DETAILED ACTION
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 .
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.
Claim 1-13 are rejected under 35 U.S.C. 101 because the claimed invention are directed to abstract ideas without significantly more.
Regarding Claim 1:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 1 which included an abstract idea (see rejection claim 1). Additionally, claim 2 recites the abstract ideas:
determining a first representation, wherein the first representation represents an embedding of the first entity: - This limitation is directed to the abstract idea of a mental process, as the process of determining the first representation is a thought process that can be performed in a human mind by observing, evaluating and judging (concepts performed in the human mind (including an observation, evaluation, judgment, opinion) (see MPEP § 2106.04(a)(2), subsection III)).
selecting a second representation from a set of representations of embeddings of entities of the knowledge graph, wherein the second representation represents an embedding of the second entity, and wherein the selecting of the second representation includes determining a prediction for the second representation, and selecting the second representation depending on the prediction for the second representation: - This limitation recites the abstract idea of mathematical concepts, as when given the broadest reasonable interpretation in light of the specification, the selecting of the second representation includes determining the prediction for the second representation. The determining of the prediction for the second representation is being executed by rotating and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation. Rotating and translating are mathematical processes, therefore, this limitation amounts to mathematical process.
wherein the determining of the first representation includes splitting the embedding of the first entity into a first set of multi-dimensional vectors, and mapping the multi-dimensional vectors of the first set to the multi-dimensional vectors of the first representation depending on the relation: - This limitation is directed to the abstract idea of a mental process, as the process of splitting the embedding of the first entity into a first set of multi-dimensional vectors, and mapping the multi-dimensional vectors of the first set to the multi-dimensional vectors of the first representation are thought processes that can be performed in a human mind by observing, evaluating and judging (concepts performed in the human mind (including an observation, evaluation, judgment, opinion) (see MPEP § 2106.04(a)(2), subsection III)).
wherein the determining of the prediction for the second representation includes rotating and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation: - This limitation recites the abstract idea of mathematical concepts, as the processes of rotating and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation, are mathematical processes, therefore, this limitation amounts to mathematical process.
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
determining the link including the first entity, the second entity, and the relation: - This limitation does no more than generally link a judicial exception to a particular technological environment as this limitation recites the use of judicial exception to the technology of knowledge graph (see MPEP 2106.05(h)).
wherein the first representation includes multi-dimensional vectors in hyperbolic spaces: - This limitation is selecting a particular data source or type of data to be manipulated as it is merely adding data to the training set, being insignificant extra-solution activity (see MEPEP n2106.05(g));
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception. The additional elements:
determining the link including the first entity, the second entity, and the relation: - This limitation does no more than generally link a judicial exception to a particular technological environment as this limitation recites the use of judicial exception to the technology of knowledge graph (see MPEP 2106.05(h));
wherein the first representation includes multi-dimensional vectors in hyperbolic spaces: - This limitation is analogous to electronic recordkeeping because it’s adding data to the training set and keeping record of it (see MPEP 2106.05(d) II (iii)).
Regarding Claim 2:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 1 which included an abstract idea (see rejection claim 1). Additionally, claim 2 recites the abstract ideas:
wherein the prediction for the second representation includes multi-dimensional vectors in the hyperbolic spaces, wherein the second representation includes multi-dimensional vectors in the hyperbolic spaces, and wherein the selecting of the second representation includes determining differences between the multi-dimensional vectors of the second representation and the multi-dimensional vectors of the prediction for the second representation that are in the same hyperbolic space, determining a distance depending on the differences, and selecting the second representation depending on the distance: - This limitation recites the abstract idea of mathematical concepts, as the process of determining differences is mathematical calculations and in the light of the specification the process of determining the distance is a mathematical calculation, therefore, this limitation amounts to mathematical concepts.
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception.
Regarding Claim 3:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 2 which included an abstract idea (see rejection claim 2). Additionally, claim 3 recites the abstract ideas:
wherein the determining the distance includes concatenating the multi-dimensional vectors of the prediction for the second representation, concatenating the multi-dimensional vectors of the second representation, and determining the distance depending on the concatenated multi-dimensional vectors of the prediction for the second representation and the concatenated multi-dimensional vectors of the second representation: - This limitation recites the abstract idea of mathematical concepts, as when given the broadest reasonable interpretation in light of the specification, the process of determining the distance is a mathematical calculations, therefore, this limitation amounts to mathematical concepts.
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception.
Regarding Claim 4:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 1 which included an abstract idea (see rejection claim 1). Additionally, claim 4 recites the abstract ideas:
wherein the determining of the first representation includes mapping the vectors of the first set to different hyperbolic spaces: - This limitation is directed to the abstract idea of a mental process, as the process of mapping the vectors of the first set to different hyperbolic spaces is a thought process that can be performed in a human mind by observing, evaluating and judging (concepts performed in the human mind (including an observation, evaluation, judgment, opinion) (see MPEP § 2106.04(a)(2), subsection III)).
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception.
Regarding Claim 5:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 1 which included an abstract idea (see rejection claim 1).
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
wherein a curvature of at least one of the hyperbolic spaces is defined by the relation: - This limitation is selecting a particular data source or type of data to be manipulated as it is merely adding data to the training set, being insignificant extra-solution activity (see MEPEP n2106.05(g));
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception.
wherein a curvature of at least one of the hyperbolic spaces is defined by the relation: - This limitation is analogous to electronic recordkeeping because it’s adding data to the training set and keeping record of it (see MPEP 2106.05(d) II (iii)).
Regarding Claim 6:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 1 which included an abstract idea (see rejection claim 1).
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
wherein an extent of the rotation in at least one of the hyperbolic spaces is defined by defined by the relation: - This limitation is selecting a particular data source or type of data to be manipulated as when given the broadest reasonable interpretation in the light of the specification the extent of the rotation is a learnable parameter, therefore, this limitation is merely adding data to the training set, being insignificant extra-solution activity (see MEPEP n2106.05(g));
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception.
wherein an extent of the rotation in at least one of the hyperbolic spaces is defined by defined by the relation: - This limitation is analogous to electronic recordkeeping because it’s adding data to the training set and keeping record of it (see MPEP 2106.05(d) II (iii)).
Regarding Claim 7:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 1 which included an abstract idea (see rejection claim 1).
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
wherein an extent of the translation in at least one of the hyperbolic spaces is defined by defined by the relation: - This limitation is selecting a particular data source or type of data to be manipulated as when given the broadest reasonable interpretation in the light of the specification the extent of the translation is a learnable parameter, therefore, this limitation is merely adding data to the training set, being insignificant extra-solution activity (see MEPEP n2106.05(g));
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception.
wherein an extent of the translation in at least one of the hyperbolic spaces is defined by defined by the relation: - This limitation is analogous to electronic recordkeeping because it’s adding data to the training set and keeping record of it (see MPEP 2106.05(d) II (iii)).
Regarding Claim 8:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 1 which included an abstract idea (see rejection claim 1).
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
providing the relation, wherein the relation includes one parameter per hyperbolic space that defines the extent of rotating, and/or the extent of translating and/or a curvature of the hyperbolic space: - This limitation is selecting a particular data source or type of data to be manipulated as it is merely adding data to the training set, being insignificant extra-solution activity (see MEPEP n2106.05(g));
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception.
providing the relation, wherein the relation includes one parameter per hyperbolic space that defines the extent of rotating, and/or the extent of translating and/or a curvature of the hyperbolic space: - This limitation is analogous to electronic recordkeeping because it’s adding data to the training set and keeping record of it (see MPEP 2106.05(d) II (iii)).
Regarding Claim 9:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 1 which included an abstract idea (see rejection claim 1). Additionally, claim 9 recites the abstract ideas:
determining the set of representations of embeddings of entities of the knowledge graph, wherein the determining of the set of representations of embeddings includes splitting each respective embedding into a set of multi-dimensional vectors, and mapping the vectors of the respective set to the multi-dimensional vectors to the respective representation depending on the relation: - This limitation is directed to the abstract idea of a mental process, as the process splitting each respective embedding into a set of multi-dimensional vectors, and mapping the vectors of the respective set to the multi-dimensional vectors to the respective representation depending on the relation is a thought process that can be performed in a human mind by observing, evaluating and judging (concepts performed in the human mind (including an observation, evaluation, judgment, opinion) (see MPEP § 2106.04(a)(2), subsection III)).
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception.
Regarding Claim 10:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 1 which included an abstract idea (see rejection claim 1).
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
training a model: (i) for mapping the first entity to the first representation depending on the relation, (ii) for mapping the second entity to the second representation depending on the relation, (iii) for rotating and translating the first representation in the hyperbolic spaces depending on the relation, and (iv) for determining the second entity depending on the prediction for the second representation and depending on the second representation, or selecting the second entity with the model: - This limitation does not integrate a judicial exception into a practical application as the training of the model is recited at a high level of generality, therefore this amounts to mere instructions to implement an abstract idea 2106.05(f).
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception. The additional elements:
training a model: (i) for mapping the first entity to the first representation depending on the relation, (ii) for mapping the second entity to the second representation depending on the relation, (iii) for rotating and translating the first representation in the hyperbolic spaces depending on the relation, and (iv) for determining the second entity depending on the prediction for the second representation and depending on the second representation, or selecting the second entity with the model: - This limitation does not amount to significantly more than the judicial exception as the training of the model is recited at a high level of generality, therefore this amounts to mere instructions to implement an abstract idea 2106.05(f).
Regarding Claim 11:
Step 1 - Is the claim directed to a process, a machine, manufacture or composition of matter? - Yes, the claim is directed to a process.
Step 2A - Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon? - Yes, the claim is dependent on claim 1 which included an abstract idea (see rejection claim 1).
Step 2A - Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application? - No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
(i) determining a control signal depending on the link, the control signal being for controlling a computer-controlled machine, the computer-controlled machine including in particular a robotic system, or a vehicle, or a domestic appliance, or a power tool, or a manufacturing machine, or a personal assistant, or an access control system, or (ii) classifying sensor data depending on the link, the classifying being for: detecting the presence of objects in the sensor data, or (iii) performing a semantic segmentation on the sensor data depending on the link the semantic segmentation being regarding traffic signs, or road surfaces, or pedestrians, or vehicles, or (iv) analyzing scalar time series data from a sensor, depending on the link, or (v) determining a state of a technical system depending on the link: - This limitation does no more than generally link a judicial exception to a particular technological environment as this limitation recites the use of judicial exception to the technologies of control signal and sensor data (see MPEP 2106.05(h)).
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception? - No, there are no additional elements that amount to significantly more than the judicial exception. The additional elements:
(i) determining a control signal depending on the link, the control signal being for controlling a computer-controlled machine, the computer-controlled machine including in particular a robotic system, or a vehicle, or a domestic appliance, or a power tool, or a manufacturing machine, or a personal assistant, or an access control system, or (ii) classifying sensor data depending on the link, the classifying being for: detecting the presence of objects in the sensor data, or (iii) performing a semantic segmentation on the sensor data depending on the link the semantic segmentation being regarding traffic signs, or road surfaces, or pedestrians, or vehicles, or (iv) analyzing scalar time series data from a sensor, depending on the link, or (v) determining a state of a technical system depending on the link: - This limitation does no more than generally link a judicial exception to a particular technological environment as this limitation recites the use of judicial exception to the technology of control signal and sensor data (see MPEP 2106.05(h)).
Regarding independent Claim 12, this claim is directed to a device and is rejected on the same basis as independent claim 1 since they are analogous claims.
Regarding independent Claim 13, this claim is directed to a non-transitory computer-readable medium and is rejected on the same basis as independent claim 1 since they are analogous claims.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-10, 12-13 are rejected under 35 U.S.C. 103 as being unpatentable over Su et al. (CN-113779219A- hereinafter Su) in view of Li et al. (CN-115563314-A - hereinafter Li) and in further view of Zhao et al. (CN-110929047-A – hereinafter Zhao).
Referring to Claim 1, Su teaches:
determining the link including the first entity, the second entity, and the relation (see Su at Pg. 60-66: “The specific operation of segmented embedding is to embed the hyperbolic relationship between the hyperbolic head entity embedding and the hyperbolic tail entity embedding, and then divide them into odd and even segments. First, assuming the hyperbolic relation embedding is d-dimensional, the d-dimensional embedding of the hyperbolic relation embedding is uniformly divided into k segments, and the dimension of each segment is d/k. The hyperbolic relation embedding can then be represented as follows. Here, is the x-th segment of the hyperbolic relation embedding. If x is odd, is the odd-numbered segment; if x is even, is the even-numbered segment of the relation embedding. At this point, the multilinear dot product formula involving piecewise segments is expressed as follows. Where x, y, w represent the number of segments in hyperbolic relation embedding, hyperbolic head entity embedding, and hyperbolic tail entity embedding, respectively.
”. Examiner interprets x, y, w representing the number of segments in hyperbolic relation embedding, hyperbolic head entity embedding, and hyperbolic tail entity embedding, respectively to be equivalent as the claimed “determining the link including the first entity, the second entity, and the relation”);
wherein the first representation includes multi-dimensional vectors in hyperbolic spaces (see Su at Pg. 60-61: “The specific operation of segmented embedding is to embed the hyperbolic relationship between the hyperbolic head entity embedding and the hyperbolic tail entity embedding, and then divide them into odd and even segments.
First, assuming the hyperbolic relation embedding is d-dimensional, the d-dimensional embedding of the hyperbolic relation embedding is uniformly divided into k segments, and the dimension of each segment is d/k. The hyperbolic relation embedding can then be represented as follows”. Examiner interprets the d-dimensional embedding of the hyperbolic relation embedding being uniformly divided into k segments, and the dimension of each segment is d/k to be equivalent as the claimed “the first representation includes multi-dimensional vectors in hyperbolic spaces”);
wherein the determining of the first representation includes splitting the embedding of the first entity into a first set of multi-dimensional vectors, and mapping the multi-dimensional vectors of the first set to the multi-dimensional vectors of the first representation depending on the relation (see Su at Pg. 60-61: “The specific operation of segmented embedding is to embed the hyperbolic relationship between the hyperbolic head entity embedding and the hyperbolic tail entity embedding, and then divide them into odd and even segments. First, assuming the hyperbolic relation embedding is d-dimensional, the d-dimensional embedding of the hyperbolic relation embedding is uniformly divided into k segments, and the dimension of each segment is d/k. The hyperbolic relation embedding can then be represented as follows. Here, is the x-th segment of the hyperbolic relation embedding. If x is odd, is the odd-numbered segment; if x is even, is the even-numbered segment of the relation embedding. At this point, the multilinear dot product formula involving piecewise segments is expressed as follows. Where x, y, w represent the number of segments in hyperbolic relation embedding, hyperbolic head entity embedding, and hyperbolic tail entity embedding, respectively.
Secondly, the hyperbolic relation is embedded into odd-numbered and even-numbered segments to maintain the relation's symmetry and antisymmetric properties. The function sx,y is defined as follows:
represents the summation of sx,y, represents the hyperbolic head entity embedding of the y-th segment, and represents the hyperbolic tail entity embedding of the w-th segment; sx,y represents the positive or negative value of each multilinear dot product term. If is an even segment, sx,y is positive, and in this case, the even sum of in the function is equal to the even sum of in the corresponding function .
Therefore, triples can be modeled as symmetric relations by embedding even segments of hyperbolic relations; if the segments are odd, the sx,y function can be positive or negative, depending on whether x+y ≥ k. If the segments are odd and x+y ≥ k, then the sx,y function is negative. In this case, the sum of odd numbers in the function is not equal to the sum of odd numbers in the function. Therefore, the function supports the embedding of hyperbolic relations into antisymmetric relations of odd-numbered segments. Finally, wx,y is used to determine the position of the candidate answer. When x is even, wxy = y, and when x is odd, wx,y = (x+y)%k. This method reduces the number of parameters in the multilinear dot product from k3 to k2, making the time complexity O(kd) and the space complexity O(d), thus achieving the goal of reducing both time and space complexity”. Examiner interprets the d-dimensional embedding of the hyperbolic relation embedding being uniformly divided into k segments, and the dimension of each segment is d/k to be equivalent as the claimed “splitting the embedding of the first entity into a first set of multi-dimensional vectors”; further Examiner interprets wx,y being used to determine the position of the candidate answer, when x (x interpreted as the relation) is even, wxy = y, and when x is odd, wx,y = (x+y)%k to be equivalent as the claimed “mapping the multi-dimensional vectors of the first set to the multi-dimensional vectors of the first representation depending on the relation”);
wherein the determining of the prediction for the second representation includes rotating the multi-dimensional vectors in the hyperbolic spaces depending on the relation (see Su at Pg. 14-23: “Furthermore, in the training of the hyperbolic geometric embedding model, an m-dimensional Poincaré sphere model with negative curvature c is used to model the knowledge graph. The distance d(X,Y) between point X and point Y on the Poincaré sphere model is expressed by the hyperbolic space distance formula as follows. Among them, arccos h. Let denote the inverse hyperbolic cosine function, ||.‖ represents the L2 norm; Knowledge graphs are represented by triples (h, r, t), where h represents the head entity, t represents the tail entity, and r represents the relationship between the head and tail entities. (h, r, t) ∈ V × R × V, where V and R represent entity datasets. The head entity h is rotated and mapped using Rotation and Reflection parameters, as shown in the following formula. Where Rot represents rotation, Ref represents mapping; P represents the Poincaré model; represents the rotation value of the hyperbolic entity embedding of the Poincaré sphere model; represents the mapping value of the hyperbolic relation embedding of the Poincaré sphere model. Θr and Φr both represent relation-specific parameters; represents the hyperbolic head entity embedding. Then, using the hyperbolic attention mechanism, and are combined and applied to the hyperbolic transformation formula, as follows. Where Q(p,r) represents the query embedding, which is the value of the logical operation between the rotation and reflection logical encoding pattern of the hyperbolic head entity embedding and the hyperbolic relation embedding; Att represents the hyperbolic attention mechanism; represents the hyperbolic relation embedding, obtained through the hyperbolic embedding model, r∈R; a<sup>r</sup> represents the carrier of the hyperbolic attention mechanism related to the relation; is a logical operator that indicates that an XOR operation is performed first, and then the complement is taken. Finally, the query embedding and the hyperbolic tail entity embedding are compared using the hyperbolic spatial distance formula to obtain the scoring function s(p,r,t), as shown in the following formula”. Examiner interprets the combination of the following statements: an m-dimensional Poincaré sphere model (interpreted as multi-dimensional hyperbolic space), the head entity h is rotated and mapped using rotation parameters, Θr and Φr both represent relation-specific parameters (interpreted as relation dependence), obtaining the scoring function s(p,r,t) (interpreted as prediction) by comparing the query embedding and the hyperbolic tail entity embedding (interpreted as second representation); to be equivalent as the claimed “the determining of the prediction for the second representation includes rotating the multi-dimensional vectors in the hyperbolic spaces depending on the relation”).
However, Su fails to teach:
wherein the determining of the prediction for the second representation includes translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation;
Li teaches, in analogous system,
wherein the determining of the prediction for the second representation includes translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation (see Li at Pg. 38-44: “This invention captures the structural information of the knowledge graph itself using a low-dimensional approach, thereby reducing the embedding dimension of the final entities and relationships.
Since transforming the relationships between entities in hyperbolic space mapping leads to high complexity overhead, and this invention is only for mining the structural information of the graph itself, in order to reduce overhead, this invention regards the hyperbolic space mapping as a translation process and uses distance and translation calculation functions in hyperbolic space to represent it, namely the hyperbolic TransE method. The following section details the process of capturing structural information from a knowledge graph using the spatial properties of hyperbolic space. The semantic translation-based method initializes entities and relations in the knowledge graph using word embedding models, then calculates the distance and translation formulas in hyperbolic space, and continuously trains to obtain hyperbolic embedding triple vectors. During model training, the distance between the head entity's vector after relational translation and the tail entity is calculated. During training, there are positive and negative samples; positive samples represent fact triples that actually exist, and negative samples represent fact triples that do not exist. Negative samples are generally obtained by randomly replacing the head or tail entity. During training, it is desirable that the distance value of real fact triples should be less than that of non-existent fact triples. When the loss function is trained to a plateau, a hyperbolic embedding triple vector is obtained”. Examiner interprets the combination of the following passages: the translation calculation functions in hyperbolic space, the distance between the head entity's vector after relational translation and the tail entity is calculated, and the hyperbolic embedding triple vector is obtained when the loss function is trained, to be equivalent as the claimed “the determining of the prediction for the second representation includes translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Su with the above teachings of Li by rotating the multi-dimensional vectors in the hyperbolic spaces depending on the relation, as taught by Su, and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation, as taught by Li. The modification would have been obvious because one of ordinary skill in the art would be motivated to capture the structural information from knowledge graph in order to reduce overhead (as suggested by Li at Pg. 39: “Since transforming the relationships between entities in hyperbolic space mapping leads to high complexity overhead, and this invention is only for mining the structural information of the graph itself, in order to reduce overhead, this invention regards the hyperbolic space mapping as a translation process and uses distance and translation calculation functions in hyperbolic space to represent it, namely the hyperbolic TransE method”).
However, Su - Li fails to teach:
determining a first representation, wherein the first representation represents an embedding of the first entity;
selecting a second representation from a set of representations of embeddings of entities of the knowledge graph, wherein the second representation represents an embedding of the second entity, and wherein the selecting of the second representation includes determining a prediction for the second representation, and selecting the second representation depending on the prediction for the second representation;
Zhao teaches, in analogous system,
determining a first representation, wherein the first representation represents an embedding of the first entity (see Zhao at Pg. 51-52: ”Step 102: Obtain the initial embedding representations of entities and relations in the knowledge graph. Knowledge graphs contain a large amount of structured knowledge, which is usually represented in the form of triples. In a triple, there are: head entity, tail entity, and entity relation. Embedding refers to representing the information in a triple in the form of a vector, which makes it easier for the terminal to recognize and calculate. Commonly used embedding models include word vector models and bag-of-words models”. Examiner interprets transforming the head entity and tail entity to a high-dimensional space according to a pre-set transformation matrix to obtain a high-dimensional embedding representation of the head entity (interpreted as first entity) to be equivalent as the claimed “determining a first representation, wherein the first representation represents an embedding of the first entity”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Su and Li with the above teachings of Zhao by rotating and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation, as taught by Su and Li, and determining a first representation, wherein the first representation represents an embedding of the first entity, as taught by Zhao. The modification would have been obvious because one of ordinary skill in the art would be motivated to represent the information in a triple in the form of a vector, which makes it easier for the terminal to recognize and calculate (as suggested by Zhao at Pg. 52: “Knowledge graphs contain a large amount of structured knowledge, which is usually represented in the form of triples. In a triple, there are: head entity, tail entity, and entity relation. Embedding refers to representing the information in a triple in the form of a vector, which makes it easier for the terminal to recognize and calculate. Commonly used embedding models include word vector models and bag-of-words models”).
selecting a second representation from a set of representations of embeddings of entities of the knowledge graph, wherein the second representation represents an embedding of the second entity, and wherein the selecting of the second representation includes determining a prediction for the second representation, and selecting the second representation depending on the prediction for the second representation (see Zhao at Pg. 55-60: “This invention belongs to the third type of reasoning problem, namely, determining the tail entity in a triple. When constructing a neighbor subgraph, the head entity can be the center, and other tail entities can be the neighbor nodes of the center, with entity relationships as edges between entity pairs. Attention probability refers to the probability of obtaining neighboring nodes through reasoning, which can be understood as assigning weights to neighboring nodes. Step 106: Obtain the feature representation of each pair of triples in the neighbor subgraph and obtain the head entity embedding representation that integrates the neighbor node information based on the attention probability and the feature representation. The feature representation is obtained by learning from the initial embedding representation. Step 108: Concatenate the head entity embedding representation and the initial embedding representation corresponding to the head entity to obtain the final head entity embedding representation. Step 110: Based on the embedding representation, evaluate and rank each triplet, and infer the tail entity in the triplet based on the evaluation and ranking results”. Examiner interprets evaluating and ranking each triplet; and inferring the tail entity in the triplet based on evaluation and ranking results to be equivalent as the claimed “selecting of the second representation includes determining a prediction for the second representation” and “selecting the second representation depending on the prediction for the second representation” respectively).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Su and Li with the above teachings of Zhao by rotating and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation, as taught by Su and Li, and selecting a second representation from a set of representations of embeddings of entities of the knowledge graph, as taught by Zhao. The modification would have been obvious because one of ordinary skill in the art would be motivated to evaluate and rank each triplet then infer the tail entity in the triplet based on the evaluation and ranking results (as suggested by Zhao at Pg. 60: “Step 110: Based on the embedding representation, evaluate and rank each triplet, and infer the tail entity in the triplet based on the evaluation and ranking results”).
Referring to Claim 2, Su teaches the method of claim 1:
wherein the prediction for the second representation includes multi-dimensional vectors in the hyperbolic spaces, wherein the second representation includes multi-dimensional vectors in the hyperbolic spaces, and wherein the selecting of the second representation includes determining differences between the multi-dimensional vectors of the second representation and the multi-dimensional vectors of the prediction for the second representation that are in the same hyperbolic space, determining a distance depending on the differences, and selecting the second representation depending on the distance (see Su at Pg. 14-23: “Furthermore, in the training of the hyperbolic geometric embedding model, an m-dimensional Poincaré sphere model with negative curvature c is used to model the knowledge graph. The distance d(X,Y) between point X and point Y on the Poincaré sphere model is expressed by the hyperbolic space distance formula as follows. Among them, arccos h. Let denote the inverse hyperbolic cosine function, ‖.‖ represents the L2 norm. Knowledge graphs are represented by triples (h, r, t), where h represents the head entity, t represents the tail entity, and r represents the relationship between the head and tail entities. (h, r, t) ∈ V × R × V, where V and R represent entity datasets. The head entity h is rotated and mapped using Rotation and Reflection parameters, as shown in the following formula. Where Rot represents rotation, Ref represents mapping; P represents the Poincaré model; represents the rotation value of the hyperbolic entity embedding of the Poincaré sphere model; represents the mapping value of the hyperbolic relation embedding of the Poincaré sphere model; Θr and Φr both represent relation-specific parameters; represents the hyperbolic head entity embedding. Then, using the hyperbolic attention mechanism, and are combined and applied to the hyperbolic transformation formula, as follows. Where Q(p,r) represents the query embedding, which is the value of the logical operation between the rotation and reflection logical encoding pattern of the hyperbolic head entity embedding and the hyperbolic relation embedding; Att represents the hyperbolic attention mechanism; represents the hyperbolic relation embedding, obtained through the hyperbolic embedding model, r∈R; ar represents the carrier of the hyperbolic attention mechanism related to the relation; is a logical operator that indicates that an XOR operation is performed first, and then the complement is taken. Finally, the query embedding and the hyperbolic tail entity embedding are compared using the hyperbolic spatial distance formula to obtain the scoring function s(p,r,t), as shown in the following formula”. Examiner interprets obtaining the scoring function s(p,r,t) by comparing the query embedding and the hyperbolic tail entity embedding (interpreted as second representation) using the hyperbolic spatial distance formula and knowledge graph being modeled using m-dimensional Poincaré sphere model to be equivalent as the claimed “the prediction for the second representation includes multi-dimensional vectors in the hyperbolic spaces, wherein the second representation includes multi-dimensional vectors in the hyperbolic spaces, and wherein the selecting of the second representation includes determining differences between the multi-dimensional vectors of the second representation and the multi-dimensional vectors of the prediction for the second representation that are in the same hyperbolic space, determining a distance depending on the differences, and selecting the second representation depending on the distance”).
Referring to Claim 3, Li teaches the method of claim 2:
the determining the distance includes concatenating the multi-dimensional vectors of the prediction for the second representation, concatenating the multi-dimensional vectors of the second representation, and determining the distance depending on the concatenated multi-dimensional vectors of the prediction for the second representation and the concatenated multi-dimensional vectors of the second representation (see Li at Pg. 52-58: “This invention constructs a representation learning model based on semantic translation, which maps entities to hyperbolic space for knowledge graph embedding. It utilizes the curvature of hyperbolic space as -1 and its characteristic of capturing self-generated structural information to enhance the representational ability of the knowledge graph itself.
The text-enhanced triple vector and the hyperbolic embedding triple vector are concatenated, and the concatenated entity vectors are fused through a multi-layer neural network to obtain the fused entity relationship triple vector. As shown in Figure 2, the multilayer neural network in this embodiment includes a feedforward neural network and a transformation neural network. First, the text-enhanced triple vector and the hyperbolic embedding triple vector are concatenated. The concatenated triple vector is then used as the input vector and passed through a feedforward neural network to obtain a self-fused triple vector. This vector possesses semantic information describing the entity and features of a low-dimensional hyperbolic space that can capture its own structural information. The self-fused triplet vectors are then reconstructed from the previously spliced triplet vectors using a transformation neural network. Finally, the mean squared error between the reconstructed triplet vector and the initially concatenated triplet vector is calculated, and this mean squared error is used as the loss function to train and obtain the entity relation representation vector that integrates entity description text and hyperbolic space information”. Examiner interprets the concatenated triple vector obtained by concatenating the text-enhanced triple vector and the hyperbolic embedding triple vector to be equivalent as the claimed “concatenating the multi-dimensional vectors of the second representation”, further Examiner interprets the reconstructed self-fused triplet vectors from the previously spliced triplet vectors using the transformation neural network to be equivalent as the claimed “concatenating the multi-dimensional vectors of the prediction for the second representation” and further more Examiner interprets the calculation of the mean squared error between the reconstructed triplet vector and the initially concatenated triplet vector to be equivalent as the claimed “determining the distance depending on the concatenated multi-dimensional vectors of the prediction for the second representation and the concatenated multi-dimensional vectors of the second representation”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Su, Li and Zhao with the above teachings of Li by selecting a representation from a set of representations of embeddings of entities of the knowledge graph, rotating and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation, as taught by Su, Li and Zhao, and determining the distance depending on the concatenated multi-dimensional vectors of the prediction for the second representation and the concatenated multi-dimensional vectors of the second representation, as taught by Li. The modification would have been obvious because one of ordinary skill in the art would be motivated to calculate the mean squared error between the reconstructed triplet vector and the initially concatenated triplet vector, and this mean squared error is used as the loss function (as suggested by Li at Pg. 58: “Finally, the mean squared error between the reconstructed triplet vector and the initially concatenated triplet vector is calculated, and this mean squared error is used as the loss function to train and obtain the entity relation representation vector that integrates entity description text and hyperbolic space information”).
Referring to Claim 4, Su teaches the method of claim 1:
wherein the determining of the first representation includes mapping the vectors of the first set to different hyperbolic spaces (see Su at Pg. 46-49: “Knowledge graphs are generally represented by triples (h, r, t), where h represents the head entity, t represents the tail entity, r represents the relationship between the head and tail entities, and (h, r, t) ∈ V × R × V, where V and R represent entity datasets. First, the head entity h is rotated and mapped using the Rotation and Reflection parameters, as shown in the following formula. Where Rot represents rotation, Ref represents mapping; p represents the Poincaré model; represents the rotation value of the hyperbolic entity embedding of the Poincaré sphere model, and represents the mapping value of the hyperbolic relation embedding of the Poincaré sphere model; Θr and Φr both represent relation-specific parameters; represents the hyperbolic head entity embedding”. Examiner interprets the head entity h (interpreted as first set) being mapped to be equivalent as the claimed “mapping the vectors of the first set”, further Examiner interprets Θr and Φr both representing relation-specific parameters to be equivalent as the claimed “different hyperbolic spaces”).
Referring to Claim 5, Su teaches the method of claim 1:
wherein a curvature of at least one of the hyperbolic spaces is defined by the relation (see Su Pg. 40 - 42: “Hyperbolic geometry is a class of non-Euclidean geometries with constant negative curvature. This invention uses an m-dimensional Poincaré sphere model with negative curvature c to model the knowledge graph. The Poincaré sphere model formula is expressed as. Where Pm,c represents the numerical value of an m-dimensional Poincaré sphere model with negative curvature c, c<0; x represents a point on the Poincaré sphere model, represents the set of entities, and represents the m-dimensional entity vector space; ‖.‖ represents the L_NER11 norm”. Examiner interprets the m-dimensional Poincaré sphere model having negative curvature c to model the knowledge graph to be equivalent as the claimed “a curvature of at least one of the hyperbolic spaces is defined by the relation”).
Referring to Claim 6, Su teaches the method of claim 1:
wherein an extent of the rotation in at least one of the hyperbolic spaces is defined by the relation (see Su at Pg. 17-19: “Knowledge graphs are represented by triples (h, r, t), where h represents the head entity, t represents the tail entity, and r represents the relationship between the head and tail entities. (h, r, t) ∈ V × R × V, where V and R represent entity datasets. The head entity h is rotated and mapped using Rotation and Reflection parameters, as shown in the following formula. Where Rot represents rotation, Ref represents mapping; P represents the Poincaré model; represents the rotation value of the hyperbolic entity embedding of the Poincaré sphere model; represents the mapping value of the hyperbolic relation embedding of the Poincaré sphere model; Θr and Φr both represent relation-specific parameters; represents the hyperbolic head entity embedding”. Examiner interprets the head entity h being rotated using rotation parameters as well as Θr and Φr representing relation-specific parameters to be equivalent as the claimed “an extent of the rotation in at least one of the hyperbolic spaces is defined by the relation”).
Referring to Claim 7, Su teaches the method of claim 1:
wherein an extent of the translation in at least one of the hyperbolic spaces is defined by defined by the relation (see Li at Pg. 39-41: “Since transforming the relationships between entities in hyperbolic space mapping leads to high complexity overhead, and this invention is only for mining the structural information of the graph itself, in order to reduce overhead, this invention regards the hyperbolic space mapping as a translation process and uses distance and translation calculation functions in hyperbolic space to represent it, namely the hyperbolic TransE method. The following section details the process of capturing structural information from a knowledge graph using the spatial properties of hyperbolic space. The semantic translation-based method initializes entities and relations in the knowledge graph using word embedding models, then calculates the distance and translation formulas in hyperbolic space, and continuously trains to obtain hyperbolic embedding triple vectors. During model training, the distance between the head entity's vector after relational translation and the tail entity is calculated”. Examiner interprets the combination of the following passages: the hyperbolic space mapping being a translation process and uses distance and translation calculation functions in hyperbolic space as well as the distance between the head entity's vector after relational translation and the tail entity is calculated, to be equivalent as the claimed “an extent of the translation in at least one of the hyperbolic spaces is defined by defined by the relation”).
Referring to Claim 8, Su – Li teaches the method of claim 1:
providing the relation, wherein the relation includes one parameter per hyperbolic space that defines the extent of rotating, and/or the extent of translating and/or a curvature of the hyperbolic space (see Su at Pg. 17-19: “Knowledge graphs are represented by triples (h, r, t), where h represents the head entity, t represents the tail entity, and r represents the relationship between the head and tail entities. (h, r, t) ∈ V × R × V, where V and R represent entity datasets. The head entity h is rotated and mapped using Rotation and Reflection parameters, as shown in the following formula. Where Rot represents rotation, Ref represents mapping; P represents the Poincaré model; represents the rotation value of the hyperbolic entity embedding of the Poincaré sphere model; represents the mapping value of the hyperbolic relation embedding of the Poincaré sphere model; Θr and Φr both represent relation-specific parameters; represents the hyperbolic head entity embedding”. Examiner interprets the head entity h being rotated using rotation parameters as well as Θr and Φr representing relation-specific parameters to be equivalent as the claimed “the relation includes one parameter per hyperbolic space that defines the extent of rotating”) further (see Li at Pg. 39-41: “Since transforming the relationships between entities in hyperbolic space mapping leads to high complexity overhead, and this invention is only for mining the structural information of the graph itself, in order to reduce overhead, this invention regards the hyperbolic space mapping as a translation process and uses distance and translation calculation functions in hyperbolic space to represent it, namely the hyperbolic TransE method. The following section details the process of capturing structural information from a knowledge graph using the spatial properties of hyperbolic space. The semantic translation-based method initializes entities and relations in the knowledge graph using word embedding models, then calculates the distance and translation formulas in hyperbolic space, and continuously trains to obtain hyperbolic embedding triple vectors.
During model training, the distance between the head entity's vector after relational translation and the tail entity is calculated”. Examiner interpreted the following passages: the hyperbolic space mapping being a translation process and uses distance and translation calculation functions in hyperbolic space, as well as the distance between the head entity's vector after relational translation and the tail entity being calculated, to be equivalent as the claimed “the relation includes one parameter per hyperbolic space that defines the extent of translation”) further more (see Su Pg. 40 - 42: “Hyperbolic geometry is a class of non-Euclidean geometries with constant negative curvature. This invention uses an m-dimensional Poincaré sphere model with negative curvature c to model the knowledge graph. The Poincaré sphere model formula is expressed as. Where Pm,c represents the numerical value of an m-dimensional Poincaré sphere model with negative curvature c, c<0; x represents a point on the Poincaré sphere model, represents the set of entities, and represents the m-dimensional entity vector space; ||.‖ represents the L_NER11 norm”. Examiner interprets the m-dimensional Poincaré sphere model having negative curvature c to model the knowledge graph, to be equivalent as the claimed “the relation includes one parameter per hyperbolic space that defines curvature of the hyperbolic space”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Su, Li and Zhao with the above teachings of Li by selecting a representation from a set of representations of embeddings of entities of the knowledge graph, rotating and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation, as taught by Su, Li and Zhao, and the relation including one parameter per hyperbolic space that defines the extent of translation, as taught by Li. The modification would have been obvious because one of ordinary skill in the art would be motivated to capture the structural information in knowledge graph (as suggested by Li at Pg. 39: “Since transforming the relationships between entities in hyperbolic space mapping leads to high complexity overhead, and this invention is only for mining the structural information of the graph itself, in order to reduce overhead, this invention regards the hyperbolic space mapping as a translation process and uses distance and translation calculation functions in hyperbolic space to represent it, namely the hyperbolic TransE method”).
Referring to Claim 9, Su teaches the method of claim 1:
determining the set of representations of embeddings of entities of the knowledge graph, wherein the determining of the set of representations of embeddings includes splitting each respective embedding into a set of multi-dimensional vectors, and mapping the vectors of the respective set to the multi-dimensional vectors to the respective representation depending on the relation (see Su at Pg. 60-61: “The specific operation of segmented embedding is to embed the hyperbolic relationship between the hyperbolic head entity embedding and the hyperbolic tail entity embedding, and then divide them into odd and even segments. First, assuming the hyperbolic relation embedding is d-dimensional, the d-dimensional embedding of the hyperbolic relation embedding is uniformly divided into k segments, and the dimension of each segment is d/k. The hyperbolic relation embedding can then be represented as follows. Here, is the x-th segment of the hyperbolic relation embedding. If x is odd, is the odd-numbered segment; if x is even, is the even-numbered segment of the relation embedding. At this point, the multilinear dot product formula involving piecewise segments is expressed as follows. Where x, y, w represent the number of segments in hyperbolic relation embedding, hyperbolic head entity embedding, and hyperbolic tail entity embedding, respectively. Secondly, the hyperbolic relation is embedded into odd-numbered and even-numbered segments to maintain the relation's symmetry and antisymmetric properties. The function sx,y is defined as follows: represents the summation of sx,y, represents the hyperbolic head entity embedding of the y-th segment, and represents the hyperbolic tail entity embedding of the w-th segment; sx,y represents the positive or negative value of each multilinear dot product term. If is an even segment, sx,y is positive, and in this case, the even sum of in the function is equal to the even sum of in the corresponding function. Therefore, triples can be modeled as symmetric relations by embedding even segments of hyperbolic relations; if the segments are odd, the sx,y function can be positive or negative, depending on whether x+y ≥ k. If the segments are odd and x+y ≥ k, then the sx,y function is negative. In this case, the sum of odd numbers in the function is not equal to the sum of odd numbers in the function. Therefore, the function supports the embedding of hyperbolic relations into antisymmetric relations of odd-numbered segments. Finally, wx,y is used to determine the position of the candidate answer. When x is even, wxy = y, and when x is odd, wx,y = (x+y)%k. This method reduces the number of parameters in the multilinear dot product from k3 to k2, making the time complexity O(kd) and the space complexity O(d), thus achieving the goal of reducing both time and space complexity”. Examiner interprets the specific operation of segmented embedding being to embed the hyperbolic relationship between the hyperbolic head entity embedding and the hyperbolic tail entity embedding, and then divide them into odd and even segments, to be equivalent as the claimed “determining the set of representations of embeddings of entities of the knowledge graph”; the passage: the d-dimensional embedding of the hyperbolic relation embedding is uniformly divided into k segments, and the dimension of each segment is d/k is interpreted to be equivalent as the claimed “the determining of the set of representations of embeddings includes splitting each respective embedding into a set of multi-dimensional vectors”; further Examiner interprets wx,y being used to determine the position of the candidate answer, when x (x interpreted as the relation) is even, wxy = y, and when x is odd, wx,y = (x+y)%k to be equivalent as the claimed “mapping the vectors of the respective set to the multi-dimensional vectors to the respective representation depending on the relation”).
Referring to Claim 10, Su teaches the method of claim 1:
training a model: (i) for mapping the first entity to the first representation depending on the relation, (ii) for mapping the second entity to the second representation depending on the relation, (iii) for rotating and translating the first representation in the hyperbolic spaces depending on the relation, and (iv) for determining the second entity depending on the prediction for the second representation and depending on the second representation, or selecting the second entity with the model (see Zhao at Pg. 55-57: “This invention belongs to the third type of reasoning problem, namely, determining the tail entity in a triple. When constructing a neighbor subgraph, the head entity can be the center, and other tail entities can be the neighbor nodes of the center, with entity relationships as edges between entity pairs. Attention probability refers to the probability of obtaining neighboring nodes through reasoning, which can be understood as assigning weights to neighboring nodes. Step 106: Obtain the feature representation of each pair of triples in the neighbor subgraph, and obtain the head entity embedding representation that integrates the neighbor node information based on the attention probability and the feature representation. The feature representation is obtained by learning from the initial embedding representation. Step 108: Concatenate the head entity embedding representation and the initial embedding representation corresponding to the head entity to obtain the final head entity embedding representation. Step 110: Based on the embedding representation, evaluate and rank each triplet, and infer the tail entity in the triplet based on the evaluation and ranking results”. Examiner interprets entity relationships being used as edges between entity pairs and obtaining the feature representation of each pair of triples in the neighbor subgraph, as well as obtaining the head entity (interpreted as first entity) embedding representation that integrates the neighbor node information to be equivalent as the claimed “mapping the first entity to the first representation depending on the relation”; Examiner interprets inferring the tail entity (second entity) in the triplet based on the evaluation and ranking results related to the final head entity embedding representation to be equivalent as the claimed “mapping the second entity to the second representation depending on the relation”) further (see Su at Pg. 14 – 19: “Furthermore, in the training of the hyperbolic geometric embedding model, an m-dimensional Poincaré sphere model with negative curvature c is used to model the knowledge graph. The distance d(X,Y) between point X and point Y on the Poincaré sphere model is expressed by the hyperbolic space distance formula as follows: Among them, arccos h. Let denote the inverse hyperbolic cosine function, ‖.‖ represents the L2 norm. Knowledge graphs are represented by triples (h, r, t), where h represents the head entity, t represents the tail entity, and r represents the relationship between the head and tail entities. (h, r, t) ∈ V × R × V, where V and R represent entity datasets. The head entity h is rotated and mapped using Rotation and Reflection parameters, as shown in the following formula. Where Rot represents rotation, Ref represents mapping; P represents the Poincaré model; represents the rotation value of the hyperbolic entity embedding of the Poincaré sphere model; represents the mapping value of the hyperbolic relation embedding of the Poincaré sphere model; Θr and Φr both represent relation-specific parameters; represents the hyperbolic head entity embedding”. Examiner interprets m-dimensional Poincaré sphere model with negative curvature c being used to model the knowledge graph and the head entity h being rotated and mapped using rotation parameters as well as Θr and Φr representing relation-specific parameters to be equivalent as the claimed “rotating the first representation in the hyperbolic spaces depending on the relation”; further (see Li at Pg. 47-48: “The final score function is: S(E)=d(Vhh ⊕ Vh r, Vh t). Where Vhh, Vh t and Vh r represent the vector representations of the head entity, tail entity and relation in hyperbolic space, respectively”. Examiner interprets translating the representation of the first entity Vhh by the relationship vector Vhr within the hyperbolic space before calculating the distance to the second entity Vht to be equivalent as the claimed “translating the first representation in the hyperbolic spaces depending on the relation”; further (see Li at Pg. 42-43: “During training, there are positive and negative samples; positive samples represent fact triples that actually exist, and negative samples represent fact triples that do not exist. Negative samples are generally obtained by randomly replacing the head or tail entity. During training, it is desirable that the distance value of real fact triples should be less than that of non-existent fact triples”. Examiner interprets making the distance value of real fact triples less than that of non-existent fact triples as well as negative samples being generally obtained by randomly replacing the head or tail entity to be equivalent as the claimed “determining the second entity depending on the prediction for the second representation and depending on the second representation, or selecting the second entity with the model”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Su, Li and Zhao with the above teachings of Li by selecting a representation from a set of representations of embeddings of entities of the knowledge graph, rotating and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation, as taught by Su, Li and Zhao, and translating the first representation in the hyperbolic spaces depending on the relation, as taught by Li. The modification would have been obvious because one of ordinary skill in the art would be motivated to calculate the loss function (as suggested by Li at Pg. 47-48: “The final score function is: S(E)=d(Vhh ⊕ Vh r, Vh t). Where Vhh, Vh t and Vh r represent the vector representations of the head entity, tail entity and relation in hyperbolic space, respectively”).
Referring to independent Claims 12 and 13, these claims are rejected on the same basis as independent claim 1 since they are analogous claims.
Claims 11 is rejected under 35 U.S.C. 103 as being unpatentable over Su et al. (CN-113779219A- hereinafter Su) in view of Li et al. (CN-115563314-A - hereinafter Li) and in further view of Zhao et al. (CN-110929047-A – hereinafter Zhao), and in further view of Stetson et al. (US-11507099-B2 – hereinafter Stetson). and in further view of Baldini et al. (US-20220303291-A1 – hereinafter Baldini).
Referring to Claim 11, Su – Li – Zhao teaches the method of claim 1:
However, Su – Li – Zhao fails to teach:
(i) determining a control signal depending on the link, the control signal being for controlling a computer-controlled machine, the computer-controlled machine including in particular a robotic system, or a vehicle, or a domestic appliance, or a power tool, or a manufacturing machine, or a personal assistant, or an access control system, or (ii) classifying sensor data depending on the link, the classifying being for: detecting the presence of objects in the sensor data, or (iii) performing a semantic segmentation on the sensor data depending on the link the semantic segmentation being regarding traffic signs, or road surfaces, or pedestrians, or vehicles;
Stetson teaches, in analogous system,
(i) determining a control signal depending on the link, the control signal being for controlling a computer-controlled machine, the computer-controlled machine including in particular a robotic system, or a vehicle, or a domestic appliance, or a power tool, or a manufacturing machine, or a personal assistant, or an access control system, or (ii) classifying sensor data depending on the link, the classifying being for: detecting the presence of objects in the sensor data, or (iii) performing a semantic segmentation on the sensor data depending on the link the semantic segmentation being regarding traffic signs, or road surfaces, or pedestrians, or vehicles (see Stetson at Cl. 3 Ln. 18-39 and Cl. 4 Ln. 33-34: “For example, an autonomous vehicle has to contend with everything from rocks on the road and snowflakes on the sensors, to GPS failures, to brute force attacks on its communication channels. In many situations, it would be difficult or impossible to store all such possible sources of risk in a separate column in a table. Moreover, there may not be a convex hull that one can define such that points inside are easier for the AI to handle than points outside. Instead, every single traffic case can be mapped onto a manifold. Here, a manifold represents the entire set of situations an AI is expected to handle. As in a space, nearby points in the manifold encode similar situations. After operating at a particular point, the AI can move or be moved to another point in the manifold either close to or far away, in many embodiments influenced by the output of the last operation. In a variety of embodiments, a similarity graph can act as a manifold. In numerous embodiments, the manifold itself does not contain all the underlying data for each individual situation, but the underlying data can be encoded in an outside data structure (e.g. a knowledge graph) and linked to the manifold, which itself can also be stored in the knowledge graph. In a further additional embodiment, the AI model is used in an autonomous vehicle control system”. Examiner interprets the AI being moved to another point in the manifold either close to or far away influenced by the output of the last operation; and the underlying data can be encoded in an outside data structure (e.g. a knowledge graph) and linked to the manifold; as well as the AI model being used in an autonomous vehicle control system to be equivalent as the claimed “(i) determining a control signal depending on the link, the control signal being for controlling a computer-controlled machine, the computer-controlled machine including in particular a robotic system”) further (see Stetson at Cl. 3 Ln. 50-58: “In another embodiment, to encode the set of training data into the knowledge graph, the graph interface application directs the processor to identify objects within each scenario in the plurality of scenarios, store the identified objects within the first knowledge graph, determine spatiotemporal features for each object, store the spatiotemporal features within the first knowledge graph, and generate spatiotemporal scenarios by clustering the first knowledge graph based on the spatiotemporal features. In another embodiment again, the training data comprises forward-mounted monocular camera footage and textual accident reports”. Examiner interprets identifying objects within each scenario in the plurality of scenarios, storing the identified objects within the first knowledge graph, determining spatiotemporal features for each object, storing the spatiotemporal features within the first knowledge graph, and generating spatiotemporal scenarios by clustering the first knowledge graph based on the spatiotemporal features as well as the training data being forward-mounted monocular camera footage to be equivalent as the claimed “(ii) classifying sensor data depending on the link, the classifying being for: detecting the presence of objects in the sensor data”) further (see Stetson at Cl. 3 Ln. 50-58: “In a further embodiment, the graph interface application further directs the processor to add semantic meaning to spatiotemporal features”. Examiner interprets an autonomous vehicle having to contend with everything from rocks on the road, snowflakes on the sensors, GPS failures, brute force attacks on its communication channels; and data being encoded in knowledge graph and linked to the manifold (see Stetson at Cl. 3 Ln. 18-39) as well as the processor adding semantic meaning to spatiotemporal features to be equivalent as the claimed “(iii) performing a semantic segmentation on the sensor data depending on the link the semantic segmentation being regarding traffic signs, or road surfaces, or pedestrians, or vehicles”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Su, Li and Zhao with the above teachings of Stetson by selecting a representation from a set of representations of embeddings of entities of the knowledge graph, rotating and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation, as taught by Su, Li and Zhao, and determining a control signal depending on the link and classifying sensor data depending on the link as well as performing a semantic segmentation on the sensor data depending on the link, as taught by Stetson. The modification would have been obvious because one of ordinary skill in the art would be motivated to map every single traffic case to a manifold, where a manifold represents the entire set of situations an AI is expected to handle (as suggested by Stetson at Cl. 3 Ln. 18-39: “For example, an autonomous vehicle has to contend with everything from rocks on the road and snowflakes on the sensors, to GPS failures, to brute force attacks on its communication channels. In many situations, it would be difficult or impossible to store all such possible sources of risk in a separate column in a table. Moreover, there may not be a convex hull that one can define such that points inside are easier for the AI to handle than points outside. Instead, every single traffic case can be mapped onto a manifold. Here, a manifold represents the entire set of situations an AI is expected to handle. As in a space, nearby points in the manifold encode similar situations. After operating at a particular point, the AI can move or be moved to another point in the manifold either close to or far away, in many embodiments influenced by the output of the last operation. In a variety of embodiments, a similarity graph can act as a manifold. In numerous embodiments, the manifold itself does not contain all the underlying data for each individual situation, but the underlying data can be encoded in an outside data structure (e.g. a knowledge graph) and linked to the manifold, which itself can also be stored in the knowledge graph”).
However, Su – Li – Zhao fails to teach:
(iv) analyzing scalar time series data from a sensor, depending on the link, or (v) determining a state of a technical system depending on the link;
Baldini teaches, in analogous system,
(iv) analyzing scalar time series data from a sensor, depending on the link, or (v) determining a state of a technical system depending on the link (see Baldini at Pg. 63-64: “In one example, the IoT devices are computing devices that perform tasks such as temperature analyzer or image capture that generally have a low-resource demand. The anomaly detection system detects an anomaly in the IoT environment. The computing infrastructure includes computing resources to store a knowledge graph, a set of IoT models, and any necessary information for the data retrieval controller to access data from the IoT devices. The computing infrastructure may be a cloud computing environment or an on-premise set of computing resources such as one having a local server. The data retrieval controller, according to some embodiments of the present invention, relies on four modules: (i) device metrics analyzer; (ii) anomaly detection analyzer; (iii) data retrieval adjustment manager; and (iv) knowledge graph manager. The device metrics analyzer determines if the IoT devices are performing according to a pre-defined service level agreement (SLA) while capturing data for the anomaly detection system. The anomaly detection analyzer monitors the function of anomaly detection services or systems to determine if the frequency of data collection, amount of data collected, and/or types of data collected meets the demand of the anomaly detection systems. The data retrieval adjustment manager operates to adjust the frequency of data collection, amount of data collected, and/or types of data collected according to configuration updates driven by supervised machine learning techniques and reinforcement learning techniques that determine how such configurations should be updated. The knowledge graph manager facilitates reuse of the learned data collection configurations for various IoT devices and/or data collection contexts”. Examiner interprets the IoT devices being computing devices that perform tasks such as temperature analyzer; and the determination of the frequency of data collection; as well as the knowledge graph manager facilitating reuse of the learned data collection configurations; to be equivalent as the claimed “(iv) analyzing scalar time series data from a sensor, depending on the link”) further (see Baldini at Pg. 42: “Some embodiments of the present invention recognize the following facts, potential problems and/or potential areas for improvement with respect to the current state of the art: (i) early detection of IoT device anomalies is an important goal for the industry; (ii) conventional anomaly detection systems ensure proper detection of anomalies by using machine learning techniques that depend on large amounts of high-quality data from the IoT devices; (iii) improper retrieval of data for anomaly detection can negatively affect an IoT device and its environment; (iv) IoT devices have limitations in terms of data transfer bandwidth/latency, battery capability, and computing power; and/or (v) when too much data is retrieved from an IoT device for anomaly detection, the functionality and/or lifespan of the IoT device can be reduced”. Examiner interprets the data retrieval controller relying on knowledge graph manager (see Baldini at Pg. 63-64) as well as recognizing potential areas for improvement with respect to the current state of the art such as early detection of IoT device anomalies to be equivalent as the claimed “(v) determining a state of a technical system depending on the link”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Su, Li and Zhao with the above teachings of Baldini by selecting a representation from a set of representations of embeddings of entities of the knowledge graph, rotating and translating the multi-dimensional vectors in the hyperbolic spaces depending on the relation, as taught by Su, Li and Zhao, and analyzing scalar time series data from a sensor, depending on the link, and determining a state of a technical system depending on the link, as taught by Baldini. The modification would have been obvious because one of ordinary skill in the art would be motivated to adjust the frequency of data collection according to configuration updates driven by supervised machine learning techniques (as suggested by Baldini at Pg. 64: “The device metrics analyzer determines if the IoT devices are performing according to a pre-defined service level agreement (SLA) while capturing data for the anomaly detection system. The anomaly detection analyzer monitors the function of anomaly detection services or systems to determine if the frequency of data collection, amount of data collected, and/or types of data collected meets the demand of the anomaly detection systems. The data retrieval adjustment manager operates to adjust the frequency of data collection, amount of data collected, and/or types of data collected according to configuration updates driven by supervised machine learning techniques and reinforcement learning techniques that determine how such configurations should be updated. The knowledge graph manager facilitates reuse of the learned data collection configurations for various IoT devices and/or data collection contexts”).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AWADAGBE G HOUNTON whose telephone number is (571)270-0670. The examiner can normally be reached Monday-Friday 8am-5pm.
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.
/AWADAGBE G HOUNTON/Examiner, Art Unit 2126
/DAVID YI/Supervisory Patent Examiner, Art Unit 2126
Prosecution Insights