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 .
Detailed Action
This action is in response to the claims filed 5/1/2023:
Claims 1 – 20 are pending.
Claims 1, 11, and 20 are independent.
Claim Rejections - 35 USC § 101
101 Rejection
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 1-6, 8, 9, 11-16, 18, 19, and 20 are rejected under 35 USC § 101 because the claimed invention is directed to non-statutory subject matter.
Regarding Claim 1: Claim 1 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 1 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: Claim 1 under its broadest reasonable interpretation is a series of mental processes. For example, but for the generic computer components language, the above limitations in the context of this claim encompass machine learning processing, including the following:
detecting and actioning user intent in natural language requests (observation, evaluation, and judgement)
identifying a candidate predicate based on the request (observation, evaluation, and judgement),
concatenating features derived from the subgraph with pretrained word embeddings to yield a set of request inputs and a set of predicate inputs; (observation, evaluation, and judgement)
calculating a matching score for the request and candidate predicate using a trained machine learning model based on the set of request inputs and the set of predicate inputs (observation, evaluation, and judgement)
selecting a matching predicate comprising user intent based on the matching score (observation, evaluation, and judgement)
performing an action to effectuate the user intent (observation, evaluation, and judgement)
Therefore, claim 1 recites an abstract idea which is a judicial exception.
Step 2A Prong Two Analysis: Claim 1 recites additional elements “a digital assistant system”. However, these additional features are computer components recited at a high-level of generality, such that they amount to no more than mere instructions to apply the judicial exception using a generic computer component. An additional element that merely recites the words “apply it” (or an equivalent) with the judicial exception, or merely includes instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea, does not integrate the judicial exception into a practical application (See MPEP 2106.05(f)). Claim 1 also recites additional elements “receiving a request from a user”, “retrieving a subgraph from a knowledge base based on the request”, and “outputting a response to the user” which amounts to insignificant extra-solution activity of gathering and outputting data which does not integrate the judicial exception into a practical application (See MPEP 2106.05(g)). Therefore, claim 1 is directed to a judicial exception.
Step 2B Analysis: Claim 1 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the lack of integration of the abstract idea into a practical application, the additional elements recited in claim 1 amount to no more than mere instructions to apply the judicial exception using a generic computer component and insignificant extra-solution activity. The gathering and outputting of data is considered well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)).
For the reasons above, claim 1 is rejected as being directed to non-patentable subject matter under §101. This rejection applies equally to independent claims 11 and 20, which recite a system and a computer program product, respectively, as well as to dependent claims 2-6, 8, 9, 12-16, 18, and 19.
Independent claim 11 recites additional instructions to apply the judicial exception using generic computer components “A digital assistant system comprising a processor and computer executable instructions, that when executed by the processor, cause the system to perform operations comprising”.
Independent claim 20 recites additional instructions to apply the judicial exception using generic computer components “A computer storage medium comprising executable instructions that, when executed by a processor of a machine, cause the machine to perform operations of”.
The additional limitations of the dependent claims are addressed briefly below:
Dependent claims 2 and 12 recite additional instructions to apply the judicial exception using generic computer components “wherein the trained machine learning model comprises a first trained bi-directional LSTM neural network and a second trained bi-directional LSTM network”
Dependent claims 3 and 13 recite additional instructions to apply the judicial exception using generic computer components “the trained machine learning model comprises a trained bi-directional matching LSTM neural network
Dependent claims 4 and 14 recite additional instructions to apply the judicial exception using generic computer components “wherein the trained machine learning model further comprises a first trained bi-directional LSTM network utilizing the set of request inputs and a second trained bi-directional LSTM network utilizing the set of predicate inputs”
Dependent claims 5 and 15 recite additional observation, evaluation, and judgement “wherein the set of request inputs comprises word embedding based on the request concatenated with a subset of the features derived from the subgraph”.
Dependent claims 6 and 16 recite additional observation, evaluation, and judgement “the set of predicate inputs comprises word embedding based on the candidate predicate concatenated with a subset of the features derived from the subgraph”
Dependent claims 8 and 18 recite additional instructions to apply the judicial exception using generic computer components “the trained machine learning model comprises a sigmoid layer” (a Bi-LSTM has sigmoid activation by definition)
Dependent claims 9 and 19 also recite additional observation, evaluation, and judgement “wherein the pretrained word embeddings for a first intent domain also apply to a second intent domain without retraining”
Therefore, when considering the elements separately and in combination, they do not add significantly more to the inventive concept. Accordingly, claims 1-6, 8, 9, 11-16, 18, 19, and 20 are rejected under 35 U.S.C. § 101.
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 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.
Claims 1-4, 6-9, 11-14, and 16-20 are rejected under U.S.C. §103 as being unpatentable over the combination of Yu (“Improved Neural Relation Detection for Knowledge Base Question Answering”, 2017) and Li (US20170109355A1).
PNG
media_image1.png
678
1238
media_image1.png
Greyscale
FIG. 2 of Yu
Regarding claim 1, Yu teaches A method of a digital assistant system detecting and actioning user intent in natural language requests, comprising: receiving a request [from a user];([Abstract] "we propose a hierarchical recurrent neural network enhanced by residual learning which detects KB relations given an input question")
identifying a candidate predicate based on the request;([p. 572] "We first identify the topic entity with entity linking and then detect the relation asked by the question with relation detection (from all relations connecting the topic entity)." Relation interpreted as synonymous with candidate predicate)
retrieving a subgraph from a knowledge base based on the request;([p. 577] "Our method can be viewed as entity linking on a KB sub-graph. It contains two steps: (1) Sub-graph generation: given the top scored query generated by the previous 3 steps 5, for each node v (answer node or the CVT node like in Figure 1(b)), we collect all the nodes c connecting to v (with relation rc) with any relation, and generate a sub-graph associated to the original query")
concatenating features derived from the subgraph ([p. 574] "We transform each token above to its word embedding then use two BiLSTMs (with shared parameters) to get their hidden representations [...] (each row vector i is the concatenation between forward/backward representations at i)." [p. 576] “Use the raw question text as input for a relation detector to score all relations in the KB that are associated to the entities in ELK(q); use the relation scores to re-rank ELK(q) and generate a shorter list EL0 K0(q) […] For each question q, after generating a score srel(r;q) for each relation using HR-BiLSTM, we use the top l best scoring relations (Rl q) to re-rank the original entity candidates. Concretely, for each entity e and its associated relations Re, given the original entity linker score”) with pretrained word embeddings ([p. 577] "All word vectors are initialized with 300-d pretrained word embeddings" Yu’s subgraph provides the candidates and then the model derives features by encoding the labels of those candidates)
to yield a set of request inputs ([p. 574] "The first-layer of BiLSTM works on the word embeddings of question words q ={q1,··· ,qN} and gets hidden representations Γ(1) 1:N =[...] The second-layer BiLSTM Γ(1) 1:N to get the second set of hidden representations Γ(2) 1:N. Since the second BiLSTM starts with the hidden vectors from the first layer, intuitively it could learn more general and abstract information compared to the first layer" set Γ interpreted as a set of request inputs)
and a set of predicate inputs;([p. 574] "their hidden representations [Bword 1:M1 : Brel 1:M2 ] (each row vector i is the concatenation between forward/backward representations at i)." Hidden representation vector B interpreted as set of predicate inputs)
calculating a matching score for the request and candidate predicate using a trained machine learning model based on the set of request inputs and the set of predicate inputs;([p. 577] "we compute a matching score between each n-gram in the input question (without overlapping the topic entity) and entity name of c (except for the node in the original query) by taking into account the maximum overlapping sequence of characters between them (see Appendix A for details and B for special rules dealing with date/answer type constraints). If the matching score is larger than a threshold (tuned on training set), we will add the constraint entity c (and rc) to the query by attaching it to the corresponding node v on the core-chain" Yu explicitly produces hq and hr and computes a cosine similarity between hq and hr which are explicitly output from the BiLSTM which generates hidden representation request inputs Γ and predicate inputs B (see FIG. 2 cosine similarity))
selecting a matching predicate comprising user intent based on the matching score;([p. 572] "Finally the highest scored query from the above steps is used to query the KB for answers" Yu matching score is result of cosine similarity between output representations of Bi-LSTM network, the scores are explicitly ranked and the highest ranked score is used to select the matching predicate comprising user intent.).
However, Yu does not explicitly teach receiving a request from a user;
performing an action to effectuate the user intent; and
outputting a response to the user.
Li, in the same field of endeavor, teaches receiving a request from a user;([¶0002] "The present disclosure relates generally to computing technologies, and more specifically to systems and methods for automating the answering of questions raised in natural language and improving human computer interfacing." [¶0046] "an input query having one or more words is received" [¶0096] "the input query may include a human inspired question")
performing an action to effectuate the user intent; and([¶0066] "a structured query is generated and sent to a KG server. Then, the KG server executes the structure query to obtain the object, i.e., answer to the question")
outputting a response to the user.([¶0092] " answer rendering module 412 that outputs and presents the results").
Yu as well as Li are directed towards using LSTM for natural language queries. Therefore, Yu as well as Li are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Yu with the teachings of Li by using the system and user interface of Li to implement the BiLSTM system in Yu. Li provides as additional motivation for combination ([p. 2] “The present disclosure relates generally to computing technologies, and more specifically to systems and methods for automating the answering of questions raised in natural language and improving human computer interfacing.”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 2, the combination of Yu, and Li teaches The method of claim 1 wherein the trained machine learning model comprises a first trained bi-directional LSTM neural network and a second trained bi-directional LSTM network.(Yu [p.575] "during training we adopt a ranking loss to maximizing the margin between the gold relation r+ and other relations r- in the candidate pool R" See also FIG. 2 which explicitly shows Bi-LSTM 1 and Bi-LSTM 2).
Regarding claim 3, the combination of Yu, and Li teaches The method of claim 1 wherein the trained machine learning model comprises a trained bi-directional matching LSTM neural network.(Yu [p.575] "during training we adopt a ranking loss to maximizing the margin between the gold relation r+ and other relations r- in the candidate pool R" [p. 5 §4.3] "Unlike the standard usage of deep BiLSTMs that employs the representations in the final layer for prediction, here we expect that two layers of question representations can be complementary to each other and both should be compared to the relation representation space (Hierarchical matching)" See also FIG. 2).
Regarding claim 4, the combination of Yu, and Li teaches The method of claim 3 wherein the trained machine learning model further comprises a first trained bi-directional LSTM network utilizing the set of request inputs (Yu [p. 574] "The first-layer of BiLSTM works on the word embeddings of question words q ={q1,··· ,qN} and gets hidden representations Γ(1) 1:N =[...] The second-layer BiLSTM Γ(1) 1:N to get the second set of hidden representations Γ(2) 1:N. Since the second BiLSTM starts with the hidden vectors from the first layer, intuitively it could learn more general and abstract information compared to the first layer" set Γ interpreted as a set of request inputs utilized by first bi-LSTM (bi-LSTM 2) network (see FIG. 2))
and a second trained bi-directional LSTM network utilizing the set of predicate inputs.(Yu [p. 574] "their hidden representations [Bword 1:M1 : Brel 1:M2 ] (each row vector i is the concatenation between forward/backward representations at i)." Hidden representation vector B interpreted as set of predicate inputs utilized by second Bi-LSTM network (Bi-LSTM 1 in FIG. 2). (Note that the second bi-LSTM utilizes both sets of inputs)).
Regarding claim 6, the combination of Yu, and Li teaches The method of claim 1 wherein the set of predicate inputs comprises word embedding based on the candidate predicate concatenated with a subset of the features derived from the subgraph.(Yu [p. 577] "Our method can be viewed as entity linking on a KB sub-graph. It contains two steps: (1) Sub-graph generation: given the top scored query generated by the previous 3 steps 5, for each node v (answer node or the CVT node like in Figure 1(b)), we collect all the nodes c connecting to v (with relation rc) with any relation, and generate a sub-graph associated to the original query" [p. 4] "We transform each token above to its word embedding then use two BiLSTMs (with shared parameters) to get their hidden representations [...] (each row vector i is the concatenation between forward/backward representations at i)." Yu explicitly constructs predicate candidates as relations/ relation chains connected to the topic entity (i.e. an entity-neighborhood slice of the KB graph) and feed those into the Bi-LSTM for scoring/matching).
Regarding claim 7, the combination of Yu, and Li teaches The method of claim 1 wherein the trained machine learning model comprises a self-attention layer.(Yu [p. 576] "Another way of hierarchical matching consists in relying on attention mechanism, e.g. (Parikh et al., 2016), to find the correspondence between different levels of representations" [p. 8] "replacing residual with attention […] For the attention-based baseline, we tried the model from (Parikh et al., 2016) and its one-way variations").
Regarding claim 8, the combination of Yu, and Li teaches The method of claim 1 wherein the trained machine learning model comprises a sigmoid layer.(Yu [Abstract] "Our method uses deep residual bidirectional LSTMs" LSTMs use sigmoid activations in gates by definition.).
Regarding claim 9, the combination of Yu, and Li teaches The method of claim 1 wherein the pretrained word embeddings for a first intent domain also apply to a second intent domain without retraining.(Yu [p. 571] "relation detection for KBQA often becomes a zero-shot learning task" [p. 575] "We initialize the relation sequence LSTMs with the final state representations of the word sequence, as a back-off for unseen relations" Unseen relations interpreted as second intent domain. Yu explicitly uses pretrained word embeddings for targeting zero shot/unseen labels.).
Regarding claims 11-14 and 16-19, claims 11-14 and 16-19 are directed towards a system for performing the methods of claims 1-4 and 6-9, respectively. Therefore, the rejections applied to claims 1-4 and 5-9 also apply to claims 11-14 and 16-19. Claim 11 also recites additional elements a processor and computer executable instructions, that when executed by the processor, cause the system to perform operations comprising: (Li [¶0102] “As illustrated in FIG. 12, system 600 includes one or more central processing units (CPU) 601 that provides computing resources and controls the computer. CPU 601 may be implemented with a microprocessor or the like, and may also include one or more graphics processing units (GPU) 617 and/or a floating point coprocessor for mathematical computations. System 600 may also include a system memory 602, which may be in the form of random-access memory (RAM), read-only memory (ROM), or both”).
Regarding claim 20, claim 20 is directed towards a computer readable media for performing the method of claim 1. Therefore, the rejection applied to claim 1 also applies to claim 20. Claim 20 also recites additional elements A computer storage medium comprising executable instructions that, when executed by a processor of a machine, cause the machine to perform operations of a digital assistant system, the operations comprising: (Li [¶0102] “As illustrated in FIG. 12, system 600 includes one or more central processing units (CPU) 601 that provides computing resources and controls the computer. CPU 601 may be implemented with a microprocessor or the like, and may also include one or more graphics processing units (GPU) 617 and/or a floating point coprocessor for mathematical computations. System 600 may also include a system memory 602, which may be in the form of random-access memory (RAM), read-only memory (ROM), or both”).
Claims 5, 10, and 15 are rejected under U.S.C. §103 as being unpatentable over the combination of Yu and Li and in further view of Sun (“Open Domain Question Answering Using Early Fusion of Knowledge Bases and Text”, 2018).
Regarding claim 5, the combination of Yu, and Li teaches The method of claim 1.
However, the combination of Yu, and Li doesn't explicitly teach wherein the set of request inputs comprises word embedding based on the request
concatenated with a subset of the features derived from the subgraph..
Sun, in the same field of endeavor, teaches the set of request inputs comprises word embedding based on the request ([p. 4] "To represent q, let wq1,…wq|q| be the words in the question. The initial representation is computed as [See Eqn. 4]" That is a request derived embedding produced from the question's words)
concatenated with a subset of the features derived from the subgraph.([p. 4] "The update for entity nodes involves a single-layer feed-forward nework (FFN) over the concatenation of four states [See Eqn. 1] The first two terms correspond to the entity representation and question representation (details below), respectively, from the previous layer" Sun explicitly teaches concatenating a request-derived representation with features derived from the question subgraph).
The combination of Yu, and Li as well as Sun are directed towards using LSTMs for natural language questions over knowledge bases/graphs. Therefore, the combination of Yu, and Li as well as Sun are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of the combination of Yu, and Li with the teachings of Sun by using Yu’s hierarchical matching alongside or as a substitution for Sun’s GRAFT-Net scoring method. Sun provides as additional motivation for combination ([p. 7 §5.3] “GRAFT-Nets (GN) shows consistent improvement over KV-MemNNs on both datasets in all settings, including KB only (-KB), text only (-EF, Text Only column), and early fusion (-EF).”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 10, the combination of Yu, and Li teaches The method of claim 1 wherein retrieving a subgraph from a knowledge base based on the request comprises: detecting an entity in the request;(Yu [p. 572] "We first identify the topic entity with entity linking")
retrieving the subgraph from the knowledge base based on the entity;(Yu [p. 572] "from all relations connecting the topic
entity" where Yu explicitly discloses that all the relations connecting the topic entity are a subgraph [p. 7]. See also FIG. 1)
deriving the features from the subgraph [using a convolutional neural network.] (Yu [p. 575] "We compute the matching score of r given q as srel(r;q)=cos(hr,hq)" Yu explicitly derives subgraph-based matching features in the next step ("Entity-linking on sub-graph nodes"), computing match scores between question n-grams and names of entities in the generated subgraph).
However, the combination of Yu, and Li doesn't explicitly teach deriving the features from the subgraph using a convolutional neural network.
Sun, in the same field of endeavor, teaches deriving the features from the subgraph using a convolutional neural network. ([p. 1] "To enable early fusion, in this paper we propose a novel graph convolution based neural network, called GRAFT-Net (Graphs of Relations Among Facts and Text Networks), specifically designed to operate over heterogeneous graphs of KB facts and text sentences").
The combination of Yu, and Li as well as Sun are directed towards using LSTMs for natural language questions over knowledge bases/graphs. Therefore, the combination of Yu, and Li as well as Sun are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of the combination of Yu, and Li with the teachings of Sun by using Yu’s hierarchical matching alongside or as a substitution for Sun’s GRAFT-Net scoring method. Sun provides as additional motivation for combination ([p. 7 §5.3] “GRAFT-Nets (GN) shows consistent improvement over KV-MemNNs on both datasets in all settings, including KB only (-KB), text only (-EF, Text Only column), and early fusion (-EF).”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 15, claim 5 is directed towards a system for performing the method of claim 5. Therefore, the rejection applied to claim 5 also applies to claim 15.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Zhang (“An Attention-Based Word-Level Interaction Model: Relation Detection for Knowledge Base Question Answering”, 2018) is directed towards a CNN and Bi-LSTM hybrid model for natural language processing.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SIDNEY VINCENT BOSTWICK whose telephone number is (571)272-4720. The examiner can normally be reached M-F 7:30am-5:00pm EST.
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, Miranda Huang can be reached on (571)270-7092. 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.
/SIDNEY VINCENT BOSTWICK/Examiner, Art Unit 2124
/VINCENT GONZALES/Primary Examiner, Art Unit 2124