DETAILED ACTION
This office action is in response to amendments filed on 02/20/2026.
Claims 1-10 and 12-16 have been canceled. Claims 11 and 17-20 have been amended. Claims 21-23 have been added. Claims 11 and 17-23 are pending.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
Objection to the Specification:
In light of applicant’s amendment to the title (pg. 2), the objection to the specification has been withdrawn.
Rejections Under 35 USC § 112:
In light of the cancellation of claims 1-10 (pg. 3), the interpretation under 35 USC § 112(f) and the resulting rejections under 35 USC § 112(a) and 112(b) have been withdrawn.
Prior Art Rejections:
Applicant's arguments regarding the prior art rejections (pg. 10-12) have been fully considered but they are not persuasive. Applicant argues that the cited references fail to disclose features of the amended claim 11, which incorporates subject matter from the original claim 16.
Applicant specifically argues that Vargas teaches a single model for recommending users to an item instead of recommending items to a user, while in amended claim 11, two different models are used: one for learning user embeddings which are close to item embeddings with a positive purchase history (i.e. recommending items to a user), and one for learning item embeddings which are close to user embeddings with a positive purchase history (i.e. recommending users to an item). Examiner respectfully notes that Vargas is not relied upon to teach the two reinforcement learning models. Rather, the two-model structure is taught by Liu: “We have two policy networks of the same structure for users and items, respectively” (Liu, pg. 2310, section 2.3.3), and the recommendation of items to users is taught by at least Wang: “Finally, we conduct the inner product to estimate the user’s preference towards the target item” (Wang, Pg. 4, section 2.3). The teachings of Vargas simply would have motivated one of ordinary skill in the art to additionally recommend users to items: “the inversion approach results in a significant increase of sales diversity” (Vargas, Pg. 145, section 1), and can be straightforwardly applied to other recommender models such as those suggested by Lai, Wang, and Liu: “we propose, firstly, to explore the application of state of the art collaborative filtering algorithms to the inverted recommendation task, that is, simply swapping the role of items and users in the algorithms” (Vargas, pg. 145, section 1).
Applicant additionally argues that Vargas fails to disclose the technical features of claim 11, such as the graph neural network. Examiner respectfully notes that Vargas is not relied upon to teach these technical features. As can be seen in the rejection below, the graph neural network is taught by Lai, and additional technical features are also taught by Wang and Liu.
Applicant finally argues that Vargas’s kNN model is structurally incompatible with the Markov decision process used by Lai. Examiner respectfully notes that the test for obviousness is not whether the features of a secondary reference may be bodily incorporated into the structure of the primary reference; nor is it that the claimed invention must be expressly suggested in any one or all of the references. Rather, the test is what the combined teachings of the references would have suggested to those of ordinary skill in the art. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981). In this case, as explained above, the teachings of Vargas would have motivated one of ordinary skill in the art to invert the relationship between users and items in the model suggested by Lai, Wang, and Liu, in order to increase sales diversity.
The prior art rejections have been updated to include the amended limitations and to clarify the reasoning given for the limitations that were not amended.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 11 and 17-23 are rejected under 35 U.S.C. 103 as being unpatentable over
Lai et al. (hereinafter Lai), “Policy-GNN: Aggregation Optimization for Graph Neural Networks” in view of
Wang et al. (hereinafter Wang), “Neural Graph Collaborative Filtering”,
Liu et al. (hereinafter Liu), “Automated Embedding Size Search in Deep Recommender Systems”, and
Vargas et al. (hereinafter Vargas), “Improving Sales Diversity by Recommending Users to Items”.
Regarding Claim 11,
Lai teaches A method of determining a number of layers, performed by an apparatus for determining a number of layers, the method comprising: (Pg. 10, Appendix A: “To develop the Policy-GNN, we implement two main components, GNN module and meta-policy module using Python as the programming language.” The use of Python programming necessitates implementation of the described method on a computer (i.e. an apparatus).)
obtaining a graph structure including information between nodes; (Pg. 6, section 4.1: “We consider the benchmark datasets [35] that are commonly used for studying node classification, including the citation graph-structured data of Cora, Citeseer, and Pubmed… In these citation graphs, the nodes and the edges correspond to the published documents and their citation relations, respectively… We follow the standard training/validation/test splits as in previous studies [8, 9, 37, 45] Specifically, we use 20 nodes per class for training, 500 nodes for validation, and 1000 nodes for testing.” A citation graph is a graph structure, and edges representing citation relations are information between nodes.)
controlling a reinforcement learning model designed to set the graph structure as an environment of reinforcement learning, set any one of nodes included in the graph structure as a state of reinforcement learning, and set a number of branches causing a highest expected value for a reward of reinforcement learning from any node as an action of reinforcement learning; (Pg. 4, section 3.1: “As discussed in Section 2.3, the key components of an MDP include states, actions and rewards, as well as a transition probability that maps the current state and action into the next state. With the above in mind, we now discuss how we define these components in the context of graph representation learning… the proposed aggregation process consists of three phases: 1) selecting a start node and obtaining its attributes as the current state
s
t
, 2) generating an action
a
t
from
π
(
s
t
)
to specify the number of hops for the current node, and…” Pg. 3, section 2.3: “We aim to search for the optimal decisions so as to maximize the expected discounted cumulative reward, i.e., we would like find a policy
π
:
S
→
A
to maximize
E
π
[
∑
t
=
0
∞
γ
t
r
t
]
.” Reinforcement learning is performed on the graph (i.e. the graph structure is the environment), with a node corresponding to the state and the node’s number of hops (i.e. branches) corresponding to the action, which is determined by policy
π
to cause the highest expected reward value.)
storing the number of branches determined by the reinforcement learning model as an action for a predetermined node input to the reinforcement learning model; and (Pg. 5, algorithm 1, step 6: “Store
s
t
and
a
t
into GNN buffer
B
a
t
” The GNN buffer stores the state
s
t
representing a node, and the associated action
a
t
representing its number of branches.)
applying the same number of layers for extracting an embedding of the predetermined node as the stored number of branches in a graph neural network designed based on the graph structure, (Pg. 5, section 3.2: “As Figure 3 shows, the Meta-Policy maps the node attributes to the number of hops, and uses the number of hops to construct a specific GNN architecture for each node… Instead of aggregating a fixed number of layers for every nodes, our framework aggregates
a
t
layers for each node attribute
s
t
at timestep
t
.” A node’s embedding is extracted by aggregating (i.e. applying) the determined number of hops (i.e. layers).)
wherein the reinforcement learning model includes:
a [first] reinforcement learning model designed to set the graph structure as an environment of reinforcement learning, set any one of the user nodes as a state of reinforcement learning, and set a number of branches causing a highest expected value for a reward of reinforcement learning from any user node as an action of reinforcement learning; and (Pg. 4, section 3.1: “As discussed in Section 2.3, the key components of an MDP include states, actions and rewards, as well as a transition probability that maps the current state and action into the next state. With the above in mind, we now discuss how we define these components in the context of graph representation learning… the proposed aggregation process consists of three phases: 1) selecting a start node and obtaining its attributes as the current state
s
t
, 2) generating an action
a
t
from
π
(
s
t
)
to specify the number of hops for the current node, and…” Pg. 3, section 2.3: “We aim to search for the optimal decisions so as to maximize the expected discounted cumulative reward, i.e., we would like find a policy
π
:
S
→
A
to maximize
E
π
[
∑
t
=
0
∞
γ
t
r
t
]
.” Reinforcement learning is performed on the graph (i.e. the graph structure is the environment), which includes user nodes as taught by Wang, with a node corresponding to the state and the node’s number of hops (i.e. branches) corresponding to the action, which is determined by policy
π
to cause the highest expected reward value.)
a [second] reinforcement learning model designed to set the graph structure as an environment of reinforcement learning, set any one of the item nodes as a state of reinforcement learning, and set a number of branches causing a highest expected value for a reward of reinforcement learning from any item node as an action of reinforcement learning, (Examiner notes that this limitation is identical to the previous limitation except that it operates on item nodes rather than user nodes. The reinforcement learning model taught by Lai, as shown above, operates on all nodes of the graph (i.e. both user and item nodes).)
wherein the storing comprises:
storing a [first] tuple list including an identifier corresponding to each of first to (n+1)-th user nodes based on the user nodes and the number of branches corresponding to each of the first to (n+1)-th user nodes in a case where the first reinforcement learning model extracts the number of (n+1)-th branches for the (n+1)-th user node reached from an n-th user node through branching by the number of n-th branches in the graph structure based on the n-th user node (n being a natural number) input to the first reinforcement learning model and the number of n-th branches determined by the first reinforcement learning model for the n-th user node; and (Pg. 5, algorithm 1, step 6: “Store
s
t
and
a
t
into GNN buffer
B
a
t
” Pg. 5, section 3.3: “In each timestep, after we obtain a batch of nodes’ attributes and a batch of actions, we store them into the action buffer and check if the buffer has reached the batch size.” The GNN buffer (i.e. storage) stores a batch (i.e. list) of tuples including state
s
t
(i.e. an identifier corresponding to each node) and action
a
t
(i.e. the number of branches corresponding to each node). Pg. 4, figure 3: “In each timestep, the meta-policy samples the number of layers (action) based on the attributes of the nodes (state). The next nodes (next state) are obtained by randomly sampling a node from the k-hop neighbors of the current nodes, where k is the output of the meta-policy (action).” A node (i.e. the (n+1)-th node) is selected as the next state (i.e. reached) by sampling from the current node’s k-hop neighbors (i.e. branching from the n-th node by the n-th number of branches).)
storing a [second] tuple list including an identifier corresponding to each of first to (n+1)-th item nodes based on the item nodes and the number of branches corresponding to each of the first to (n+1)-th item nodes in a case where the second reinforcement learning model extracts the number of (n+1)-th branches for the (n+1)-th item node reached from an n-th item node through branching by the number of n-th branches in the graph structure based on the n-th item node (n being a natural number) input to the second reinforcement learning model and the number of n-th branches determined by the second reinforcement learning model for the n-th item node, (Examiner notes that this limitation is identical to the previous limitation except that it includes item nodes rather than user nodes. The reinforcement learning model and tuple list taught by Lai, as shown above, includes all nodes of the graph (i.e. both user and item nodes).)
wherein the controlling comprises:
setting a [first] reward of the [first] reinforcement learning model based on information on item nodes included in the [second] tuple list; and (Pg. 4, figure 3: “The meta-policy is updated based on the feedback from the GNN.” Pg. 4, section 3.1: “The key factor in guiding the deep Q-learning is the reward signal. We employ a baseline in the reward function, defined as
R
s
t
,
a
t
=
λ
(
M
s
t
,
a
t
-
∑
i
=
t
-
b
t
-
1
M
s
i
,
a
i
b
-
1
)
,” where
M
s
t
,
a
t
represents the model performance on the state-action tuple list at timestep
t
(i.e. the reward is based on node information in the tuple list). The tuple list includes all nodes of the graph (i.e. both user and item nodes).)
setting a [second] reward of the [second] reinforcement learning model based on information on user nodes included in the [first] tuple list, (Examiner notes that this limitation is identical to the previous limitation except that the reward setting is based on user nodes rather than item nodes. The reward setting taught by Lai, as shown above, is based on all nodes of the graph (i.e. both user and item nodes).)
Lai does not appear to explicitly disclose the remaining claim 11 limitations.
However, Wang teaches wherein the graph structure includes user nodes and item nodes, and the information between nodes includes purchase information indicating whether or not users of user nodes purchase items of item nodes, (Pg. 2, figure 1: “An illustration of the user-item interaction graph and the high-order connectivity,” where nodes labeled
u
represent users and nodes labeled
i
represent items. Pg. 1, section 1: “At its core is estimating how likely a user will adopt an item based on the historical interactions like purchases and clicks.”)
wherein the [first] reward applied to a first input node of the [first] reinforcement learning model includes a function for assigning a higher score as an embedding distance between a positive node, which has purchase information regarding the first input node among the item nodes included in the [second] tuple list, and the first input node is closer, and a function for deducting a higher score as an embedding distance between a negative node, which has no purchase information regarding the first input node among the item nodes included in the [second] tuple list, and the first input node is closer, and (Pg. 4, section 2.3: “Finally, we conduct the inner product to estimate the user’s preference towards the target item:
y
^
N
G
C
F
u
,
i
=
e
u
*
T
e
i
*
.” Pg. 4, section 2.4: “Specifically, BPR assumes that the observed interactions, which are more reflective of a user’s preferences, should be assigned higher prediction values than unobserved ones. The objective function is as follows,
L
o
s
s
=
∑
(
u
,
i
,
j
)
∈
O
-
ln
σ
y
^
u
i
-
y
^
u
j
+
λ
Θ
2
2
, where
O
=
{
(
u
,
i
,
j
)
|
(
u
,
i
)
∈
R
+
,
(
u
,
j
)
∈
R
-
}
denotes the pairwise training data,
R
+
indicates the observed interactions, and
R
-
is the unobserved interactions…”
y
^
u
i
is the inner product of embeddings for a positive pair, which will contribute a higher score when the embeddings are closer together.
y
^
u
j
is the inner product of embeddings for a negative pair, which will deduct a higher score when the embeddings are closer together. The resulting loss is used as the reward for reinforcement learning, as taught by Lai: “The meta-policy is updated based on the feedback from the GNN” (Lai, pg. 4, figure 3).)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lai and Wang. Lai teaches adapting the number of aggregations performed on each node of a GNN based on a meta-policy trained via reinforcement learning. Wang teaches a recommendation framework based on GNN which learns node embeddings to model user-item affinity. One of ordinary skill would have motivation to combine Lai and Wang because Lai’s “proposed algorithm significantly outperforms all the state-of-the-art alternatives that we are aware of” (Lai, pg. 2, section 1) on graph representation learning tasks, which are “pervasive in many real-world applications, such as… recommendation systems” (Lai, pg. 1, section 1), and Wang’s recommendation system “demonstrate[ed] significant improvements over several state-of-the-art models” (Wang, pg. 1, abstract).
Lai and Wang do not appear to explicitly disclose a first reinforcement learning model for user nodes and second reinforcement learning model for item nodes.
However, Liu teaches a first reinforcement learning model for user nodes and second reinforcement learning model for item nodes. (Pg. 2310, section 2.3: “The policy network serves as an RL agent that dynamically adjusts the embedding sizes of users and items. It maintains the current embedding sizes of all the users and items. Given the state of a user or an item, it takes an action to adjust the embedding size following a policy. It interacts with the environment by feeding the adjusted embeddings into the recommendation model and receives the corresponding reward based on the prediction… We have two policy networks of the same structure for users and items, respectively.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lai, Wang, and Liu. Lai teaches adapting the number of aggregations performed on each node of a GNN based on a meta-policy trained via reinforcement learning. Wang teaches a recommendation framework based on GNN which learns node embeddings to model user-item affinity. Liu teaches a recommendation system which adapts user and item embedding sizes using two reinforcement learning models for users and items, respectively. One of ordinary skill would have motivation to combine Lai, Wang, and Liu because Lai’s strategy of adapting the number of aggregations in a graph is applicable to “many real-world applications, such as… recommendation systems” (Lai, pg. 1, section 1), and according to Liu, in “real-world recommendation problems where the frequencies of users and items are highly varied and change dynamically” (Liu, pg. 2315, section 5), adapting the generation of user and item embeddings using a separate RL model for each “outperforms representative baselines” (Liu, pg. 2307, abstract).
Lai, Wang, and Liu do not appear to explicitly disclose the remaining claim 11 limitations.
However, Vargas teaches wherein the [second] reward applied to a second input node of the [second] reinforcement learning model includes a function for assigning a higher score as an embedding distance between a positive node, which has purchase information regarding the second input node among the user nodes included in the [first] tuple list, and the second input node is closer, and a function for deducting a higher score as an embedding distance between a negative node, which has no purchase information regarding the second input node among the user nodes included in the [first] tuple list, and the second input node is closer. (Examiner notes that this limitation is identical to the previous limitation taught by Wang, except that it compares the embedding of a selected item node to positive and negative user nodes, rather than comparing the embedding of a selected user node to positive and negative item nodes. Pg. 145, section 1: “we propose, firstly, to explore the application of state of the art collaborative filtering algorithms to the inverted recommendation task, that is, simply swapping the role of items and users in the algorithms.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lai, Wang, Liu, and Vargas. Lai teaches adapting the number of aggregations performed on each node of a GNN based on a meta-policy trained via reinforcement learning. Wang teaches a recommendation framework based on GNN which learns node embeddings to model user-item affinity. Liu teaches a recommendation system which adapts user and item embedding sizes using two reinforcement learning models for users and items, respectively. Vargas teaches inverting a recommendation task by selecting which users an item should be recommended to rather than selecting which items to recommend to a user. One of ordinary skill would have motivation to combine Lai, Wang, Liu, and Vargas because “the inversion approach results in a significant increase of sales diversity while retaining a good trade-off on top-N item recommendation precision” in recommendation systems (Pg. 145, section 1).
Regarding Claim 17, Lai, Wang, Liu, and Vargas teach The method of claim 11, as shown above.
Wang also teaches wherein the controlling comprises:
setting, as the [first] reward, a difference between an inner product of an embedding vector of the first input node and an embedding vector of a positive node in the [second] tuple list and an inner product of the embedding vector of the first input node and an embedding vector of a negative node in the [second] tuple list; and (Pg. 4, section 2.3: “Finally, we conduct the inner product to estimate the user’s preference towards the target item:
y
^
N
G
C
F
u
,
i
=
e
u
*
T
e
i
*
.” Pg. 4, section 2.4: “Specifically, BPR assumes that the observed interactions, which are more reflective of a user’s preferences, should be assigned higher prediction values than unobserved ones. The objective function is as follows,
L
o
s
s
=
∑
(
u
,
i
,
j
)
∈
O
-
ln
σ
y
^
u
i
-
y
^
u
j
+
λ
Θ
2
2
, where
O
=
{
(
u
,
i
,
j
)
|
(
u
,
i
)
∈
R
+
,
(
u
,
j
)
∈
R
-
}
denotes the pairwise training data,
R
+
indicates the observed interactions, and
R
-
is the unobserved interactions…”
y
^
u
i
is the inner product of embeddings for a positive pair,
y
^
u
j
is the inner product of embeddings for a negative pair, and the normalized difference between the inner products is the loss, which is used as the reward for reinforcement learning, as taught by Lai: “The meta-policy is updated based on the feedback from the GNN” (Lai, pg. 4, figure 3).)
Vargas teaches setting, as the [second] reward, a difference between an inner product of an embedding vector of the second input node and an embedding vector of a positive node in the [first] tuple list and an inner product of the embedding vector of the second input node and an embedding vector of a negative node in the [first] tuple list. (Examiner notes that this limitation is identical to the previous limitation except that it computes the inner product between the embeddings of a selected item node and positive and negative user nodes, rather than computing the inner product between the embeddings of a selected user node and positive and negative item nodes. Pg. 145, section 1: “we propose, firstly, to explore the application of state of the art collaborative filtering algorithms to the inverted recommendation task, that is, simply swapping the role of items and users in the algorithms.”)
Liu teaches a separate first and second reinforcement learning pipeline for user nodes and item nodes, respectively. (Pg. 2310, section 2.3: “The policy network serves as an RL agent that dynamically adjusts the embedding sizes of users and items. It maintains the current embedding sizes of all the users and items. Given the state of a user or an item, it takes an action to adjust the embedding size following a policy. It interacts with the environment by feeding the adjusted embeddings into the recommendation model and receives the corresponding reward based on the prediction… We have two policy networks of the same structure for users and items, respectively.”)
Regarding Claim 18, Lai, Wang, Liu, and Vargas teach The method of claim 11, as shown above.
Wang also teaches wherein the controlling comprises:
setting the [first] reward by sampling the same number of positive nodes and negative nodes in the [second] tuple list; and setting the [second] reward by sampling the same number of positive nodes and negative nodes in the [first] tuple list. (Pg. 6, section 4.1: “For each observed user-item interaction, we treat it as a positive instance, and then conduct the negative sampling strategy to pair it with one negative item that the user did not consume before.” Each positive user-item pair has a corresponding negative pair, i.e., there are the same number of positive and negative nodes sampled.)
Liu teaches a separate first and second reinforcement learning pipeline for user nodes and item nodes, respectively. (Pg. 2310, section 2.3: “The policy network serves as an RL agent that dynamically adjusts the embedding sizes of users and items. It maintains the current embedding sizes of all the users and items. Given the state of a user or an item, it takes an action to adjust the embedding size following a policy. It interacts with the environment by feeding the adjusted embeddings into the recommendation model and receives the corresponding reward based on the prediction… We have two policy networks of the same structure for users and items, respectively.”)
Regarding Claim 19, Lai, Wang, Liu, and Vargas teach The method of claim 11, as shown above.
Lai also teaches wherein the applying comprises training the graph neural network based on the first tuple list or the second tuple list by setting the same number of layers of the graph neural network as the number of k-th branches when a k-th node (k being a natural number equal to or greater than 1 and equal to or less than n+1) among tuples included in the first tuple list or the second tuple list is used for learning, and using an embedding that has finally passed the number of layers from the k-th node as a final embedding of the k-th node. (Pg. 5, section 3.2: “As Figure 3 shows, the Meta-Policy maps the node attributes to the number of hops, and uses the number of hops to construct a specific GNN architecture for each node… Instead of aggregating a fixed number of layers for every nodes, our framework aggregates
a
t
layers for each node attribute
s
t
at timestep
t
.” Pg. 5, section 3.3: “If the buffer of a specific action is full, then we construct the GNN with selected layers, and train the GNN with the data in the buffer of the action.” For an element (i.e. k-th node) in the GNN buffer (i.e. tuple list), the GNN is constructed with the associated number of hops (i.e. number of k-th branches) and trained.)
Claims 20-23 are product claims containing substantially the same elements as method claims 11 and 17-19. Lai, Wang, Liu, and Vargas teach the elements of claims 11 and 17-19, as shown above.
Lai also teaches A non-transitory computer-readable recording medium storing computer executable instructions, wherein the instructions, when executed by a processor, cause the processor to perform a method of determining a number of layers (Examiner notes that this limitation is interpreted as implementation of the disclosed method in a computing environment. Pg. 10, Appendix A: “To develop the Policy-GNN, we implement two main components, GNN module and meta-policy module using Python as the programming language.” The use of Python programming necessitates implementation on a computer.)
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BENJAMIN M ROHD whose telephone number is (571)272-6445. The examiner can normally be reached Mon-Thurs 8:00-6:00 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, Viker Lamardo can be reached at (571) 270-5871. 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.
/B.M.R./Examiner, Art Unit 2147 /VIKER A LAMARDO/Supervisory Patent Examiner, Art Unit 2147
Prosecution Insights