DETAILED ACTION
Receipt of Applicant’s Amendment, filed March 12, 2026 is acknowledged.
Claims 1, 2, 5, 6, 8, 11, 12, 15, 16, 18, and 20 were amended.
Claims 7 and 17 were cancelled.
Claims 1-6, 8-16, 18-20 are pending in this office action.
Claim Interpretation
Within claims 1, 11, and 20, claim 1 recites “wherein clusters of the words represent features”. Claims 11 and 20 appear to recite substantially similar language. Paragraph [0076] of the original specification recites “The clusters each thereby represent a different feature”. Paragraph [0062] recites “Accordingly, a plurality of clusters may be established, that each include two or more words determined to have a relatively high correlation with one another… Each cluster of the words represents a feature of at least one of the documents. For context, a feature defines a plurality of words that have a relatively high correlation with one another in one or more of the documents” One of ordinary skill in the art would therefore recognize the term ‘feature’ and ‘cluster’ as describing the same thing, e.g. a collection of “two or more words determined to have a relatively high correlation with each other”.
Claim Objections
Claims 5-7 and 15-17 are objected to because of the following informalities. Appropriate correction is required.
With regard to claims 5-6 and 15-16, claim 5 recites “the first matrix icnldues elements”. The remaining claims recite similar language with claims 6 and 16 referring to the second matrix. This claim limitation lacks antecedent basis. Claim 1 recites “elements” as being an aspect of the third matrix not the first. Each unique claim element is expected to have a unique claim label and each unique claim label is expected to refer to the same claim label. The use of the single label “element” to refer to two distinct claim elements (e.g. something within the third matrix verse something within the first matrix) raises an antecedent basis issue. Distinct labels should be used (e.g. first elements and third elements for the respective first and third matrix).
For examination purposes the elements of claims 5 and 15 have been construed as referring to –first elements--, the elements of claims 6 and 16 have been construed as referring to –second elements--.
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 5, 6, 15, and 16 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.
With regard to claims 5, 6, 15, and 16, the claims recite:
Claim 5 recites “wherein the elements of the first matrix indicate a frequency that a given word …appears in the document that the given word is extracted from”. Claim 15 recites substantially similar language.
Claim 6 recites “wherein the elements of the second matrix indicate a frequency that a given word … appears in the deduplication chunks.” Claim 16 recites substantially similar language.
Claim 1 recites “a second matrix based on deduplication chunks by determining, for each deduplication chunk, a frequency that a first word occurs … wherein the elements of the third matrix indicate frequencies that the features appear in the at least one of the documents” Claim 11 recites substantially similar language.
These claim limitations lack antecedent basis. It is unclear how many frequencies are being recited. Each unique claim label is expected to refer to a unique claim element. The instant claim uses the same label (e.g. ‘frequency’) to apparently refer to three distinct elements (e.g. first matrix value, second matrix value, third matrix value). The descriptions of said frequencies also raise the question of what if any of the frequencies are actually the same calculation or not. One description is describing frequencies ‘that a given word appears in the document’, while a second description is for frequencies ‘that a given word appears in the deduplication chunk’.
For examination purposes this claim limitation has been construed recite two distinct frequencies (e.g. first frequency, second frequency); the first frequency that a given word appears in the document; the second frequency that a given word appears in the deduplication chunk’; wherein the first matrix includes the first frequency; the second matrix includes the second frequency; the third matrix includes the first frequency.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-6, 8-16, 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Holt [6701305] in view of Baker [2015/0339288].
With regard to claim 1 Holt teaches A computer-implemented method, comprising:
generating a first matrix as the term-by-document matrix D (Holt, Column 6, lines 16-19 “The text data collection is represented by a term-by-document matrix having a plurality of entries with each entry representing the frequency of occurrence of a term in a respective document”; Column 3, lines 20-21) based on words extracted (Holt, Column 9, lines 59-62 “the logic of generating a term list. The logic of FIG. 2 moves from a start block to block 130 where terms are tokenized according to a tokenizing policy”) from documents (Holt, Column 9, lines 5-8 “If so, the logic moves to block 104 where a term list is generated from the initial document collection. Generating a term list from the initial document collection is illustrated in detail FIG. 2,”);
generating a second matrix as working matrix Ak (Holt, Column 11, 53-60 “For example, the working matrix A can be projected into a k dimensional subspace, thereby defining the subspace representation Ak. While the working matrix A can be projected into the subspace according to a variety of techniques including a variety of orthogonal decompositions, the projection of A into the subspace is typically performed via a two-sided orthogonal matrix decomposition”) based on [[ as sets (Holt, Column 11, lines 66 – Column 12, line 5 “Statistically, the effect of the TURV is to combine the original large set of variables into a smaller set of more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features”) by determining, for each [[ as for the set (Holt, Column 11, lines 66 – Column 12, line 5 “Statistically, the effect of the TURV is to combine the original large set of variables into a smaller set of more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features” ; Please see the 103 combination detailed bellow), a frequency that a first word occurs(Holt, Column 3, lines 22-28 “The term-by-document matrix can then be preprocessed to define a working matrix A by normalizing the columns of the term-by-document matrix D to have a unit sum, stabilizing the variance of the term frequencies via a nonlinear function and then centering the term frequencies with respect to the mean vector of the columns”);
determining a total count of the [[(Holt, Column 7, lines 25-29 “each term is weighted by determining an inverse one-norm of the term, i.e., the inverse of the sum of the absolute values of the entries of the row of the subspace representation Ak corresponding to the term.” ; Please see the 103 combination detailed bellow); and
determining a count of the [[as mean and average calculates involve summing the total and dividing by the number of occurrences (Holt, Column 3, lines 25-31 “stabilizing the variance of the term frequencies via a nonlinear function and then centering the term frequencies with respect to the mean vector of the columns. This preprocessing is denoted as A=f(D)-ceT in which c is the mean of the columns of f(D) and e is a d-vector whose components are all 1, so that the average of the columns of A is now” ; Please see the 103 combination detailed bellow), wherein the [[ as terms of the document (Holt, Column 12, lines 3-5 “more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features”);
performing word clustering (Holt, Column 18, lines 26-30 “In this regard and as with any classification method, there is a training phase where a training sample is used to determine a classifier and a classification phase that uses this classifier to determine the manner in which new documents will be classified into classes.”) based on results of as the document being classified into classes (Holt, Column 18, lines 26-31 “In this regard and as with any classification method, there is a training phase where a training sample is used to determine a classifier and a classification phase that uses this classifier to determine the manner in which new documents will be classified into classes”) an analysis as the classification (Id; Holt, Column 18, lines 36-41 “A transformation for generating a subspace representation of the classes is then generated from the matrix by using a two-sided orthogonal 40 decomposition, analogous to the indexing of a term-by document matrix D for information retrieval”) performed on the second matrix as the subspace representation Ak (Holt, Column 18, lines 49-51 “Those portions of the subspace representation Ak of the term-by-class matrix that relate to the terms of the document to be classified”), wherein clusters of the words (Holt, Column 6, lines 60-63 “A term-by-class matrix is formed from this training set having a plurality of entries with each entry representing the frequency of occurrence of a term in all the documents assigned to a class”) represent features as the class represents related terms of the document (Holt, Column 18, lines 49-51 “Those portions of the subspace representation Ak of the term-by-class matrix that relate to the terms of the document to be classified”; Column 12, lines 1-5 “The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features.” Please note this claim limitation has been interpreted in light of paragraph [0062] which recites “a feature defines a plurality of words that have a relatively high correlation with one another in one or more of the documents”) of at least one of the documents as the document (Id), wherein the analysis comprises calculating, based on word vectors, distances between the words (Column 17 line 66 - Column 18 line 2, “The similarity between the query vector and the document vectors is then determined by measuring the distance there between”);
generating a third matrix (Holt, Column 6, lines 60-63 “A term-by-class matrix is formed from this training set having a plurality of entries with each entry representing the frequency of occurrence of a term in all the documents assigned to a class”) based on the first matrix as the information stored in the term-by-document matrix is the term to document frequencies used to generate the term-by-class matrix (Holt, Column 18, lines 35-37 “The entries of this matrix are the frequencies of the terms in the documents that belong to a given class.”; Column 6, lines 16-19) and the clusters as the class (Holt, Column 18, lines 15-17 “documents can be classified into none, one or more of a plurality of predefined classes as shown in FIGS. 7”), wherein the third matrix (Holt, Column 6, lines 60-63 “A term-by-class matrix”) includes elements of the third matrix that define rows and columns of the third matrix as a matrix with the row and columns being the term and the assigned class, and the entry being the frequency there between (Holt, Column 6, lines 6-63 “A term-by-class matrix is formed from this training set having a plurality of entries with each entry representing the frequency of occurrence of a term in all the documents assigned to a class.”), wherein the elements of the third matrix indicate frequencies that the features appear in the at least one of the documents as the frequency of occurrence of that term in the documents (Id); and
performing text mining (Holt, Column 9, lines 47-50 “If so, the logic moves to block 118 for performance of a text mining operation, namely, an information retrieval operation as depicted in FIG. 6.”; Column 19, lines 33-35 “A display 66 is provided for viewing text mining data, and interacting with a user interface to request text mining operations.”) using the third matrix (Holt, Column 6, lines 60-63 “A term-by-class matrix”).
Holt does not explicitly teach deduplication chunks. Baker teaches deduplication chunks (Baker, ¶40 “Once this has been done, for each pair of articles, deduplication module 210 may extract the uni grams, bi grams and trigrams from each pair of preprocessed bodies of text and converted into sets of tokens.”; ¶46 “the article that is contained by the superset article is classified as a duplicate and removed”) by determining, for each deduplication chunk, a frequency that a first word occurs ( Baker, ¶92 “TF/IDF algorithm”);
determining a total count of the deduplication chunks such as for example, 5 unigrams (Baker, ¶31 “The sentence "The dog chased the cat" has 5 unigrams: "the", "dog", "chased", "the", "cat", that is each individual word token”); and
determining a count of the deduplication chunks that the first word occurs in as for example the unigram “the” has a count of 2 in the example ‘The dog chased the cat’ (Baker, ¶32 “From the previous example, the sentence "The dog chased the cat" may be represented as a vector of the following form: [2, 1, 1, 1]. This vector has 4 dimensions.”).
It would have been obvious to one of ordinary skill to which said subject matter pertains at the time the invention was filed to have implemented the projection of the data into sets of a reduced number of dimension using the deduplication analysis taught by Baker as it yields the predictable results of reducing the storage space for the article (Baker, ¶35 “Deduplication module 210 may first use the titles of each article in the set as a filtering stage to reduce the search space to be explored for article deduplication”)
With regard to claims 2 and 12, the proposed combination further teaches wherein the first matrix is a relatively higher dimension representation of the documents (Holt, Column 2, lines 30-33 “individual documents are treated as vectors in a high-dimensional vector space in which each dimension corresponds to some feature of a document.”), wherein the third matrix is a relatively lower dimension representation of the documents (Holt, Column 11, lines 51-53 “the matrix A is projected into a lower dimensional subspace.”).
With regard to claims 3 and 13, the proposed combination further teaches
determining, subsequent to performing the text mining (Holt, Column 9, lines 47-50 “If so, the logic moves to block 118 for performance of a text mining operation, namely, an information retrieval operation as depicted in FIG. 6.”; Column 19, lines 33-35 “A display 66 is provided for viewing text mining data, and interacting with a user interface to request text mining operations.”), whether the elements of the third matrix as the entire subspace (Holt, Column 12, lines 13-18 “As will be described hereinafter, the entire subspace representation Ak need not always be determined. Instead, only those portions, i.e., those rows, of the subspace representation Ak that correspond to the terms included within the query must be determined”) exceed as only the portions necessary (Id) a predetermined number of elements as the rows that correspond to the terms that must be determined (Id), and in response to a determination that the elements of the third matrix do not exceed the predetermined number of elements: increasing a threshold used for performing the word clustering as expanding the document subspace (Holt, Column 12, lines 37-41 “Next, in block 172, the existing term subspace Uk is augmented with the normalized residual, which is orthogonal to the original term subspace, and the document subspace, Vk, is expanded by adding a small identity matrix accordingly.”), updating the clusters and re-generating the third matrix based on the updated clusters (Column 12, lines 27 “Still referring to FIG. 4, the logic then moves to block 164 where a new subspace representation is determined by updating the existing subspace with new documents and terms”).
With regard to claims 4 and 14, the proposed combination further teaches wherein performing text mining (Holt, Column 9, lines 47-50 “If so, the logic moves to block 118 for performance of a text mining operation, namely, an information retrieval operation as depicted in FIG. 6.”; Column 19, lines 33-35 “A display 66 is provided for viewing text mining data, and interacting with a user interface to request text mining operations.”) using the third matrix (Holt, Column 6, lines 60-63 “A term-by-class matrix”) includes running a text mining program as the text mining operation (Holt, Column 9, lines 47-50 “If so, the logic moves to block 118 for performance of a text mining operation, namely, an information retrieval operation as depicted in FIG. 6.”; Column 19, lines 33-35 “A display 66 is provided for viewing text mining data, and interacting with a user interface to request text mining operations.”) on the relatively lower dimension representation of the documents (Holt, Column 11, lines 51-53 “the matrix A is projected into a lower dimensional subspace.”).
With regard to claims 5 and 15, the proposed combination further teaches wherein the first matrix includes elements of the first matrix that define rows and columns of the first matrix as the term-by-document matrix D, wherein one of ordinary skill in the art would recognize the term and document as the row and column, with the element as the frequency of those (Holt, Column 6, lines 16-19), wherein the elements of the first matrix indicate a frequency that a given word of the words extracted from the documents (Holt, Column 9, lines 59-62 “the logic of generating a term list. The logic of FIG. 2 moves from a start block to block 130 where terms are tokenized according to a tokenizing policy”) appears in the documents (Holt, Column 6, lines 16-19 “The text data collection is represented by a term-by-document matrix having a plurality of entries with each entry representing the frequency of occurrence of a term in a respective document”).
With regard to claims 6 and 16, the proposed combination further teaches wherein the second matrix includes elements of the second matrix that define rows and columns of the second matrix as working matrix A, wherein one of ordinary skill in the art would recognize the matrix as having rows and columns that define the matrix (Holt, Column 3, lines 22-28), wherein the elements of the second matrix as working matrix A (Holt, Column 3, lines 22-28 “The term-by-document matrix can then be preprocessed to define a working matrix A by normalizing the columns of the term-by-document matrix D to have a unit sum, stabilizing the variance of the term frequencies via a nonlinear function and then centering the term frequencies with respect to the mean vector of the columns.”) indicate a frequency that a given word of the words extracted from the documents (Holt, Column 9, lines 59-62 “the logic of generating a term list. The logic of FIG. 2 moves from a start block to block 130 where terms are tokenized according to a tokenizing policy”) appears as the normalized and stabilized term frequencies (Id) in a the deduplication chunks as the sets (Holt, Column 11, lines 66 – Column 12, line 5 “Statistically, the effect of the TURV is to combine the original large set of variables into a smaller set of more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features”; Please note this claim limitation has been construed in light of ¶46 which recites “These groups are used as features, which are used to represent a document instead of words, with lower dimension.” Wherein the chunks instead of the original words themselves).
With regard to claims 8 and 18, the proposed combination further teaches
wherein the deduplication chunks also include punctuation as the punctuation that determines the boundary (¶47 “Preprocessing module 212 may then perform sentence splitting on the tokenized article text by splitting the text into individual sentences by using punctuation as anchors to determine sentence
boundaries.”) and/or strings of numbers as the string of digits representing money (Holt, ¶55 “Another positive scoring factor may be the number of money expressions present in the sentence (e.g., "$20", "$2", "$1 ,000"). … followed by one of more digits”) in the documents (Holt, Column 9, lines 5-8).
With regard to claims 9 and 19, the proposed combination further teaches causing the documents to be stored into a deduplication storage (Holt, Column 19, lines 30-31 “The computer also includes nonvolatile storage 64, such as a hard disk drive, where data is stored”), wherein content of the documents is split into the deduplication chunks as the sets (Holt, Column 11, lines 66 – Column 12, line 5 “Statistically, the effect of the TURV is to combine the original large set of variables into a smaller set of more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features”) during the documents being stored into the deduplication storage (Holt, Column 19, lines 30-31 “The computer also includes nonvolatile storage 64, such as a hard disk drive, where data is stored”).
With regard to claim 10, the proposed combination further teaches wherein data normalization is not performed to generate matrixes (Holt, Column 10, lines 19-20 “This policy may be not to perform any term normalization, thereby making this an optional step”).
With regard to claim 11, the proposed combination teaches A computer program product, the computer program product comprising a computer readable storage medium (Holt, Column 19, lines 27-30 “The computer 50 includes a processing unit 60 and a system memory 62 which includes random access memory (RAM) and read-only memory (ROM)”) having program instructions (Holt, Column 19, lines 46-51 “computer program instructions may be loaded onto the computer or other programmable apparatus to produce a machine, such that the instructions which execute on the computer or other programmable apparatus create means for implementing the functions specified in the block diagram, flowchart or control flow block(s) or step(s).”) embodied therewith, the program instructions readable and/or executable by a computer to cause the computer to:
Generate, by the computer (Holt, Column 19, line 21 “general purpose computer 50”), a first matrix (Holt, Column 6, lines 16-19 “The text data collection is represented by a term-by-document matrix having a plurality of entries with each entry representing the frequency of occurrence of a term in a respective document”) based on words extracted (Holt, Column 9, lines 59-62 “the logic of generating a term list. The logic of FIG. 2 moves from a start block to block 130 where terms are tokenized according to a tokenizing policy”) from documents (Holt, Column 9, lines 5-8 “If so, the logic moves to block 104 where a term list is generated from the initial document collection. Generating a term list from the initial document collection is illustrated in detail FIG. 2,”);
generate, by the computer (Holt, Column 19, line 21 “general purpose computer 50”), a second matrix as working matrix Ak(Holt, Column 11, 53-60 “For example, the working matrix A can be projected into a k dimensional subspace, thereby defining the subspace representation Ak. While the working matrix A can be projected into the subspace according to a variety of techniques including a variety of orthogonal decompositions, the projection of A into the subspace is typically performed via a two-sided orthogonal matrix decomposition”) based on [[ as sets (Holt, Column 11, lines 66 – Column 12, line 5 “Statistically, the effect of the TURV is to combine the original large set of variables into a smaller set of more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features”) by: determining, for each [[ as for the set (Holt, Column 11, lines 66 – Column 12, line 5 “Statistically, the effect of the TURV is to combine the original large set of variables into a smaller set of more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features”; Please see the 103 combination detailed bellow), a frequency that a first word occurs(Holt, Column 3, lines 22-28 “The term-by-document matrix can then be preprocessed to define a working matrix A by normalizing the columns of the term-by-document matrix D to have a unit sum, stabilizing the variance of the term frequencies via a nonlinear function and then centering the term frequencies with respect to the mean vector of the columns”);
determining a total count of the [[(Holt, Column 7, lines 25-29 “each term is weighted by determining an inverse one-norm of the term, i.e., the inverse of the sum of the absolute values of the entries of the row of the subspace representation Ak corresponding to the term.”; Please see the 103 combination detailed bellow); and
determining a count of the [[as mean and average calculates involve summing the total and dividing by the number of occurrences (Holt, Column 3, lines 25-31 “stabilizing the variance of the term frequencies via a nonlinear function and then centering the term frequencies with respect to the mean vector of the columns. This preprocessing is denoted as A=f(D)-ceT in which c is the mean of the columns of f(D) and e is a d-vector whose components are all 1, so that the average of the columns of A is now”; Please see the 103 combination detailed bellow), wherein the [[ as terms of the document (Holt, Column 12, lines 3-5 “more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features”);
perform, by the computer (Holt, Column 19, line 21 “general purpose computer 50”), word clustering (Holt, Column 18, lines 26-30 “In this regard and as with any classification method, there is a training phase where a training sample is used to determine a classifier and a classification phase that uses this classifier to determine the manner in which new documents will be classified into classes.”) based on results of as the document being classified into classes (Holt, Column 18, lines 26-31 “In this regard and as with any classification method, there is a training phase where a training sample is used to determine a classifier and a classification phase that uses this classifier to determine the manner in which new documents will be classified into classes”) an analysis as the classification (Id; Holt, Column 18, lines 36-41 “A transformation for generating a subspace representation of the classes is then generated from the matrix by using a two-sided orthogonal 40 decomposition, analogous to the indexing of a term-by document matrix D for information retrieval”) performed on the second matrix as the subspace representation Ak (Holt, Column 18, lines 49-51 “Those portions of the subspace representation Ak of the term-by-class matrix that relate to the terms of the document to be classified”), wherein clusters of the words (Holt, Column 6, lines 60-63 “A term-by-class matrix is formed from this training set having a plurality of entries with each entry representing the frequency of occurrence of a term in all the documents assigned to a class”) represents features as the class represents related terms of the document (Holt, Column 18, lines 49-51 “Those portions of the subspace representation Ak of the term-by-class matrix that relate to the terms of the document to be classified”; Column 12, lines 1-5 “The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features.” Please note this claim limitation has been interpreted in light of paragraph [0062] which recites “a feature defines a plurality of words that have a relatively high correlation with one another in one or more of the documents”) of at least one of the documents as the document (Id), wherein the analysis comprises calculating, based on word vectors, distances between the words (Column 17 line 66 - Column 18 line 2, “The similarity between the query vector and the document vectors is then determined by measuring the distance there between”);
generate, by the computer (Holt, Column 19, line 21 “general purpose computer 50”), a third matrix (Holt, Column 6, lines 60-63 “A term-by-class matrix is formed from this training set having a plurality of entries with each entry representing the frequency of occurrence of a term in all the documents assigned to a class”) based on the first matrix as the information stored in the term-by-document matrix is the term to document frequencies used to generate the term-by-class matrix (Holt, Column 18, lines 35-37 “The entries of this matrix are the frequencies of the terms in the documents that belong to a given class.”; Column 6, lines 16-19) and the clusters as the class (Holt, Column 18, lines 15-17 “documents can be classified into none, one or more of a plurality of predefined classes as shown in FIGS. 7”), wherein the third matrix (Holt, Column 6, lines 60-63 “A term-by-class matrix”) includes elements of the third matrix that define rows and columns of the third matrix as a matrix with the row and columns being the term and the assigned class, and the entry being the frequency there between (Holt, Column 6, lines 6-63 “A term-by-class matrix is formed from this training set having a plurality of entries with each entry representing the frequency of occurrence of a term in all the documents assigned to a class.”), wherein the elements of the third matrix indicate frequencies that the features appear in the at least one of the documents as the frequency of occurrence of that term in the documents (Id); and
perform, by the computer (Holt, Column 19, line 21 “general purpose computer 50”), text mining (Holt, Column 9, lines 47-50 “If so, the logic moves to block 118 for performance of a text mining operation, namely, an information retrieval operation as depicted in FIG. 6.”; Column 19, lines 33-35 “A display 66 is provided for viewing text mining data, and interacting with a user interface to request text mining operations.”) using the third matrix (Holt, Column 6, lines 60-63 “A term-by-class matrix”).
Holt does not explicitly teach deduplication chunks. Baker teaches deduplication chunks (Baker, ¶40 “Once this has been done, for each pair of articles, deduplication module 210 may extract the uni grams, bi grams and trigrams from each pair of preprocessed bodies of text and converted into sets of tokens.”; ¶46 “the article that is contained by the superset article is classified as a duplicate and removed”) by determining, for each deduplication chunk, a frequency that a first word occurs ( Baker, ¶92 “TF/IDF algorithm”);
determining a total count of the deduplication chunks such as for example, 5 unigrams (Baker, ¶31 “The sentence "The dog chased the cat" has 5 unigrams: "the", "dog", "chased", "the", "cat", that is each individual word token”); and
determining a count of the deduplication chunks that the first word occurs in as for example the unigram “the” has a count of 2 in the example ‘The dog chased the cat’ (Baker, ¶32 “From the previous example, the sentence "The dog chased the cat" may be represented as a vector of the following form: [2, 1, 1, 1]. This vector has 4 dimensions.”).
It would have been obvious to one of ordinary skill to which said subject matter pertains at the time the invention was filed to have implemented the projection of the data into sets of a reduced number of dimension using the deduplication analysis taught by Baker as it yields the predictable results of reducing the storage space for the article (Baker, ¶35 “Deduplication module 210 may first use the titles of each article in the set as a filtering stage to reduce the search space to be explored for article deduplication”).
With regard to claim 20, the proposed combination teaches A system, comprising:
a processor (Holt, Column 19, lines 27-30 “The computer 50 includes a processing unit 60 and a system memory 62 which includes random access memory (RAM) and read-only memory (ROM)”); and
logic (Holt, Column 19, lines 46-51 “computer program instructions may be loaded onto the computer or other programmable apparatus to produce a machine, such that the instructions which execute on the computer or other programmable apparatus create means for implementing the functions specified in the block diagram, flowchart or control flow block(s) or step(s).”) integrated with the processor, executable by the processor, or integrated with and executable by the processor, the logic being configured to:
Generate, a first matrix (Holt, Column 6, lines 16-19 “The text data collection is represented by a term-by-document matrix having a plurality of entries with each entry representing the frequency of occurrence of a term in a respective document”) based on words extracted (Holt, Column 9, lines 59-62 “the logic of generating a term list. The logic of FIG. 2 moves from a start block to block 130 where terms are tokenized according to a tokenizing policy”) from documents (Holt, Column 9, lines 5-8 “If so, the logic moves to block 104 where a term list is generated from the initial document collection. Generating a term list from the initial document collection is illustrated in detail FIG. 2,”);
generate, a second matrix as working matrix Ak(Holt, Column 11, 53-60 “For example, the working matrix A can be projected into a k dimensional subspace, thereby defining the subspace representation Ak. While the working matrix A can be projected into the subspace according to a variety of techniques including a variety of orthogonal decompositions, the projection of A into the subspace is typically performed via a two-sided orthogonal matrix decomposition”) based on [[ as sets (Holt, Column 11, lines 66 – Column 12, line 5 “Statistically, the effect of the TURV is to combine the original large set of variables into a smaller set of more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features”) by: determining, for each [[ as for the set (Holt, Column 11, lines 66 – Column 12, line 5 “Statistically, the effect of the TURV is to combine the original large set of variables into a smaller set of more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features”; Please see the 103 combination detailed bellow), a frequency that a first word occurs(Holt, Column 3, lines 22-28 “The term-by-document matrix can then be preprocessed to define a working matrix A by normalizing the columns of the term-by-document matrix D to have a unit sum, stabilizing the variance of the term frequencies via a nonlinear function and then centering the term frequencies with respect to the mean vector of the columns”);
determining a total count of the [[(Holt, Column 7, lines 25-29 “each term is weighted by determining an inverse one-norm of the term, i.e., the inverse of the sum of the absolute values of the entries of the row of the subspace representation Ak corresponding to the term.”; Please see the 103 combination detailed bellow); and
determining a count of the [[as mean and average calculates involve summing the total and dividing by the number of occurrences (Holt, Column 3, lines 25-31 “stabilizing the variance of the term frequencies via a nonlinear function and then centering the term frequencies with respect to the mean vector of the columns. This preprocessing is denoted as A=f(D)-ceT in which c is the mean of the columns of f(D) and e is a d-vector whose components are all 1, so that the average of the columns of A is now”; Please see the 103 combination detailed bellow), wherein the [[ as terms of the document (Holt, Column 12, lines 3-5 “more semantically significant features. The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features”);
perform, word clustering (Holt, Column 18, lines 26-30 “In this regard and as with any classification method, there is a training phase where a training sample is used to determine a classifier and a classification phase that uses this classifier to determine the manner in which new documents will be classified into classes.”) based on results of as the document being classified into classes (Holt, Column 18, lines 26-31 “In this regard and as with any classification method, there is a training phase where a training sample is used to determine a classifier and a classification phase that uses this classifier to determine the manner in which new documents will be classified into classes”) an analysis as the classification (Id; Holt, Column 18, lines 36-41 “A transformation for generating a subspace representation of the classes is then generated from the matrix by using a two-sided orthogonal 40 decomposition, analogous to the indexing of a term-by document matrix D for information retrieval”) performed on the second matrix as the subspace representation Ak (Holt, Column 18, lines 49-51 “Those portions of the subspace representation Ak of the term-by-class matrix that relate to the terms of the document to be classified”), wherein clusters of the words (Holt, Column 6, lines 60-63 “A term-by-class matrix is formed from this training set having a plurality of entries with each entry representing the frequency of occurrence of a term in all the documents assigned to a class”) represents features as the class represents related terms of the document (Holt, Column 18, lines 49-51 “Those portions of the subspace representation Ak of the term-by-class matrix that relate to the terms of the document to be classified”; Column 12, lines 1-5 “The coordinates of the projected data in the reduced number of dimensions can be used to characterize the documents, and therefore represent the effect of thousands or tens of thousands of terms in a few hundred or more significant features.” Please note this claim limitation has been interpreted in light of paragraph [0062] which recites “a feature defines a plurality of words that have a relatively high correlation with one another in one or more of the documents”) of at least one of the documents as the document (Id), wherein the analysis comprises calculating, based on word vectors, distances between the words (Column 17 line 66 - Column 18 line 2, “The similarity between the query vector and the document vectors is then determined by measuring the distance there between”);
generate, a third matrix (Holt, Column 6, lines 60-63 “A term-by-class matrix is formed from this training set having a plurality of entries with each entry representing the frequency of occurrence of a term in all the documents assigned to a class”) based on the first matrix as the information stored in the term-by-document matrix is the term to document frequencies used to generate the term-by-class matrix (Holt, Column 18, lines 35-37 “The entries of this matrix are the frequencies of the terms in the documents that belong to a given class.”; Column 6, lines 16-19) and the clusters as the class (Holt, Column 18, lines 15-17 “documents can be classified into none, one or more of a plurality of predefined classes as shown in FIGS. 7”), wherein the third matrix (Holt, Column 6, lines 60-63 “A term-by-class matrix”) includes elements of the third matrix that define rows and columns of the third matrix as a matrix with the row and columns being the term and the assigned class, and the entry being the frequency there between (Holt, Column 6, lines 6-63 “A term-by-class matrix is formed from this training set having a plurality of entries with each entry representing the frequency of occurrence of a term in all the documents assigned to a class.”), wherein the elements of the third matrix indicate frequencies that the features appear in the at least one of the documents as the frequency of occurrence of that term in the documents (Id); and
perform, text mining (Holt, Column 9, lines 47-50 “If so, the logic moves to block 118 for performance of a text mining operation, namely, an information retrieval operation as depicted in FIG. 6.”; Column 19, lines 33-35 “A display 66 is provided for viewing text mining data, and interacting with a user interface to request text mining operations.”) using the third matrix (Holt, Column 6, lines 60-63 “A term-by-class matrix”).
Holt does not explicitly teach deduplication chunks. Baker teaches deduplication chunks (Baker, ¶40 “Once this has been done, for each pair of articles, deduplication module 210 may extract the uni grams, bi grams and trigrams from each pair of preprocessed bodies of text and converted into sets of tokens.”; ¶46 “the article that is contained by the superset article is classified as a duplicate and removed”) by determining, for each deduplication chunk, a frequency that a first word occurs ( Baker, ¶92 “TF/IDF algorithm”);
determining a total count of the deduplication chunks such as for example, 5 unigrams (Baker, ¶31 “The sentence "The dog chased the cat" has 5 unigrams: "the", "dog", "chased", "the", "cat", that is each individual word token”); and
determining a count of the deduplication chunks that the first word occurs in as for example the unigram “the” has a count of 2 in the example ‘The dog chased the cat’ (Baker, ¶32 “From the previous example, the sentence "The dog chased the cat" may be represented as a vector of the following form: [2, 1, 1, 1]. This vector has 4 dimensions.”).
It would have been obvious to one of ordinary skill to which said subject matter pertains at the time the invention was filed to have implemented the projection of the data into sets of a reduced number of dimension using the deduplication analysis taught by Baker as it yields the predictable results of reducing the storage space for the article (Baker, ¶35 “Deduplication module 210 may first use the titles of each article in the set as a filtering stage to reduce the search space to be explored for article deduplication”).
Response to Arguments
Applicant's arguments filed March 12, 2026 have been fully considered but they are not persuasive. All the arguments regarding the newly added limitations are addressed in the above rejections.
With regard to claim 1, applicant argues that Baker does not teach generating a second matrix based on deduplication chunks, and that Holt does not teach deduplication chunks.
In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986). Holt details that the data can be organized into sets, wherein the matrix are build using those sets. Baker teaches a means of reducing the number of dimensions of chunks of data by performing deduplication. One of ordinary skill in the art would readily recognize that the ‘sets of terms’ taught by Hold are similar enough to the ‘chunks’ taught by Baker to enable the ‘chunk deduplication’ techniques taught by Baker to be able to be performed on the sets of terms taught by Holt. Within the proposed combination, after the sets of terms have been deduplciated via Bakers techniques (e.g. the generation of the deduplicated chunks) the system may continue using those deduplciated chunks (e.g. deduplciated sets of terms) to perform the matrix generation as taught by Holt.
Applciant’s arguments do not address the proposed combination on record, and instead appear to attack the references individually.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AMANDA WILLIS whose telephone number is (571)270-7691. The examiner can normally be reached Monday-Friday 8am-2pm.
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, Ajay Bhatia can be reached at 571-272-3906. 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.
/AMANDA L WILLIS/ Primary Examiner, Art Unit 2156
Prosecution Insights