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 .
Claims 1-20 are pending and claims 1 and 20 are independent claims.
Claim Objections
Claim 8 is objected to because of the following informalities:
Claim 8 states “cannot be derive” which seems to state “cannot be derived”
Appropriate correction is required.
Claims 2-19 are objected to because of the following informalities: “An apparatus according to claim …”
“An” should be replaced by “The”.
Appropriate correction is required.
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The independent claims 1 and 20 recite “receive input text… determine a sequence of tokens… sequence of tokens represents a text output… selecting a next token … token selection process … selecting a plurality of candidate tokens … filter the plurality of candidate tokens to exclude one or more of the candidate tokens …” as drafted cover an abstract idea of data analysis/retrieval and mental steps. More specifically, the “receive input text; apply a trained model to the input text to determine a sequence of tokens, wherein the sequence of tokens represents a text output, wherein the trained model is configured to perform a repeating next token selection process comprising selecting a next token in the sequence based on at least one previous token in the sequence, the next token selection process comprises selecting a plurality of candidate tokens and the processing circuitry is configured to filter the plurality of candidate tokens to exclude one or more of the candidate tokens if they match at least one exclusion criterion” which requires just data analysis / retrieval step and mental process. For instance, one can receive input text using pen and pencil and may form tokens using a spreadsheet. One may look over the text in the spreadsheet in different creative ways to determine a sequence of tokens, select some words or phrases which are considered as tokens that should be excluded, for example, based on previous history or criterion, context, etc., or based on whether those terms match which in effect results in filtering the plurality of candidate tokens. Once a person finishes identifying/selecting the tokens, which can also be done mentally, and one may put together the list of excluded/banned tokens that can be used for future use. Those output results can be put together in a spreadsheet to generate the list of excluded output text. The trained model in these claims can just be considered as an additional element. The claimed invention is, therefore, directed to an abstract idea and a mental process without significantly more and thus, claims 1 and 20 are rejected under 35 U.S.C. 101.
Similarly, the dependent claims 2-19 recite similar claim language as in claims 1 and 20.
Claim 2 recites “wherein at least one exclusion criterion is set based on the input text or upon input tokens,” which requires just a data analysis/retrieval and mental step of setting the input text or input tokens as one exclusion criterion. Thus, claim 2 is directed to an abstract idea.
Claim 3 recites “the at least one exclusion criterion comprises a match with a banned list of items, wherein the banned list of items is determined based on the input text,” which requires just a data analysis/retrieval and mental step of matching with a banned list of items as one exclusion criterion. Thus, claim 3 is directed to an abstract idea.
Claim 4 recites “the banned list of items comprises labels or tokens,” which requires just a data analysis/retrieval and mental step of collecting banned list of items as labels or tokens. Thus, claim 4 is directed to an abstract idea.
Claim 5 recites “the processing resource is configured to process the input text to select items that match or can be derived from the input text, and to generate the banned list of items based on items that do not match or are not derived from the input text,” which requires just a data analysis/retrieval and mental step of processing the input text to select items that match or can be derived from the input text, and to generate the banned list of items based on items that do not match or are not derived from the input text. Thus, claim 5 is directed to an abstract idea.
Claim 6 recites “identifying a plurality of tokens or other text items from the input text and the selecting of items that match or can be derived from the input text comprises selecting items from a dictionary that match the identified plurality of tokens or other text items,” which requires just a data analysis/retrieval and mental step of identifying a plurality of tokens or other text items from the input text and the step of selecting of items that match or can be derived from the input text. Thus, claim 6 is directed to an abstract idea.
Claim 7 recites “the selecting of items from the dictionary includes selecting synonyms for the identified tokens or other text items,” which requires just a data analysis/retrieval and mental step of selecting of items from the dictionary that includes selecting synonyms for the identified tokens or other text items. Thus, claim 7 is directed to an abstract idea.
Claim 8 recites “the processing circuitry is configured to select items for the banned list of items from items in the dictionary that do not match or cannot be derive from the input text,” which requires just a data analysis/retrieval and mental step of selecting items for the banned list of items from items in the dictionary that do not match or cannot be derived from the input text. Thus, claim 8 is directed to an abstract idea.
Claim 9 recites “the generation of the banned list of items comprises selecting items from a dictionary,” which requires just a data analysis/retrieval and mental step of generating or creating the banned list of items which is basically selecting items from a dictionary. Thus, claim 9 is directed to an abstract idea.
Claim 10 recites “the dictionary comprises a set of medical terms and associated synonyms,” which requires just a data analysis/retrieval and mental step of creating the dictionary consisting of a set of medical terms and associated synonyms. Thus, claim 10 is directed to an abstract idea.
Claim 11 recites “the processing resource is configured to navigate a knowledge graph based on the input text thereby to obtain at least one context-specific sub-graph and to use the least one context-specific sub-graph to expand the banned list,” which requires just a data analysis/retrieval and mental step of navigating a knowledge graph of input text to obtain context-specific sub-graph. Thus, claim 11 is directed to an abstract idea.
Claim 12 recites “the input text comprises at least one of text from physician notes, text from a medical record, or text associated with at least one scan, medical investigation or other medical procedure,” which requires just a data analysis/retrieval and mental step in which the input text consists of text from physician notes, text from a medical record, or text associated with at least one scan, medical investigation or other medical procedure. Thus, claim 12 is directed to an abstract idea.
Claim 13 recites “the banned list represents at least one of a pathology, disease, medical condition or symptom,” which requires just a data analysis/retrieval and mental step in which the banned list represents at least one of a pathology, disease, medical condition or symptom. Thus, claim 13 is directed to an abstract idea.
Claim 14 recites “output text represents at least one of a status of a patient, a medical condition, a diagnosis or a summary of at least one of physician notes, a medical record, or a description of outcome of a scan, medical investigation or other medical procedure,” which requires just a data analysis/retrieval and mental step in which the output text represents at least one of a status of a patient, a medical condition, a diagnosis or a summary of at least one of physician notes, a medical record, or a description of outcome of a scan, medical investigation or other medical procedure. Thus, claim 14 is directed to an abstract idea.
Claim 15 recites “the trained model comprises an encoder – decoder model,” which requires just additional elements of “encoder – decoder model.” This is because the indicated “encoder – decoder model” is a general-purpose neural network model (Spec page 4, para 2) and there is no specific training procedure or scheme provided in the claim(s) that might make their claimed invention novel. Thus, claim 15 is directed to an abstract idea.
Claim 16 recites “the trained model comprises a large language model (LLM) or other language model,” which requires just additional elements of “a large language model (LLM) or other language model.” The Spec indicates that the “large language model (LLM) or other language model” in the claims is a general-purpose generative large language model (LLM) or other language model (Spec page 4, para 4) and well known models like generative LLM. Thus, claim 16 is directed to an abstract idea.
Claim 17 recites “the model comprises at least one of GPT-2, GPT-3.5, GPT-4, PaLM, LLaMa, BLOOM, Ernie, T5, Claude or Claude 2, or any suitable derivatives or developments thereof,” which requires just additional elements of “GPT-2, GPT-3.5, GPT-4, PaLM, LLaMa, BLOOM, Ernie, T5, Claude or Claude 2.” The indicated models “GPT-2, GPT-3.5, GPT-4, PaLM, LLaMa, BLOOM, Ernie, T5, Claude or Claude 2” are just general-purpose (Spec page 4, para 4) and well known models like chat GPT which are available to the public. Thus, claim 17 is directed to an abstract idea.
Claim 18 recites “the dictionary comprises or is derived from a hierarchical ontology,” which requires just a data analysis/retrieval and mental step of creating a dictionary that comprises of or derived from a hierarchical ontology. Thus, claim 18 is directed to an abstract idea.
Claim 19 recites “the hierarchical ontology comprises the International Classification of Disease (ICD), SNOMED CT, Radlex or other diagnostic code ontology,” which requires just a data analysis/retrieval and mental step of the hierarchical ontology that comprises the International Classification of Disease (ICD), SNOMED CT, Radlex or other diagnostic code ontology. Thus, claim 19 is directed to an abstract idea.
Thus, claims 1-20 as drafted cover a mental process and abstract idea of data gathering/retrieval and analysis/processing steps, and they are mental processes directed to an abstract idea of using pen and pencil, spreadsheet, or implementing data processing and data analysis using a conventional/generic (general-purpose) computer (Spec., page 3, 1st paragraph) as well and thus, all the claims are directed to an abstract idea.
This judicial exception is not integrated into a practical application. In particular, claims 1 and 12 recite additional element of “processor”, “memory” and “trained model” as per the independent claims. Similarly, dependent claims 15-17 recite additional elements of “encoder – decoder model,” “large language model (LLM) or other language model,” “GPT-2, GPT-3.5, GPT-4, PaLM, LLaMa, BLOOM, Ernie, T5, Claude or Claude 2, or any suitable derivatives or developments thereof,” “GPT-2, GPT-3.5, GPT-4, PaLM, LLaMa, BLOOM, Ernie, T5, Claude or Claude 2.” As indicated above, the claimed invention makes use of a general-purpose “encoder – decoder model” neural network model, a general-purpose and well known GPT model(s) like chat GPT, a general-purpose and well known generative large language model (LLM) or other language model and no specific training scheme or procedure has been provided for the training of their models. Accordingly, these additional elements do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
Thus, taken alone, the additional elements do not amount to significantly more than the above-identified judicial exception (the abstract idea). Looking at the limitations as an ordered combination adds nothing that is not already present when looking at the elements taken individually. There is no indication that the combination of elements improves the functioning of a computer or improves any other technology. Their collective functions merely provide conventional general purpose computer implementation. Claims 1-20, are therefore not drawn to patent eligible subject matter as they are directed to an abstract idea without significantly more. Thus, the claimed invention is directed to an abstract idea and a mental process without significantly more and thus, claims 1-20 are rejected under 35 U.S.C. 101.
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 the integration of the abstract idea into a practical application, the additional element of using a computer is noted as a general computer as noted. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept (Spec., page 3, 1st paragraph). The claims are not patent eligible.
Dependent claims 2-19 are also directed toward an abstract idea and do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements when considered both individually and as an ordered combination do not amount to significantly more than the abstract idea. Therefore, claims 1-20 do not contain patent eligible subject matter that has been identified by the courts.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1-14 and 20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Ferrandez et al. Pat App No. US 20180373844 A1 (Ferrandez).
Regarding Claim 1, Ferrandez discloses an apparatus comprising processing circuitry configured to (Ferrandez, para 0029, a CAC system utilizing a clinical concept relevance (CCR) component configured to facilitate reducing false positive rates in suggesting medical codes to a customer):
receive input text (Ferrandez, para 0087, presented with an input text);
apply a trained model to the input text to determine a sequence of tokens (Ferrandez, para 0087, the statistical entity detection model may be trained to be able to probabilistically label new texts (e.g., texts not included in the training corpus) with automatic entity labels using the same feature extraction technique that was applied to the training corpus. When later presented with an input text without manual entity labels, the statistical model may then apply the same feature extraction techniques to extract features from the input text, and may apply the learned probabilistic relationships to automatically determine the most likely entity labels for word sequences in the input text. Any suitable statistical modeling technique may be used to learn such probabilistic relationships, as aspects of the invention are not limited in this respect. Non-limiting examples of suitable known statistical modeling techniques include machine learning techniques such as maximum entropy modeling, support vector machines, and conditional random fields, among others; [i.e., trained “statistical (entity detection) model” (including “machine learning” /”maximum entropy”/ “ support vector machines”/”conditional random fields”/… models” among others) as “a trained model”]Ferrandez, para 0182-0184, As one example, for a given portion of text 1410 that has been annotated by NLU engine 1420 with one or more facts, feature extractor 1963 may extract the same set of features 1995 from the relevant portions of text 1410 that were extracted from the training data and provide the set of features as input to the trained CCR model 1965. According to some embodiments, CCR model 1969 includes a classifier that, in response to receiving features 1995, classifies the fact(s) to which the features pertain as a false positive or a true positive…a white list 1967 is utilized to facilitate training and, more particularly, to train the CCR model with training data that corresponds to facts (e.g., as represented by corresponding medical codes) for which the CAC system generally performs poorly. For example, white list 1967 may include a list of medical codes that have been identified as problematic, for example, by examining customer feedback 1495 to identify medical codes that frequently serve as a source for false positive billing code suggestions…white list 1967 may include a list of medical codes that are not problematic and that should not contribute to the training data 1895 for which feature extraction is performed. Accordingly, white list 1967 may operate as a pass filter or a blocking filter), wherein the sequence of tokens represents a text output (Ferrandez, para 0185, FIG. 20 illustrates a CCR component comprising a machine learning CCR model 2069 that receives feature set 2095 (e.g., features 2095a-h) as an input and produces an output 2097. Based on output 2097, one or more modifications 2099 to parameters of CCR model 2095 may be adjusted. For example, output 2097 may be compared with “ground truth” (e.g., customer feedback as to whether a suggested billing code was correct or not); [i.e., the list of “ground truth” (e.g., customer feedback)” to be compared with the “output” is text, i.e., the CCR model produces a text output]),
wherein the trained model is configured to perform a repeating next token selection process comprising selecting a next token in the sequence based on at least one previous token in the sequence (Ferrandez, para 0087-0090, Any suitable statistical modeling technique may be used to learn such probabilistic relationships, as aspects of the invention are not limited in this respect. Non-limiting examples of suitable known statistical modeling techniques include machine learning techniques such as maximum entropy modeling, support vector machines, and conditional random fields, among others… In some embodiments, when an unlabeled text is input to the trained statistical entity detection model, the model may process the text to extract features and determine probabilities for individual tokens of being associated with various entity (e.g., fact type) labels. In some embodiments, the most probable label (including the non-entity label, if it is most probable) may be selected for each token in the input text. In other embodiments, labels may be selected through more contextual analysis, such as at the phrase level or sentence level, rather than at the token level. Any suitable technique, such as Viterbi techniques, or any other suitable technique, may be used, as aspects of the invention are not limited in this respect. In some embodiments, a lattice may be constructed of the associated probabilities for all entity types for all tokens in a sentence, and the best (e.g., highest combined probability) path through the lattice may be selected to determine which word sequences in the sentence are to be automatically labeled with which entity (e.g., fact type) labels. In some embodiments, not only the best path may be identified, but also the (N-1)-best alternative paths with the next highest associated probabilities. In some embodiments, this may result in an N-best list of alternative hypotheses for fact type labels to be associated with the same input text; [i.e., “statistical modeling techniques include machine learning techniques…” as “the trained model”; “most probable label (including the non-entity label, if it is most probable) may be selected for each token in the input text… the associated probabilities for all entity types for all tokens in a sentence, and the best (e.g., highest combined probability) path through the lattice may be selected to determine which word sequences in the sentence are to be automatically labeled with which entity (e.g., fact type) labels…not only the best path may be identified, but also the (N-1)-best alternative paths with the next highest associated probabilities. In some embodiments, this may result in an N-best list of alternative hypotheses for fact type labels to be associated with the same input text” as “selecting a next token (with the next highest associated probabilities) in the sequence based on at least one previous token in the sequence ”]),
the next token selection process comprises selecting a plurality of candidate tokens and the processing circuitry is configured to filter the plurality of candidate tokens to exclude one or more of the candidate tokens if they match at least one exclusion criterion (Ferrandez, para 0170-0175, annotations 1660 that include medical codes that do appear in whitelist 1667 are provided as annotations 1660a to CCR model 1669 (potentially with additional information derived from text 1610) for further evaluation to determine which, if any, of the annotations that include whitelisted medical codes should be excluded from further processing by CAC application 1675… Only facts represented by medical codes that have historically been problematic from a false positive rate perspective are even considered as candidates for exclusion from further processing. Alternatively, the whitelist may comprise medical codes on which the CAC system performs well, so that annotations including matched medical codes are provided to CAC application 1675 as a basis for billing code suggestions, while annotations including medical codes that do not appear in the whitelist are first evaluated by CCR model 1669 as candidates for exclusion from further consideration by CAC application 1675... In act 1720, at least one of the medical facts extracted from the text is identified for exclusion from being evaluated in providing billing code suggestions to a customer. For example, at least one medical fact may be identified as having a high likelihood of giving rise to a false positive billing code suggestion and therefore be excluded from evaluation to avoid erroneous billing codes being assigned to the text. … If a medical code included in the annotations produced by the NLU appears in the list, it may be selected as a candidate for exclusion from the process of suggesting billing codes for the corresponding medical facts assigned the medical code. Annotations including or associated with the candidate medical codes may then be processed by the CCR component to identify medical fact(s) that should not be considered when suggesting billing codes to a customer. Annotations that include medical codes that do not appear in the list can be further processed as a basis of suggesting billing codes to a customer (e.g., annotations including medical codes that do not appear in the list may bypass the CCR component for consideration for billing code suggestion). In this manner, only medical facts that have been identified as producing relatively high false positive rates are even considered for exclusion, thereby preventing the exclusion of annotations including or associated with medical facts enjoying high true positive rates and avoiding degrading performance in this respect. As discussed above, as an alternative to a whitelist comprising problematic medical codes, a whitelist may instead comprise medical codes associated with high positive rates. In such circumstances, only annotations including medical codes that do not appear in the list are candidates for potential exclusion as a basis for suggesting billing codes to a customer (e.g., annotations including medical codes that are not listed are provided to the CCR component for further evaluation). Thus, the technique of filtering candidates that are even considered for possible exclusion as a basis for suggesting billing codes (e.g., via a whitelist) may be implemented using either type of list; [“the annotations that include whitelisted medical codes should be excluded from further processing by CAC application 1675” as “exclusion criterion for filtering candidate tokens”]).
Regarding Claim 2, Ferrandez discloses the apparatus according to claim 1, wherein at least one exclusion criterion is set based on the input text or upon input tokens (para 0170, Figure 16, input text 1610, CCR component 1665 (including Whiltelist 1667, CCR model 1669 and processor 1625), annotations 1660 that include medical codes that do appear in whitelist 1667 are provided as annotations 1660a to CCR model 1669 (potentially with additional information derived from text 1610) for further evaluation to determine which, if any, of the annotations that include whitelisted medical codes should be excluded from further processing by CAC application 1675. Annotations that are not flagged for exclusion are provided as annotations 1660b that, along with the annotations not implicated by whitelist 1667, are provided to CAC application 1675 as a basis for suggesting billing codes. In this manner, only facts represented by medical codes that have historically been problematic from a false positive rate perspective are even considered as candidates for exclusion from further processing. Alternatively, the whitelist may comprise medical codes on which the CAC system performs well, so that annotations including matched medical codes are provided to CAC application 1675 as a basis for billing code suggestions, while annotations including medical codes that do not appear in the whitelist are first evaluated by CCR model 1669 as candidates for exclusion from further consideration by CAC application 1675 ;[the “Whitelist” in Figure 16 as a major “exclusion criterion”; “additional information derived from text 1610 (i.e. the input text) for further evaluation to determine which, if any, of the annotations that include whitelisted medical codes” as “additional exclusion criterion”]).
Regarding Claim 3, Ferrandez discloses the apparatus according to claim 2, wherein the at least one exclusion criterion comprises a match with a banned list of items, wherein the banned list of items is determined based on the input text (para 0169, Figure 16, whitelist 1667 comprises the list of medical codes that have historically given rise to significant false positive billing code suggestions relative to true positive billing suggestions. In operation, each medical code included in annotations 1660 may be compared to whitelist 1667 and, if the medical code appears in whitelist 1667, the corresponding annotation is further evaluated by CCR model 1669 to assess whether it should be excluded from further consideration; [“excluded” as “banned”; “compared to” as “match” based on input text 1610]).
Regarding Claim 4, Ferrandez discloses the apparatus according to claim 3, wherein the banned list of items comprises labels or tokens (para 0172-0173, Figure 17, In act 1710, text is processed to extract a plurality of facts. For example, free-form text documenting a patient encounter may be processed by an NLU engine to extract a plurality of facts that, along with other pertinent information such as medical codes associated with at least some of the facts, semantic labels of the facts, relationships between facts and/or labels, etc., form annotations for the free-form text. As discussed above, the text may have resulted from transcribing physician dictation, either automatically, manually or combination of both, or the text may have resulted from another source, as method 1700 may be performed on any suitable text independent of the source. In act 1720, at least one of the medical facts extracted from the text is identified for exclusion from being evaluated in providing billing code suggestions to a customer. For example, at least one medical fact may be identified as having a high likelihood of giving rise to a false positive billing code suggestion and therefore be excluded from evaluation to avoid erroneous billing codes being assigned to the text; [“excluded” as “banned”]).
Regarding Claim 5, Ferrandez discloses the apparatus according to claim 3, wherein the processing resource is configured to process the input text to select items that match or can be derived from the input text, and to generate the banned list of items based on items that do not match or are not derived from the input text (para 0175, Figure 17, as an alternative to a whitelist comprising problematic medical codes, a whitelist may instead comprise medical codes associated with high positive rates. In such circumstances, only annotations including medical codes that do not appear in the list are candidates for potential exclusion as a basis for suggesting billing codes to a customer (e.g., annotations including medical codes that are not listed are provided to the CCR component for further evaluation). Thus, the technique of filtering candidates that are even considered for possible exclusion as a basis for suggesting billing codes (e.g., via a whitelist) may be implemented using either type of list. It should be further appreciated that act 1720 may be performed without using a whitelist. For example, all annotations produced by the NLU engine may be provided to a CCR component to identify those that are problematic with respect to high false positive rates and that therefore should be excluded from consideration when suggesting billing codes to the customer).
Regarding Claim 6, Ferrandez discloses the apparatus according to claim 5, comprising identifying a plurality of tokens or other text items from the input text and the selecting of items that match or can be derived from the input text comprises selecting items from a dictionary that match the identified plurality of tokens or other text items (para 0081-0085, In some embodiments, each training text (e.g., free-form clinician narration) may be tokenized to break it down into various levels of syntactic substructure. For example, in some embodiments, a tokenizer module may be implemented to designate spans of the text as representing structural/syntactic units such as document sections, paragraphs, sentences, clauses, phrases, individual tokens, words, sub-word units such as affixes, etc. In some embodiments, individual tokens may often be single words, but some tokens may include a sequence of more than one word that is defined, e.g., in a dictionary, as a token. For example, the term “myocardial infarction” could be defined as a token, although it is a sequence of more than one word. In some embodiments, a token's identity (i.e., the word or sequence of words itself) may be used as a feature of that token. In some embodiments, the token's placement within particular syntactic units in the text (e.g., its section, paragraph, sentence, etc.) may also be used as features of the token… In some embodiments, other types of features may be extracted, i.e., identified and associated with tokens in the training text. For example, in some embodiments, an N-gram feature may identify the previous (N-1) words and/or tokens in the text as a feature of the current token. In another example, affixes (e.g., suffixes such as -ectomy, -oma, -itis, etc.) may be used as features of tokens. In another example, one or more predefined dictionaries and/or ontologies may be accessed, and a token's membership in any of those dictionaries may be used as a feature of that token. For example, a predefined dictionary of surgical procedures may be accessed, and/or a dictionary of body sites, and/or a dictionary of known diseases, etc.).
Regarding Claim 7, Ferrandez discloses the apparatus according to claim 6, wherein the selecting of items from the dictionary includes selecting synonyms for the identified tokens or other text items (para 0098, in some embodiments, the tokens identified in the text as corresponding to a medical fact may be matched to corresponding terms in an ontology. In some embodiments, a list of closest matching terms may be generated, and may be ranked by their similarity to the tokens in the text; [“closest matching terms” as “synonyms”]).
Regarding Claim 8, Ferrandez discloses the apparatus according to claim 6, wherein the processing circuitry is configured to select items for the banned list of items from items in the dictionary that do not match or cannot be derive from the input text (Ferrandez, para 0170, whitelisted medical codes should be excluded from further processing by CAC application 1675. Annotations that are not flagged for exclusion are provided as annotations 1660b that, along with the annotations not implicated by whitelist 1667, are provided to CAC application 1675 as a basis for suggesting billing codes…Alternatively, the whitelist may comprise medical codes on which the CAC system performs well, so that annotations including matched medical codes are provided to CAC application 1675 as a basis for billing code suggestions).
Regarding Claim 9, Ferrandez discloses the apparatus according to claim 5, wherein the generation of the banned list of items comprises selecting items from a dictionary (Ferrandez, para 0199-0200, According to some embodiments, feature(s) 2095h comprise one or more dictionary-based features. For example, feature(s) 2095h may include a representation indicating which dictionaries a target fact is located in, such as a dictionary for disorders, findings, medications and/or procedures. In this respect, feature(s) 2095h may comprise a binary representation (e.g., a vector) having a component associated with each dictionary being utilized and wherein a value of 1 is set for each component in the vector corresponding to the respective dictionary in which the target fact was found… the above discussed features … the features used for training may also be extracted during operation for evaluation by the trained CCR component to assess whether a fact being evaluated is consequential from a billing perspective or whether it should be excluded from consideration as a basis for suggesting one or more medical codes (e.g., medical billing codes); [“the features used … during operation for evaluation by the trained CCR component to assess whether a fact being evaluated is consequential from a billing perspective or whether it should be excluded” as “selecting items from a dictionary”; according to the limitation, “the generation of the banned list of items” comprises “selecting items [to be excluded] from a dictionary”]).
Regarding Claim 10, Ferrandez discloses the apparatus according to claim 6, wherein the dictionary comprises a set of medical terms and associated synonyms (Ferrandez, para 0097-0098, For example, given the input text, “His sinuses are constantly inflamed,” in some embodiments, an entity detection model together with a relation model (or a single model performing both functions) may identify the tokens “sinuses,” “constantly” and “inflamed” as representing a medical fact. In some embodiments, a normalization/coding process may then be applied to identify the standard form for documenting “constantly inflamed sinuses” as “sinusitis, chronic.” …Any suitable coding system may be used, as aspects of the invention are not limited in this respect. Exemplary standard codes include ICD (International Classification of Diseases) codes, CPT (Current Procedural Terminology) codes, E&M (Evaluation and Management) codes, MedDRA (Medical Dictionary for Regulatory Activities) codes, SNOMED codes, LOINC (Logical Observation Identifiers Names and Codes) codes, RxNorm codes, NDC (National Drug Code) codes and RadLex codes. In some embodiments, a normalization/coding process may be rule-based (e.g., using lists of possible ways of phrasing particular medical facts, and/or using an ontology of medical terms and/or other language units to normalize facts extracted from input text to their standard forms). For example, in some embodiments, the tokens identified in the text as corresponding to a medical fact may be matched to corresponding terms in an ontology. In some embodiments, a list of closest matching terms may be generated, and may be ranked by their similarity to the tokens in the text; [“closest matching terms” as “synonyms”]).
Regarding Claim 11, Ferrandez discloses the apparatus according to claim 3, wherein the processing resource is configured to navigate a knowledge graph based on the input text thereby to obtain at least one context-specific sub-graph and to use the least one context-specific sub-graph to expand the banned list (Ferrandez, para 0098, Figure 1, Fact extraction component 104, In some embodiments, a normalization/coding process may be rule-based (e.g., using lists of possible ways of phrasing particular medical facts, and/or using an ontology of medical terms and/or other language units to normalize facts extracted from input text to their standard forms). For example, in some embodiments, the tokens identified in the text as corresponding to a medical fact may be matched to corresponding terms in an ontology… it should be appreciated that in some embodiments multiple alternative hypotheses for a medical fact to be extracted from a portion of input text may be identified by fact extraction component 104; para 0066, Concepts in the formal ontology linked to the medical terms that appear in the free-form narration may then be identified, and concept relationships in the formal ontology may be traced to identify further relevant concepts. Through these relationships, as well as the linguistic knowledge represented in the formal ontology, one or more medical facts may be extracted. For example, if the free-form narration includes the medical term “hypertension” and the linguistic context relates to the patient's past, the fact extraction component may automatically extract a fact indicating that the patient has a history of hypertension [“ontology of medical terms” as “knowledge graph”, according to Spec, 5th page, 5th para; “ontology linked to the medical terms… linguistic knowledge represented in the formal ontology, one or more medical facts … extracted… context”; ]).
Regarding Claim 12, Ferrandez discloses the apparatus according to claim 1, wherein the input text comprises at least one of text from physician notes, text from a medical record, or text associated with at least one scan, medical investigation or other medical procedure (0047-0048, To allow clinicians and other healthcare personnel to enter medical documentation data directly into an EHR in its discrete structured data format, many EHRs are accessed through user interfaces that make extensive use of point-and-click input methods. While some data items, such as the patient's name, may require input in (structured) textual or numeric form… While some clinicians may appreciate the ability to directly enter structured data into an EHR through a point-and-click interface, many clinicians may prefer being unconstrained in what they can say and in what terms they can use in a free-form note, and many may be reluctant to take the time to learn where all the boxes and buttons are and what they all mean in an EHR user interface; para 0045, an Electronic Health Record (EHR) is an electronic medical record that generally is maintained by a specific healthcare institution and contains data documenting the care that a specific patient has received from that institution over time).
Regarding Claim 13, Ferrandez discloses the apparatus according to claim 1, wherein the banned list represents at least one of a pathology, disease (para 0008-0010, identifying at least one of the plurality of facts to be excluded from consideration when assigning medical codes to the text, and evaluating each of the plurality of facts, except for the identified at least one fact, to assign one or more medical codes to the text), medical condition or symptom.
Regarding Claim 14, Ferrandez discloses the apparatus according to claim 1, wherein the output text represents at least one of a status of a patient (Ferrandez, para 0129, FIG. 7A, these include two History & Physical reports, a Discharge Summary, an Emergency Room Record, a Consultation report, a Progress Note, and an Operative Report. Indicator 712 shows that the current document being viewed is the Discharge Summary dated 6/18/2014, and this document appears in panel 720 where the user can view the text of the document), a medical condition, a diagnosis or a summary of at least one of physician notes, a medical record, or a description of outcome of a scan, medical investigation or other medical procedure.
Regarding Claim 20, Ferrandez discloses a computer-implemented method (Ferrandez, para 0036, Computer-Assisted Medical Coding (CAC) systems) comprising:
receiving input text (Ferrandez, para 0087, presented with an input text);
applying a trained model to the input text to determine a sequence of tokens (Ferrandez, para 0087, the statistical entity detection model may be trained to be able to probabilistically label new texts (e.g., texts not included in the training corpus) with automatic entity labels using the same feature extraction technique that was applied to the training corpus. When later presented with an input text without manual entity labels, the statistical model may then apply the same feature extraction techniques to extract features from the input text, and may apply the learned probabilistic relationships to automatically determine the most likely entity labels for word sequences in the input text. Any suitable statistical modeling technique may be used to learn such probabilistic relationships, as aspects of the invention are not limited in this respect. Non-limiting examples of suitable known statistical modeling techniques include machine learning techniques such as maximum entropy modeling, support vector machines, and conditional random fields, among others; [i.e., trained “statistical (entity detection) model” (including “machine learning” /”maximum entropy”/ “ support vector machines”/”conditional random fields”/… models” among others) as “a trained model”]; Ferrandez, para 0182-0184, As one example, for a given portion of text 1410 that has been annotated by NLU engine 1420 with one or more facts, feature extractor 1963 may extract the same set of features 1995 from the relevant portions of text 1410 that were extracted from the training data and provide the set of features as input to the trained CCR model 1965. According to some embodiments, CCR model 1969 includes a classifier that, in response to receiving features 1995, classifies the fact(s) to which the features pertain as a false positive or a true positive…a white list 1967 is utilized to facilitate training and, more particularly, to train the CCR model with training data that corresponds to facts (e.g., as represented by corresponding medical codes) for which the CAC system generally performs poorly. For example, white list 1967 may include a list of medical codes that have been identified as problematic, for example, by examining customer feedback 1495 to identify medical codes that frequently serve as a source for false positive billing code suggestions…white list 1967 may include a list of medical codes that are not problematic and that should not contribute to the training data 1895 for which feature extraction is performed. Accordingly, white list 1967 may operate as a pass filter or a blocking filter), , wherein the sequence of tokens represents a text output that is generated based on the input text (Ferrandez, para 0185, FIG. 20 illustrates a CCR component comprising a machine learning CCR model 2069 that receives feature set 2095 (e.g., features 2095a-h) as an input and produces an output 2097. Based on output 2097, one or more modifications 2099 to parameters of CCR model 2095 may be adjusted. For example, output 2097 may be compared with “ground truth” (e.g., customer feedback as to whether a suggested billing code was correct or not); [i.e., the list of “ground truth” (e.g., customer feedback)” to be compared with the “output” is text, i.e., the CCR model produces a text output; the input to the CCR is also a text]),
wherein the trained model is configured to perform a repeating next token selection process comprising selecting a next token in the sequence based on at least one previous token in the sequence (Ferrandez, para 0087-0090, Any suitable statistical modeling technique may be used to learn such probabilistic relationships, as aspects of the invention are not limited in this respect. Non-limiting examples of suitable known statistical modeling techniques include machine learning techniques such as maximum entropy modeling, support vector machines, and conditional random fields, among others… In some embodiments, when an unlabeled text is input to the trained statistical entity detection model, the model may process the text to extract features and determine probabilities for individual tokens of being associated with various entity (e.g., fact type) labels. In some embodiments, the most probable label (including the non-entity label, if it is most probable) may be selected for each token in the input text. In other embodiments, labels may be selected through more contextual analysis, such as at the phrase level or sentence level, rather than at the token level. Any suitable technique, such as Viterbi techniques, or any other suitable technique, may be used, as aspects of the invention are not limited in this respect. In some embodiments, a lattice may be constructed of the associated probabilities for all entity types for all tokens in a sentence, and the best (e.g., highest combined probability) path through the lattice may be selected to determine which word sequences in the sentence are to be automatically labeled with which entity (e.g., fact type) labels. In some embodiments, not only the best path may be identified, but also the (N-1)-best alternative paths with the next highest associated probabilities. In some embodiments, this may result in an N-best list of alternative hypotheses for fact type labels to be associated with the same input text; [i.e., “statistical modeling techniques include machine learning techniques…” as “the trained model”; “most probable label (including the non-entity label, if it is most probable) may be selected for each token in the input text… the associated probabilities for all entity types for all tokens in a sentence, and the best (e.g., highest combined probability) path through the lattice may be selected to determine which word sequences in the sentence are to be automatically labeled with which entity (e.g., fact type) labels…not only the best path may be identified, but also the (N-1)-best alternative paths with the next highest associated probabilities. In some embodiments, this may result in an N-best list of alternative hypotheses for fact type labels to be associated with the same input text” as “selecting a next token (with the next highest associated probabilities) in the sequence based on at least one previous token in the sequence”]),
the next token selection process comprises selecting a plurality of candidate tokens and the processing circuitry is configured to filter the plurality of candidate tokens to exclude one or more of the candidate tokens if they match at least one exclusion criterion (Ferrandez, para 0170-0175, annotations 1660 that include medical codes that do appear in whitelist 1667 are provided as annotations 1660a to CCR model 1669 (potentially with additional information derived from text 1610) for further evaluation to determine which, if any, of the annotations that include whitelisted medical codes should be excluded from further processing by CAC application 1675… Only facts represented by medical codes that have historically been problematic from a false positive rate perspective are even considered as candidates for exclusion from further processing. Alternatively, the whitelist may comprise medical codes on which the CAC system performs well, so that annotations including matched medical codes are provided to CAC application 1675 as a basis for billing code suggestions, while annotations including medical codes that do not appear in the whitelist are first evaluated by CCR model 1669 as candidates for exclusion from further consideration by CAC application 1675... In act 1720, at least one of the medical facts extracted from the text is identified for exclusion from being evaluated in providing billing code suggestions to a customer. For example, at least one medical fact may be identified as having a high likelihood of giving rise to a false positive billing code suggestion and therefore be excluded from evaluation to avoid erroneous billing codes being assigned to the text. … If a medical code included in the annotations produced by the NLU appears in the list, it may be selected as a candidate for exclusion from the process of suggesting billing codes for the corresponding medical facts assigned the medical code. Annotations including or associated with the candidate medical codes may then be processed by the CCR component to identify medical fact(s) that should not be considered when suggesting billing codes to a customer. Annotations that include medical codes that do not appear in the list can be further processed as a basis of suggesting billing codes to a customer (e.g., annotations including medical codes that do not appear in the list may bypass the CCR component for consideration for billing code suggestion). In this manner, only medical facts that have been identified as producing relatively high false positive rates are even considered for exclusion, thereby preventing the exclusion of annotations including or associated with medical facts enjoying high true positive rates and avoiding degrading performance in this respect. As discussed above, as an alternative to a whitelist comprising problematic medical codes, a whitelist may instead comprise medical codes associated with high positive rates. In such circumstances, only annotations including medical codes that do not appear in the list are candidates for potential exclusion as a basis for suggesting billing codes to a customer (e.g., annotations including medical codes that are not listed are provided to the CCR component for further evaluation). Thus, the technique of filtering candidates that are even considered for possible exclusion as a basis for suggesting billing codes (e.g., via a whitelist) may be implemented using either type of list; [“the annotations that include whitelisted medical codes should be excluded from further processing by CAC application 1675” as “exclusion criterion for filtering candidate tokens”]).
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 15 is rejected under 35 U.S.C. 103 as being unpatentable over Ferrandez et al. Pat App No. US 20180373844 A1 (Ferrandez) in view of Dreyer et al. Pat App No. US 20250166609 A1 (Dreyer).
Regarding Claim 15, Ferrandez discloses the apparatus according to claim 1
Ferrandez does not specifically disclose the wherein the trained model comprises an encoder – decoder model.
However, Dreyer, in the same field of endeavor, discloses wherein the trained model comprises an encoder – decoder model (Dreyer, 0081, The ML model 510 may have an architecture including an encoder and a decoder. The encoder and the decoder may be trained simultaneously).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Dreyer in the method of Ferrandez because this would enable generating different summaries using multiple documents relating to a particular entity, and one of the different summaries may be selected for output in response to a user input (Dreyer, Abstract).
Claim 16 is rejected under 35 U.S.C. 103 as being unpatentable over Ferrandez et al. Pat App No. US 20180373844 A1 (Ferrandez) in view of Ataei et al. Pat App No. US 20250028882 A1 (Ataei).
Regarding Claim 16, Ferrandez discloses the apparatus according to claim 14.
Ferrandez does not specifically disclose wherein the trained model comprises a large language model (LLM) or other language model.
However, Ataei, in the same field of endeavor, discloses wherein the trained model comprises a large language model (LLM) or other language model (Ataei, para 0155, the trained language model comprises a trained large language model (LLM)).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Ataei in the method of Ferrandez because this would enable determining user intent by receiving user input that comprises first natural language text, performing one or more operations to map the user input to one or more classes of intents included in a plurality of classes of intents, and responsive to determining that the user input does not map to any class of intents, generating, via a first trained language model, second natural language text requesting additional user input (Ataei, Abstract).
Claim 17 is rejected under 35 U.S.C. 103 as being unpatentable over Ferrandez et al. Pat App No. US 20180373844 A1 (Ferrandez) in view of Poulis et al. Pat No. US 12182678 B1 (Poulis).
Regarding Claim 17, Ferrandez discloses the apparatus according to claim 14.
Ferrandez does not specifically disclose wherein the model comprises at least one of GPT-2, GPT-3.5, GPT-4, PaLM, LLaMa, BLOOM, Ernie, T5, Claude or Claude 2, or any suitable derivatives or developments thereof.
However, Poulis, in the same field of endeavor, discloses wherein the model comprises at least one of GPT-2, GPT-3.5, GPT-4, PaLM, LLaMa, BLOOM, Ernie, T5, Claude or Claude 2, or any suitable derivatives or developments thereof (Poulis, col 2, ln 5-10, Language models, including LLMs or LMMs, work by taking an input text and repeatedly predicting the next token or word. Notable current examples of LLM systems include OpenAI's GPT models (e.g., GPT-3.5 and GPT-4, used in ChatGPT), Google's PaLM (used in Bard), and Meta's LLaMa, as well as BLOOM, Ernie 3.0 Titan, and Anthropic's Claude 2).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Poulis in the method of Ferrandez because this would have enabled different types of GenAI that may include large language models (LLMs) that are text content-based or large multimodal models (LMM) that use all types of content and modalities so that this can be useful for aligning the generative artificial intelligence system to a plurality of different domains (Poulis, col 1, ln 44-48 and Abstract).
Claim 18 is rejected under 35 U.S.C. 103 as being unpatentable over Ferrandez et al. Pat App No. US 20180373844 A1 (Ferrandez) in view of Wang et al. Pat App No. CN 113204960 A (Wang).
Regarding Claim 18, Ferrandez discloses the apparatus according to claim 9.
Ferrandez does not specifically disclose wherein the dictionary comprises or is derived from a hierarchical ontology.
However, Wang, in the same field of endeavor, discloses wherein the dictionary comprises or is derived from a hierarchical ontology (Wang, 7th page, 5th para, the invention uses the data to be analyzed for the term extraction, semantic analysis, concept extraction, creating a data dictionary, determining the field and range of the main body; extracting the hierarchical relationship and the non-hierarchical relationship).
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Wang in the method of Ferrandez because the invention of creating a data dictionary by extracting the hierarchical relationship and the non-hierarchical relationship between the concepts would have enabled the realization of fast inquiry of information, enterprise information error correction, and effectively improving the enterprise management efficiency in high technology, while at the same time, filtering part of the enterprise in the field knowledge ontology (Wang, 7th page, 5th para).
Claim 19 is rejected under 35 U.S.C. 103 as being unpatentable over Ferrandez in view of Wang, and further in view of Gupta et al. Pat No. US 12099480 B1 (Gupta).
Regarding Claim 19, Ferrandez in view of Wang disclose the apparatus according to claim 18.
Ferrandez in view of Wang do not specifically disclose wherein the hierarchical ontology comprises the International Classification of Disease (ICD), SNOMED CT.
However, Gupta, in the same field of endeavor, discloses wherein the hierarchical ontology comprises the International Classification of Disease (ICD), SNOMED CT (Gupta, col 4, ln 36-40, FIG. 2 shows an illustrative generalized directed acyclic graph (DAG) 200 that provides an organization for a universal SNOMED-CT clinical concept ontology that uses a subtype classification system to enable aggregation of information for statistical analysis based on a hierarchy of types; Gupta, col 10, ln 37 – col 11, ln 18, Figure 14, In step 1405, a SNOMED CT (Systematized Nomenclature of Medicine Clinical Terminology) ontology is organized in a directed acyclic graph (DAG)… FIG. 15 is a flowchart of an illustrative alternative method 1500 performed by a computing device to implement the present graph-based clinical concept mapping algorithm. In step 1505, mappings between ICD (International Classification of Disease) codebases and a SNOMED CT ontology that is organized as a directed acyclic graph (DAG) are obtained… FIG. 16 is a flowchart of an illustrative alternative method 1600 performed by a computing device to implement the present graph-based clinical concept mapping algorithm. In step 1605, information is extracted from a database of ICD-9 (International Classification of Disease, Revision 9), ICD-10 (International Classification of Disease, Revision 10) and SNOMED (Systematized Nomenclature of Medicine) vocabulary concepts. In step 1610, mappings between ICD (International Classification of Disease) codebases and a SNOMED CT ontology are obtained that is represented as a directed acyclic graph (DAG)), Radlex or other diagnostic code ontology.
Therefore, it would have been obvious for one having ordinary skill in the art before the effective filing date of the claimed invention to incorporate the method of Gupta in the method of Ferrandez in view of Wang because this has enabled the graph-based clinical concept mapping algorithm is further advantageously utilized to group ICD-9/10 codes into higher order, more prevalent SNOMED concepts to support clinical interpretation (Gupta, Abstract).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MULUGETA T. DUGDA whose telephone number is (703)756-1106. The examiner can normally be reached Mon - Fri, 4:30am - 7:00pm.
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.
/MULUGETA TUJI DUGDA/Examiner, Art Unit 2653
/Paras D Shah/Supervisory Patent Examiner, Art Unit 2653
05/30/2026
Prosecution Insights