DETAILED ACTION
This action is responsive to the Application filed on 12/12/2023. Claims 1-20 are pending in the case. Claims 1, 11 and 20 are independent claims.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim 1-20 are rejected under 35 U.S.C. 101 because the claim are directed to an abstract idea without significantly more.
Regarding Claim 1/11/20:
Under step 1, claim 1 is directed to a method which is directed to a process, one of the statutory categories.
Under step 1, claim 11 is directed to a system which is directed to a machine, one of the statutory categories.
Under step 1, claim 20 is directed to a non transitory computer readable medium which is directed to a product of manufacture, one of the statutory categories.
Under Step 2A Prong 1, the claim recites the following limitations which are considered mental evaluations:
identifying a field from a data source, the field including content data and metadata associated with an item;
determining a first attribute of the item based on the content data and the metadata;
matching the first attribute with a second attribute in a product taxonomy;
and based on matching the first attribute, generating a classifier that represents a category of the item.
Each of these amount to mental evaluation because they describe manipulation of abstract data. Identification, determination and matching based on abstract features are all data evaluations performed in the mind. Generating a classifier without any additional detail does not suggest any particular computer technology but rather a heuristic rule for data classification.
Under step 2A Prong 2, The claim recites the following additional element(s):
[claim 11] a memory storing instructions; and one or more hardware processors communicatively coupled to the memory and configured by the instructions to perform operations…[claim 20] A non-transitory computer-readable medium comprising instructions that, when executed by a hardware processor of a device, cause the device (which amounts to descriptions which makes use of or applies the abstract idea because under 2106.05(f)(1) “the claim fails to recite details of how a solution to a problem is accomplished”, as no details of the functioning of the training or encoder are claimed.)
Therefore the claim is directed to a judicial exception.
Under step 2B, the recited additional elements when considered alone or in combination neither integrates the abstract idea into a practical application nor provides significantly more than the abstract idea itself.
Regarding Claim 2/12, 3/13, 5/15, 6/16, 7/17
The rejection of claim 1/11 is incorporated and further:
Each of the limitations described in the claim, under Step 2A Prong 1, only serve to describe the abstract ideas addressed in the independent claim, in particular the limitations describe mental evaluations.
Each of these limitations are descriptions of the features of the abstract data, or further mental evaluations on such data.
Regarding Claim 4/14
The rejection of claim 1/11 is incorporated and further,
Under Step 2A Prong 1, the claim recites the following limitations which are considered mental evaluations:
wherein the field comprises one of a text field, a video field, or an image field, and wherein the field comprises a plurality of attributes associated with the item, and wherein the determining of the first attribute of the item… to infer the first attribute of the item based on the content data and the metadata.
Under step 2A Prong 2, The claim recites the following additional element(s):
using a machine learning model (which amounts to descriptions which makes use of or applies the abstract idea because under 2106.05(f)(1) “the claim fails to recite details of how a solution to a problem is accomplished”, as no details of the functioning of the training or encoder are claimed.)
Under step 2B, the recited additional elements when considered alone or in combination neither integrates the abstract idea into a practical application nor provides significantly more than the abstract idea itself.
Regarding Claim 8/18
The rejection of claim 1/11 is incorporated and further,
Under Step 2A Prong 1, the claim recites the following limitations which are considered mental evaluations:
determining that the field is an image field that includes an image… to identify the first attribute associated with the item based on the image.
Under step 2A Prong 2, The claim recites the following additional element(s):
using a convolutional neural network (CNN) based machine learning model (which amounts to descriptions which makes use of or applies the abstract idea because under 2106.05(f)(1) “the claim fails to recite details of how a solution to a problem is accomplished”, as no details of the functioning of the training or encoder are claimed.)
Under step 2B, the recited additional elements when considered alone or in combination neither integrates the abstract idea into a practical application nor provides significantly more than the abstract idea itself.
Regarding Claim 9/19
The rejection of claim 1/11 is incorporated and further,
Under Step 2A Prong 1, the claim recites the following limitations which are considered mental evaluations:
determining that the field is a text field that includes a plurality of words… to identify the first attribute associated with the item based on the plurality of words.
Under step 2A Prong 2, The claim recites the following additional element(s):
using a Bidirectional Encoder Representations from Transformers (BERT) machine learning model (which amounts to descriptions which makes use of or applies the abstract idea because under 2106.05(f)(1) “the claim fails to recite details of how a solution to a problem is accomplished”, as no details of the functioning of the training or encoder are claimed.)
Under step 2B, the recited additional elements when considered alone or in combination neither integrates the abstract idea into a practical application nor provides significantly more than the abstract idea itself.
Regarding Claim 10
The rejection of claim 1 is incorporated and further,
Under Step 2A Prong 1, the claim recites the following limitations which are considered mental evaluations:
generating a graph based at least on content data and metadata associated with a plurality of items, the content data including at least one of structured data and unstructured data … to extract a plurality of features based on the graph; and identifying correlations of the plurality of items based on the plurality of features.
Under step 2A Prong 2, The claim recites the following additional element(s):
using a machine learning model (which amounts to descriptions which makes use of or applies the abstract idea because under 2106.05(f)(1) “the claim fails to recite details of how a solution to a problem is accomplished”, as no details of the functioning of the training or encoder are claimed.)
Under step 2B, the recited additional elements when considered alone or in combination neither integrates the abstract idea into a practical application nor provides significantly more than the abstract idea itself.
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.
Claims 7 and 17 are 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.
The term “shortened text format” in claim 7 and 17 is a relative term which renders the claim indefinite. The term “shortened” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 1-5, 8, 11-15, 18 and 20 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Ristoski et al. “A machine learning approach for product matching and categorization: Use case: Enriching product ads with semantic structured data”
Claim 1/11/20
Ristoski teaches, A method comprising… A system comprising: a memory storing instructions; and one or more hardware processors communicatively coupled to the memory and configured by the instructions to perform operations comprising… A non-transitory computer-readable medium comprising instructions that, when executed by a hardware processor of a device, cause the device to perform operations (Section 5.4 “All experiments were run on a MacBook Pro with 8 GB of RAM and 2.3 GHz Intel Core i7 CPU processor”)
identifying a field from a data source, the field including content data and metadata associated with an item; (Section 3.1 “We have a database A of structured products and a dataset of unstructured product descriptions P extracted from the Web. Every record consists of a title, description, URL, and a set of attribute-value pairs extracted from the title of the product, where the attributes are numeric, categorical, or free-text (an example of such a product record is given in Table 1)… More precisely, we use the structured information as a supervision for building a feature extraction model able to extract attribute-value pairs” the data source contains fields including various content data (i.e structured products) and metadata (unstructured products which describe a product), such fields are extracted or identified for further processing) determining a first attribute of the item based on the content data and the metadata (section 3.1 “More precisely, we use the structured information as a supervision for building a feature extraction model able to extract attribute-value pairs from the unstructured product descriptions in P.” attributes are extracted from the items associated with the fields. This is also shown in figure 1 as the extracted features which are first attributes
PNG
media_image1.png
460
881
media_image1.png
Greyscale
) matching the first attribute with a second attribute in a product taxonomy; (Section 3.2 “Our approach for products matching consists of three main steps: (i) feature extraction, (ii) calculating similarity feature vectors and (iii) classification… Next, we manually label a small training set of matching and non-matching unstructured pairs of product descriptions. Subsequently, we calculate the similarity feature vectors for the labeled training product pairs (Section 3.4)” calculating similarity vectors between a first and second attribute is determining a matching degree between attributes) and based on matching the first attribute, generating a classifier that represents a category of the item. (Section 3.2 “In the final step, the similarity feature vectors are used to train a classification model for distinguishing matching and non-matching pairs (Section 3.5). After the training phase is over, we have a trained feature extraction model and a classification model.” Here the final classification generated by the classification model generates either a matching or non-matching category based on the attributes which are used to create the similarity vectors as previously noted.)
Claim 2/12
Ristoski teaches claim 1
Ristoski teaches, identifying a further item that is associated with the second attribute (Section 3.2 “After the training phase is over, we have a trained feature extraction model and a classification model…. In the application phase, we generate a set M of all possible candidate matching pairs, which leads to a large number of candidates… Then, we extract the attribute-value pairs using the feature extraction model” Section 3.3 “ identification of candidate is identifying further items associates with the second attributes) and generating a title of the further item based on the second attribute. (Section 3.3.1 “We use the database A of structured products to generate a dictionary of attributes and values. Let F represent all the attributes present in the product database A. The dictionary represents an inverted index D from A such that returns the attribute name associated with a string value v. In the example depicted in Table 1, f(gold) would yield color, f(apple) would yield brand, etc.…. Then, to extract features from a given product description , we generate all possible token n-grams” See figure 2 caption
PNG
media_image2.png
155
614
media_image2.png
Greyscale
Fig. 2. Example of attribute extraction from a product title. Examiner notes that the unstructured text title is used to generate further items such as various attributes (brand, size, color, etc.). Under BRI, identification of the brand attribute value is generation of a title of the further item.)
Claim 3/13
Ristoski teaches claim 2
Ristoski teaches, generating an item description of the further item based on the second attribute. (Section 3.3.1 “We use the database A of structured products to generate a dictionary of attributes and values. Let F represent all the attributes present in the product database A. The dictionary represents an inverted index D from A such that returns the attribute name associated with a string value v. In the example depicted in Table 1, f(gold) would yield color, f(apple) would yield brand, etc…. Then, to extract features from a given product description , we generate all possible token n-grams” See figure 2 caption
PNG
media_image2.png
155
614
media_image2.png
Greyscale
Fig. 2. Example of attribute extraction from a product title. Examiner notes that the unstructured text title is used to generate further items such as various attributes (brand, size, color, etc.). Under BRI, identification of the brand attribute value is generation of a description of the further item. Titles and descriptions are not considered so narrowly to describe particular entities, but rather broadly include any textual or numerical description of an item.)
Claim 4/14
Ristoski teaches claim 1
Ristoski teaches, wherein the field comprises one of a text field, a video field, or an image field, and wherein the field comprises a plurality of attributes associated with the item, (Section 3 “we use various feature extraction methods to derive a set of useful features for the product matching task
PNG
media_image3.png
245
765
media_image3.png
Greyscale
” the fields in the data source include structured text records consisting of attributes associated with the item or record.) and wherein the determining of the first attribute of the item comprises using a machine learning model to infer the first attribute of the item based on the content data and the metadata. (Section 3.1 “Our objective is to use the structured information from the product set A as supervision for identifying duplicate records in P, in combination with neural text embeddings extracted from all the records in P. More precisely, we use the structured information as a supervision for building a feature extraction model able to extract attribute-value pairs from the unstructured product descriptions in P.” the supervised system (i.e machine learning model) identifies attributes which are duplicates according to the provided structured and unstructured data which is content data and meta data.)
Claim 5/15
Ristoski teaches claim 1
Ristoski teaches, wherein the data source includes a plurality of fields that includes at least one of a text field, a video field, and an image field, and wherein the text field includes at least one of a color field, a title field, and a product description field (Section 3 “we use various feature extraction methods to derive a set of useful features for the product matching task
PNG
media_image3.png
245
765
media_image3.png
Greyscale
” the fields in the data source include structured text records consisting of attributes associated with the item or record. Which includes a title description and color.)
Claim 8/18
Ristoski teaches claim 1
Ristoski teaches, determining that the field is an image field that includes an image and using a convolutional neural network (CNN) based machine learning model to identify the first attribute associated with the item based on the image. (Section 3.1 “Every record consists of a title and a description as unstructured textual fields, and a product image” Section 4.1.3 “We use the same CNN model for extracting image embeddings as described in Section 3.3.5. In this case, we use the image vectors as such, i.e., we use the complete image vectors for the task of image classification.” the data source includes images and text, in the case where data is determined to be images the extract attributes using a CNN based on the image)
Claim Rejections - 35 U.S.C. § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. §§ 102 and 103 (or as subject to pre-AIA 35 U.S.C. §§ 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. § 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102 of this title, 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) 6/16 and 7/17 are rejected under 35 U.S.C. § 103 as being unpatentable over Ristoski further in view of Tanglersampan US Document ID US 20190042976 A1
Claim 6/16
Ristoski teaches claim 5/15
Ristoski teaches, wherein the field is the text field (Section 3.1 “We have a database A of structured products and a dataset of unstructured product descriptions P extracted from the Web. Every record consists of a title, description, URL, and a set of attribute-value pairs extracted from the title of the product, where the attributes are numeric, categorical, or free-text (an example of such a product record is given in Table 1)
Ristoski does not explicitly teach, and wherein the metadata comprises a length value of the content data, further comprising: determining that a format of the content data is a descriptive format based on the length value of the content data, the descriptive format indicating that the content data includes a product description.
Tanglersampan however when addressing contextual recommendations based on length of content metadata teaches, and wherein the metadata comprises a length value of the content data (para 0045 “The content selection module 208 can determine content for a card… features can include content attributes. Content attributes can include any attributes associated with content included in cards. Some examples of content attributes can include a type of media (e.g., an image, a video, an audio, text, etc.), a length of content,”) determining that a format of the content data is a descriptive format based on the length value of the content data, the descriptive format indicating that the content data includes a product description. (para 0043 “The card generation module 206 can generate a card for a ranked card type… In some embodiments, the card generation module 206 can create a header, content, and a footer for each card…. Multiple cards can be organized in a contextual recommendation unit, and the unit can be displayed to a user. In some embodiments, the user can scroll (e.g., horizontally or vertically) within the contextual recommendation unit to view different cards included in the contextual recommendation unit” 0045 “In some embodiments, content for a card can be determined based on machine learning techniques…. The machine learning model can be trained based on training data… The training data can include… Some examples… a length of content,…” para 0047 “FIG. 3A illustrates an example user interface 300 for providing contextual recommendations” various types of content in a determined format which can be considered product descriptions are presented to a user based in part on the length of content data in the training data.)
Accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify the text processing system of Ristoski to comprise determinations of attributes based on the length of content in training data items described by Tanglersampan. One would have been motivated to make such a combination because both reference address the use of textual attributes for information retrieval relevant to users. Further, Tanglersampan notes “the disclosed technology can provide recommendations that are relevant to what a user is trying to accomplish on a page and can help the user engage with the page in accordance with the desires of the user.” (paragraph 0029 Tanglersampan)
Claim 7/17
Ristoski/Tanglersampan teaches claim 6/16
Tanglersampan teaches, determining that a format of the content data is a shortened text format based on the length value of the content data, the shortened text format indicating that the content data includes a title. (para 0043 “The card generation module 206 can generate a card for a ranked card type… In some embodiments, the card generation module 206 can create a header, content, and a footer for each card…. Multiple cards can be organized in a contextual recommendation unit, and the unit can be displayed to a user. In some embodiments, the user can scroll (e.g., horizontally or vertically) within the contextual recommendation unit to view different cards included in the contextual recommendation unit” 0045 “In some embodiments, content for a card can be determined based on machine learning techniques…. The machine learning model can be trained based on training data… The training data can include… Some examples… a length of content,…” para 0047 “FIG. 3A illustrates an example user interface 300 for providing contextual recommendations” various types of content in a determined format including titles are presented to a user based in part on the length of content data in the training data.)
Ristoski/Tanglersampan are combined for the reasons set forth in the rejection of claim 6
Claim(s) 9/19 are rejected under 35 U.S.C. § 103 as being unpatentable over Ristoski further in view of Gur et al. “Cross-Modal Retrieval Augmentation for Multi-Modal Classification”
Claim 9
Ristoski teaches claim 1
Ristoski teaches, determining that the field is a text field that includes a plurality of words (Section 4.1.1 “Supervised text-based feature extraction …To generate the feature vectors for each instance, after the features from the text are extracted, the value of each feature is tokenized, lower-cased, and eliminated tokens shorter than 3 characters” the system first determines that data is structured which is determining that the data field is text includes plurality of words. Section 4.1.2 “Unsupervised text-based feature extraction we use neural language modeling to extract text embeddings from the unstructured product descriptions” the unsupervised text fields are also determined to be text fields processed in a different manner and includes a plurality of words)
Ristoski does not explicitly teach, and using a Bidirectional Encoder Representations from Transformers (BERT) machine learning model to identify the first attribute associated with the item based on the plurality of words.
Gur when addressing information retrieval from text teaches, and using a Bidirectional Encoder Representations from Transformers (BERT) machine learning model to identify the first attribute associated with the item based on the plurality of words. (Figure 1 caption Cross-modal alignment architecture. We use a pre-trained ResNet-152 and BERT as feature extractors with an in-batch hinge loss.”
PNG
media_image4.png
192
961
media_image4.png
Greyscale
the system describes a BERT model to identify projection attributes associates with the text provided to the model which is a plurality of words.)
Accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify the text processing system of Ristoski to comprise a BERT model for processing text in an information retrieval environment as in described by Gur. One would have been motivated to make such a combination because both references describe processing textual information with machine learning models Further, Gur notes that in information retrieval BERT models are known models to leverage strengths of text based neural network models “Retrieval components are promising because they allow for easy revision and expansion of their memory… In the multi modal setting, retrieval augmentation allows for leveraging the strengths of text-based models—as evidenced by the strong performance of BERT based models in vision-and-language… Being able to seamlessly “hot swap” knowledge sources without the need for re-training the model affords a unique scalability not typically seen in the traditional deep learning literature.” (Gur Introduction)
Claim(s) 10 are rejected under 35 U.S.C. § 103 as being unpatentable over Ristoski further in view of Karamanolakis et al. “TXtract: Taxonomy-Aware Knowledge Extraction for Thousands of Product Categories”
Claim 10
Ristoski teaches claim 1
Ristoski teaches, the content data including at least one of structured data and unstructured data; (Section 3.1 “We define the problem of product matching similarly… We have a database A of structured products and a dataset of unstructured product descriptions P extracted from the Web.”)
Ristoski does not explicitly teach, generating a graph based at least on content data and metadata associated with a plurality of items, using a machine learning model to extract a plurality of features based on the graph; and identifying correlations of the plurality of items based on the plurality of features.
Karamanolakis when addressing using a taxonomical hierarchy for conditional machine learning knowledge extraction teaches, generating a graph based at least on content data and metadata associated with a plurality of items; and identifying correlations of the plurality of items based on the plurality of features. (Section 3.2 “We represent the product taxonomy as a tree C, where the root node is named “Product” and each taxonomy node corresponds to a distinct product category” the data is structured as a tree which amounts to generating a graph, the tree is made up of the named metadata features such as “product” and the categories which is the content data.) using a machine learning model to extract a plurality of features based on the graph (Figure 2 pg 4
PNG
media_image5.png
538
923
media_image5.png
Greyscale
the taxonomy or graph is provided as input the machine learning model to extract embeddings or features, which are correlated in the CondSelfAtt stage of the model)
Accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify the text processing system of Ristoski to comprise encodings conditional on a taxonomy graph described by Karamanolakis. One would have been motivated to make such a combination because both reference address the use of textual attributes for information retrieval relevant to users of e-commerce platforms. Further, Karamanolakis notes “Experiments on products from a taxonomy with 4,000 categories show that TXtract outperforms state-of-the-art approaches by up to 10% in F1 and 15% in coverage across all categories (Karamanolakis abstract)
Conclusion
Prior art not relied upon:
Chen et al. “Cross-Modal Knowledge Adaptation for Language-Based Person Search” describes a system for matching images and unstructured text.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOHNATHAN R GERMICK whose telephone number is (571)272-8363. The examiner can normally be reached M-F 9:30-4:30.
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, Kakali Chaki can be reached on 571-272-3719. 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.
/J.R.G./
Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights