DETAILED ACTION
This communication is in response to the Amendments and Arguments filed on 1/28/2026.
Claims 1-2,4-11 and 13-22 are pending and have been examined.
All previous objections / rejections not mentioned in this Office Action have been withdrawn by the examiner.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendments
Regarding the Applicant’s arguments for the rejections under 35 U.S.C. § 101, applicant has amended independent claim 1, 10, and 19 and added dependent claims 21 and 22. Applicant asserts that the amended claims recites a “specific technical solution to technical problems in natural language processing for workflow generation”. (Remarks P0012) Specifically, applicant asserts that conventional systems use both encoder and decoder models and the steps recited in the claim provides a solution that does not involve a decoder that can be large that needs “to train billions of parameters”. Examiner respectfully disagrees. During patent examination, pending claims must be “given their broadest reasonable interpretation consistent with the specification.” MPEP 2111. Also, claims should not be interpreted by reading limitations of the specification into the claim, to narrow the scope of the claim, by implicitly adding disclosed limitations that have no express basis in the claim language. In re Prater, 415 F.2d 1393. First, applicant asserts that the claimed steps is an improvement over conventional systems because the steps does not involve a decoder. However, the specification states that the encoder utilizes a transformer head that “includes feed forward neural network (FFNN), softmax function and argmax function”. (Spec. P0129) A person of ordinary skill in the art would recognize that a decoder can include a feed forward neural network and a softmax function. The term decoder is defined in the specification to “converts the hidden state vectors into a human-understandable result similar to human-like language description”. (Spec. P0061) Decoder is further defined by the statement “in order to generate descriptions that humans can understand, the decoder in the Transformer is very large and needs to train billions of parameters”. (Spec. P0063) Examiner interprets decoder, in the specification, narrowly as a neural network with billions of parameters that converts that hidden state vectors into human-understandable results. The steps to split sentences into sub-sentences involve the use of a neural network and softmax function as described in the specification that a person of ordinary skill in the art can understand as a decoder to classify tokens into tags. Second, the steps recited in the claim limitation can be performed in the mind. Specifically, the human mind can read text, think of phrases in the text, think of workflow label for the phrases, think of the task workflow that the phrases pertain, and write on paper using a pen or pencil the workflow. The use of an encoder-only transformer model can be interpreted as a set of rules or instructions and the human mind can follow determined rules or instructions. The claim encompasses mental observations or evaluations that can be practically performed in the human mind. Third, MPEP 2106.05(f) provides the following considerations for determining whether a claim simply recites a judicial exception with the words “apply it” (or an equivalent), such as mere instructions to implement an abstract idea on a computer: (1) whether the claim recites only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished; (2) whether the claim invokes computers or other machinery merely as a tool to perform an existing process; and (3) the particularity or generality of the application of the judicial exception. Here, the steps to use a transformer model to split sentences into sub-sentences does not describe any specific improvement to that would demonstrate integration of the abstract idea into a practical application. Specifically, the claim invokes computers as a tool to perform an existing process. The steps recited in the claim limitation does not provide the improvement in generating workflow from input text. Therefore, the claims as currently recited does not overcome the 35 U.S.C. § 101 abstract idea rejection.
Regarding the Applicant’s arguments for the rejections under 35 U.S.C. § 103, applicant has amended independent claim 1, 10, and 19 and added dependent claims 21 and 22. Applicant asserts that prior art references dos not teach the encoder-only transformer model for obtaining the first sub-sentence. Examiner respectfully disagrees. Maes teaches that the natural language processing engine may parse text represented by the input data (Maes P0044) and that the natural language processing engine may be a neural machine learning model (Maes P0048). Figure 6 of the specification includes 2 layers when obtaining a sub-sentence that include the encoder part and the head part. The head includes components that a person of ordinary skill in the art would consider to be part of a decoder. Therefore, prior art reference Maes teaches an encoder and decoder that corresponds to the encoder and head of the disclosure.
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.
Claim 1-2,4-11 and 13-22 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1, 10, and 19 the limitations of “receiving, as input to a workflow generation application executing on the computing device, text for describing a process of a task for which a workflow is to be generated”, “obtaining, using one or more encoder-only transformer models, a first sub-sentence group from the text for describing the process of the task, wherein the first sub-sentence group includes a plurality of sub-sentences arranged in an order in which the process occurs, each of the plurality of sub-sentences in the first sub-sentence group includes a workflow component, and the workflow component is configured to indicate information of a corresponding functional unit of a plurality of functional units for performing the task, wherein the obtaining the first sub-sentence group includes”, “decomposing the text for describing the process of the task to obtain a third sub-sentence group, wherein each of a plurality of sub-sentences included in the third sub-sentence group includes at most one workflow component, wherein the decomposing the text includes”, “performing sentence boundary detection on the text to obtain a plurality of sentences of the text”, “splitting, by a first encoder-only transformer model of the one or more encoder- only transformer models, each of the plurality of sentences of the text into one or more sub- sentences to obtain the third sub-sentence group”, “obtaining the first sub-sentence group based on the third sub-sentence group”, “determining a workflow label corresponding to each sub-sentence in the first sub-sentence group, wherein workflow labels corresponding to the plurality of sub-sentences in the first sub-sentence group include a first workflow label for indicating a first workflow component or a second workflow label for indicating a workflow pattern”, “generating the workflow of the task based on the workflow labels corresponding to the plurality of sub-sentences in the first sub-sentence group”, and “generating a visual representation of the workflow for display by the workflow generation application”, as drafted, are processes that, under broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components. More specifically, the mental process of a human reading text, thinking of phrases in the text, thinking of workflow label for the phrases, thinking of the task workflow that the phrases pertain, and writing on paper using a pen or pencil the workflow. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation in the mind but for the recitation of generic computer components, then it falls within the --Mental Processes-- grouping of abstract ideas. Accordingly, the claims recite an abstract idea.
This judicial exception is not integrated into a practical application because the recitation of an application on a computing device in claim 1 and 10 and a non-transitory computer readable storage medium in claims 19, reads to generalized computer components, based upon the claim interpretation wherein the structure is interpreted using P0087-P0108 and P0213-225 in the specification. 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 claims are directed to an abstract idea.
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract idea into a practical application, the additional element of using generalized computer components to read text, think of phrases in the text, think of workflow label for the phrases, think of the task workflow that the phrases pertain, and write on paper using a pen or pencil the workflow amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. The claims are not patent eligible.
With respect to claim 2, 11, and 20, the claim recites “filtering the plurality of sub-sentences in the third sub-sentence group to obtain a fourth sub-sentence group, wherein each of a plurality of sub-sentences included in the fourth sub-sentence group includes one workflow component”, “reordering the plurality of sub-sentences in the fourth sub-sentence group to obtain the first sub-sentence group”, “the obtaining the first sub-sentence group from the text for describing the process of the task further includes”, “filtering the plurality of sub-sentences in the third sub-sentence group to obtain the fourth sub-sentence group, wherein each of the plurality of sub-sentences included in the fourth sub-sentence group includes one workflow component”, “reordering the plurality of sub-sentences in the fourth sub-sentence group to obtain a second sub-sentence group”, and “inserting at least one identifier of at least one workflow pattern into the second sub-sentence group to obtain the first sub-sentence group, wherein the at least one workflow pattern is configured to indicate a structure of the process of the task and a relationship between the plurality of functional units”, which reads on a human segmenting, filtering, reordering, and labeling words read in the mind. No additional limitations are present.
With respect to claim 4 and 13, the claim recites “determining, by a second encoder-only transformer model of the one or more encoder-only transformer models, whether each sub-sentence of the plurality of sub-sentences in the third sub-sentence group includes one workflow component” and “if the sub-sentence in the third sub-sentence group includes zero workflow component, removing the sub-sentence in the third sub-sentence group, so as to obtain the fourth sub-sentence group”, which reads on a human filtering words in a sentence in the mind. No additional limitations are present.
With respect to claim 5 and 14, the claim recites “reordering, by a third encoder-only transformer model of the one or more encoder-only transformer models, the plurality of sub-sentences in the fourth sub-sentence group, wherein an input of the third encoder-only transformer model is two sub-sentences in the fourth sub-sentence group, and an output of the third encoder-only transformer model is information for indicating whether an order of the two sub-sentences is correct”, which reads on a human reordering phrases read in the mind. No additional limitations are present.
With respect to claim 6 and 15, the claim recites “determining the at least one workflow pattern of the second sub-sentence group; for each workflow pattern of the at least one workflow pattern, obtaining the first sub-sentence group by performing, on the second sub-sentence group, at least one of the following”, “if no sub-sentence for indicating a boundary of the workflow pattern exists in the second sub-sentence group, inserting the workflow pattern boundary indicating sentence of the workflow pattern into the second sub-sentence group”, and “if no sub-sentence serving as the workflow pattern indicator exists in the second sub-sentence group, inserting the workflow pattern indicator of the workflow pattern into the second sub-sentence group”, which reads on a human putting labels on words read in the mind. No additional limitations are present.
With respect to claim 7 and 16, the claim recites “performing a pattern keyword detection on a plurality of sub-sentences in the second sub-sentence group” and “determining the at least one workflow pattern according to at least one pattern keyword that is detected, wherein each of the at least one pattern keyword is configured to indicate a workflow pattern of the at least one respective workflow pattern”, which reads on a human detecting keywords from read text in the mind. No additional limitations are present.
With respect to claim 8 and 17, the claim recites “matching each sub-sentence in the first sub-sentence group to a plurality of workflow labels included in a label dictionary, so as to determine the workflow label corresponding to each sub-sentence in the first sub-sentence group from the label dictionary”, which reads on a human detecting keywords from read text that match words on a list in the mind. No additional limitations are present.
With respect to claim 9 and 18, the claim recites “performing semantic similarity processing to match each sub-sentence in the first sub-sentence group to the pre-defined values in the label dictionary” and “determining a workflow label corresponding to a pre-defined value with a highest similarity score with respect to the sub-sentence in the pre-defined values as the workflow label corresponding to the sub-sentence”, which reads on a human detecting keywords from read text that are similar words on a list in the mind. No additional limitations are present. No additional limitations are present.
With respect to claim 21, the claim recites “wherein the first encoder-only transformer model is a pre-trained transformer model comprising”, “a plurality of sequentially connected encoders, each encoder of which having a plurality of input terminals and a plurality of output terminals, and a plurality of classification heads, each classification head including”, and “a feed-forward neural network (FFNN),a softmax function configured to normalize probability values, and an argmax function configured to select a highest probability value.”, which reads on a human utilizing a set of instructions or computations on paper using a pen or pencil. No additional limitations are present. No additional limitations are present.
With respect to claim 22, the claim recites “wherein the first encoder-only transformer model comprises”, “a plurality of sequentially connected encoders, each encoder of which configured to process tokens of an input sentence and output hidden state vectors”, “a plurality of classification heads, each classification head of which connected to an output of a corresponding encoder and configured to assign a Begin-Inside-Outside-End (BIOE) tag to each token to identify sub-sentence boundaries, wherein the splitting comprises” and “splitting each sentence based on the BIOE tags.”, which reads on human utilizing a set of instructions or computations on paper using a pen or pencil to assign tags. No additional limitations are present. No additional limitations are present.
These claims further do not remedy the judicial exception being integrated into a practical application and further fail to include additional elements that are sufficient to amount to significantly more than the judicial exception.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 8, 10, 17, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Brunswig et al. (U.S. PG Pub No. 20170185255), hereinafter Brunswig, in view of Maes et al. (U.S. PG Pub No. 20200192975), hereinafter Maes, and in further view of Burton (U.S. PG Pub No. 20250165717).
Regarding claim 1, 10 and 19 Brunswig teaches:
(Claim 1) A text processing method, performed by a computing device, comprising: (A text processing method, performed by a computing device, comprising:)
(Claim 10) A computing device, comprising: a memory; and at least one processor connected to the memory; wherein the memory is configured to store computer instructions that, when executed by the at least one processor, cause the computing device to implement: (A computing device, comprising: a memory; and at least one processor connected to the memory; wherein the memory is configured to store computer instructions that, when executed by the at least one processor, cause the computing device to implement:)
(Claim 19) A non-transitory computer-readable storage medium having stored computer instructions thereon, wherein when executed by a computer, the computer instructions cause the computer to implement: (A non-transitory computer-readable storage medium having stored computer instructions thereon, wherein when executed by a computer, the computer instructions cause the computer to implement:)
receiving, as input to a workflow generation application executing on the computing device, text for describing a process of a task for which a workflow is to be generated; (P0061, FIG. 6 is a diagram of example screens illustrating extraction of data elements or data objects from user input, and the use of the data elements or data objects in generating an assistant item. The screen can be at least generally similar to the screen, including presenting a text entry field. The text entry field is shown as containing text entered by a user. The text can represent, for example, a note that the user wishes to record with the workflow assistant. The user can enter the text into the workflow assistant for processing (e.g., for analysis or storage) by selecting a send icon.)
obtaining, using one or more encoder-only transformer models, a first sub-sentence group from the text for describing the process of the task, wherein the first sub-sentence group includes a plurality of sub-sentences arranged in an order in which the process occurs, each of the plurality of sub-sentences in the first sub-sentence group includes a workflow component, and the workflow component is configured to indicate information of a corresponding functional unit of a plurality of functional units for performing the task, wherein the obtaining the first sub-sentence group includes: (P0062, As part of the processing of the text, the workflow assistant, or a component in communication with the workflow assistant, can parse the text, such as to associate text elements (e.g., words or phrases) with reference keywords.)
decomposing the text for describing the process of the task to obtain a third sub-sentence group, wherein each of a plurality of sub-sentences included in the third sub-sentence group includes at most one workflow component, wherein the decomposing the text includes: (P0062, As part of the processing of the text, the workflow assistant, or a component in communication with the workflow assistant, can parse the text, such as to associate text elements (e.g., words or phrases) with reference keywords. Reference keywords, in some cases, can represent data objects or data elements maintained in a computing system associated with the workflow assistant.)
determining a workflow label corresponding to each sub-sentence in the first sub-sentence group, wherein workflow labels corresponding to the plurality of sub-sentences in the first sub-sentence group include a first workflow label for indicating a workflow component or a second workflow label for indicating a workflow pattern; (P0062, As part of the processing of the text, the workflow assistant, or a component in communication with the workflow assistant, can parse the text, such as to associate text elements (e.g., words or phrases) with reference keywords. Reference keywords, in some cases, can represent data objects or data elements maintained in a computing system associated with the workflow assistant.)
generating the workflow of the task based on the workflow labels corresponding to the plurality of sub-sentences in the first sub-sentence group; and (P0064, An analysis component of, or in communication with, the workflow assistant can also analyze the text to try and predict actions a user may wish to take, or suggest actions that may be helpful to a user. For example, the occurrence of a particular data object or data element associated with the text, or a particular collection of such objects or elements, may often be associated with one, or more, subsequent actions, such as the creation of a new data object, assistant item, or collection. In a specific case, the occurrence of the name of a company, or the name of an individual associated with a company, in connection with other keywords (e.g., order, invoice, sale), the name of a product, other data objects or data elements, or combinations thereof, may indicate that a user may wish to generate a new sales quote, or that the user may wish to consider doing so.)
generating a visual representation of the workflow for display by the workflow generation application. (P0067, The screen can show the user the result of taking the suggested action after selection of the action interface element, or guide the user in completing the suggested action. The screen can includes fields that can be populated with information from the text, context information of the base application, or data elements or data objects associated with a current assistant item or collection.)
Brunswig does not specifically teach:
performing sentence boundary detection on the text to obtain a plurality of sentences of the text, and
splitting, by a first encoder-only transformer model of the one or more encoder- only transformer models, each of the plurality of sentences of the text into one or more sub- sentences to obtain the third sub-sentence group; and
obtaining the first sub-sentence group based on the third sub-sentence group;
Maes, however, teaches:
splitting, by a first encoder-only transformer model of the one or more encoder- only transformer models, each of the plurality of sentences of the text into one or more sub- sentences to obtain the third sub-sentence group; and (P0044, The natural language processing engine may parse text represented by the input data to separate the text into words associated with a particular language construct and then further encode the parsed text into tokens.; P0048, Natural language processing engine may be a neural machine learning model, such as a recurrent neural network (RNN) model.)
obtaining the first sub-sentence group based on the third sub-sentence group; (P0045, The input sequence for the example depicted in FIG. 1A may be parsed into the following units of text: Migrate; VM-1; from; DataCenter1; to; and DataCenter2. Based on the units of text and other criteria, the natural language processing engine may then encode the units into corresponding tokens.)
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to split sentences into sub-sentences. It would have been obvious to combine the references because the input data that represents a sequence of steps to solve a particular problem may not be sufficient to solve the problem and modification may be required by the natural learning generation engine. (Maes P0022)
Brunswig in view of Maes does not specifically teach:
performing sentence boundary detection on the text to obtain a plurality of sentences of the text, and
Burton, however, teaches:
performing sentence boundary detection on the text to obtain a plurality of sentences of the text, and (P0015, Sentence boundary detection.)
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to perform sentence boundary detection. It would have been obvious to combine the references because sentence boundary detection is a known technique to yield a predictable result of annotating or labeling sentences or phrases in sentences for further language processing. (Burton P0015)
Regarding claim 8 and 17 Brunswig, in view of Maes, and further view of Burton teach claim 1 and 10.
Brunswig further teaches:
matching each sub-sentence in the first sub-sentence group to a plurality of workflow labels included in a label dictionary, so as to determine the workflow label corresponding to each sub-sentence in the first sub-sentence group from the label dictionary. (P0092, A data object can be associated with a dictionary template, such that potential keywords or search terms can be extracted from the data object. The search client can determine whether words, phrases, numbers, or other content of the text analysis results may be associated with, or related to, data objects or data elements used in the architecture and provide search results. A matching component can access the search results and text analysis results and, using matching rules, determine whether the elements of the text analysis results should be associated with a corresponding element of the search results.)
Claims 2, 4-7, 11, 13-16, 20 are rejected under 35 U.S.C. 103 as being unpatentable over Brunswig in view of Maes, in view of Burton, and further view of Suparna et al. (U.S. PG Pub No. 20180113850), hereinafter Suparna.
Regarding claim 2, 11, and 20 Brunswig in view of Maes and further view of Burton teach claim 1, 10, and 19.
Brunswig further teaches:
obtaining the first sub-sentence group from the text for describing the process of the task further includes: (P0062, As part of the processing of the text, the workflow assistant, or a component in communication with the workflow assistant, can parse the text, such as to associate text elements (e.g., words or phrases) with reference keywords. Reference keywords, in some cases, can represent data objects or data elements maintained in a computing system associated with the workflow assistant.)
inserting at least one identifier of at least one workflow pattern into the second sub-sentence group to obtain the first sub-sentence group, wherein the at least one workflow pattern is configured to indicate a structure of the process of the task and a relationship between the plurality of functional units. (P0064, The workflow assistant can also analyze the text to try and predict actions a user may wish to take, or suggest actions that may be helpful to a user. For example, the occurrence of a particular data object or data element associated with the text, or a particular collection of such objects or elements, may often be associated with one, or more, subsequent actions, such as the creation of a new data object, assistant item, or collection.; P0092, The search client can determine whether words, phrases, numbers, or other content of the text analysis results may be associated with, or related to, data objects or data elements used in the architecture and provide search results.; P0093, If a search result is identified as relevant, it can be identified for the user. For example, the name or other identifier of the search result can be displayed to the user.)
Brunswig does not specifically teach:
filtering the plurality of sub-sentences in the third sub-sentence group to obtain a fourth sub-sentence group, wherein each of a plurality of sub-sentences included in the fourth sub-sentence group includes one workflow component; and
reordering the plurality of sub-sentences in the fourth sub-sentence group to obtain the first sub-sentence group; or
filtering the plurality of sub-sentences in the third sub-sentence group to obtain the fourth sub-sentence group, wherein each of the plurality of sub-sentences included in the fourth sub-sentence group includes one workflow component;
reordering the plurality of sub-sentences in the fourth sub-sentence group to obtain a second sub-sentence group; and
Maes, however, teaches:
reordering the plurality of sub-sentences in the fourth sub-sentence group to obtain the first sub-sentence group; or (P0047, For the example of FIG. 2, the machine translation produces a single output cloud orchestration workflow that includes the following ordered sequence of operations.)
reordering the plurality of sub-sentences in the fourth sub-sentence group to obtain a second sub-sentence group; and (P0047, For the example of FIG. 2, the machine translation produces a single output cloud orchestration workflow that includes the following ordered sequence of operations.)
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to reorder subsentences. It would have been obvious to combine the references because the input data that represents a sequence of steps to solve a particular problem may not be sufficient to solve the problem and modification may be required by the natural learning generation engine. (Maes P0022)
Brunswig in view of Maes does not specifically teach:
filtering the plurality of sub-sentences in the third sub-sentence group to obtain a fourth sub-sentence group, wherein each of a plurality of sub-sentences included in the fourth sub-sentence group includes one workflow component; and
filtering the plurality of sub-sentences in the third sub-sentence group to obtain the fourth sub-sentence group, wherein each of the plurality of sub-sentences included in the fourth sub-sentence group includes one workflow component;
Suparna, however, teaches:
filtering the plurality of sub-sentences in the third sub-sentence group to obtain a fourth sub-sentence group, wherein each of a plurality of sub-sentences included in the fourth sub-sentence group includes one workflow component; and (P0024, Each word in the natural language statement may be classified, for example, as one of a noun, a verb, an adjective and a preposition, Words that do not fit one of the classifications may be ignored.; P0028, Second word mapper may remove any prepositions from the group of unmapped words.)
filtering the plurality of sub-sentences in the third sub-sentence group to obtain the fourth sub-sentence group, wherein each of the plurality of sub-sentences included in the fourth sub-sentence group includes one workflow component; (P0024, Each word in the natural language statement may be classified, for example, as one of a noun, a verb, an adjective and a preposition, Words that do not fit one of the classifications may be ignored.; P0028, Second word mapper may remove any prepositions from the group of unmapped words.)
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to filter sub-sentences. It would have been obvious to combine the references because the filtering process is a known technique to yield a predictable result of disregarding words that are not part of potential workflows. (Suparna P0033)
Regarding claim 4 and 13 Brunswig, in view of Maes, in view of Burton, and further view of Suparna teach claim 2 and 11.
Brunswig does not specifically teach:
determining, by a second encoder-only transformer model of the one or more encoder-only transformer models, whether each sub-sentence of the plurality of sub-sentences in the third sub-sentence group includes one workflow component; and
if the sub-sentence in the third sub-sentence group includes zero workflow component, removing the sub-sentence in the third sub-sentence group, so as to obtain the fourth sub-sentence group.
Maes, however, teaches:
determining, by a second encoder-only transformer model of the one or more encoder-only transformer models, whether each sub-sentence of the plurality of sub-sentences in the third sub-sentence group includes one workflow component; and (P0045, The input sequence for the example depicted in FIG. 1A may be parsed into the following units of text: Migrate; VM-1; from; DataCenter1; to; and DataCenter2. Based on the units of text and other criteria, the natural language processing engine may then encode the units into corresponding tokens and apply the machine translation to these tokens based on the natural language processing model to generate one or multiple cloud orchestration workflows.)
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to determine workflow component from sub-sentence. It would have been obvious to combine the references because determining workflow component from sub-sentence is a known technique to yield a predictable result of generating a workflow from a sentence. (Maes P0022)
Suparna, however, teaches:
if the sub-sentence in the third sub-sentence group includes zero workflow component, removing the sub-sentence in the third sub-sentence group, so as to obtain the fourth sub-sentence group. (P0033, If no matching implementations are found, the noun may marked as blank.)
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to remove sub-sentences. It would have been obvious to combine the references because the filtering process is a known technique to yield a predictable result of disregarding words that are not part of potential workflows. (Suparna P0033)
Regarding claim 5 and 14 Brunswig, in view of Maes, in view of Burton, and further view of Suparna teach claim 2 and 11.
Brunswig does not specifically teach:
reordering, by a third encoder-only transformer model of the one or more encoder-only transformer models, the plurality of sub-sentences in the fourth sub-sentence group, wherein an input of the third encoder-only transformer model is two sub-sentences in the fourth sub-sentence group, and an output of the third encoder-only transformer model is information for indicating whether an order of the two sub-sentences is correct.
Maes, however, teaches:
reordering, by a third encoder-only transformer model of the one or more encoder-only transformer models, the plurality of sub-sentences in the fourth sub-sentence group, wherein an input of the third encoder-only transformer model is two sub-sentences in the fourth sub-sentence group, and an output of the third encoder-only transformer model is information for indicating whether an order of the two sub-sentences is correct. (P0047, For the example of FIG. 2, the machine translation produces a single output cloud orchestration workflow that includes the following ordered sequence of operations.; P0049, Translate a sequence of words or tokens describing the input sequence to a corresponding output sequence.)
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to reorder subsentences. It would have been obvious to combine the references because the input data that represents a sequence of steps to solve a particular problem may not be sufficient to solve the problem and modification may be required by the natural learning generation engine. (Maes P0022)
Regarding claim 6 and 15 Brunswig, in view of Maes, in view of Burton, and further view of Suparna teach claim 2 and 11.
Brunswig further teaches:
determining the at least one workflow pattern of the second sub-sentence group; for each workflow pattern of the at least one workflow pattern, obtaining the first sub-sentence group by performing, on the second sub-sentence group, at least one of the following: if no sub-sentence for indicating a boundary of the workflow pattern exists in the second sub-sentence group, inserting the workflow pattern boundary indicating sentence of the workflow pattern into the second sub-sentence group; or if no sub-sentence serving as the workflow pattern indicator exists in the second sub-sentence group, inserting the workflow pattern indicator of the workflow pattern into the second sub-sentence group. (P0064, The workflow assistant can also analyze the text to try and predict actions a user may wish to take, or suggest actions that may be helpful to a user. For example, the occurrence of a particular data object or data element associated with the text, or a particular collection of such objects or elements, may often be associated with one, or more, subsequent actions, such as the creation of a new data object, assistant item, or collection.; P0092, The search client can determine whether words, phrases, numbers, or other content of the text analysis results may be associated with, or related to, data objects or data elements used in the architecture and provide search results.; P0093, If a search result is identified as relevant, it can be identified for the user. For example, the name or other identifier of the search result can be displayed to the user.)
Regarding claim 7 and 16 Brunswig, in view of Maes, in view of Burton, and further view of Suparna teach claim 6 and 15.
Brunswig further teaches:
performing a pattern keyword detection on a plurality of sub-sentences in the second sub-sentence group; and determining the at least one workflow pattern according to at least one pattern keyword that is detected, wherein each of the at least one pattern keyword is configured to indicate a respective workflow pattern of the at least one workflow pattern. (P0064, The workflow assistant can also analyze the text to try and predict actions a user may wish to take, or suggest actions that may be helpful to a user. For example, the occurrence of a particular data object or data element associated with the text, or a particular collection of such objects or elements, may often be associated with one, or more, subsequent actions, such as the creation of a new data object, assistant item, or collection.; P0092, The search client can determine whether words, phrases, numbers, or other content of the text analysis results may be associated with, or related to, data objects or data elements used in the architecture and provide search results.; P0093, If a search result is identified as relevant, it can be identified for the user. For example, the name or other identifier of the search result can be displayed to the user.)
Claims 9 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Brunswig Brunswig, in view of Maes, in view of Burton, and in further view of Carus et al. (U.S. PG Pub No. 20160350283), hereinafter Carus.
Regarding claim 9 and 18 Brunswig, in view of Maes, amd in further view of Burton teach claim 8 and 10.
Brunswig further teaches:
performing semantic similarity processing to match each sub-sentence in the first sub-sentence group to the pre-defined values in the label dictionary; and (P0092, A data object can be associated with a dictionary template, such that potential keywords or search terms can be extracted from the data object. The search client can determine whether words, phrases, numbers, or other content of the text analysis results may be associated with, or related to, data objects or data elements used in the architecture and provide search results. A matching component can access the search results and text analysis results and, using matching rules, determine whether the elements of the text analysis results should be associated with a corresponding element of the search results.)
Brunswig, in view of Maes, and in further view of Burton does not specifically teach:
determining a workflow label corresponding to a pre-defined value with a highest similarity score with respect to the sub-sentence in the pre-defined values as the workflow label corresponding to the sub-sentence.
Carus, however, teaches:
determining a workflow label corresponding to a pre-defined value with a highest similarity score with respect to the sub-sentence in the pre-defined values as the workflow label corresponding to the sub-sentence. (P0052, Said terms in the at least two texts are allowed by contributing their respective distances only once and only with their respective highest matching scores, where the at least two texts are ranked for similarity.)
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to determine semantic similarity based on similarity score. It would have been obvious to combine the references because the measure of semantic similarity of words, phrases, and texts is a known technique to yield a predictable result of obtaining similar words or phrases. (Carus P0004)
Claims 21 and 22 are rejected under 35 U.S.C. 103 as being unpatentable over Brunswig Brunswig, in view of Maes, in view of Burton, and in further view of Gao et al. (U.S. PG Pub No. 20200065374), hereinafter Gao.
Regarding claim 21 Brunswig, in view of Maes, and in further view of Burton teach claim 1.
Brunswig, in view of Maes, and in further view of Burton does not specifically teach:
wherein the first encoder-only transformer model is a pre-trained transformer model comprising:a plurality of sequentially connected encoders, each encoder of which having a plurality of input terminals and a plurality of output terminals, and a plurality of classification heads, each classification head including:a feed-forward neural network (FFNN),a softmax function configured to normalize probability values, and an argmax function configured to select a highest probability value.
Gao, however, teaches:
wherein the first encoder-only transformer model is a pre-trained transformer model comprising: a plurality of sequentially connected encoders, each encoder of which having a plurality of input terminals and a plurality of output terminals, and a plurality of classification heads, each classification head including: a feed-forward neural network (FFNN), a softmax function configured to normalize probability values, and an argmax function configured to select a highest probability value. (P0034, NER network for predicting named entity tags and a RE network for predicting relation labels. In some embodiments, NER network may include several layers, such as a word representing layer where several types of representations are determined, a sequential encoder layer, and a decoding layer. In some embodiments, word representation layer may further include an attention-based subword encoder sublayer, a dilation-based subword encoder sublayer, a capitalization embedding sublayer, and a word embedding sublayer. In some embodiments, as shown by FIG. 2, RE network may also include several layers, such as a position embedding layer, a relation encoder layer, and a softmax layer. In some embodiments, RE network may share word representation layer with NER.)
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention to use an encoder with multiple layers. It would have been obvious to combine the references because the use of a neural network with softmax and argmax are a known technique to yield a predictable result of predicting labels for input text. (Gao P0043, P0044)
Regarding claim 22 Brunswig, in view of Maes, and in further view of Burton teach claim 1.
Brunswig, in view of Maes, and in further view of Burton does not specifically teach:
wherein the first encoder-only transformer model comprises: a plurality of sequentially connected encoders, each encoder of which configured to process tokens of an input sentence and output hidden state vectors, and a plurality of classification heads, each classification head of which connected to an output of a corresponding encoder and configured to assign a Begin-Inside-Outside-End (BIOE) tag to each token to identify sub-sentence boundaries, wherein the splitting comprises: splitting each sentence based on the BIOE tags.
Gao, however, teaches:
wherein the first encoder-only transformer model comprises: a plurality of sequentially connected encoders, each encoder of which configured to process tokens of an input sentence and output hidden state vectors, and a plurality of classification heads, each classification head of which connected to an output of a corresponding encoder and configured to assign a Begin-Inside-Outside-End (BIOE) tag to each token to identify sub-sentence boundaries, wherein the splitting comprises: splitting each sentence based on the BIOE tags. (P0043, Relation classification component further include relation encoder layer.; P0044, Relation classification component further includes softmax layer to predict the relation labels based on the output of relation encoder layer.; Fig. 3, Sequentially connected layers.; P0026-P0030, NER can use the BIOES (Begin, Inside, Outside, End, Single) tagging scheme. For example, if there are two entity tags T1 and T2, all available labels are {B-T1, I-T1, O-T1, E-T1, S-T1, B-T2, I-T2, O-T2, E-T2, S-T2}, and each word will be assigned such a label. Training text “Jim had a bypass heart surgery at the Massachusetts General Hospital in 2010” can be tagged as: … corresponding relation labels may be as follows: (operated on, Bypass Heart Surgery, Jim) (treated in, Jim, Massachusetts General Hospital) (happened during, Bypass Heart Surgery, 2010))
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention to use an encoder to assign tags. It would have been obvious to combine the references because the use of a neural network with softmax and argmax are a known technique to yield a predictable result of predicting labels for input text. (Gao P0043, P0044)
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DANIEL WONSUK CHUNG whose telephone number is (571)272-1345. The examiner can normally be reached Monday - Friday (7am-4pm)[PT].
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, PIERRE-LOUIS DESIR can be reached at (571)272-7799. 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.
/DANIEL W CHUNG/Examiner, Art Unit 2659
/PIERRE LOUIS DESIR/Supervisory Patent Examiner, Art Unit 2659
Prosecution Insights