DETAILED ACTION
This communication is in response to the Applicant Arguments/Remarks filed 1/2/2026. Claims 1, 3-8, 10-15, 17-20 are pending in the application.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1, 3-8, 10-15, 17-20 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention. In relating to the amended limitation “wherein the likelihood is based in part on a length of each sentence” in the independent claims 1, 8, 15, the specification does not teach especially the underlined limitation. Specification para. 7 or 29 does not teach the limitation “length” or usage of “length” but recognized that “The reason the "collapse" phenomenon occurs is because the sentence vector representation method, which is based on semantic vectors, has not undergone "difference" marking training, which leads to its tendency to coalesce”.
The “difference marking” or contrastive learning which means the applied language vector model was not explicitly trained to separate similar concepts from dissimilar ones. Claims 3-7, 10-14, and 17-20 are rejected for incorporating the errors of claims 1, 8, and 15 by dependency.
Response to Arguments
Applicant's arguments filed 1/2/2026 have been fully considered.
Regarding the arguments on pages 10-13 that the cited references fail to teach the amended steps including determining…on a length of each sentence, and dividing… is less than the threshold, please see the newly cited columns and lines below.
Specification, para. 53-54 disclose “Word embedding is a term used in Natural Language Processing (NLP) for the learned representation of words for text analysis, typically in the form of a real-valued vector that encodes the meaning of the word such that the words that are closer in the vector space are similar in meaning and context. Word embeddings can be obtained using a set of language modeling and feature learning techniques where words or phrases from the vocabulary are mapped to vectors of real numbers.” Thus, words in sentences of the received set of text are vectorized in to vectors in the vector space or spatial region.
Ghoshal et al. teaches at para. 100: representation of weighted vectors of knowledge base concepts may correspond to topic modeling, in which each sentence or word first may be converted into a feature vector, and then the feature difference or similarity may be calculated in the high-dimensional vector space;
para. 122: the similarity estimate quantifies the degree of semantic relatedness between a grammatical unit and the input text from which the grammatical unit was extracted. For example, the similarity estimate may be a value between 1 and 0 inclusive with values closer to 1 representing a higher degree of semantic relatedness and values closer to 0 representing a lower degree of semantic relatedness which is equivalent to semantically unrelated category; para. 108-109;
figs. 23-24: more words are tagged in relating to “crane” which disambiguates the meaning of “crane” thus, overcomes the likelihood of collapsing.
Ghoshal also teaches at para. 82: all of the feature vectors satisfying a particular closeness threshold (e.g., distance between vectors<threshold(T)) may be selected; para. 95, 108-109: determine the grammatical unit having the highest degree of semantic relatedness to the given text document and then selecting all other granunatical units where the difference in the degree of semantic relatedness to the given text document of the grammatical unit and the highest degree is below a predefined threshold. Using a dissimilarity measure to select grammatical units for inclusion in the text summary can prevent multiple similar grammatical units from being included in the same text summary.
Regarding the argument that “Ghoshal fails to teach or suggest "dividing," "tagging," "filtering," and "vectorizing" for "sentences with a number of words that could be encoded to a spatial region that is semantically unrelated to the set of text", examiner respectfully disagrees.
Ghoshal teaches dividing/parsing the received text in fig.12: parse/divide and process original authored content, extract and identify keywords/topics/features from content; para. 79: parsing, tagging; para. 100: representation of weighted vectors of knowledge base concepts may correspond to topic modeling, in which each sentence or word first may be converted into a feature vector, and then the feature difference/similarity may be calculated in the high-dimensional vector space; para. 122: the similarity estimate quantifies the degree of semantic relatedness between a grammatical unit and the input text from which the grammatical unit was extracted. For example, the similarity estimate may be a value between 1 and 0 inclusive with values closer to 1 representing a higher degree of semantic relatedness and values closer to 0 representing a lower degree of semantic relatedness which is equivalent to semantically unrelated category; para. 193-194: content tagger; para. 84: in filtered feature space comparisons, the vector space may first be filtered based on tags (and/or other properties, such as resource media type, creation date, etc.), to identify subset of the feature vectors within the vector space(s) having tags (and/or other properties) that match those of the feature vector generated in step. Then, the feature vector generated in step 1206 may be compared to only those feature vectors in the subset having matching tags/properties. Dasgupta was applied to teach “a canopy clustering algorithm” and Tomkins for “node in the graph structure” as in ontology – See para. 13. The combination of references does teach the argued limitations.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-3, 6, 8-10, 13, 15-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ghoshal et al. (US 20200125575) in view of Dasgupta (US 10719301) and further in view of Tomkins et al. (20210295822).
Specification, para. 48 discloses “The confidence score refers to a probability that each sentence to be classified is considered to be a certain category.”
As per claims 1, 8, 15, Ghoshal et al. (US 20200125575) teaches
A computer-implemented method comprising:
receiving, by one or more processors, a set of text, wherein the set of text contains one or more sentences (para. 74: the received text items with content features extracted and saved as tags associated with the content items; para. 79: the tagging processes described herein for tagging of text blocks, images, audio/video data, and/or other content with tags corresponding to any identified content topics, categories, or features; para. 113: extracted from the given text can be a sentence, a phrase, a paragraph, a word);
determining, for each sentence of the one or more sentences, a likelihood that vectorizing each sentence will produce a vectorized representation that collapses to a spatial region that is semantically unrelated to the set of text wherein the likelihood is based in part on a length of each sentence (para. 58: interactions between client device and content recommendation engine may be Internet-based web browsing sessions, or client-server application sessions, during which users access may input original authored content via the client device, and receive content recommendations from content recommendation engine in the form of additional content (equivalent to any content or other materials from the content repository and not in the received set of text) that is retrieved from the content repository and linked or embedded into the content authoring user interface at the client device; para. 62, 100: the text summary approach using explicit semantic analysis operates generally as follows: (1) grammatical units (e.g., sentences or words) are extracted from the given text document using any known technique for identifying and extracting such units, (2) each of the extracted grammatical units and the text document are represented as weighted vectors of knowledge base concepts; para. 108-110: using a dissimilarity measure/difference/semantically unrelated to select grammatical units for inclusion in the text summary can prevent multiple similar grammatical units from being included in the same text summary; para. 117: forming the input vector may also include unit length normalization such as described above with respect to the training data item vectors; para. 122);
dividing, by one or more processors, each sentence of the set of text into one of i) a first category where the likelihood is higher than a threshold, (fig.12: parse/divide and process original authored content, extract and identify keywords/topics/features from content; para. 79: parsing, tagging; para. 100: representation of weighted vectors of knowledge base concepts may correspond to topic modeling, in which each sentence or word first may be converted into a feature vector, and then the feature difference/similarity may be calculated in the high-dimensional vector space; para. 122: the similarity estimate quantifies the degree of semantic relatedness between a grammatical unit and the input text from which the grammatical unit was extracted. For example, the similarity estimate may be a value between 1 and 0 inclusive with values closer to 1 representing a higher degree of semantic relatedness and values closer to 0 representing a lower degree of semantic relatedness which is equivalent to semantically unrelated category);
and ii) a second category where the likelihood is less than the threshold (para. 62, 82: all of the feature vectors satisfying a particular closeness threshold (e.g., distance between vectors<threshold(T)) may be selected; para. 95, 108-109: determine the grammatical unit having the highest degree of semantic relatedness to the given text document and then selecting all other granunatical units where the difference in the degree of semantic relatedness to the given text document of the grammatical unit and the highest degree is below a predefined threshold. Using a dissimilarity measure to
select grammatical units for inclusion in the text summary can prevent multiple similar grammatical units from being included in the same text summary);
tagging, by the one or more processors, each sentence of the set of text with one or more tags using a plurality of open-source text classification models, wherein the first category comprises additional tags for overcoming the higher likelihood of collapsing (para. 79: tagging; para.113: extracted from the given text can be a sentence, a phrase, a paragraph, a word; para. 4-5: analyze text and/or visual input, extract keywords or topics from the input, classify and tag the input content, and store the classified/tagged content in one or more content repositories; para. 65, 94-95: text classification techniques may be used to explicitly represent the meaning of any text in terms of Wikipedia-based concepts/open-source model; para. 181-182: tagging; para. 193: various machine learning techniques using pre-trained machine learning models and/or other artificial intelligence based tools, including AI-based text or image classification systems (open source), topic or feature extractions, and/or any other combination of techniques described above, may be used for determining the tags to be associated with a content item and the corresponding tag values; figs. 23-24: more words are tagged in relating to “crane” which disambiguates the meaning of “crane” thus, overcomes the likelihood of collapsing; para. 193-194).
subsequent to tagging each sentence of the set of text with the one or more tags, filtering, by one or more processors, the one or more tags using a confidence score (para. 4-5: analyze text and/or visual input, extract keywords or topics from the input, classify and tag the input content, and store the classified/tagged content in one or more content repositories, comparisons may include thorough and exhaustive deep searches and/or more efficient tag-based filtered searches; para. 84, 100: the text summary approach using explicit semantic analysis operates generally as follows: (1) grammatical units (e.g., sentences or words) are extracted from the given text document using any known technique for identifying and extracting such units, (2) each of the extracted grammatical units and the text document are represented as weighted vectors of knowledge base concepts; para. 181: the content items that are available for searching by the recommendation system are tagged and the content items along with the tags may be stored in one or more repositories. The tagging may be performed by a content item tagging service/application. For a content item, the one or more tags associated with the content item are indicative of the content contained by the content item. A value (also sometimes referred to as a tag probability) may also be associated with each tag, where the value provides a measure (e.g., a probability) of the content indicated by the tag occurring in the content item);
subsequent to filtering the one or more tags, representing, by one or more processors, each remaining sentence of the filtered set of text as a first node in a graph structure (para. 56: analyze text and/or visual input, extract keywords or topics from the input, classify and tag the input content, and store the classified/tagged content in one or more content repositories; para. 94-95: given a text snippet, the nearest Wikipedia page title may be returned (e.g., "Mount Everest," "Stephen Hawking," "Car Accident," etc.), which may be used as the class/category of the text. (Thus, a category is equivalent to a first node of its own graph structure corresponding to input text). Input texts therefore may be represented as weighted vectors of concepts, called interpretation vectors. Semantic relatedness of texts then may be computed by comparing their vectors in the space defined by the concepts, for example, using the cosine metric. To speed up semantic interpretation, an inverted index, which maps each word into a list of concepts in which it appears, may be used. The inverted index also may be used to discard/filter insignificant associations between words and concepts, by removing those concepts whose weights for a given word are below a certain threshold; para. 100-101: representation of weighted vectors of knowledge base concepts may correspond to topic modeling, in which each sentence or word first may be converted
into a feature vector, and then the feature difference/similarity may be calculated in the high-dimensional vector space; para. 119-120: the weights in the attribute vector quantify the strength of association between respective concepts of the knowledge base and the attribute mapped to the attribute vector by the inverted index; para. 193-194: content tagger; para. 84: in filtered feature space comparisons, the vector space may first be filtered based on tags (and/or other properties, such as resource media type, creation date, etc.), to identify subset of the feature vectors within the vector space(s) having tags (and/or other properties) that match those of the feature vector generated in step. Then, the feature vector generated in step 1206 may be compared to only those feature vectors in the subset having matching tags/properties);
vectorizing, by one or more processors, each sentence and each tag of the first node (fig. 20; para. 58: a content retrieval and embedding service. Additionally, system 400 includes one or more content repositories storing content files/resources, and one or more vector spaces. As described in more detail below, a vector space may refer to a multi-dimensional data structure configured to store one or more feature vectors; para. 64: provide and monitor the classification and vectorization of content resources, as well as to manage the underlying storage/server/network resources; para. 77, 94-95: given a text snippet, the nearest Wikipedia page title may be returned (e.g., "Mount Everest," "Stephen Hawking," "Car Accident," etc.), which may be used as the class/category of the text. (Thus, a category is equivalent to a first node of its own graph structure corresponding to the text). Input texts therefore may be represented as weighted vectors of concepts, called interpretation vectors. Semantic relatedness of texts then may be computed by comparing their vectors in the space defined by the concepts, for example, using the cosine metric. To speed up semantic interpretation, an inverted index, which maps each word into a list of concepts in which it appears, may be used. The inverted index also may be used to discard/filter insignificant associations between words and concepts, by removing those concepts whose weights for a given word are below a certain threshold; para. 100,119: the weights in the attribute vector quantify the strength of association between respective concepts of the knowledge base and the attribute mapped to the attribute vector by the inverted index);
wherein the vectorizing uses each tag as auxiliary information for each of the one or more sentences with a number of words that could be encoded to a spatial region that is semantically unrelated to the set of text (figs. 5-6: words are extracted from the content – See fig. 8: keyword extraction from the content including one or more sentences; fig. 12: generate vector for content resource based on extracted keywords/topics/features. Perform vector space comparison to identify closest content vectors from repository. Retrieve corresponding content resources and embed within author user interface; fig. 20: word vectors: feature space(3D); fig. 22: unrelated set of text/words or two different “Crane” tags; para. 57: converting the original content (e.g., input text and/or images) into vectors within a multi-dimensional vector space; para. 122: the similarity estimate may be a value between 1 and 0 inclusive with values closer to 1 representing a higher degree of semantic relatedness and values closer to 0 representing a lower degree of semantic relatedness);
performing, by one or more processors, a preliminary clustering the vectorized first node (fig. 14: clusters of vectorized first nodes: coffee, mountain, bird etc. in relating to fig. 20: vectorized keywords in the word vector space based on similar features or closeness threshold/distance);
Ghoshal does not each a canopy clustering algorithm.
Dasgupta teaches
performing, by one or more processors, a preliminary clustering the vectorized first node under strict conditions using a canopy clustering algorithm (figs. 14, 20, item 2030-2040: perform clustering of feature vectors; col. 33:47-51: these feature vectors are then used to obtain a set of diversified examples from the image set as the seed images. For example, a clustering technique may be used such as k-medoids centroids are used to choose the seed images; col. 9:19-67: manual and/or automated tagging of input data. In the media modeling context, the models in question typically comprise neural network models. The ML media models may be used to perform a variety of media analysis tasks, such as image classification, object detection, semantic segmentation, or video, text, or speech processing. the MDE 130 may provide a library of model architectures for ML media models, some of which may have already been trained extensively to make a large variety of decisions about media data; fig. 25: classes and probability score for each class).
Thus, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Ghoshal and Dasgupta in order to effectively classify text/image contents which allow users to better view, interact, and/or analyze the available data.
Even if Ghoshal and Dasgupta do not explicitly teach as a first node in the graph structure.
Tomkins teaches said limitation at para. 74: obtain a plurality of initial ontology graphs and update each of the initial plurality graphs as new data is received or analyzed. For example, some embodiments may obtain a first ontology graph storing embedding vectors representing aeronautical engineering concepts and a second ontology graph storing embedding vectors representing airplane pilot concepts; para. 53: the data may include a set of existing ontology data 211, a set of natural-language text documents 212, or a set of structured data 214. For example, the set of existing ontology data 210 may include an existing knowledge graph structured in an existing ontology data model
Thus, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Ghoshal, Dasgupta with a first node in the graph structure of Tomkins in order to effectively analyze and classify the received text contents and display to the users the graph structure of similar context which allow users to further access, view, interact with the displayed visual graph.
As per claims 3, 10, 17, Ghoshal et al. teaches
wherein the filtering comprises retaining, by one or more processors, each tag of the one or more tags with a confidence score greater than 0.5; removing, by the one or more processors, each tag of the one or more tags with a confidence score less than 0.5 (para. 181, 192: the tag values may be represented using different formats. For example, in some implementations, that tag values may be expressed as floating point numbers between 0.0 and 1.0. Because values closer to 1 representing a higher degree of being in a certain category or relatedness and values closer to 0 representing a lower degree of relatedness. It is helpful to retain a tag with a score of greater than average or 0.5 and discarding tags with scores less than average or less than 0.5) - para. 122-123).
Spec. para. 51: represents each of the tags associated with each sentence of the set of text as attributes of the node in the graph structure (e.g., as shown in Figure 3B and 3E).
As per claims 4, 11, 18, Ghoshal et al. teaches
representing, by the one or more processors, each tag of the one or more tags as an attribute of the first node in the graph structure (fig. 5 shows words that are extracted from the content are tagged by at least a classification model and a probability value is computed for said word/feature vector; para. 64: provide and monitor the classification and vectorization of content resources, as well as to manage the underlying storage/server/network resources; para. 94-95: given a text snippet, the nearest Wikipedia page title may be returned (e.g., "Mount Everest," "Stephen Hawking," "Car Accident," etc.), which may be used as the class/category of the text; para. 100: representation of weighted vectors of knowledge base concepts may correspond to topic modeling, in which each sentence or word first may be converted into a feature vector, and then the feature difference/similarity may be calculated in the high-dimensional vector space; para. 119: the weights in the attribute vector quantify the strength of association between respective concepts of the knowledge base and the attribute mapped to the attribute vector by the inverted index).
Even if Ghoshal and Dasgupta do not explicitly teach as a first node in the graph structure.
Tomkins teaches said limitation at para. 74: obtain a plurality of initial ontology graphs and update each of the initial plurality graphs as new data is received or analyzed. For example, some embodiments may obtain a first ontology graph storing embedding vectors representing aeronautical engineering concepts and a second ontology graph storing embedding vectors representing airplane pilot concepts; para. 53: the data may include a set of existing ontology data 211, a set of natural-language text documents 212, or a set of structured data 214. For example, the set of existing ontology data 210 may include an existing knowledge graph structured in an existing ontology data model
Thus, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Ghoshal, Dasgupta with a first node in the graph structure of Tomkins in order to effectively analyze and classify the received text contents and display to the users the graph structure of similar context which allow users to further access, view, interact with the displayed visual graph.
As per claims 6, 13, Ghoshal et al. teaches
wherein performing the preliminary clustering of the one or more nodes under strict conditions using the canopy clustering algorithm further comprises: enabling, by one or more processors, a part of the one or more nodes to organize to form a settlement.
Ghoshal does not each said claims.
Dasgupta teaches said limitation at col. 9:56-61: in the media modeling context, the models in question typically comprise neural network models. The ML media models may be used to perform a variety of media analysis tasks, such as image classification, object detection, semantic segmentation, or video, text, or speech processing; col. 41:19-29: a classification model is initialized based on the user's annotations of the seed samples, the initialization assigns the classification models with an initial set of parameters, which may be further tune in successive rounds of active learning; col. 56:27-29: cluster the images or media samples in the data set in question using a clustering technique such as canopy clustering; fig. 25: classes and probability score for each class; col. 35:59-63: data features are organized and vectorized into feature vectors on two or more dimensional space which is equivalent to a form of settlement; col. 43:12-13: the specified thresholds enable users to export only the most accurate and trustworthy classifications).
Thus, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Ghoshal and Dasgupta in order to effectively classify text/image contents which allow users to better view, interact, and/or analyze the available data.
Claim(s) 5, 7, 12, 14, 19-20, is/are rejected under 35 U.S.C. 103 as being unpatentable over Ghoshal et al. (US 20200125575) in view of Dasgupta (US 10719301) and further in view of Tomkins et al. (20210295822) and Paik (US 20210216822).
As per claims 5, 12, 19, Ghoshal et al. teaches
subsequent to representing each tag of the one or more tags as an attribute of the first node in the graph structure, embedding, by one or more processors, each sentence and each tag of the first node (fig. 20; para. 58: a content retrieval and embedding service. Additionally, system 400 includes one or more content repositories 440 storing content files/resources, and one or more vector spaces. As described in more detail below, a vector space may refer to a multi-dimensional data structure configured to store one or more feature vectors; para. 94-95: given a text snippet, the nearest Wikipedia page title may be returned (e.g., "Mount Everest," "Stephen Hawking," "Car Accident," etc.), which may be used as the class/category of the text; para. 100: representation of weighted vectors of knowledge base concepts may correspond to topic modeling, in which each sentence or word first may be converted into a feature vector, and then the feature difference/similarity may be calculated in the high-dimensional vector space; para. 119: the weights in the attribute vector quantify the strength of association between respective concepts of the knowledge base and the attribute mapped to the attribute vector by the inverted index).
Ghoshal, Dasgupta, Tomkins do not explicitly teach using a one-hot encoding method.
Paik et al. teaches said limitation at para. 106-108, 194-195: these attributes are inferred by a machine learning classifier that takes into consideration one or more sources of information, for example, the imaging order, clinical text etc., categorical data are one-hot encoded while continuous values are used directly to create a metadata vector that is passed into the classifier; para. 271-272: when the user selects a particular finding from the list, a structured representation of the finding is generated. This finding is represented through a knowledge graph that represents various concepts including the anatomic location and the type of observation. Thus, it would have been obvious to one or ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Ghoshal, Dasgupta, Tomkins and the one-hot encoding method of Paik in order to effectively classify the received text contents and display to the users the graph structure of categorical context which allow users to further access, view, interact with the displayed visual graph.
As per claims 7, 14, 20, Tomkins teaches at para. 37: generating ontology graphs arranged in a hierarchical ontology data model based on ingested documents. Ontology
graphs may be associated with their own domain categories or may be arranged into subgraphs having vertices associated with specific domain categories; para. 84: vertices or other subgraph components of a first ontology graph may be associated with a first set of documents, and vertices or other subgraph components of a second ontology graph may be associated with a second set of documents.
Ghoshal, Dasgupta, Tomkins do not explicitly teach said claims.
Paik et al. teaches
determining, by one or more processors, the first node is a subordinate of a cluster using an algorithm; organizing, by one or more processors, the first node into a miniature graph; calculating, by one or more processors, a degree of similarity between the miniature graph and a second node; assigning, by one or more processors, a clustering relationship to the miniature graph and the second node using link prediction (para. 106-108, 194-195: these attributes are inferred by a machine learning classifier that takes into consideration one or more sources of information, for example, the imaging order, clinical text etc., categorical data are one-hot encoded while continuous values are used directly to create a metadata vector that is passed into the classifier; para. 271-272: when the user selects a particular finding from the list, a structured representation of the finding is generated. This finding is represented through a knowledge graph that represents various concepts including the anatomic location and the type of observation. Thus, a finding is equivalent to the first node into a miniature graph which similarity or relating to other concepts – See para. 186: machine learning methods are used to determine series similarity by analyzing a vector of standardized technical acquisition parameters from the DICOM images; para. 172: machine learning functions that may be trained to learn the combination of these inputs that most accurately predicts actual quality and efficiency for a given set of imaging studies).
Thus, it would have been obvious to one or ordinary skill in the art before the
effective filing date of the claimed invention to combine the teachings of Ghoshal, Dasgupta, Tomkins and Paik in order to effectively classify the received text contents and display to the users the graph structure of categorical context which allow users to further access, view, interact with the displayed visual graph.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Andreoli et al. (US 20160124944) teaches at para. 86: source or target sentence probability.
Brin et al. (US 20230111047) teaches at para. 150: The similarities between units of text can be calculated based on their embeddings. For instance, an embedding in n-space may be calculated for each sentence of training and/or test data using paragraph vectors or transformer-derived sentence embedders. Then, for each embedding, a dot product (e.g., a similarity measure in n-space that may be equivalent to cosine similarity when vectors are normalized) can be calculated with some or all other embeddings. These dot products may be represented as a similarity score that takes on values between −1 and 1, for example, where values close to 1 indicate that their units of text are nearly semantically identical, values close to 0 indicate minimal similarity, and values close to −1 indicate that their units of text are semantically unrelated.
Wenger et al. (US 20200213605) teaches at para. 69: semantically distinct content and semantically different spatial regions.
Goel et al. (US 20180308487) teaches at para. 40: in order to assign an action tag to every piece of word string, the semantic engine 104 utilizes a pre-trained knowledge that consists relevant meaning and context in which the words are used. An intermediate output of the speech to semantic mapper 116 is a word2vec model 302 which encodes the meaning of words. The word2vec model converts the words into a vector that represents semantic meaning of the particular word. We could call this a semantic vector space because sentences with similar meaning have a small Euclidian distance between vectors. Zhao et al. (US 20220147836) teaches at para. 18: forming the text representation vectors further comprises: establishing an attention weight matrix, and computing attention weight values of the sentences based on structure entity vectors in the knowledge graph and relation representation vectors of the sentences using the attention weight matrix. Thereby, vectors of different sentences can be aggregated to form association-discriminated text relation representation vectors. Hertz (US 11222052) discloses in fig. 2: generate directed graph, identify first entity and second entity, apply criteria to first association, weight critical values; fig. 8: sentence evidence classifier.
THIS ACTION IS MADE FINAL. 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 LINH BLACK whose telephone number is (571)272-4106. The examiner can normally be reached 9AM-5PM EST M-F.
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, Tony Mahmoudi can be reached on 571-272-4078. 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.
/LINH BLACK/Examiner, Art Unit 2163 5/13/2026
/TONY MAHMOUDI/Supervisory Patent Examiner, Art Unit 2163
Prosecution Insights