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 Objections
Claim 20 objected to because of the following informalities: the method as claimed in claim 1 should be the method as claimed in claim 12. Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim’s 1-21 rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Where applicant acts as his or her own lexicographer to specifically define a term of a claim contrary to its ordinary meaning, the written description must clearly redefine the claim term and set forth the uncommon definition so as to put one reasonably skilled in the art on notice that the applicant intended to so redefine that claim term. Process Control Corp. v. HydReclaim Corp., 190 F.3d 1350, 1357, 52 USPQ2d 1029, 1033 (Fed. Cir. 1999). The term “train, by a machine learning engine, the one or more content inputs” in claim 1 is used by the claim to mean “training inputs by a machine learning model,” while the accepted meaning is “training a machine learning model on inputs.” The term is indefinite because the specification does not clearly redefine the term.
Claim Interpretation
Regarding claims 1,11, and 12, the claim recites “train, by a machine learning engine (214), the one or more content inputs received based on the second and the third set of attributes and a predefined dataset obtained from a knowledge graph associated with the domain” with nothing in the spec pointing towards clear definition as to how the inputs would get trained by a machine learning model.
Given the indefiniteness of the wording, and in the interest of simplicity, the examiner has chosen to interpret “train, by a machine learning engine (214), the one or more content inputs received based on the second and the third set of attributes and a predefined dataset obtained from a knowledge graph associated with the domain” as training the machine learning model on the one or more content inputs.
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 text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
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 1-5, 7, 11-16, and 18 are rejected under U.S.C 103 over Bougiatiotis et al (NPL: Enhanced Movie Content Similarity Based on Textual, Auditory and Visual Information) (“Bougiatiotis”) in view of Prayaga et al (US20200065347A1) (“Prayaga”), Kumar et al (NPL: A Movie Recommender System: MOVREC) (“Kumar”)
Regarding Claim 1, Bougiatiotis teaches A system (110) for generating recommendation for providing input services, said system (110) comprising; one or more processors (202) operatively coupled to a plurality of computing devices (104), the one or more processors (202) coupled with a memory (204), wherein said memory (204) stores instructions which when executed by the one or more processors (202) causes said system (110) to ([Abstract] teaches a system used for movie recommendations)
based on the extracted first set of attributes, divide, the one or more content inputs into a plurality of independent blocks ([3.2.1-Page 6] teaches each movie being represented as a vector with each movie having an associated value based on number of occurrences for corresponding words from the bag of words. The bag of words here are functioning as an attribute with the movies themselves being the independent blocks teaching the limitation)
extract a second set of attributes from the plurality of independent blocks, the second set of attributes pertaining to the predefined domain and a predefined event present in the plurality of independent blocks ([3.4-Page 14] teaches their visual analysis technique where they are extracting visual features from movies. The predefined domain here being the movies with the predefined event present being the extracted scenes and the independent blocks are the different movies teaching this limitation)
extract a third set of attributes from the plurality of independent blocks, the third set of attributes pertaining to predefined information associated with each independent block ([4.2-Page 24] teaches 1100 unique tags for every movie in their data set. The different tags can be associated with adjectives for the movie, directors, nouns associated with the movie, etc. Those tags are being interpreted as a third set of attributes from the independent blocks. The movies these tags are coming from are interpreted as independent blocks teaching this limitation)
generate, a trained model based on the trained one or more content inputs ([3.3- Page 12] teaches two trained models, that are trained to classify movie audio segments, those audio segments are the content inputs)
Bougiatiotis does not teach receive one or more content inputs from the plurality of computing devices (104), the one or more content inputs associated with a predefined domain;
extract a first set of attributes from the received one or more content inputs, the first set of attributes pertaining to one or more contextual parameters associated with the one or more content input;
based on the extracted second and third set of attributes, determine a weight to be assigned to each said independent blocks, the weight pertaining to an importance of each block with respect to the second and the third set of attributes extracted
and auto-recommend, a final contextual block, based on the generated trained model, the final contextual block comprising one or more independent blocks with the highest weight
However, Prayaga does teach receive one or more content inputs from the plurality of computing devices (104), the one or more content inputs associated with a predefined domain ([0003] teaches input vectors containing raw data about contextual aspects about an associated and different user which is interpreted as content inputs. [0039] teaches extracting topics from a user response and using text categorization to classify documents in predefined categories)
extract a first set of attributes from the received one or more content inputs, the first set of attributes pertaining to one or more contextual parameters associated with the one or more content input ([0003] teaches extracting attributes from the input vector. The attributes are associated with users and weights corresponding to each attribute that is particular to a user, those attribute associations is being interpreted as contextual parameters teaching this limitation)
Bougiatiotis and Prayaga are analogous art because they both deal with machine learning and recommendations
It would have been obvious to combine Bougiatiotis with the contextual parameters content input of Prayaga. Doing so would lead to a more cost efficient model when it comes to handling larger number of users for user recommendations ([Prayaga-0002])
The combination of Bougiatiotis and Prayaga does not teach based on the extracted second and third set of attributes, determine a weight to be assigned to each said independent blocks, the weight pertaining to an importance of each block with respect to the second and the third set of attributes extracted
and auto-recommend, a final contextual block, based on the generated trained model, the final contextual block comprising one or more independent blocks with the highest weight
However, Kumar does teach based on the extracted second and third set of attributes, determine a weight to be assigned to each said independent blocks, the weight pertaining to an importance of each block with respect to the second and the third set of attributes extracted ([Page 3] teaches a total weight formula that would be assigned to a particular movie in the movie recommendation system. The values used to come up with each movie’s weighted score is decided by a weighted average formula that includes weighted average scores that are calculated for each attribute for that particular movie. With these weighted values being used in conjunction with their recommendation system to recommend movies it can be reasonably interpreted that the weight here does pertain to the importance of each movie based on their attribute values teaching this limitation.
and auto-recommend, a final contextual block, based on the generated trained model, the final contextual block comprising one or more independent blocks with the highest weight (Page 4-Conclusion] teaches movie recommendations being made based on their cumulative weight value after the user inputs a set of attributes. Step 2 in their algorithm teaches that if the number of suggested movies exceeds 20 then only the top 20 movies would be selected based on ratings. With the total weight being calculated before their clustering algorithm and being used to displaying the cluster with the highest rating it can be reasonably interpreted here that the movies with the highest rating would also have the highest weight [Page 3-Column 1: 30-57]. The movie titles are the independent blocks, and the list of recommendations returned to the user is being interpreted as the final contextual block, teaching this limitation)
Bougiatiotis, Prayaga and Kumar are analogous art because they both deal with machine learning and recommendations
It would have been obvious to combine Bougiatiotis with the contextual parameters content input of Prayaga, and the weighted recommendation techniques of Kumar. Doing so would lead to a more user friendly recommendation experience ([Kumar-Abstract])
Regarding Claim 2, Bougiatiotis, Prayaga and Kumar teaches all the limitations of Claim 1
Bougiatiotis also teaches wherein the one or more content inputs pertain to any or a combination of images, video streams, audio streams and textual content ([3.3-Page 13] teaches analyzing audio streams)
Regarding Claim 3, Bougiatiotis, Prayaga and Kumar teaches all the limitations of Claim 1
Bougiatiotis also teaches wherein the knowledge graph is provided with predefined markers using time-based split for the combination of video streams and audio streams and location-based split for the textual content ([5.3-Page 31] teaches a graph of the different movie titles and the similar movies to each title. The graph here is being interpreted as a knowledge graph. The graph is an interactive demo of their trained models which considers audio analysis, visual analysis and metadata analysis. It can be reasonably interpreted that audio analysis includes analysis of audio streams and visual analysis includes video streams along with metadata analysis including location based split for textual content)
Regarding Claim 4, Bougiatiotis, Prayaga and Kumar teaches all the limitations of Claim 1
Bougiatiotis also teaches wherein the predefined information pertains to a plurality of sentiment and mood, language and dialect parameters across a plurality of regions and users ([5.3-Page 31] teaches node size corresponding to the movie score from IMDB. It can be reasonably interpreted that that score represents users’ overall sentiment and feelings pertaining to the movie across different users and regions teaching this limitation)
Regarding Claim 5, Bougiatiotis and Prayaga teaches all the limitations of Claim 1
Bougiatiotis also teaches wherein the importance of the predefined weight assigned to each independent block is determined based on position of occurrence, co-occurrence with pre-determined information of the domain. ([3.2.2-Page 7] teaches assigning weight to bag of words associated with the different movies based on two different factors, with one of those factors being the frequency of the words appearance.)
Regarding Claim 7, Bougiatiotis, Prayaga and Kumar teaches all the limitations of Claim 1
Bougiatiotis also teaches wherein the one or more content inputs are inserted as a node. ([Figure 14] teaches movies being represented as nodes)
Regarding Claim 11, see the analysis of Claim 1
Regarding Claim 12, see the analysis of Claim 1
Regarding Claim 13, see the analysis of Claim 2
Regarding Claim 14, see the analysis of Claim 3
Regarding Claim 15, see the analysis of Claim 4
Regarding Claim 16, see the analysis of Claim 5
Regarding Claim 18, see the analysis of Claim 7
Claim 6 and 17 are rejected under U.S.C 103 over Bougiatiotis et al (NPL: Enhanced Movie Content Similarity Based on Textual, Auditory and Visual Information) (“Bougiatiotis”) in view of Prayaga et al (US20200065347A1) (“Prayaga”), Kumar et al (NPL: A Movie Recommender System: MOVREC) (“Kumar”), and Elkahky (NPL: A Multi-View Deep Learning Approach for Cross Domain User Modeling in Recommendation Systems) (“Elkahky”)
Regarding Claim 6, Bougiatiotis, Prayaga and Kumar teaches all the limitations of Claim 1
Bougiatiotis also teaches determine an affinity of the received information with the third set of attributes ([4.1] teaches unique tags for every movie, the tags here are interpreted as an affinity for the movie)
The combination of Bougiatiotis, Prayaga and Kumar does not teach wherein the system (110) is configured to: receive information from a plurality of information sources
and aggregate the affinity associated with the plurality of information sources to obtain the final contextual block.
However, Elkahky does teach wherein the system (110) is configured to: receive information from a plurality of information sources ([Page 3 Column 2: Lines 37-42] teaches four different datasets)
and aggregate the affinity associated with the plurality of information sources to obtain the final contextual block ([Page 8 Column 2: Lines 48-58] teaches cross domain recommendation used for recommending authors. The data set contains data from five different domains with those five different domains being interpreted as a plurality of information sources. The recommended authors are representing the final contextual block, and the research fields are being interpreted as the affinity teaching this limitation.)
Bougiatiotis, Prayaga, Kumar and Elkahky are analogous art because they both deal with machine learning and recommendations
It would have been obvious to combine Bougiatiotis with the contextual parameters content input of Prayaga, and the plurality of sources processing of Elkahky. Doing so would lead to higher quality recommendations ([Elkahky-Abstract])
Regarding Claim 17, see the analysis of Claim 6
Claim 8 and 19 are rejected under U.S.C 103 over Bougiatiotis et al (NPL: Enhanced Movie Content Similarity Based on Textual, Auditory and Visual Information) (“Bougiatiotis”) in view of Prayaga et al (US20200065347A1) (“Prayaga”), Kumar et al (NPL: A Movie Recommender System: MOVREC) (“Kumar”), and Guo et al (NPL: DA-GCN: A Domain-aware Attentive Graph Convolution Network for Shared-account Cross-domain Sequential Recommendation) (“Guo”)
Regarding Claim 8, Bougiatiotis, Prayaga and Kumar teaches all the limitations of Claim 1
The combination of Bougiatiotis, Prayaga and Kumar does not teach wherein the system is further configured to: calculate similarity, by a graph traversal and embeddings-based engine, between a plurality of content inputs of the pre-defined domain and a plurality of cross-domains, wherein the plurality of cross-domains refers to nature of the cross- domains different from the predefined domain.
However, Guo does teach wherein the system is further configured to: calculate similarity, by a graph traversal and embeddings-based engine, between a plurality of content inputs of the pre-defined domain and a plurality of cross-domains, wherein the plurality of cross-domains refers to nature of the cross- domains different from the predefined domain ([abstract] teaches DA-GCN a domain-aware graph convolutional network, with the GCN here being used to link different users in different domains for shared-account cross domain recommendations. The GCN is used to account for user’s domain specific preferences, with those user specific preferences functioning as the predefined domain. The content input is the users and items in each domain. As stated in the spec a graph traversal and embedding based algorithm includes graph convolutional methods [0038] so the DA-GCN teaches the graph traversal and embedding based engine, teaching this limitation)
Bougiatiotis, Prayaga, Kumar and Guo are analogous art because they both deal with machine learning and recommendations
It would have been obvious to combine Bougiatiotis with the contextual parameters content input of Prayaga, and the similarity calculations of Guo. Doing so would lead to a better performing recommendation system ([Guo-Abstract])
Regarding Claim 19, see the analysis of Claim 8
Claim 9 and 20 are rejected under U.S.C 103 over Bougiatiotis et al (NPL: Enhanced Movie Content Similarity Based on Textual, Auditory and Visual Information) (“Bougiatiotis”) in view of Prayaga et al (US20200065347A1) (“Prayaga”), Kumar et al (NPL: A Movie Recommender System: MOVREC) (“Kumar”), Elkahky (NPL: A Multi-View Deep Learning Approach for Cross Domain User Modeling in Recommendation Systems) (“Elkahky”) and Ma et al (US 20220245705 A1) (“Ma”)
Regarding Claim 9 Bougiatiotis, Prayaga and Kumar teaches all the limitations of Claim 1
The combination of Bougiatiotis, Prayaga and Kumar does not teach wherein the system is further configured to: filter the plurality of content inputs based on a user interaction such as watch history, click, detail page visit, summary viewed, added to wish list, preferred language, location of the users and preferences
re-rank the plurality cross-domains to be recommended to an independent block associated with the predefined domain
However, Elkahky does teach wherein the system is further configured to: filter the plurality of content inputs based on a user interaction such as watch history, click, detail page visit, summary viewed, added to wish list, preferred language, location of the users and preferences ([Page 1 Column 2: Lines 6-9] teaches collaborative filtering using the user’s history on a website to generate recommendations)
Bougiatiotis, Prayaga, Kumar, and Elkahky are analogous art because they both deal with machine learning and recommendations
It would have been obvious to combine Bougiatiotis with the contextual parameters content input of Prayaga, and the plurality of sources processing of Elkahky. Doing so would lead to higher quality recommendations ([Elkahky-Abstract])
The combination of Bougiatiotis, Prayaga, Kumar and Elkahky does not teach re-rank the plurality cross-domains to be recommended to an independent block associated with the predefined domain.
However, Ma teaches to re-rank the plurality cross-domains to be recommended to an independent block associated with the predefined domain ([0070] teaches re-ranking items associated with cross-domain diversity in a recommendation pool. The recommendation pool here is the independent block. The re-ranking process can be done based off a complementary score of the recommended items; it can be reasonably interpreted that the presence of a complementary score would mean that there is a predefined domain on what the user would be interested in teaching the predefined domain part of the limitation. With this being done with items associated with cross-domain diversity the re-ranking of the items here is being interpreted as re-ranking domains teaching this part of the limitation)
Bougiatiotis, Prayaga, Kumar, Elkahky, and Ma are analogous art because they both deal with machine learning and recommendations
It would have been obvious to combine Bougiatiotis with the contextual parameters content input of Prayaga, the plurality of sources processing of Elkahky, and the reranking of Ma. Doing so would lead to more personalized recommendations ([Ma-0002])
Regarding Claim 20, See the analysis of Claim 9
Claim 10 and 21 are rejected under U.S.C 103 over Bougiatiotis et al (NPL: Enhanced Movie Content Similarity Based on Textual, Auditory and Visual Information) (“Bougiatiotis”) in view of Prayaga et al (US20200065347A1) (“Prayaga”), Kumar et al (NPL: A Movie Recommender System: MOVREC) (“Kumar”), Galai et al (US10693981) (“Galai”) and Zhen et al (CN112541132) (“Zhen”)
Regarding Claim 10 Bougiatiotis, Prayaga, and Kumar teaches all the limitations of Claim 1
The combination of Bougiatiotis, Prayaga, and Kumar does not teach wherein the system is further configured to: assign appropriate weight to an independent block according to a domain associated with an entity;
extract, from a plurality of information sources, information associated with the independent block;
and, determine, a domain specific affinity between the independent block and the plurality of information sources;
and, based on the affinity determined, connect the independent block with the plurality of cross- domains using the knowledge graph.
However, Galai does teach wherein the system is further configured to: assign appropriate weight to an independent block according to a domain associated with an entity ([Column 9: Lines 39 to 51] teaches assigning weights to different recommendation pools with their example consisting of a user choosing evergreen content over ctr driven content leading to evergreen content recommendation pool being seen as a potential recommendations for the user. The pools here are functioning as independent blocks with the evergreen content being the domain and the users are the entity. The favoring of the evergreen pool over the ctr is being interpreted as weights being assigned teaching this limitation)
extract, from a plurality of information sources, information associated with the independent block ([Column 9: Lines 39 to 51] teaches recommending user content from the evergreen content recommendation pool vs the ctr driven content recommendation due to them choosing more evergreen content over ctr driven content. It can be reasonably interpreted that in order for the system to know that the user interacts with evergreen content over ctr driven content that they are extracting information with the source being the users and the independent blocks being the content recommendation pool teaching this limitation)
and, determine, a domain specific affinity between the independent block and the plurality of information sources ([Column 9: Lines 39 to 51] teaches recommending evergreen content to a user because of their interest in evergreen content over ctr driven content. Evergreen is the domain specific affinity with the recommendation pool being the independent block and the user’s choice of content is the plurality of sources teaching this limitation)
Bougiatiotis, Prayaga, Kumar, and Galai are analogous art because they both deal with machine learning and recommendations
It would have been obvious to combine Bougiatiotis with the contextual parameters content input of Prayaga, and the pool recommendation technique of Galai. Doing so would lead to higher user engagement in the recommended content ([Galai-Column 1: Lines 23 to 34])
The combination of Bougiatiotis, Prayaga, Kumar, and Galai does not teach and, based on the affinity determined, connect the independent block with the plurality of cross- domains using the knowledge graph.
However, Zhen does teach and, based on the affinity determined, connect the independent block with the plurality of cross- domains using the knowledge graph ([0074] teaches integrating similar attributes present in different fields into a graph structure, the graph structure here is being interpreted as a knowledge graph and the different fields is interpreted as cross domains. With the items with similar attribute relation being a product of a cross -domain recommendation method based on multi-view knowledge the items are being interpreted as independent blocks. [0020-0021])
Bougiatiotis, Prayaga, Kumar Galai, and Zhen are analogous art because they both deal with machine learning and recommendations
It would have been obvious to combine Bougiatiotis with the contextual parameters content input of Prayaga, and the pool recommendation technique of Galai and the connecting of the knowledge graph to cross domains of Zhen. Doing so would lead to a better performing recommendation systems ([Zhen-0010])
Regarding Claim 21, See the analysis of Claim 10
Conclusion
The prior arts are made of record and relied upon is considered to applicant’s disclosure
Moon et al US11442992B1 (2019-08-30) ([Abstract] “In one embodiment, a method includes receiving a query from a user from a client system associated with the user, accessing a knowledge graph comprising a plurality of nodes and edges connecting the nodes, wherein each node corresponds to an entity and each edge corresponds to a relationship between the entities corresponding to the connected nodes, determining one or more initial entities associated with the query based on the query, selecting one or more candidate nodes by a conversational reasoning model from the knowledge graph corresponding to one or more candidate entities, respectively, wherein each candidate node is selected based on the nodes corresponding to the initial entities, dialog states associated with the query, and a context associated with the query, generating a response based on the initial entities and the candidate entities, and sending instructions for presenting the response to the client system in response to the query”
Jingjie et al EP4181026A1 (2020-07-24) ([Abstract] “This application provides a recommendation model training method, a recommendation method, an apparatus, and a computer-readable medium in the artificial intelligence (AI) field. The training method includes: obtaining a first recommendation model, where a model parameter of the first recommendation model is obtained through training based on n first training samples; determining an impact function value of each first training sample with respect to a verification loss of m second training samples in the first recommendation model; determining, based on the impact function value of each first training sample with respect to the verification loss, a weight corresponding to each first training sample; and training the first recommendation model based on the n first training samples and the weights corresponding to the n first training samples, to obtain a target recommendation model. According to technical solutions of this application, impact of data distribution inconsistency between training samples on training of a recommendation model can be alleviated, and accuracy of the recommendation model can be improved.”)
Goodsitt et al US12061862B2 (2020-08-13) ([Abstract] “Embodiments disclosed herein provide for dynamically modifying image and/or video content based on user preferences. Embodiments provide a server, a database containing user preference information and one or more content templates, and a content generator comprising a plurality of generative adversarial networks, wherein each generative adversarial network is associated with a corresponding style transfer, and the content generator is configured to apply one or more style transfers to a base image to convert the base image into a desired style. Once created, the customized content is sent to a client device to for display.”)
Any inquiry concerning this communication or earlier communications from the examiner should be directed to URIAH V MOORE whose telephone number is (571)384-8341. 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, Mariela Reyes can be reached at (571)270-1006. 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.
/URIAH VENDELL MOORE/Examiner, Art Unit 2142 /Mariela Reyes/Supervisory Patent Examiner, Art Unit 2142
