graph embedding model (Para 0056, the graph autoencoder model) in the second iteration (Para 0056, iterative learning) with the set of triples (Para 0056, inject them into the knowledge graph, and then) for the second iteration [Para 0056, We then select axioms with high scores to generate new triplets and inject them into the knowledge graph, and then return to the graph autoencoder model for iterative learning].
Xin teaches a number of claim 1’s limitations including the ontology and the plurality of triples.
However, Xin does not teach a device for automatically generating negative samples for training a knowledge graph embedding model, comprising: a storage configured to provide a knowledge graph and/or an ontology including constraints that characterize correct triples; a machine learning system configured to provide at least one first triple, wherein the first triple is a true triple of a knowledge graph, provide at least one second triple, determine at least one triple that violates at least one constraint of the constraints or that violates a combination of at least some of the constraints, a generator, and wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent.
Cai teaches,
automatically generating negative samples (Sect 1, pg. 2, col 1, para 1, KBGAN, for generating better negative examples) for training a knowledge graph embedding model (Sect 1, pg. 2, col 1, para 1, to train knowledge graph embedding models), comprising: a knowledge graph (Sect 1, pg. 1-2, para 1, A common representation of knowledge graph beliefs) and/or an ontology including constraints that characterize correct triples; a machine learning system (Sect 1, pg. 2, col 1, para 1, KBGAN) [Sect 1, pg. 2, col 1, para 1, we propose a novel adversarial learning framework, namely, KBGAN, for generating better negative examples to train knowledge graph embedding models; Sect 1, pg. 1-2, para 1, A common representation of knowledge graph beliefs is in the form of a discrete relational triple such as LocatedIn(NewOrleans,Louisiana)] configured to provide at least one first triple, wherein the first triple is a true triple of a knowledge graph (Fig. 1, Box 1; Fig. 1 caption, discriminator (D) receives… the ground truth triple); providing at least one second triple (Fig. 2, Box 2; Fig. 1 caption, discriminator (D) receives the generated negative triple) [Fig. 1 caption, discriminator (D) receives the generated negative triple as well as the ground truth triple];
PNG
media_image2.png
540
1082
media_image2.png
Greyscale
generator [Fig. 1, generator (G)] configured to determine (Sect 3.3, pg. 6, col 2, para 3, generate) at least one triple (Sect 3.3, pg. 6, col 2, para 3, Neg(h; r; t)) that violates (Sect 3.3, pg. 6, col 2, para 3, replacing the head entity or the tail entity of a positive triple with a random entity) at least one constraint of the constraints (Sect 3.3, pg. 6, col 2, para 3, head entity or the tail entity of a positive triple) or that violates a combination of at least some of the constraints [Sect 3.3, pg. 6, col 2, para 3, we instead generate Neg(h; r; t) by uniformly sampling of Ns entities (a small number compared to the number of all possible negatives) from
ε
to replace h or t; Sect 3.1, pg. 3, col 2, para 2, The negative triple is generated by replacing the head entity or the tail entity of a positive triple with a random entity in the knowledge graph].
Cai is analogous to the claimed invention as they both relate to the knowledge graph embedding models. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Xin’s teachings to incorporate the teachings of Cai and provide a method of generating negatives samples for knowledge graph embedding model including providing a plurality of triples, wherein one of the triples violates an ontological constraint [Cai, Sect 1, pg. 2, col 1, para 1] for generating better negative examples to reliably train knowledge graph embedding models.
Xin-Cai teach the above limitations of claim 1 including the at least one triple violates the at least one constraint or the combination of at least some of the constraints (Cai, Sect 3.1, pg. 3, col 2, para 2 and Sect 3.3, pg. 6, col 2, para 3) and automatically training the knowledge graph embedding model (Cai, Sect 1, pg. 2, col 1, para 1).
Xin-Cai do not teach a device, storage, and a union of at least one triple with at least one constraint or combination is inconsistent.
Jian teaches,
a device and storage [Para 0025, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method];
a union (Para 0138, concatenated) of the at least one triple (Para 0138, triple positive sample) with the at least one constraint or the combination (Para 0138, third subject and the third relation) is inconsistent (Para 0137, Obtain a negative sample candidate set according to a query sentence formed by the third subject and the third relationship of each triple positive sample) [Para 0136, Each triplet positive sample includes the third subject, the third relation and the third object constraints. Among them, the third object constraint refers to the conditions that the third object must meet; Para 0137, Obtain a negative sample candidate set according to a query sentence formed by the third subject and the third relationship of each triple positive sample; Para 0138, After determining the triple positive samples, a negative sample candidate set is obtained based on the triple positive samples. Specifically, the third subject and the third relation contained in each triple positive sample are concatenated to obtain a query statement].
Jian is analogous to the claimed invention as they both relate to the utilization of knowledge graphs. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Xin-Cai’s teachings to incorporate the teachings of Jian and provide a union of at least one triple with at least one constraint or combination is inconsistent [Jian, para 0143] in order to assess the prediction accuracy of machine learning models.
Claims 8-12 are device claims that recite the same limitations as claims 2-6, respectively. Therefore, claims 8-12 are rejected using the same rationale as claims 2-6, respectively.
Regarding claim 13, Xin teaches,
Training (Para 0056, iterative learning) the knowledge graph embedding model (Para 0056, graph autoencoder model) to predict triples (Para 0056, generate new triplets) of the knowledge graph depending on a set of triples (Para 0011, training triples, valid triples, and test triples) [Para 0056, We then select axioms with high scores to generate new triplets and inject them into the knowledge graph, and then return to the graph autoencoder model for iterative learning; Para 0011, Step 1: Input knowledge graph data: This step reads the knowledge graph data, including entity sets, relationship sets, training triples, valid triples, and test triples, and then initializes the read data];
determining (Para 0041, obtain) vector representations of entities (Para 0041, vector representation of the entity) and relations (Para 0041, embedding matrix representation of the relationship) with the knowledge graph embedding model [Para 0041, The graph autoencoder layer mainly consists of two parts: an entity encoder and a decoder. The relational graph convolutional neural network is used as the entity encoder to obtain the embedded vector representation of the entity in the knowledge graph. Secondly, the linear mapping-based embedding method DistMult is used as the decoder to decode the entity embedding vector to obtain the embedding matrix representation of the relationship.];
determining a plurality of triples (Para 0026, generate new triples) with the vector representations of entities (Para 0019, entity embedding E) and relations (Para 0022, relation embedding) [Para 0019, Step 5, Relation Embedding R: This step decodes the entity embedding E through DistMult; Para 0022, the axiom satisfied by the relation can be obtained by calculating the similarity between the relation embedding and the rule conclusion; Para 0026, Then, axioms with high scores can be selected to generate new triples];
providing an ontology (Para 0031, ontology semantic information) including constraints (Para 0031, rule constraints) that characterize correct triples (Para 0052, a new triple (Bob, hasFriend, Alice)) [Para 0031, IterG proposed ten OWL-like attribute axioms and formalized the axiom conditions and rule constraints to extract the ontology semantic information in the knowledge graph; Para 0052, consider the axiom SymmetricOP(hasFriend). If a knowledge graph contains the triple (Alice, hasFriend, Bob), then according to the rule form and rule conclusion of the symmetric axiom in Figures 3 and 4, a new triple (Bob, hasFriend, Alice) can be inferred];
determining the at least one triple (Para 0056, generate new triplets) in a first iteration, adding the at least one triple to the set of triples (Para 0056, inject them into the knowledge graph) for a second iteration (Para 0056, and then return to the graph autoencoder model for iterative learning) [Para 0056, We then select axioms with high scores to generate new triplets and inject them into the knowledge graph, and then return to the graph autoencoder model for iterative learning];
and training the knowledge graph embedding model (Para 0056, the graph autoencoder model) in the second iteration (Para 0056, iterative learning) with the set of triples (Para 0056, inject them into the knowledge graph, and then) [Para 0056, We then select axioms with high scores to generate new triplets and inject them into the knowledge graph, and then return to the graph autoencoder model for iterative learning].
Xin teaches a number of claim 1’s limitations including the ontology and the plurality of triples.
However, Xin does not teach A non-transitory computer-readable storage medium on which is stored a computer program for automatically generating negative samples for training a knowledge graph embedding model, the computer program, when executed by a computer, causing the computer to perform the following steps: providing at least one first triple, wherein the first triple is a true triple of a knowledge graph; providing at least one second triple, and determining at least one triple that violates at least one constraint of the constraints or that violates a combination of at least some of the constraints, wherein the at least one triple violates the at least one constraint or the combination of at least some of the constraints when a union of the at least one triple with the at least one constraint or the combination is inconsistent.
Cai teaches,
a computer program (Sect 1, pg. 2, col 1, KBGAN) for automatically (Sect 1, pg. 2, col 1, GAN) generating negative samples (Sect 1, pg. 2, col 1, generating better negative examples) for training a knowledge graph embedding model (Sect 1, pg. 2, col 1, to train knowledge graph embedding models) [Sect 1, pg. 2, col 1, para 1, we propose a novel adversarial learning framework, namely, KBGAN, for generating better negative examples to train knowledge graph embedding models], the computer program, when executed by a computer, causing the computer to perform the following steps: providing at least one first triple, wherein the first triple is a true triple of a knowledge graph (Fig. 1, Box 1; Fig. 1 caption, discriminator (D) receives… the ground truth triple); providing at least one second triple (Fig. 2, Box 2; Fig. 1 caption, discriminator (D) receives the generated negative triple) [Fig. 1 caption, discriminator (D) receives the generated negative triple as well as the ground truth triple];
PNG
media_image2.png
540
1082
media_image2.png
Greyscale
determining (Sect 3.3, pg. 6, col 2, para 3, generate) at least one triple (Sect 3.3, pg. 6, col 2, para 3, Neg(h; r; t)) that violates (Sect 3.3, pg. 6, col 2, para 3, replacing the head entity or the tail entity of a positive triple with a random entity) at least one constraint of the constraints (Sect 3.3, pg. 6, col 2, para 3, head entity or the tail entity of a positive triple) or that violates a combination of at least some of the constraints [Sect 3.3, pg. 6, col 2, para 3, we instead generate Neg(h; r; t) by uniformly sampling of Ns entities (a small number compared to the number of all possible negatives) from
ε
to replace h or t; Sect 3.1, pg. 3, col 2, para 2, The negative triple is generated by replacing the head entity or the tail entity of a positive triple with a random entity in the knowledge graph].
Cai is analogous to the claimed invention as they both relate to the knowledge graph embedding models. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Xin’s teachings to incorporate the teachings of Cai and provide a method of generating negatives samples for knowledge graph embedding model including providing a plurality of triples, wherein one of the triples violates an ontological constraint [Cai, Sect 1, pg. 2, col 1, para 1] for generating better negative examples to reliably train knowledge graph embedding models.
Xin-Cai teach the above limitations of claim 1 including the at least one triple violates the at least one constraint or the combination of at least some of the constraints (Cai, Sect 3.1, pg. 3, col 2, para 2 and Sect 3.3, pg. 6, col 2, para 3) and automatically training the knowledge graph embedding model (Cai, Sect 1, pg. 2, col 1, para 1).
Xin-Cai do not teach a non-transitory computer-readable storage medium on which is stored a computer program and a union of at least one triple with at least one constraint or combination is inconsistent.
Jiang teaches,
A non-transitory computer-readable storage medium on which is stored a computer program [Para 0025, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method];
a union (Para 0138, concatenated) of the at least one triple (Para 0138, triple positive sample) with the at least one constraint or the combination (Para 0138, third subject and the third relation) is inconsistent (Para 0137, Obtain a negative sample candidate set according to a query sentence formed by the third subject and the third relationship of each triple positive sample) [Para 0136, Each triplet positive sample includes the third subject, the third relation and the third object constraints. Among them, the third object constraint refers to the conditions that the third object must meet; Para 0137, Obtain a negative sample candidate set according to a query sentence formed by the third subject and the third relationship of each triple positive sample; Para 0138, After determining the triple positive samples, a negative sample candidate set is obtained based on the triple positive samples. Specifically, the third subject and the third relation contained in each triple positive sample are concatenated to obtain a query statement].
Jian is analogous to the claimed invention as they both relate to the utilization of knowledge graphs. Therefore, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Xin-Cai’s teachings to incorporate the teachings of Jian and provide a union of at least one triple with at least one constraint or combination is inconsistent [Jian, para 0143] in order to assess the prediction accuracy of machine learning models.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SYED RAYHAN AHMED whose telephone number is (571)270-0286. The examiner can normally be reached Mon-Fri ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, David Yi can be reached at (571) 270-7519. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/SYED RAYHAN AHMED/Examiner, Art Unit 2126
/DAVID YI/Supervisory Patent Examiner, Art Unit 2126