DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
35 U.S.C. §103
2. Applicant's arguments, see Remarks pp. 10 -16, filed April 1st 2026, with
respect to the rejections of claims 1, 18 and 19 under 35 U.S.C. §103 have been fully considered and they are not persuasive.
First, applicant argues that the Brinker reference does not in fact disclose or suggest the features of a document score comprising “one or more confidence scores indicative of the model’s confidence in accurately determining corresponding one or more attributes of the document” or the one or more cluster identifiers being “indicative of a low confidence score class of documents for which one or more low confidence scores have been allocated”
Examiner respectfully disagrees. The said paragraph [0034] of the Brinkler reference teaches, “a feature for each word in a document may be determined. These features may then be used to generate a feature vector for the document” The words in a document are part composition of the attributes of a document. Indeed, word-related data is a core part of a document's attributes, usually referred to as metadata or document properties. Therefore the feature vectors of these words are indicative of how attributes of the document may be determined. Further in paragraph [0036] Brinkler teaches “Using the single feature vector for each document generated in step 210, a single index can be used to determine the closest cluster among the set of candidate clusters without performing an exhaustive search, improving the
clustering speed over methods multiple feature vectors.” Determining how close candidate clusters are based on the feature vectors is mathematical comparison that results in a determination that those that a closer will tend to be given a mathematical representation albeit a score indicative of their relevance.
Second, applicant argues that a skilled person would not have combined Brinkler and Lewis in the manner suggested by the Examiner without the benefit of hindsight.
Examiner respectfully disagrees. In response to applicant's argument that the
examiner’s conclusion of obviousness is based upon improper hindsight reasoning, it must be recognized that any judgment on obviousness is in a sense necessarily a reconstruction based upon hindsight reasoning. But so long as it takes into account only knowledge which was within the level of ordinary skill at the time the claimed invention was made, and does not include knowledge gleaned only from the applicant's disclosure, such a reconstruction is proper. See In re McLaughlin, 443 F.2d 1392, 170 USPQ 209 (CCPA 1971).
Third applicant argues, that, nothing in the Brinkler reference discloses or suggest any value that can be considered a confidence score “indicative of the model’s confidence in accurately determining corresponding one ore more attributes of the document” as claimed.
Examiner respectfully disagrees. Brinkler in paragraph [0043] teaches, “In
step 304, a feature vector of a document (received from feature extractor 116 or memory 102) or information indicative of the feature vector is sent to cluster hash 110 from cluster module 118. Cluster hash 110, which contains information (e.g., cluster identification information) for clusters stored in database 104, returns a set of candidate clusters to the cluster module 118 in step 306. That is, based on the feature vector of the document a hash value is calculated. The hash value is used to identify a set of cluster identifiers stored in the cluster hash 110. The feature vectors ( e.g., centroids) of the candidate clusters are selected to be similar to the feature vector of the document with high probability”. From the paragraph a hash value is calculated that forms the base of a probability determination of the model’s confidence.
Fourth, applicant argues that the cited documents also fail to disclose or suggest, “associating each document with its respective cluster identifier; wherein one or more cluster identifiers are indicative of a low confidence score class of document for which one or more low confidence scores have been allocated" (emphasis added).”
Examiner respectfully disagrees. The Brinkler reference teaches inter alia, , “In at least one embodiment, the predetermined threshold is approximately 0.75. If the distance measure between the document's feature vector and any cluster centroid is below a predetermined threshold, the document is associated with a cluster in step 314.” paragraph [0047])”
The statutory rejection is maintained.
Claim Rejections – 35 U.S.C. §103
3. 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.
4. The factual inquiries set forth in Graham v John Deere Co., 383 U.S. 1, 148 USPQ
459 (1966), that are applied for establishing a background for determining obviousness
under 35 U.S.C. 103 are summarized as follows:
a. Determining the scope and contents of the prior art
b. Ascertaining the differences between the prior art and the claims at issue
c. Resolving the level of ordinary skill in the pertinent art
d. Considering objective evidence present in the application indicating
obviousness or nonobviousness
Claims 1, 18 and 19 are rejected under 35 U.S.C. 103 as being unpatentable Brinker et al. (United States Patent Publication Number 2008/0205775 ), hereinafter Brinker, in view of Lewis et al. (United States Patent Number 9367814) hereinafter Lewis
Regarding claim 1 Brinker teaches a method (Fig. 2, flowchart of method of document clustering [0015]) comprising: a) determining a plurality of documents associated with an entity; (Documents may be news articles in a data stream from one or more online news services, article servers, or other document databases and/or servers. [0025]) b) for each document: (each document [0030]) i) providing the document to a numerical representation generation model; (Term Frequency Inverse Document Frequency (TFIDF) [0030]) such as “numerical representation generation model” ii) generating, (In calculation of TFIDF, N is the number of documents processed, Document Length (DLk) is the length of the kth document in words, and Document Frequency (DF) is the number of documents having each word. [0032]) by the numerical representation generation model, (Term Frequency Inverse Document Frequency (TFIDF) [0030]) such as “numerical representation generation model” a numerical representation of the document; (a feature for each word in a document may be determined. [0034]) such as “numerical representation of the document” iii) providing the numerical representation (a feature for each word in a document may be determined. [0034]) such as “numerical representation of the document” to a prediction model; (the feature extractor 116 [0030]) such as “prediction model” and iv) generating, (generated by the feature extractor 116 [0030]) by the prediction model, (feature extractor 116 [0030]) a document score (feature vector [0034]) such as “document score” for the document (Documents may be news articles in a data stream from one or more online news services, article servers, or other document databases and/or servers. [0025]) based on the numerical representation, (a feature for each word in a document may be determined. [0034]) such as “numerical representation of the document” wherein the document score (feature vector [0034]) such as “document score” comprises one or more confidence scores indicative of confidence of the prediction model in accurately determining corresponding one or more attributes of the document; (a mathematical representation … based on information about the words in the document [0034]) such as “confidence indication/mathematical certainty” c) providing document scores (feature vector [0034]) such as “document score” for the documents (Documents may be news articles in a data stream from one or more online news services, article servers, or other document databases and/or servers. [0025]) to a clustering module; (documents are clustered by cluster module 118. [0035]) d) determining, by the clustering module, (cluster module 118. [0035]) one or more clusters, (candidate clusters [0036]) each cluster being associated with a class of the documents; (Clusters may also each have a centroid- a feature vector indicative of the documents in that cluster. That is, a feature vector describing and/or generally representative of all the documents in each cluster may be generated and/or used as a cluster centroid. [0042]) e) outputting, by the clustering module, (cluster module 118. [0035])a cluster identifier (cluster identifiers [0043]) indicative of the class of each document;) (ABS., the cluster identifiers each indicative of a respective cluster centroid) (cluster centroid [0042]) and f) associating (If the distance measure between the document's feature vector and any cluster centroid is below a predetermined threshold, the document is associated with a cluster in step 314. [0047])each document(Documents may be news articles in a data stream from one or more online news services, article servers, or other document databases and/or servers. [0025]) with its respective cluster identifier, (cluster identifiers [0043]) wherein one or more cluster identifiers (cluster identifiers [0043])are indicative of a low confidence score class of document for which one or more low confidence scores have been allocated; (In step 312, a determination is made as to whether the distance between the feature vector of the document and the centroid of the cluster is less than a threshold. In at least one embodiment, the predetermined threshold is approximately 0.75. If the distance measure between the document's feature vector and any cluster centroid is below a predetermined threshold, the document is associated with a cluster in step 314. [0047]) )
Brinker does not fully disclose and wherein the method further comprises: selecting the documents of the one or more low confidence score classes for label review; and/or retraining the prediction model used to generate the confidence scores using documents from the one or more low confidence score classes of documents.
Lewis teaches selecting the documents of the one or more low confidence score classes (ABS., One or more classified documents that are associated
with a classification confidence level below a predetermined threshold value are selected to create a set of low confidence documents. ) ((3) selecting
one or more classified documents that are associated with a 65 classification confidence level below a predetermined threshold value to create a set of low-confidence documents; Col 4 ln 63 - 66) for label review; (ABS., A user is prompted to enter a classification within the taxonomy for at least one low-confidence document.) (and (5) prompting a user to enter a classification within the taxonomy for at least one low-confidence document, wherein the low-confidence document
is associated with the entered classification Col 5 ln 1 - 4) such as “label review” and/or retraining the prediction model used to generate the confidence scores using documents from the one or more low confidence score classes of documents (If the confidence level is lower than a predefined threshold, the document is declassified and passed to a manual review process. A manual reviewer classifies the document (e.g., by tagging with the correct label), and the document is used as "golden" data (a document with a high classification confidence level) in the next iteration of the training process. Accordingly, operation of the classifiers may be gradually adjusted over time based on manual classification of low-confidence documents. Col 4 ln 34 - 42)
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker to incorporate the teachings of Lewis wherein the method further comprises: selecting the documents of the one or more low confidence score classes for label review; and/or retraining the prediction model used to generate the confidence scores using documents from the one or more low confidence score classes of documents. By doing so the low-confidence document is associated with the entered classification and with a predetermined confidence level to create a newly classified document. Lewis Col 5 ln 43 - 45
Claims 18 and 19 corresponds to claim 1 and are rejected accordingly
Claims 2, 3, 5, 11 – 13 and 15 are rejected under 35 U.S.C. 103 as being unpatentable Brinker et al. (United States Patent Publication Number 2008/0205775 ), hereinafter Brinker, in view of Lewis et al. (United States Patent Number 9367814) hereinafter Lewis and in further view of Enuka et al. (United States Patent Publication Number 20200311414 ), hereinafter Enuka,
Regarding claim 2 Brinker in view of Lewis teaches the method (method [0025]) of claim 1,
Brinker as modified teaches and performing steps b) to e)
Brinker does not fully disclose wherein each document is associated with a corresponding label of a plurality of labels, and the method further comprises: determining a dataset for each label of the plurality of labels, the dataset comprising the documents associated with the label;
Enuka as modified further teaches wherein each document is associated with a corresponding label of a plurality of labels, ((each document is labeled based on content categories such as 'sensitive', 'marketing', 'financial', etc. (plurality of labels); [0089]),) and the method further comprises: determining a dataset for each label of the plurality of labels, the dataset comprising the documents associated with the label; ((the label is propagated (determining a dataset) to all documents in
cluster (associated with the label); para [00911)) for each dataset separately ((document labeling is performed by a clustering algorithm performed iteratively (for each dataset); [0075), [0089)).)
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Enuka wherein each document is associated with a corresponding label of a plurality of labels, and the method further comprises: determining a dataset for each label of the plurality of labels, the dataset comprising the documents associated with the label. By doing so one or more documents
722 in a given cluster 720a may be selected and manually labeled 760 by a user, and such label may then be propagated to all documents (e.g., document 721) in the corresponding cluster 720b. Enuka [0091]
Regarding claim 3 Brinker in view of Lewis teaches the method (method [0025]) of claim 1,
Brinker as modified does not fully disclose wherein the numerical representation is a multi-dimensional vector, and wherein the method further comprises: providing the numerical representation to a dimensionality reduction model to determine the document score.
Enuka teaches wherein the numerical representation(document feature vector [0027]) is a multi-dimensional vector, (unique IDs are assigned to all tokens in a current document, and the number of occurrences of each token is recorded to create a document feature vector, in step 220; Fig 2; para [0042))) and wherein the method further comprises: providing the numerical representation (document feature vector [0027]) to a dimensionality reduction model to determine the document score ((a software module (dimensionality reduction model) to determine a similarity score between the current document feature vector (numerical representation) and the cluster feature vector of the current cluster; Fig 3; [0061)).)
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Enuka wherein the numerical representation is a multi-dimensional vector, and wherein the method further comprises: providing the numerical representation to a dimensionality reduction model to determine the document score. By doing so the feature vector of one document may be selected as the representative cluster feature vector. The selected feature vector may be a feature vector most representative of the documents in the cluster. Enuka [0062]
Regarding claim 5 Brinker in view of Lewis teaches the method (method [0025]) of claim 1,
Brinker as modified does not fully disclose wherein the one or more attributes comprise: (i) amount; (ii) entity; (iii) due date; (iv) bill date; (v) invoice number; (vi) tax amount; and/or (vii) currency.
Enuka teaches wherein the one or more attributes(document information such as textual context and/or any associated metadata information. Exemplary textual content may include, but is not limited to, characters, words, sequences, symbols, etc. And exemplary metadata information may include, but is not limited to, date created, date modified, date last opened, tags, author, custodian, recipient, copyees, assignee, signatories, party names, audience, brand, language, personal identity information present, word count, page count, source, tone/sentiment, security level, attachment range, file type/extension, path name, hash value, signature date, effective date, and/or expiration date [0028]) comprise: (i) amount; (ii) entity; (iii) due date; (effective date, and/or expiration date. [0028])(iv) bill date; (v) invoice number; (vi) tax amount; and/or (vii) currency.
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Enuka wherein the one or more attributes comprise: (i) amount; (ii) entity; (iii) due date; (iv) bill date; (v) invoice number; (vi) tax amount; and/or (vii) currency. By doing so a document may comprise or otherwise be represented by document information such as textual context and/or any associated metadata information. Enuka [0028]
Regarding claim 11 Brinker in view of Lewis teaches the method (method [0025]) of claim 1,
Brinker as modified does not fully disclose wherein determining, by the clustering module, one or more clusters, comprises performing k-means clustering, density-based spatial clustering (DBSCAN) or hierarchical clustering.
Enuka teaches wherein determining, by the clustering module, (Fig. 4, (412) combined method, (414) vectorized method [0076]); such as “clustering module” one or more clusters, (the inventive clustering methods were employed to cluster 7,438 unstructured documents, including 38 Non-Disclosure Agreements ("NDAs"). [0075])comprises performing k-means clustering, (Fig. 4A K-Means method [0020], [0076]) density-based spatial clustering (DBSCAN) or hierarchical clustering.
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Enuka wherein determining, by the clustering module, one or more clusters, comprises performing k-means clustering, density-based spatial clustering (DBSCAN) or hierarchical clustering. By doing so the K-means algorithm works iteratively to optimize the position of the centroids and stops once the
centroids have stabilized or the number of iterations has been reached. Enuka [0075]
Regarding claim 12 Brinker in view of Lewis teaches the method (method [0025]) of claim 1,
Brinker as modified does not fully disclose wherein each document comprises character data, and the method further comprising: for each document: providing the character data to an attribute determination model; determining an attribute associated with the document based on the character data; and associating the document with the determined attribute as a label.
Enuka teaches, wherein each document (documents 110 [0026]) comprises character data, (A document may comprise or otherwise be represented by document information such as textual context and/or any associated metadata information. Exemplary textual content may include, but is not limited to, characters, words, sequences, symbols, etc. [0028]) and the method (method [0025]) further comprising: for each document: (documents 110 [0026]) providing the character data (short words (e.g., words comprising less than 3 characters or less than 4 characters), [0033]) to an attribute determination model; (At step 210, a current document is preprocessed. Generally, preprocessing of a document may include one or more of: transforming document text to lowercase, removing ( e.g., filtering) HTML, XML and/or other programming language tags, removing excess whitespace, removing punctuation, removing stop words, removing short words (e.g., words comprising less than 3 characters or less than 4 characters), removing numeric characters, and/or word
stemming or lemmatization. [0033]) determining an attribute (any one of (date created, date modified, date last opened, tags, author, custodian, recipient, copyees, assignee, signatories, party names, audience, brand, language, personal identity information present, word count, page count, source, tone/sentiment, security level, attachment range, file type/extension, path name, hash value, signature date, effective date, and/or expiration date. [0028]) associated with the document (documents 110 [0026]) based on the character data; (short words (e.g., words comprising less than 3 characters or less than 4 characters), [0033]) and associating the document(documents 110 [0026]) with the determined attribute (any one of (date created, date modified, date last opened, tags, author, custodian, recipient, copyees, assignee, signatories, party names, audience, brand, language, personal identity information present, word count, page count, source, tone/sentiment, security level, attachment range, file type/extension, path name, hash value, signature date, effective date, and/or expiration date. [0028]) as a label (categorizing documents based on content ( e.g. "sensitive," "marketing," "financial," etc.). [0089])
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Enuka wherein each document comprises character data, and the method further comprising: for each document: providing the character data to an attribute determination model; determining an attribute associated with the document based on the character data; and associating the document with the determined attribute as a label. By doing so a current document is preprocessed.
Generally, preprocessing of a document may include one or more of: transforming document text to lowercase, removing ( e.g., filtering) HTML, XML and/or other programming language tags, removing excess whitespace, removing punctuation,
removing stop words, removing short words (e.g., words comprising less than 3 characters or less than 4 characters), removing numeric characters, and/or word
stemming or lemmatization. It will be appreciated that the system may preprocess individual documents and/or may preprocess batches of documents retrieved from database. Enuka [0033]
Regarding claim 13 Brinker in view of Lewis teaches the method (method [0025]) of claim 1,
Brinker as modified does not fully disclose wherein each document comprises image data, and the method further comprises: for each document: extracting the image data from the document; providing the image data to an attribute determination model; determining, by the attribute determination model, an attribute associated with the document based on the image data; and associating the document with the determined attribute as a label.
Enuka teaches, wherein each document (documents 110 [0026]) comprises image data, (symbols [0028]) such as “image data” and the method (method [0025]) further comprises: for each document: (documents 110 [0026]) extracting (extraction method [0017]) the image data (symbols [0028]) such as “image data” from the document; (documents 110 [0026]) providing the image data (symbols [0028]) such as “image data” to an attribute determination model; (At step 210, a current document is preprocessed. Generally, preprocessing of a document may include one or more of: transforming document text to lowercase, removing ( e.g., filtering) HTML, XML and/or other programming language tags, removing excess whitespace, removing punctuation, removing stop words, removing short words (e.g., words comprising less than 3 characters or less than 4 characters), removing numeric characters, and/or word stemming or lemmatization. [0033]) determining, by the attribute determination model, (At step 210, a current document is preprocessed. Generally, preprocessing of a document may include one or more of: transforming document text to lowercase, removing ( e.g., filtering) HTML, XML and/or other programming language tags, removing excess whitespace, removing punctuation, removing stop words, removing short words (e.g., words comprising less than 3 characters or less than 4 characters), removing numeric characters, and/or word stemming or lemmatization. [0033]) an attribute (any one of (date created, date modified, date last opened, tags, author, custodian, recipient, copyees, assignee, signatories, party names, audience, brand, language, personal identity information present, word count, page count, source, tone/sentiment, security level, attachment range, file type/extension, path name, hash value, signature date, effective date, and/or expiration date. [0028]) associated with the document(documents 110 [0026]) based on the image data; (symbols [0028]) such as “image data” and associating the document (documents 110 [0026]) with the determined attribute (any one of (date created, date modified, date last opened, tags, author, custodian, recipient, copyees, assignee, signatories, party names, audience, brand, language, personal identity information present, word count, page count, source, tone/sentiment, security level, attachment range, file type/extension, path name, hash value, signature date, effective date, and/or expiration date. [0028]) as a label (categorizing documents based on content ( e.g. "sensitive," "marketing," "financial," etc.). [0089])
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Enuka wherein each document comprises image data, and the method further comprises: for each document: extracting the image data from the document; providing the image data to an attribute determination model; determining, by the attribute determination model, an attribute associated with the document based on the image data; and associating the document with the determined attribute as a label. By doing so the container images can be managed, version
controlled and downloaded from a container hub, or loaded from compressed files in case the organization's environment does not allow hub access. Enuka [0094]
Regarding claim 15 Brinker in view of Lewis and Enuka teaches the method (method [0025]) of claim 12,
Brinker as modified further teaches wherein the attribute (a mathematical representation … based on information about the words in the document [0034]) such as “confidence indication/mathematical certainty” is an entity associated with the document. (Documents may be news articles in a data stream from one or more online news services, article servers, or other document databases and/or servers. [0025]
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable Brinker et al. (United States Patent Publication Number 2008/0205775 ), hereinafter Brinker, in view of Lewis et al. (United States Patent Number 9367814) hereinafter Lewis in view of Enuka et al. (United States Patent Publication Number 20200311414 ), hereinafter Enuka, and in further view of Van et al. (AU 2009227778), hereinafter referred to as Van
Regarding claim 4 Brinker in view of Lewis and Enuka teaches the method (method [0025]) of claim 3,
Brinker as modified does not fully disclose wherein the dimensionality reduction model performs Principal Component Analysis (PCA) to generate a dimensionally reduced numerical representation of the document, and the method further comprises: multiplying the dimensionally reduced numerical representation by a variance-ratio to determine the document score.
Enuka as modified further teaches wherein the dimensionality reduction model ((a software module (dimensionality reduction model) [0061]) to determine the document score (similarity scores between documents [0074])
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Enuka wherein the dimensionality reduction model to determine the document score. By doing so the feature vector of one document may be selected as the representative cluster feature vector. The selected feature vector may be a feature vector most representative of the documents in the cluster. Enuka [0062]
Van teaches performs Principal Component Analysis (PCA) to generate a dimensionally reduced numerical representation of the document, (Principle Component Analysis (PCA) is a well known approach for dimensional reduction. Given a set of data points in high-dimensional space, PCA operates by computing a small set of basis vectors along which the distribution of data points in the set has the largest spread Pages 7, 10)and the method further comprises: multiplying the dimensionally reduced numerical representation by a variance-ratio (multiplying the determined dimensional feature vector by a known scaling operator (variance-ratio) to produce a single transformation matrix; Page 15, lines 5-12).
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis and Enuka to incorporate the teachings of Van whereby performs Principal Component Analysis (PCA) to generate a dimensionally reduced numerical representation of the document, and the method further comprises: multiplying the dimensionally reduced numerical representation by a variance-ratio. By doing so the benefit of overcoming deficiencies of document retrieval operations by improving the ability to identify documents and quickly search and retrieve contents. Van Page 5, lines 6-8; Page 7, lines 17-18),
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable Brinker et al. (United States Patent Publication Number 2008/0205775 ), hereinafter Brinker, in view of Lewis et al. (United States Patent Number 9367814) hereinafter Lewis and in further view of Thomas et al., (United States Patent Publication Number 20200394509) hereinafter Thomas
Regarding claim 6 Brinker in view of Lewis teaches the method (method [0025]) of claim 1,
Brinker as modified teaches the numerical representation generation model (Term Frequency Inverse Document Frequency (TFIDF) [0030]) such as “numerical representation generation model”
Brinker as modified does not fully disclose comprising: applying a filter to the document to blur image data of the document before providing the filtered document to
Thomas teaches applying a filter to the document to blur image data of the document before providing the filtered document to (the convolutional neural network was originally developed for image processing (e.g., implemented as image filters to blur or sharpen images). [0028])
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Thomas wherein applying a filter to the document to blur image data of the document before providing the filtered document. By doing so a convolution layer captures features in a plurality of small regions of data (e.g., a particular n-gram feature) and at every location the corresponding feature is converted to a low-dimensional vector with information relevant to the task being preserved, which is termed convolutional mapping herein. The mapping is shared among all the locations, so that useful features can be detected irrespective of their locations. Thomas [0028]
Claims 9 and 10 are rejected under 35 U.S.C. 103 as being unpatentable Brinker et al. (United States Patent Publication Number 2008/0205775 ), hereinafter Brinker, in view of Lewis et al. (United States Patent Number 9367814) hereinafter Lewis and in further view of Schuetze et al., (United States Patent Publication Number 2003/0074368) hereinafter Schuetze
Regarding claim 9 Brinker in view of Lewis teaches the method (method [0025]) of claim 1,
Brinker as modified does not fully disclose wherein determining, by the clustering module, one or more clusters, comprises: determining a plurality of histogram bins based on the document scores; determining a bin score for each document; and grouping histogram bins into clusters.
Schuetze teaches wherein determining, by the clustering module, (Fig. 9, wavefront clustering [0056]; Fig. 10, k-means clustering [0057]) one or more clusters, (Figs. 12 23 various text and image clusters [0060] – [0070]) comprises: determining a plurality of histogram bins (To embed images into vector space, two modalities have been successfully used: color histogram and complexity. For the color histogram feature, image documents are embedded into Rn", where nh is the number of "bins" in the histogram (twelve, in a presently preferred embodiment of the invention). Preferably, a single color histogram is used as the color feature. [0112]) based on the document scores; (documents probability scores [0110]) determining a bin score (Given N bins, and a maximum horizontal dimension of ~, any horizontal run rx
longer than ~/4 is placed into the Nth (or last) bin. Shorter runs rx are placed into the bin indexed by floor(rx(N-1)/(nx/4))+1 (where the "floor" function rounds its argument down to the nearest integer). [0118]) for each document; (image document [0112]) and grouping histogram bins (For the histogram feature, it is sufficient to calculate the proper histogram bin for only the subsampled pixels. For the complexity feature, it is also necessary to determine the lengths of runs, both horizontal and vertical that subsampled pixels belong to. [0123]) into clusters. (clusters [0132])
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Schuetze wherein determining, by the clustering module, one or more clusters, comprises: determining a plurality of histogram bins based on the document scores; determining a bin score for each document; and grouping histogram bins into clusters. By doing so the image features are used independently of text features to create multiple clusterings which the human user can navigate between, using text (e.g., section headings, abstract title, "ALT" tags in image anchors) when it is perceived to be more appropriate, and image features when they are more so. Schuetze [0149]
Regarding claim 10 Brinker in view of Lewis and Schuetze teaches the method of claim 9,
Brinker as modified does not fully disclose wherein grouping histogram bins into clusters comprises determining the clusters as respective local minima of the histogram bins.
Schuetze teaches wherein grouping histogram bins (color histogram and complexity. For the color histogram feature, image documents are embedded into Rn", where nh is the number of "bins" in the histogram (twelve, in a presently preferred embodiment of the invention). Preferably, a single color histogram is used as the color feature. [0112]) into clusters (Fig. 9, wavefront clustering [0056]; Fig. 10, k-means clustering [0057]) comprises determining the clusters as respective local minima (Each pixel in the image being processed is then categorized (step 612): its hue, saturation, and value will fall into one of the four bins for each dimension, so the corresponding vector element is incremented (step 614). In a preferred embodiment of the invention, the color histogram for each document is normalized (step 616) so that all of the bin values sum to one-the result is then stored as the histogram vector (step 618). [0113]) of the histogram bins(color histogram and complexity. For the color histogram feature, image documents are embedded into Rn", where nh is the number of "bins" in the histogram (twelve, in a presently preferred embodiment of the invention). Preferably, a single color histogram is used as the color feature. [0112])
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Schuetze wherein grouping histogram bins into clusters comprises determining the clusters as respective local minima of the histogram bins. By doing so the distance between histograms can computed via an intersection measure with normalization by the largest bin value. Schuetze [0114]
Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable Brinker et al. (United States Patent Publication Number 2008/0205775 ), hereinafter Brinker, in view of Lewis et al. (United States Patent Number 9367814) hereinafter Lewis and in further view of Hu et al., (United States Patent Publication Number 20210303939) hereinafter Hu
Regarding claim 14 Brinker in view of Lewis teaches the method of claim 1,
Brinker as modified does not fully disclose wherein each document comprises image data, and the method further comprises: for each document: extracting the image data from the document; providing the image data to an image-based numerical representation generation module; determining by the image-based numerical representation generation module, an image-based numerical representation of the document; providing character data to a character-based numerical representation generation module; determining by the character-based numerical representation generation module, a character-based numerical representation of the document; providing, to a consolidated numerical representation generation module, the image-based numerical representation of the document and the character-based numerical representation of the document; generating, by the consolidated numerical representation generation module, a combined numerical representation of the character data and the image data of the document; providing the combined numerical representation to an attribute prediction module; determining, by the attribute prediction module, an attribute associated with the document; and associating the document with the determined attribute as a label.
Hu teaches wherein each document (Fig. 1, (110) electronic document, e.g. web page [0035]) comprises image data, (Fig. 1, (104) input image [0035]) and the method (method [0023]) further comprises: for each document: (Fig. 1, (110) electronic document, e.g. web page [0035]) extracting (extracting [0037]) the image data (Fig. 1, (104) input image [0035]) from the document; (Fig. 1, (110) electronic document, e.g. web page [0035]) providing (providing [0104]) the image data(Fig. 1, (104) input image [0035]) to an image-based numerical representation generation module; (image-processing logic of the item name identifier system 102 [0037]) such as “image-based numerical representation generation module” determining (determining [0047]) by the image-based numerical representation generation module, (image-processing logic of the item name identifier system 102 [0037]) such as “image-based numerical representation generation module” an image-based numerical representation (first vector [0038]) such as “image-based numerical representation” of the document; (Fig. 1, (110) electronic document, e.g. web page [0035]) providing (providing [0104]) character data (Fig. 1, (112) external text [0037]) such as “character data” to a character-based numerical representation generation module; (context-processing logic of the item name identifier system 102, [0037]) such as “character-based numerical representation generation module” determining (determining [0047]) by the character-based numerical representation generation module, (context-processing logic of the item name identifier system 102, [0037]) such as “character-based numerical representation generation module” a character-based numerical representation (second vector [0038]) such as “character-based numerical representation” of the document; (Fig. 1, (110) electronic document, e.g. web page [0035]) providing, (providing [0104]) to a consolidated numerical representation generation module, (Fig. 1. (142) Fusion Logic [0046]) such as “consolidated numerical representation generation module” the image-based numerical representation(first vector [0038]) such as “image-based numerical representation” of the document(Fig. 1, (110) electronic document, e.g. web page [0035]) and the character-based numerical representation (second vector [0038]) such as “character-based numerical representation” of the document; (Fig. 1, (110) electronic document, e.g. web page [0035]) generating, (generating [0075]) by the consolidated numerical representation generation module, (Fig. 1. (142) Fusion Logic [0046]) such as “consolidated numerical representation generation module” a combined numerical representation (combined fusion information [0046]) such as “combined numerical representation” of the character data (Fig. 1, (112) external text [0037]) such as “character data” and the image data (Fig. 1, (104) input image [0035]) of the document; (Fig. 1, (110) electronic document, e.g. web page [0035]) providing (providing [0104]) the combined numerical representation (combined fusion information [0046]) such as “combined numerical representation” to an attribute prediction module; (fully-connected (FC) neural networks [0046]) such as “attribute prediction module” determining, (determining [0047]) by the attribute prediction module, (fully-connected (FC) neural networks [0046]) such as “attribute prediction module” an attribute (text information [0089]) such as “attribute” associated with (can associate [0089]) the document; (Fig. 1, (110) electronic document, e.g. web page [0035]) and associating (can associate [0089]) the document(Fig. 1, (110) electronic document, e.g. web page [0035]) with the determined attribute (text information [0089]) such as “attribute” as a label (The example-producing system 1104 can associate each labeled image with text information [0089])
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Hu wherein each document comprises image data, and the method further comprises: for each document: extracting the image data from the document; providing the image data to an image-based numerical representation generation module; determining by the image-based numerical representation generation module, an image-based numerical representation of the document; providing character data to a character-based numerical representation generation module; determining by the character-based numerical representation generation module, a character-based numerical representation of the document; providing, to a consolidated numerical representation generation module, the image-based numerical representation of the document and the character-based numerical representation of the document; generating, by the consolidated numerical representation generation module, a combined numerical representation of the character data and the image data of the document; providing the combined numerical representation to an attribute prediction module; determining, by the attribute prediction module, an attribute associated with the document; and associating the document with the determined attribute as a label. By doing so the technique is said to adopt a fusion approach because it fuses the multi-modal evidence into an output conclusion that identifies at least one item name associated with the input image. Hu [0002]
Claims 16 and 17 are rejected under 35 U.S.C. 103 as being unpatentable Brinker et al. (United States Patent Publication Number 2008/0205775 ), hereinafter Brinker, in view of Lewis et al. (United States Patent Number 9367814) hereinafter Lewis and in further view of Leavitt et al., (United States Patent Publication Number 20050075979) hereinafter Leavitt
Regarding claim 16 Brinker in view of Lewis teaches the method (method [0025]) of claim 1,
Brinker as modified does not fully disclose wherein the documents are derived from previously reconciled accounting documents of an accounting system, each of which has been associated with a respective entity, and wherein a label of each document is indicative of the respective entity.
Lewis teaches, wherein a label of each document (label to each document Col 7 ln 27) is indicative of the respective entity (account management, billing, campaign management, performance, policy, etc. Col 3 ln 27 - 28)
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker to incorporate the teachings of Lewis wherein a label of each document is indicative of the respective entity. By doing so a product line manager may want to understand the more fine-grained breakdown of issues. Lewis Col 3 ln 29 - 31
Leavitt teaches wherein the documents (Fig. 5 (510) XML documents, (520) EDI documents [0093]) are derived from previously reconciled accounting documents (ABS., adjustment documents) (Fig. 4 (425) adjustment document [0088]) such as “reconciled accounting documents SEE ALSO [0060] of an accounting system,(accounting system [0070]) each of which has been associated with a respective entity (The term “buyer” is broadly drawn to include any entity submitting payment including distributors, purchasing groups, independent third parties, franchisees, transporters and other entities that are involved in the purchasing process. [0232]) (Similarly, the term "seller" has also been used throughout, but may actually represent a third party or hosted application or other arrangement. Consequently, the
term seller may be broadly drawn to include any entity receiving payment for goods or services [0233])
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Leavitt wherein the documents are derived from previously reconciled accounting documents of an accounting system, each of which has been associated with a respective entity. By doing so the invoice information 305 may include an invoice number to be used by the seller 330 for identification and tracking purposes. For example, the invoice information 305 may include an invoice number so that the seller 330 may be able to track which goods and/or services have been delivered or provided to buyer 310.Leavitt [0078]
Regarding claim 17 Brinker in view of Lewis teaches the method of claim 1,
Brinker as modified does not fully disclose wherein the document is an accounting document and the class of documents includes one or more of: (i) an invoice; (ii) a credit note; (iii) a receipt; (iv) a purchase order; and (v) a quote.
Leavitt teaches wherein the document (Fig. 5 (510) XML documents, (520) EDI documents [0093]) is an accounting document (ABS., adjustment documents) (Fig. 4 (425) adjustment document [0088]) such as “reconciled accounting documents SEE ALSO [0060] and the class of documents includes one or more of: (i) an invoice; (invoices [0235]) (ii) a credit note; (disbursement or credit [0061]) (iii) a receipt; (the financial institution 320 may electronically note the date of receipt, amount, payer, payee, and
any account, MICR, or invoice numbers on the check. [0080]) (iv) a purchase order; (purchase order [0118]) and (v) a quote. (price quote [0078])
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Brinker in view of Lewis to incorporate the teachings of Leavitt wherein the document is an accounting document and the class of documents includes one or more of: (i) an invoice; (ii) a credit note; (iii) a receipt; (iv) a purchase order; and (v) a quote. By doing so the invoice information 305 may include an invoice number to be used by the seller 330 for identification and tracking purposes. For example, the invoice information 305 may include an invoice number so that the seller 330 may be able to track which goods and/or services have been delivered or provided to buyer 310.Leavitt [0078]
Conclusion
5. 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.
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.
6. Any inquiry concerning this communication or earlier communications from the
examiner should be directed to Kweku Halm whose telephone number is (469)295-
9144. The examiner can normally be reached on 9:00AM - 5:30PM Mon - Thur. If
attempts to reach the examiner by telephone are unsuccessful, the examiner's
supervisor, Sanjiv Shah can be reached on (571) 272 - 4098. 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 the
Patent Application Information Retrieval (PAIR) system. Status information for published
applications may be obtained from either Private PAIR or Public PAIR. Status information
for unpublished applications is available through Private PAIR only. For more
information about the PAIR system, see http://pair-direct.uspto.gov. Should you have
questions on access to the Private PAIR system, contact the Electronic Business Center
(EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer
Service Representative or access to the automated information system, call 800-786-
9199 (IN USA OR CANADA) or 571-272-1000.
/KWEKU WILLIAM HALM/Examiner, Art Unit 2166
/SANJIV SHAH/Supervisory Patent Examiner, Art Unit 2166
Prosecution Insights