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 Amendment
In response to the office action from 3/25/2026, the applicant has submitted an amendment, filed 6/24/2026, amending claims 1, 2, 4, 5, 8 and 9, cancelling claims 3 and 6, adding claims 10-18, while arguing to traverse the prior art and 101 rejections. Applicant’s arguments have been fully considered but are moot with respect to new grounds of rejections further in view of CHOSA KATSUKI et al. (JP 2024123669), and Goyal et al. (US 2018/0285326) mandated by the latest amendments, and for the reasons explained in the response to arguments.
Response to Arguments
In what follows applicant’s arguments will be addressed in the order presented.
Pages 8-9 provide broad overviews of the latest amendments, “Preliminary Matters”, the interview of 6/19/2026, and some broad quotations of MPEP 2106.04 without any arguments.
Page 10, the first ¶ provides a broad overview of the amendments to claim 1, and in the second ¶ lines 3+, it is concluded: “amended claim 1 recites generation of a target vector “using the machine-learned model that has been optimized.” Applicant respectfully submits the previous position regarding replacement with mapping tables no longer applies, as a human cannot use a machine-learned model”.
Respectfully machine learning is a replica of how humans learn; i.e., learning by storing a repetitive feature in a human memory. Nonetheless as relates to the amendments here, the addition of the “optimization” step when claim is understood in light of the specification teachings quoted above, corresponds to a standard training process which uses ground truth. As the claim (and the specification) are void of any details on what the training data comprises that is being used along with the ground truth label, as well as any clear definition of the “ground truth label”, the added limitations amount to nothing more than a common practice in the art of using ground truth along with a target text to optimize the model. What is claimed is very broad without defining what is comprised in this target text and what is the ground truth label and connecting it to a tone which according to Sp. ¶ 0067 S1: “For example, reference data” “accumulated” “a past time” “which indicates the tones of past”. Therefore, this feature can be replicated by the human editor who could use some past experience with a similar article (training data from a similar case in a past training indicating a tone) which is determined relevant (has a ground truth label of e.g. relevance) as aid in modifying or as mentioned above replacing (correcting) the article. The use of experience with a past article could expedite his article edition (optimizing his training).
Furthermore, using “machine-learning” to convert “target data” into “target vector” amount to using an additional element (i.e., the “machine-learning”) precluding which will have no impact on the claim as the claim is silent on how the “machine-learning” is used to make the conversion; i.e., conversion of the words in natural language to vector format amounts to mapping of text to numerical format which can be achieved by using mapping tables which is completely within human domain of activities.
Following broad overviews of the latest amendments from the end of page 12 to the beginning of page 13, it is concluded that: “SHI-CONG does not disclose correcting the news text itself” “SHI CONG does not disclose reference data that indicates a tone of text”.
For the “correcting” feature amendment, the new office action relies on features of Fig. 1C of Goyal et al. in which “moon” is corrected to “Moon”. Please visit the new office action for further details.
As regards to “tone”, according to Sp. ¶ 0067 S1: “For example, reference data” “accumulated” “a past time” “which indicates the tones of past”. Therefore the “News” in SHI CONG which indicate something in the past map to the claim’s tones.
At the bottom of page 13, it is asserted that “SHI CONG does not disclose” basically the amended limitation starting with “an optimization process …”.
This basically corresponds to the cancelled claim 6, for which in the first action as well as the instant action, the office action has relied on CHOSA KATSUKI et al.; please visit the new office action for further details.
Given these facts, respectfully the examiner cannot agree to the contention that “independent” “claim 1” and “claims 8 and 9” are “patentable”.
Finally, it is asserted that the dependent claims 2, 4, 5, and 7 are patentable over the cited references for at least the reasons discussed above due to their respective dependencies”.
Since applicants have not argued the merits of these dependent claims, but assert patentability solely through their dependence on the allegedly patentable parent claims, they stand or fall with said parent claims and hence no further response to applicant’s arguments is necessary.
For the new claims (discussed on page 14), please visit the new office action.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-2, 4-5, 7-18 stand rejected:
Claims 1, 8 and 9 correspond to “apparatus”, “method” and “non-transitory recording medium” respectively that are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
The said claims recite “acquiring” a “target data” (Sp. ¶ 0024 S2: “may be” “an article and an editorial”), which is converted to a “target vector”, followed by a “comparison” of the “target vector” with a “reference vector” (Sp. ¶ 0028 S2+: “may be” “data which indicates the tones of past article and editorial created in the mass media” for “correcting” of e.g., “predetermined keyword, predetermined wording” (Sp. Par. 0029 S3+; Sp. ¶ 0034-35: “a process of making correction to an article created in the mass media” “Such that degree of difference between a feature of the article and a feature of data indicating a social trend or the like is within a predetermined range”.
The method further involves the “correct[ion]” step implemented by “optimizing a machine-learned model with reference to training data which contains the target text and a ground truth label associated with the target text”, and “according to a result of a comparison between the target vector and a reference vector generated from reference data” “making a correction to the target text or suggesting the correction to the target text” (Sp.¶ 0120: “optimize a machine-learned model with reference to training data which contains the target data and a ground truth label associated with the target data, the ground truth label being assigned to the target vector, and [0121] generate the target vector with use of the machine-learned model optimized”); as regards to the “ground truth label” Sp. ¶ 0051, 0095, 0118, 0140, 0165 all recite “a ground truth label associated with the target data, the ground truth label being assigned to the target vector” which simply do not constitute a specific definition of the ground truth label but merely a particular association of the label. In general, as understood in the art a “ground truth” can be an explicit answer, and a label could be a token assigned to that answer.
These limitations, under their broadest reasonable interpretation, cover performance of the limitations in the mind but for the recitation of “processor” (claim 1) and “computer” (claim 9). That is, other than reciting by the “processor” (claim 1) and/or “computer” (claim 9), nothing in the claim elements precludes the steps from practically being performed in the mind. For example, a human e.g. an editor of a magazine in which an article (target data) is written could take that and if he sees anything not favored by its base audience (including words, sentences, phrases …etc.) and/or what they call conforming to an acceptable trend, and replace it with something more acceptable. In so doing he could use other similar articles in the past pertaining to the same subject (i.e., claim’s “reference data”) and use them for “comparison”. If a claim limitation and/or limitations, under their broadest reasonable interpretation, cover performance of the limitations in the mind but for the recitation of generic computer components, then they fall within the “Mental Processes” grouping of abstract ideas. Accordingly, the claims recite an abstract idea.
The addition of the “optimization” step when claim is understood in light of the specification teachings quoted above, corresponds to a standard training process which uses ground truth. As the claim (and the specification) are void of any details on what the training data comprises that is being used along with the ground truth label, as well as any clear definition of the “ground truth label”, the added limitations amount to nothing more than a common practice in the art of using ground truth along with a target text to optimize the model. What is claimed is very broad without defining what is comprised in this target text and what is the ground truth label and connecting it to a tone which according to Sp. ¶ 0067 S1: “For example, reference data” “accumulated” “a past time” “which indicates the tones of past”. Therefore, this feature can be replicated by the human editor who could use some past experience with a similar article (training data from a similar case in a past training indicating a tone) as aid in modifying or as mentioned above replace (correct) the article. The use of experience with a past article could expedite his article edition (optimizing his training).
Furthermore, using “machine-learning” to convert “target data” into “target vector” amount to using an additional element (i.e., the “machine-learning”) precluding which will have no impact on the claim; i.e., conversion of the words in natural language to vector format amounts to mapping of text to numerical format which can be achieved by using mapping tables which is completely within human domain of activities.
This judicial exception is not integrated into a practical application. In particular, the claims recite only one additional element, i.e., the “processor” in claim 1 and “computer” in claim 9 to do the “acquiring” “generating” and “correcting” processes. The “processor” and/or the “computer” are recited at a high-level of generality (i.e., as a generic processor and/or computer performing generic computer functions associated with the limitations such that they amount no more than mere instructions to apply the exception using a generic computer component. Furthermore, according to ¶ 0100: “Examples of the processor C1 can include a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a tensor processing unit (TPU), a quantum processor, a microcontroller, and a combination thereof”; and/or ¶ 0099: “The computer C includes at least one processor C1 and at least one memory C2”. Accordingly, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claims are thus directed to an abstract idea.
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of using a “processor” (claim 1) and/or “computer” (claim 9) to perform the “acquiring” the “generating of” “vector” and the “correcting process” amount to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept.
Regarding claim 2, the human editor could take anything not favored by its base audience (including words, sentences, phrases …etc.) and/or what they call conforming to an acceptable trend based on past experience, and replace it (make a correction) with something more acceptable.
Regarding claim 4, the conversion of words in natural language (e.g., a sentence) to vector format amount to mapping of text to numerical format which can be achieved by using mapping tables.
Regarding claim 5, the article (target data) would be available as an input to the human editor the moment the human editor begins reading it.
Regarding claim 7, the human editor would intuitively search for words, phrases and sentences (reference data) in any category consistent with the word, phrase and or sentence (target data) under investigation including a plurality of categories for each particular target text.
Regarding claim 10, determination of an “inner product” falls under a “mathematical concepts” of the abstract ideas.
Regarding claim 11, “computing a difference” and its comparison against a “predetermined range” falls under a “mathematical concepts” of abstract ideas.
Regarding claim 12, to “evaluate a difference” and its comparison against “a numerical value” falls under a “mathematical concepts” of abstract ideas.
Regarding claim 13, determining a “degree of difference” and comparing it against a “predetermined threshold” falls under a “mathematical concepts” of abstract ideas.
Regarding claim 14, to “input” “text” “through a user interface”, without having any reverse impact on the performance of the “user interface” amounts to engaging with a device by an activity which is “well-understood, routine, conventional activity in the field” (MPEP 2106.05(I)(A)(v)).
Regarding claim 15, determination of a “difference” and its comparison against a “numerical value” falls under a “mathematical concepts” of abstract ideas.
Regarding claim 16, the other similar articles (reference data) used for making comparison with the article (target text data), are predetermined articles (predetermined wording).
Regarding claim 17, the human editor could display any changes (corrections) he suggests to an article (target text) on a sheet of paper (via a display).
Regarding claim 18, the human editor must obtain other similar articles (reference data) before (in advance of) any changes (corrections) he may suggest.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-2, 4-5, 7-18 is/are rejected under 35 U.S.C. 103 as being unpatentable over SHI CONG et al. (CN114780712A), and in view of CHOSA KATSUKI et al. (JP 2024123669), and further in view of Goyal et al. (US 2018/0285326).
Regarding claim 1, SHI CONG et al. do teach an information processing apparatus comprising at least one processor (¶ n0006: “an electronic device” (an information processing apparatus) “including a memory, a processor” (comprising at least one processor) “and a computer program stored in the memory and executable on the processor”),
the at least one processor carrying out:
an acquiring process of acquiring target text (¶ n0051 S1: “suppose the summary embedding vector and keyword embedding vector of news W1” (acquiring a target data which is in text) “are E11 and E12”, where the “news” “refers to text-based reports” (¶ n0021 S1));
a target vector generation process of generating, using [a] machine-learned model, a target vector from the target text (¶ n0051 S1: “suppose the summary embedding vector” (generating a target vector of) “and keyword embedding vector of news W1” (the target text) “are E11” (as the target vector) “and E12” (another target vector of the target data) “using machine learning algorithm” (using a machine learned model (¶ n0024 lines 4-5)); and
a correcting process of, according to a result of a comparison between the target vector and a reference vector generated from reference data indicating a tone of text to be referenced for correction of the target text (¶ n0051 S1-S2 and last S: “suppose the summary embedding vector” “and keyword embedding vector of news W1” “are E11” “and E12” (the target vector) “and the summary embedding vector and keyword embedding vector of news W2” (a reference data) “are E21 and E22” (and their associated reference vectors) “Then, the cosine similarity” (a comparison operation) “between news W1 and W2 can be the cosine similarity between vectors E11 and E21” (between “E11” (the target vector) and “E21” (the reference vector) “can further improve the accuracy” (used for a correcting process and/or correction-related processing referencing target text) “of the similarity between news items”; the “news W2” comprises of “news” (pieces of textual data associated with past events (i.e., indicating a tone of the text)).
SHI CONG et al. do not specifically disclose an optimization process of optimizing a machine-learned model with reference to training data which contains the target text and a ground truth label associated with the target text; a target vector generation process of generating, using the machine -learned model that has been optimized, a target vector from the target text, wherein the ground truth label is assigned to the target vector.
CHOSA KATSUKI et al. do teach an optimization process of optimizing a machine-learned model with reference to training data which contains the target text and a ground truth label associated with the target text; a target vector generation process of generating, using the machine -learned model that has been optimized, a target vector from the target text, wherein the ground truth label is assigned to the target vector (¶ 0120 lines 23+: “the processor takes sentences contained in a given pair of documents as input” (target text) “converts the sentences into sentence vectors” (is converted into target vector) “representing the features of the sentences using a third machine learning model” (with use of a machine learning model) “with a third parameter set, and learn the first parameter, the second parameter, and the third parameter to minimize the error” (which is optimized); ¶ 0101 “training data” “extracted using” “ground truth data” (a ground truth is associated with the target text)).
It would have therefore been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the “machine learning” method for conversion of “sentences” into “sentence vectors” of CHOSA KASUKI et al. into the “embedding vector” procedures of SHI CONG et al. would enable the combined systems and their associated methods to perform in combination as they do separately and to further enable generation of vectors such that it “minimize[ed] the error” in the vector generation as disclosed in CHOSA KATSUKI et al. ¶ 0120 last sentence.
SHI CONG et al. in view of CHOSA KATSUKI et al. do not specifically disclose:
making a correction to the target text or suggesting the correction to the target text.
Goyal et al. do teach:
making a correction to the target text or suggesting the correction to the target text (¶0040 lines 1+: “FIG. 1C illustrates a document comparison 128 that groups together changes” (makes corrections) “to a repeated term” (i.e., change everywhere in the “first digital version 100a” (target data (Fig. 1a)) in which the word “moon” is corrected to “Moon” which is based on comparison with Fig. 1B (a reference data)).
It would have therefore been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the “document-comparison” techniques of Goyal et al. into “news” (document) “similarity” (comparison) techniques of SHI CONG et al. in SHI CONG et al. in view of CHOSA KATSUKI et al. would enable the combined systems and their associated methods to perform in combination as they do separately and to further enable SHI CONG et al. to benefit of an “efficient way to review document changes” as disclosed in Goyal et al. ¶ 0028 last sentence.
Regarding claim 2, SHI CONG et al. in view of CHOSA KATSUKI et al. do not specifically disclose the information processing apparatus according to claim 1, wherein the reference data contains at least one selected from the group consisting of: respective pieces of trend data created by a plurality of persons, the pieces of trend data being at least one selected from the group consisting of sentences, voices, and images; and a plurality of pieces of data each accumulated as the target text of a past time.
Goyal et al. do teach
wherein the reference data contains at least one selected from the group consisting of: respective pieces of trend data created by a plurality of persons, the pieces of trend data being at least one selected from the group consisting of sentences, voices, and images (for a “document-comparison” (¶ 0033 line 2) example, according to ¶ 0034 lines 1+: “FIGS. 1A and 1B respectively illustrate a first digital version 100a” (target text is compared with) “and second digital version 100b” (reference data) “include a plurality of sentences” (consisting of sentences); ¶0040 lines 1+: “FIG. 1C illustrates a document comparison 128 that groups together changes” (makes corrections) “to a repeated term” (i.e., change everywhere in the target data the word “moon” to “Moon” which is based on comparison with reference data, which is a change based on a trend practiced by most people or persons to represent a planet using capital letter as its first letter));
and a plurality of pieces of data each accumulated as the target text of a past time (selection of “Moon” as the correct spelling instead of “moon” is based on common practice from past).
It would have therefore been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the “document-comparison” techniques of Goyal et al. into “news” (document) “similarity” (comparison) techniques of SHI CONG et al. in SHI CONG et al. in view of CHOSA KATSUKI et al. would enable the combined systems and their associated methods to perform in combination as they do separately and to further enable SHI CONG et al. in view of CHOSA KATSUKI et al. to benefit of an “efficient way to review document changes” as disclosed in Goyal et al. ¶ 0028 last sentence.
Regarding claim 4, SHI CONG et al. do teach the information processing apparatus according to claim 2, wherein the target text contains one or more sentences (¶ 0058: “The sentences” (sentences included in) “in the article” (e.g., “NEWS” “W1” (target text)) “are sorted” “and the top M sentences are selected as the article summary”),
and in the vector generation process, the target vector is generated with reference to at least one of the one or more sentences (¶ n0040: “the text vector and the vector” (vector generated) “of the sentence” (with reference to at least one or more sentences) “can be the embedding vector of the sentence”).
Regarding claim 5, SHI CONG et al. do teach the information processing apparatus according to claim 1, wherein the at least one processor further carries out a target text generating process of generating the target text from input data (Abstract S2: “The method comprises” “acquiring a first news set, wherein the first news set comprises a plurality of pieces of news” (the target text) “retrieved based on a user input condition” (generated based on input data)).
Regarding claim 7, SHI CONG et al. in view of CHOSA KATSUKI et al. do not specifically disclose the information processing apparatus according to claim 1, wherein in the correcting process, the target vector is optimized by correcting the target vector with reference to data which is contained in the reference data and which belongs to a first category, and
correcting again the target vector with reference to data which is contained in the reference data and which belongs to a second category.
Goyal et al. do teach
in the correcting process, the target vector is optimized by correcting the target vector with reference to data which is contained in the reference data and which belongs to a first category (¶ 0075 S3: “Specifically, by applying the deterministic classification algorithm, the document-comparison system identifies changes as part of an information-insert category” (using a first category associated with e.g. reference data) “for changes” (to make corrections in the target data with respect to the reference data) “that insert information”; e.g., see Fig. 1A also called “first digital version” (target text) versus Fig. 1B also called “second digital version” (reference data), and Fig. 1C (in which the target data undergoes change using the reference data)), and
correcting again the target vector with reference to data which is contained in the reference data and which belongs to a second category (¶ 0075 lines 13+: “a lexical-paraphrase category”(using a second category) “for changes” (to make corrections again in the target data with respect to reference data) “that replace a term or phrase with a synonym or that modify a style of terms or phrases”); ¶ 0046 S1: “To generate mapped-sentence combinations, in some embodiments, the document-comparison system applies a sentence-alignment algorithm to generate a sentence vector for each sentence within a first version of a document and a sentence vector for each sentence within a second version of a document” (in the “document-comparison” used for making “changes” (corrections) both the “first digital version” (target data) and the “second digital version” (reference data) are converted into “sentence vector” (i.e. target vector and reference vector respectively)).
It would have therefore been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the “document-comparison” techniques of Goyal et al. into “news” (document) “similarity” (comparison) techniques of SHI CONG et al. in SHI CONG et al. in view of CHOSA KATSUKI et al. would enable the combined systems and their associated methods to perform in combination as they do separately and to further enable SHI CONG et al. to benefit of an “efficient way to review document changes” as disclosed in Goyal et al. ¶ 0028 last sentence.
Regarding claim 8, SHI CONG et al. do teach an information processing method (¶ n0006: “an electronic device” (an information processing apparatus and its associated methods) “including a memory, a processor” “and a computer program stored in the memory and executable on the processor”),
comprising:
acquiring target text (¶ n0051 S1: “suppose the summary embedding vector and keyword embedding vector of news W1” (acquiring a target data which is in text) “are E11 and E12”, where the “news” “refers to text-based reports” (¶ n0021 S1));
generating, using [a] machine-learned model, a target vector from the target text (¶ n0051 S1: “suppose the summary embedding vector” (generating a target vector of) “and keyword embedding vector of news W1” (the target text) “are E11” (as the target vector) “and E12” (another target vector of the target data) “using machine learning algorithm” (using a machine learned model (¶ n0024 lines 4-5)); and
according to a result of a comparison between the target vector and a reference vector generated from reference data indicating a tone of text to be referenced for correction of the target text (¶ n0051 S1-S2 and last S: “suppose the summary embedding vector” “and keyword embedding vector of news W1” “are E11” “and E12” (the target vector) “and the summary embedding vector and keyword embedding vector of news W2” (a reference data) “are E21 and E22” (and their associated reference vectors) “Then, the cosine similarity” (a comparison operation) “between news W1 and W2 can be the cosine similarity between vectors E11 and E21” (between “E11” (the target vector) and “E21” (the reference vector) “can further improve the accuracy” (used for a correcting process and/or correction-related processing referencing target text) “of the similarity between news items”; the “news W2” comprises of “news” (pieces of textual data associated with past events (i.e., indicating a tone of the text)).
SHI CONG et al. do not specifically disclose optimizing a machine-learned model with reference to training data which contains the target text and a ground truth label associated with the target text; generating, using the machine -learned model that has been optimized, a target vector from the target text, wherein the ground truth label is assigned to the target vector.
CHOSA KATSUKI et al. do teach optimizing a machine-learned model with reference to training data which contains the target text and a ground truth label associated with the target text; generating, using the machine -learned model that has been optimized, a target vector from the target text, wherein the ground truth label is assigned to the target vector (¶ 0120 lines 23+: “the processor takes sentences contained in a given pair of documents as input” (target text) “converts the sentences into sentence vectors” (is converted into target vector) “representing the features of the sentences using a third machine learning model” (with use of a machine learning model) “with a third parameter set, and learn the first parameter, the second parameter, and the third parameter to minimize the error” (which is optimized); ¶ 0101 “training data” “extracted using” “ground truth data” (a ground truth is associated with the target text)).
It would have therefore been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the “machine learning” method for conversion of “sentences” into “sentence vectors” of CHOSA KASUKI et al. into the “embedding vector” procedures of SHI CONG et al. would enable the combined systems and their associated methods to perform in combination as they do separately and to further enable generation of vectors such that it “minimize[ed] the error” in the vector generation as disclosed in CHOSA KATSUKI et al. ¶ 0120 last sentence.
SHI CONG et al. in view of CHOSA KATSUKI et al. do not specifically disclose:
making a correction to the target text or suggesting the correction to the target text.
Goyal et al. do teach:
making a correction to the target text or suggesting the correction to the target text (¶0040 lines 1+: “FIG. 1C illustrates a document comparison 128 that groups together changes” (makes corrections) “to a repeated term” (i.e., change everywhere in the “first digital version 100a” (target data (Fig. 1a)) in which the word “moon” is corrected to “Moon” which is based on comparison with Fig. 1B (a reference data)).
It would have therefore been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the “document-comparison” techniques of Goyal et al. into “news” (document) “similarity” (comparison) techniques of SHI CONG et al. in SHI CONG et al. in view of CHOSA KATSUKI et al. would enable the combined systems and their associated methods to perform in combination as they do separately and to further enable SHI CONG et al. to benefit of an “efficient way to review document changes” as disclosed in Goyal et al. ¶ 0028 last sentence.
Regarding claim 9, SHI CONG et al. do teach a non-transitory recording medium having recorded thereon a program (¶ n0007: “a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method”)
For causing a computer to carrying out:
an acquiring process of acquiring target text (¶ n0051 S1: “suppose the summary embedding vector and keyword embedding vector of news W1” (acquiring a target data which is in text) “are E11 and E12”, where the “news” “refers to text-based reports” (¶ n0021 S1));
a target vector generation process of generating, using [a] machine-learned model, a target vector from the target text (¶ n0051 S1: “suppose the summary embedding vector” (generating a target vector of) “and keyword embedding vector of news W1” (the target text) “are E11” (as the target vector) “and E12” (another target vector of the target data) “using machine learning algorithm” (using a machine learned model (¶ n0024 lines 4-5)); and
a correcting process of, according to a result of a comparison between the target vector and a reference vector generated from reference data indicating a tone of text to be referenced for correction of the target text (¶ n0051 S1-S2 and last S: “suppose the summary embedding vector” “and keyword embedding vector of news W1” “are E11” “and E12” (the target vector) “and the summary embedding vector and keyword embedding vector of news W2” (a reference data) “are E21 and E22” (and their associated reference vectors) “Then, the cosine similarity” (a comparison operation) “between news W1 and W2 can be the cosine similarity between vectors E11 and E21” (between “E11” (the target vector) and “E21” (the reference vector) “can further improve the accuracy” (used for a correcting process and/or correction-related processing referencing target text) “of the similarity between news items”; the “news W2” comprises of “news” (pieces of textual data associated with past events (i.e., indicating a tone of the text)).
SHI CONG et al. do not specifically disclose an optimization process of optimizing a machine-learned model with reference to training data which contains the target text and a ground truth label associated with the target text; a target vector generation process of generating, using the machine -learned model that has been optimized, a target vector from the target text, wherein the ground truth label is assigned to the target vector.
CHOSA KATSUKI et al. do teach an optimization process of optimizing a machine-learned model with reference to training data which contains the target text and a ground truth label associated with the target text; a target vector generation process of generating, using the machine -learned model that has been optimized, a target vector from the target text, wherein the ground truth label is assigned to the target vector (¶ 0120 lines 23+: “the processor takes sentences contained in a given pair of documents as input” (target text) “converts the sentences into sentence vectors” (is converted into target vector) “representing the features of the sentences using a third machine learning model” (with use of a machine learning model) “with a third parameter set, and learn the first parameter, the second parameter, and the third parameter to minimize the error” (which is optimized); ¶ 0101 “training data” “extracted using” “ground truth data” (a ground truth is associated with the target text)).
It would have therefore been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the “machine learning” method for conversion of “sentences” into “sentence vectors” of CHOSA KASUKI et al. into the “embedding vector” procedures of SHI CONG et al. would enable the combined systems and their associated methods to perform in combination as they do separately and to further enable generation of vectors such that it “minimize[ed] the error” in the vector generation as disclosed in CHOSA KATSUKI et al. ¶ 0120 last sentence.
SHI CONG et al. in view of CHOSA KATSUKI et al. do not specifically disclose:
making a correction to the target text or suggesting the correction to the target text.
Goyal et al. do teach:
making a correction to the target text or suggesting the correction to the target text (¶0040 lines 1+: “FIG. 1C illustrates a document comparison 128 that groups together changes” (makes corrections) “to a repeated term” (i.e., change everywhere in the “first digital version 100a” (target data (Fig. 1a)) in which the word “moon” is corrected to “Moon” which is based on comparison with Fig. 1B (a reference data)).
It would have therefore been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the “document-comparison” techniques of Goyal et al. into “news” (document) “similarity” (comparison) techniques of SHI CONG et al. in SHI CONG et al. in view of CHOSA KATSUKI et al. would enable the combined systems and their associated methods to perform in combination as they do separately and to further enable SHI CONG et al. to benefit of an “efficient way to review document changes” as disclosed in Goyal et al. ¶ 0028 last sentence.
Regarding claim 10, SHI GONG et al. do teach the information processing apparatus according to claim 1, wherein the comparison is performed based on an index obtained through computation of at least one of an inner product of the target vector and the reference vector or norms of the target vector and the reference vector (¶ n0050 S1: “cosine similarity” (the comparison operation) “measures the similarity between two news items by the cosine of the angle” (is determined as an index based on the inner product of) “between their embedding vectors” (the target and the reference vector)).
Regarding claim 11, SHI GONG et al. do teach the information processing apparatus according to claim 1. Wherein the comparison includes computing a difference between the target vector and the reference vector in a feature space and determining whether a magnitude of the computed difference is within a predetermined range (The “cosine similarity” (a difference) computed in ¶ n0051 between the “E11” (the target vector) and “E21” (the reference vector) in “embedding vector” domain (a feature space), is compared against a “preset threshold” (¶ n0044), or it is compared a predetermined range of zero to the “preset threshold”, and/or to a predetermined range of the “preset threshold” to one).
Regarding claim 12, SHI GONG et al. do teach the information processing apparatus according to claim 1, wherein in the correcting process, the at least one processor evaluates a difference between the target vector and the reference vector in terms of a numerical value (The “cosine similarity” (a difference) responsible to “improve the accuracy” (the correcting process (¶ n0051 last S)) according to ¶ n0044 last line is compared against a “preset threshold”, wherein the said “threshold” according to ¶ n0045 S1 “can be a preset value” (is a numerical value)).
Regarding claim 13, SHI GONG et al. do teach the information processing apparatus according to claim 1, wherein in the correcting process, the at least one processor makes the correction to the target text when a degree of difference between the target vector and the reference vector exceeds a predetermined threshold (The “cosine similarity” (a degree of difference) responsible to “improve the accuracy” (the correcting process (¶ 0050 last S)) according to ¶ n0044 last S is assessed as “if the similarity between the two news items is higher than a preset threshold” (determined if it exceeds a predetermined threshold)).
Regarding claim 14, SHI GONG et al. do teach the information processing apparatus according to claim 1, wherein the target text is text input by a user through a user interface (¶ 0029 S1: “obtain” “news” (i.e. the target text) “based on user input” (obtained by text input by a user), wherein according to ¶ n0022: “input criteria are” “search criteria” “[e.g.] search keywords” “Users can enter one or more keywords”, which is done using the “Users” “electronic device[s]” which according to ¶ n0070 last S: “electronic device may include input/output devices” (a user interface enabling the user to input a request for the “news W1” (target text))).
Regarding claim 15, SHI GONG et al. in view of CHOSA KATSUKI et al. do not specifically disclose the information processing apparatus according to claim 12, wherein in the correcting process, the at least one processor does not make the correction to the target text in a case where the numerical value indicating the difference is within a correction criterion.
Goyal et al. do teach the information processing apparatus according to claim 12, wherein in the correcting process, the at least one processor does not make the correction to the target text in a case where the numerical value indicating the difference is within a correction criterion (¶ 0080 last S: “If the document-comparison system determines that the symmetric difference” (if a numerical difference between a target and a reference text) “is zero” (is within a correction criterion) “between terms of a first grouped set” (i.e., target text) “and terms of a second grouped set” (the reference text) “the document-comparison system determines” “no factual changes or paraphrasing changes” (does not make any corrections); note according to ¶ 0080 S2: “The term “symmetric difference” refers to a number of terms that are present in either of two grouped sets but are not common between the two grouped sets” (i.e., the “symmetric difference” is a numerical value)).
It would have therefore been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the “document-comparison” techniques of Goyal et al. into “news” (document) “similarity” (comparison) techniques of SHI CONG et al. in SHI CONG et al. in view of CHOSA KATSUKI et al. would enable the combined systems and their associated methods to perform in combination as they do separately and to further enable SHI CONG et al. to determine on the fly if there are any “factual changes or paraphrasing changes” between e.g. “news W1” (target text) and “news W2” (reference text) as disclosed in Goyal et al. last sentence.
Regarding claim 16, SHI GONG et al. do teach the information processing apparatus according to claim 1, wherein the reference vector represents one or more elements extracted from the reference data in form of a vector in a feature space, the one or more elements including at least one of a predetermined keyword, predetermined wording, or a token (“E21” (the reference vector) is defined as “summary embedding vector” (¶ n0051 S1) which by definition represents a “vector” (vector) of a “summary” (a predetermined wording) in an “embedding vector” (feature) space).
Regarding claim 17, SHI CONG et al. in view of CHOSA KATSUKI et al. do not specifically disclose the information processing apparatus according to claim 1, wherein in the correcting process, the at least one processor presents details of the correction to a user via a display.
Goyal et al. do teach the information processing apparatus according to claim 1, wherein in the correcting process, the at least one processor presents details of the correction to a user via a display (¶0040 lines 1+: “FIG. 1C illustrates a document comparison 128 that groups together changes” (makes the corrections) “to a repeated term within a graphical user interface” (to a user via a display)).
For obviousness to combine SHI CONG et al. in view of CHOSA KATSUKI et al. and Goyal et al. see claim 1.
Regarding claim 18, SHI CONG et al. in view of CHOSA KATSUKI et al. do not specifically disclose the information processing apparatus according to claim 1, wherein the reference data is acquired in advance of the correcting process.
Goyal et al. do teach the information processing apparatus according to claim 1, wherein the reference data is acquired in advance of the correcting process (the actions corresponding to Fig. 1C (i.e., “changes to a repeated term” (the correction in the correcting process (¶ 0040 lines 1+))) come after obtaining content associated with Fig. 1B (i.e., display of the “second digital version 100b” (the reference data)), or the reference data is acquired in advance of the correcting process).
For obviousness to combine SHI CONG et al. in view of CHOSA KATSUKI et al. and Goyal et al. see claim 1.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. CELLA et al. (US 2023/0214925) ¶ 2049: “feature vectors that help predict or anticipate news” “e.g., using predictive models” “that help identify relevant events in real time” “and data sources that indicate historical trends” (using trend data which indicate a tone) “such as reactions to similar news or events in the past” (a plurality of pieces of data each accumulated as the target data of the past). The reference though was silent on doing any document comparisons and associated corrections.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) 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 mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to FARZAD KAZEMINEZHAD whose telephone number is (571)270-5860. The examiner can normally be reached 10:30 am to 11:30 pm.
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, Paras D. Shah can be reached at (571) 270-1650. 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.
/Farzad Kazeminezhad/
Art Unit 2653
September 26th 2026.
