DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA . This action is responsive to pending claims 1, 4-11, 14-30 filed 7/14/2026.
Allowable Subject Matter
Claim 15 is allowable, but objected to for dependency on a rejected claim.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
Claim(s) 29-30 are rejected under 35 U.S.C. 112(b) as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, or for pre-AIA the applicant regards as the invention.
Claim 29 recites a system for “performing or facilitating performing” of the computer-implemented method.
First, a reader would not know what is claimed by such a system in “performing”. Are all the limitations of claim 1 necessarily performed, or is it merely that the computer processor configuration is enabled to or capable of performing step 1?
Second, a reader would not know what it means to “facilitate performance”; for example, does facilitate performance mere merely to aid in performing (e.g., data preparation, booting up, etc.) or does every step of claim 1 have to be performed?
Hence, the scope of the claim is unclear and the claim is indefinite.
Claim 30 recites a method for operating the neural network trained using method 1, comprising processing a query and document using the model to determine a result. A reader would not be sure whether the scope is a method necessarily positively incorporates all the limitations of claim 1, or would be broad enough to include any method suitable for operating a neural network trained in the manner of claim 1. It is unclear or ill-defined how a method for operating the system would be limited by the limitations of the system of claim 1.
Hence, the scope of the claim is unclear and the claim is indefinite.
Claim Rejections - 35 USC § 101
The 101 rejections are withdrawn.
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.
Claim(s) 1, 4-11, 30 are rejected under 35 U.S.C. 103 as being unpatentable over Gao ("Smoothing clickthrough data for web search ranking", published 2009) in view of Yu ("Data augmentation based on adversarial autoencoder handling imbalance for learning to rank", published 2019) with Chawla ("SMOTE: synthetic minority over-sampling technique", published 2002) and Qin ("Introducing LETOR 4.0 Datasets", published 2013) incorporated by reference.
For claim 1, Gao discloses: a computer-implemented method for training a neural network based ranking model (§3 ¶1-3 gives problem overview along with two smoothing techniques, with §4.1 disclosing computer-implemented experimental trials on big data) in order to improve a tail query performance of the neural network based ranking model when solving a long-tailed learning-to-rank (LTR) problem (§1 ¶3 gives overview of data imbalance such as incomplete click and missing click problem, such as resulting in the imbalanced distribution of fig.4, hence, relevance judgment click data is unevenly distributed in training data including many samples with zero or fewer hits), comprising:
performing a training data augmentation operation on a set of training data to generate a set of synthesized training data (§3.1 gives overview of clickthrough smoothing technique for generating clickthrough data for queries with fewer data, while §3.2 contemplates discounting in order to generate click probability for non-clicked documents, hence, training data is augmented, yielding the improved distribution shown in fig.4), wherein the data augmentation comprises; and
training a neural network based ranking model using the set of training data and the set of synthesized training data (§2.3 ¶2-3 contemplates training neural net ranker LambdaRank);
wherein the set of training data comprises, for each of a plurality of queries, respective query-document data and respective relevance judgement data (figs.1-2, §2.1 ¶1-3 gives overview of training data including query-document data and relevance data comprising user click features), the query-document data for a query comprises data associated with a plurality of query-documents pairs for the query (ibid: text data associated with query document pairs, see also §4.1 ¶1-4, contemplating additional query-document feature data, e.g., human-labeled relevance data), and the relevance judgement data for a query comprises one or more sets of user feedback data associated with the query (ibid: clickthrough data, se also §4.1 ¶3-4 contemplating clickthrough feature data),
wherein the set of training data has an imbalanced training data distribution such that amounts of relevance judgement data available for at least some the plurality of queries are different (§1 ¶3 gives overview of data imbalance such as incomplete click and missing click problem, such as resulting in the imbalanced distribution of fig.4, hence, relevance judgment click data is unevenly distributed in training data), and
wherein the set of synthesized training data is arranged for use to reduce training data distribution imbalance of the set of training data (ibid: generating smoothed and reducing distribution imbalance via random walk clustering in §3.1, discounting in §3.2).
Gao does not disclose the remaining limitations (see below).
Yu discloses: wherein he data augmentation comprises
i) for each of the plurality of queries, determining a respective representation of the query (§Experiments: Datasets contemplates the use of 3 LTR datasets MQ2007, MQ2008, MSLR-WEB10K, each with a query and respective Q-D pairs per query as shown in the table, each labeled with a relevance (0-5), each with an associated feature vector (46 dimensions, 136), hence, vector representations are determined for each query);
ii) for each of the plurality of queries, determining one or more respective semantic neighbor queries in a semantic space, based on the determined representations of the plurality of queries (§Experiments: Experiment Settings: SMOTE (Synthetic Minority Over-Sampling Technique) is used to oversample minority data via neighbors in the vector semantic space based on the representations, see Chawla §4.2 ¶1, Algorithm SMOTE); and
iii) for one or more of the plurality of queries, generating synthesized training databased on relevance judgement data associated with the query and relevance judgement data associated with one or more of the semantic neighbor queries of the query, the synthesized training data comprising a plurality of synthesized samples (ibid: SMOTE is used to interpolate synthetic data samples in the feature space based on the queries of nearest query-document pairs);
wherein the method further comprises, for each of the plurality of queries, steps of:
iv) defining a tailness measure based on a frequency of occurrence or relative frequency of occurrence of the query (Ibid: minority samples are defined for SMOTE for minority oversampling, such as the high-relevance minority samples of the datasets (§Experiments: Datasets: table showing r=2, r=4 high-rank samples), or, in combination, the minority samples of Guo such as shown in fig.4);
v) determining a number of the synthesized samples to be generated, based on a weighting factor that is determined on the tailness measures of the plurality of queries (§4.2 ¶1: SMOTE contemplates oversampling based on the amount of over-sampling required, hence, determining a weighting factor that determines number of samples to interpolate based on the tailness measures); the weighting factor indicating dynamic weightings of head part and tail part of the set of training data that generally follows along-tail distribution or heavy-tail distribution (ibid: only minority tail is oversampled, hence, weights vary dynamically according to head and tailness).
It would have been obvious before the effective filing date to a person of ordinary skill in the art to modify the method of Gao by incorporating the synthetic data generation technique of Yu. Both concern the art of synthetic data generation for LTR for data imbalance, and the incorporation would have, according to Yu, improved LTR model performance (Abs, §Introduction ¶1-4).
For claim 4, Gao modified by Yu discloses the method of claim 1, as described above. Gao modified by Yu further discloses: wherein each of the plurality of queries respectively corresponds to a plurality of query-document pairs with corresponding features (Gao §3.1 ¶4 contemplates query-document pair relations in data, such as encoded by the graph, Yu §Experiments: Datasets); and
wherein the determining of the respective representation of the query is based on the corresponding features of the query (ibid: determining graph representation based on said features).
For claim 5, Gao modified by Yu discloses the method of claim 4, as described above. Gao modified by Yu further discloses: wherein the determining of the respective representation of the query is based on a statistical measure of the corresponding features of the query (Gao §3.1 ¶4: determining graph representation of the query, based on number of edges and probability of edges with documents, such as via a normalized transition matrix, Yu §Experiments: Datasets contemplates using the LETOR datasets including various statistical measures, see Qin p.5 ).
For claim 6, Gao modified by Yu discloses the method of claim 5, as described above. Gao modified by Yu further discloses: wherein the determining of the respective representation of the query is based on a mean of the corresponding features of the query (Gao §3.2 Discounting discloses introducing a small positive value that would be the average of the single clicks amount distributed over all the non-clicks, see eq.7, hence, a mean value; Qin p.5: TF is an average appearance per token; BM25 and LMIR both also include an average metrics).
For claim 7, Gao modified by Yu discloses the method of claim 1, as described above. Gao modified by Yu further discloses: wherein the determining of the one or more respective neighbor queries is based on k-nearest-neighbor (KNN) method (Chawla §4.2 ¶1 gives overview of generating synthetic data via k nearest neighbors algorithm, see also p.329 alg.1 and p.330 table 1).
For claim 8, Gao modified by Yu discloses the method of claim 1, as described above. Gao modified by Yu further discloses: wherein the generating of synthesized training data for a query comprises:
(a) sampling, from a plurality of sets of user feedback data associated with the query, one set of user feedback data associated with the query to obtain a first data sample (Gao §3.1 discloses augmentation based on random walk style clustering, hence, sampling an initial query for the clustering algorithm);
(b) selecting one of the neighbor queries associated with the query (Gao §3.1 last ¶: selecting similar queries, such as up to 8, with Yu §Experiments: Experiments Settings disclosing SMOTE comprising selecting nearest neighbors for interpolation), and sampling, from a plurality of sets of user feedback data associated with the selected neighbor query, one set of user feedback data associated with the selected neighbor query to obtain a second data sample (ibid: combination with SMOTE would select user feedback data for generating interpolation); and
(c) synthesizing a data sample based on the first data sample and the second data sample (ibid: generating synthetic samples via SMOTE interpolation).
For claim 9, Gao modified by Yu discloses the method of claim 1, as described above. Gao modified by Yu further disclose: wherein the synthesizing of the data sample is based on:
l’ = lambda*l_q_i + (1 – lambda)l_q_j
where l' is the synthesized data sample, l_q_i is the first data sample, l_q_j is the second data sample, and lambda is a hyper-parameter (Chowla p.329 alg.1:17-25, p.330 table 1 discloses interpolation between nearest neighbors, with lambda based on the random number hyperparameter range in alg.1:21, see also p.330 Table 1).
For claim 10, Gao modified by Yu discloses the method of claim 9, as described above. Chawla further discloses: wherein 0 < lambda < 1 (p.329 alg.1:17-25, p.330 table 1).
For claim 11, Gao modified by Yu discloses the method of claim 8, as described above. Gao modified by Yu further discloses: wherein the generating of synthesized training data for one or more of the plurality of queries respectively further comprises:
repeating steps (a) to (c) to synthesize multiple data samples (Yu §Experiments: Experiments Settings: multiple data samples are generated via SMOTE).
For claim 30, Gao modified by Yu discloses the method of claim 1, as described above. Gao modified by Yu further discloses: processing a query and a set of document data using the neural network based ranking model to determine a result (Gao §2.3 ¶2-3 contemplates training neural net ranker LambdaRank with §4.1-4.2 disclosing operating model to determine results, Yu §Experiments Settings: RankNet).
Claim(s) 14 are rejected under 35 U.S.C. 103 as being unpatentable over Gao ("Smoothing clickthrough data for web search ranking", published 2009) in view of Yu ("Data augmentation based on adversarial autoencoder handling imbalance for learning to rank", published 2019) with Chawla ("SMOTE: synthetic minority over-sampling technique", published 2002) and Qin ("Introducing LETOR 4.0 Datasets", published 2013) incorporated by reference, in view of Zhang ("Handling Imbalanced Data, Skewed Data, and Outliers: Preprocessing Techniques in ML ", published 5/24/2022).
For claim 14, Gao modified by Yu discloses the method of claim 13, as described above. Gao modified by Yu does not disclose: wherein the relative frequency of occurrence of a query is determined based on:
T_q_i= log(t_i + 1)
where T_q_i corresponds to a tailness measure of the query, and t_i is a number of sets of user feedback data for the query.
Zhang discloses: wherein the relative frequency of occurrence of an event is determined based on:
T_q_i= log(t_i + 1)
where T_q_i corresponds to a tailness measure of the query, and t_i is a number of sets of user feedback data for the query (p.5: “Techniques to handle skewed data” bullets 1-3 and figure (p.6) discloses that user feedback data, such as web visits, can be normalized via the log(1+x) transformation in order to improve tailness, see figure on p.6, hence, combination with Gao yielding consideration of the tailness measure log1p of user clicks when considering click frequency).
It would have been obvious before the effective filing date to a person of ordinary skill in the art to modify the method of Gao modified by Yu by incorporating the log normalization technique of Zhang. Both concern the art of data imbalance for web events, and the incorporation would have, according to Zhang, reduced negative impact when considering implementations of clustering models (p.5 “Impact on Machine Learning” ¶1), such SMOTE (p.2 “Resampling Methods”).
Claim(s) 16-20 are rejected under 35 U.S.C. 103 as being unpatentable over Gao ("Smoothing clickthrough data for web search ranking", published 2009) in view of Yu ("Data augmentation based on adversarial autoencoder handling imbalance for learning to rank", published 2019) with Chawla ("SMOTE: synthetic minority over-sampling technique", published 2002) and Qin ("Introducing LETOR 4.0 Datasets", published 2013) incorporated by reference, in view of Zhou ("Bbn: Bilateral-branch network with cumulative learning for long-tailed visual recognition", published 2020).
For claim 16, Gao modified by Yu discloses the method of claim 1, as described above. Gao modified by Yu does not disclose the limitations of claim 16.
Zhou discloses: wherein the neural network based ranking model comprises:
a first model branch with a first multilayer perceptron and a first predictor operably coupled with the first multilayer perceptron (fig.3 gives overview of bilateral-branch network for achieving improved long-tailed performance, with the top branch using a conventional uniform sampler and the bottom using a tail-heavy sampling, see §4.1; hence, incorporation of Zhou into Gao would yield a technique wherein the conventional distributions of Gao, e.g., the before-smoothing distribution of Gao fig.4, the unsmoothed datasets of §4.1, table 4 would be used as the conventional head-heavy sampling while the smoothed samples would correspond to the reverse or rebalanced samples);
a second model branch with a second multilayer perceptron and a second predictor operably coupled with the second multilayer perceptron (ibid); and
a combiner for combining an output of the first predictor and an output of the second predictor (fig.3: the results are combined for cumulative learning, see §4.3).
It would have been obvious before the effective filing date to a person of ordinary skill in the art to modify the method of Gao modified by Yu by incorporating the bilateral branch classifier of Zhou. Both concern the art of improving classification on long-tailed data, and the incorporation would have, according to Zhou, improve representation learning and long-tailed recognition performance when compared to traditional models (p.9721 col.1 bullet 2), e.g., the LambdaRank model of Gao.
For claim 17, Gao modified by Yu modified by Zhao discloses the method of claim 16, as described above. Gao modified by Yu modified by Zhao further discloses: wherein the combiner is arranged to apply a weighting to the output of the first predictor and/or a weighting to the output of the second predictor (§4.4: equal weighting during inference).
For claim 18, Gao modified by Yu modified by Zhao discloses the method of claim 16, as described above. Gao modified by Yu modified by Zhao further discloses: wherein the training of the neural network based ranking model comprises:
performing a ranking or scoring operation based on the set of training data and the set of synthesized training data using the neural network based ranking model (Gao §4.1: performing ranking via trials based on real world data, see also §2.3 ¶2, Yu §Experiment Settings: combination with Zhao yielding application to conventional branch and rebalancing branch learning).
For claim 19, Gao modified by Yu modified by Zhao discloses the method of claim 18, as described above. Gao modified by Yu modified by Zhao further discloses: wherein the ranking or scoring operation comprises:
processing the set of training data with the first model branch (Zhao fig.3 top branch: processing conventional training data with top branch); and
processing a combination of the set of training data and the set of synthesized training data with the second model branch (fig.3: bottom branch: augmented data comprises combination of conventional and synthesized data).
For claim 20, Gao modified by Yu modified by Zhao discloses the method of claim 19, as described above. Gao modified by Zhao further discloses: determining ranking or scoring loss L_LTR based on the performing of the ranking or scoring operation (Zhou §4.1 gives overview of loss function, see eq.3).
Claim(s) 21-28 are rejected under 35 U.S.C. 103 as being unpatentable over Gao ("Smoothing clickthrough data for web search ranking", published 2009) in view of Yu ("Data augmentation based on adversarial autoencoder handling imbalance for learning to rank", published 2019) with Chawla ("SMOTE: synthetic minority over-sampling technique", published 2002) and Qin ("Introducing LETOR 4.0 Datasets", published 2013) incorporated by reference, in view of Zhou ("Bbn: Bilateral-branch network with cumulative learning for long-tailed visual recognition", published 2020) in view of Anand ("Supervised contrastive learning approach for contextual ranking", published 7/7/2022).
For claim 21, Gao modified by Yu modified by Zhao discloses the method of claim 18, as described above. Gao modified by Yu modified by Zhao does not disclose: wherein the training of the neural network based ranking model further comprises:
performing a contrastive learning operation based on the set of training data and/or the set of synthesized training data using the neural network based ranking model.
Anand discloses: wherein the training of the neural network based ranking model further comprises:
performing a contrastive learning operation based on the set of training data and/or the set of synthesized training data using the neural network based ranking model (§3.2 gives overview of the supervised contrastive loss method for query-document pairs, with eq.5-6 disclosing the combining of the SCL loss with the ranking loss for a total loss, see also §3.3.1 disclosing various techniques of producing augmenting data d^+_a for positive pairs, hence, performing contrastive loss on the set of training data and / or the set of synthesized data based on a ranking model via loss functions).
It would have been obvious before the effective filing date to a person of ordinary skill in the art to modify the method of Gao modified by Yu modified by Zhao by incorporating the supervised contrastive loss technique of Anand. Both concern the art of query-document ranking, and the incorporation would have, according to Anand, improve ranking performance for a wide variety of scenarios, such as limited data settings (§1 ¶4, §1.2)
For claim 22, Gao modified by Yu modified by Zhao modified by Anand discloses the method of claim 21, as described above. Gao modified by Yu modified by Zhao modified by Anand further discloses: wherein the contrastive learning operation comprises:
for one or more of the queries: for data associated with each of the plurality of query-documents pairs of the query, performing a data perturbation operation to generate respective augmented data for the data of each of the plurality of query-documents pairs (Anand §3.3.1: generating augmented data for each of the query document pairs); and
processing the augmented data using the neural network based ranking model (Anand §3.2 eq.5-6: contrastive loss is generated for the training samples to train the neural network).
For claim 23, Gao modified by Yu modified by Zhao modified by Anand discloses the method of claim 22, as described above. Gao modified by Yu modified by Zhao modified by Anand further discloses: wherein the data perturbation operation comprises:
generating, for each of the data of each of the plurality of query-documents pairs:
a first set of augmented data with a first extent of noise injection (§3.3.1 ¶1: generating augmented data via random noise, e.g., removing all but a small sample of data, see ¶3)); and
a second set of augmented data with a second extent of noise injection different from the first extent (ibid: the random second sample from a negative corpus constitutes addition noise injection).
For claim 24, Gao modified by Yu modified by Zhao modified by Anand discloses the method of claim 22, as described above. Gao modified by Yu modified by Zhao modified by Anand further discloses: wherein the neural network based ranking model further comprises:
a first projector operably coupled with the first multilayer perceptron (Zhou fig.3 shows projector coupled to the shared region to generate the feature vector for the first top branch); and
a second projector operably coupled with the second multilayer perceptron (ibid: likewise for bottom branch); and
wherein the contrastive learning operation further comprises:
processing the augmented data using the first multilayer perceptron and the first projector (ibid: augmented data, such as that of Anand §3.3.1 ¶1 is processed via a uniform sampler in the top branch); and
processing the augmented data using the second multilayer perceptron and the second projector (ibid: likewise for the bottom branch ).
For claim 25, Gao modified by Yu modified by Zhao modified by Anand discloses the method of claim 22, as described above. Gao modified by Yu modified by Zhao modified by Anand further discloses: further comprises:
determining a contrastive loss L_CL based on the performing of the contrastive learning operation (Anand §3.2 eq.5).
For claim 26, Gao modified by Yu modified by Zhao modified by Anand discloses the method of claim 22, as described above. Gao modified by Yu modified by Zhao modified by Anand further discloses: performing a joint optimization operation based on the performing of the ranking or scoring operation and the performing of the contrastive learning operation (Anand §3.2 eq.6).
For claim 27, Gao modified by Yu modified by Zhao modified by Zhao modified by Anand discloses the method of claim 26, as described above. Gao modified by Yu modified by Zhao modified by Anand further discloses: wherein the joint optimization operation comprises:
jointly optimizing a ranking or scoring loss L_LTR associated with the performing of the ranking or scoring operation and the contrastive loss L_CL associated with the performing of the contrastive learning operation (Anand §3.2 eq.6).
For claim 28, Gao modified by Yu modified by Zhao modified by Anand discloses the method of claim 27, as described above. Gao modified by Zhao modified by Anand further discloses: wherein the jointly optimizing comprises: optimizing an overall loss L_overall that equals to L_LTR + gamma*L_CL, where gamma is hyper-parameter (Anand §3.2 eq.6 shows a overall loss that is has some proportion of L_SCL to L_LTR, i.e., lambda / (1 – lambda) ).
Claim(s) 29 are rejected under 35 U.S.C. 103 as being unpatentable over Gao ("Smoothing clickthrough data for web search ranking", published 2009) in view of Yu ("Data augmentation based on adversarial autoencoder handling imbalance for learning to rank", published 2019) with Chawla ("SMOTE: synthetic minority over-sampling technique", published 2002) and Qin ("Introducing LETOR 4.0 Datasets", published 2013) incorporated by reference in view of Liu (US 20100169323 A1).
Claim 29 recites a system analogous to claim 1 and hence is rejected for the same reasons. Gao modified by Yu does not disclose: a system for training a neural network based ranking model, comprising: one or more processors; and memory storing one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for performing or facilitating performing of to the computer-implemented method.
Liu discloses: a system for training a neural network based ranking model, comprising: one or more processors (fig.9:920, 0044); and memory storing one or more programs configured to be executed by the one or more processors (fig.9:931-937, 0046), the one or more programs including instructions for performing or facilitating performing of to the computer-implemented method (fig.9:937, 0046).
It would have been obvious before the effective filing date to a person of ordinary skill in the art to modify the method of Gao modified by Yu by incorporating the computer embodiment of Liu. Both concern the art of query ranking via machine learning, and the incorporation would have, according to Liu, allow a suitable computing environment for implementation of such a method (0041).
Response to Arguments
Applicant’s arguments have been fully considered. In the remarks, Applicant argued:
1. The claims are statutory under 101.
Examiner agrees and the rejections withdrawn.
2. Regarding the 102 / 103 rejections: the cited art does not disclose the newly added limitations.
Applicant’s arguments have been fully considered but are moot in view of the newly cited art.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Zhang (US 20230098656 A1)discloses data augmentation based on group subsampling.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee 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 date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LIANG LI whose telephone number is (303)297-4263. The examiner can normally be reached Mon-Fri 9-12p, 3-11p MT (11-2p, 5-1a ET).
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice. The examiner is available for interviews Mon-Fri 6-11a, 2-7p MT (8-1p, 4-9p ET).
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor Jennifer Welch can be reached on (571)272-7212. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from Patent Center and the Private Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from Patent Center or Private PAIR. Status information for unpublished applications is available through Patent Center or Private PAIR to authorized users only. Should you have questions about access to Patent Center or the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
/LIANG LI/
Primary examiner AU 2143
