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 Arguments
The applicant's arguments/remarks, filed 12/24/2025, see pages 8 – 9, with respect to 35 U.S.C 102 and 103 rejections of Claims 1-120 have been fully considered but are moot in view of the new ground(s) of rejection. The arguments/remarks are essentially directed towards the newly introduced limitations and they are addressed in this Office Action, below.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 2, 9, 10, 11, 12, 18, 19 and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kelkar et al. (US PG PUB 20230274095), hereinafter "Kelkar" in views of Kannan et al. (US PG PUB 20180129484)l, hereinafter "Kannan".
Regarding Claim1, Kelkar discloses:
A method comprising:
inputting a conversation, including a sequence of messages, into a machine learning model (i.e. conversations including sentences, utterances, etc. [i.e. a sequence of messages], may be inputted into a Generative DNN model [i.e. a machine learning model]) (Fig. 4, ¶ 0025 and ¶ 0029);
generating, via execution of the machine learning model, a plurality of annotations comprising a plurality of canonical forms corresponding to the sequence of messages (i.e. method/system, e.g. Annotation module, may generate, via execution of the Generative DNN model [i.e. the machine learning model], annotations [i.e. a plurality of annotations] comprised of auto-intents, e.g. “cancel subscription”, “cancel reason”, etc. [i.e. a plurality of standard/canonical forms], corresponding to the user utterances, e.g. “I want to cancel my plan”, “I just don’t need anymore”, etc. [i.e. the sequence of messages]) (Fig. 4, Fig. 6, ¶ 0029 – 0030 and ¶ 0032),
individual canonical forms including a constrained semantic representation of a respective message included in the sequence of messages (i.e. auto-intents, e.g. “cancel subscription”, “cancel reason”, etc. [i.e. individual standard/canonical forms], includes intents that represents [i.e. a constrained semantic representation] messages, e.g. “I want to cancel my plan” and “I just don’t need anymore” [i.e. a respective message], included in the user utterances [i.e. the sequence of messages) (Fig. 6 and ¶ 0031 - 0032);
generating one or more dialogue flows using the plurality of canonical forms (i.e. Once all the conversations are annotated, AutoFlows Generator 209 generates the conversation flows [i.e. one or more dialogue flows] per auto-topic and auto-subtopic in a graph where the root node denotes the start of all conversation flows having the same auto-topic and auto-subtopic. A conversation flow [i.e. dialogue flows] is a sequence of sentence-level auto-intents [i.e. using the plurality of canonical forms] and turn-level auto-responses) (¶ 0039 - 0040); and
causing a conversational output to be generated (i.e. the method/system can automatically have dialog with users [i.e. causing a conversational output to be generated] using ACAI system trained based on the conversation flows) (Abstract, Fig. 7, ¶ 0041 and ¶ 0077 - 0078).
However, Kelkar does not explicitly disclose:
each dialogue flow comprising one or more executable instructions that specify one or more next steps of the conversation; and
On the other hand, in the same field of endeavor, Kannan teaches:
each dialogue flow comprising one or more executable instructions that specify one or more next steps of the conversation (i.e. developer may author/generate a machine conversation dialog flow comprising scripts, e.g. if BusinessName = empty then Prompt “Where would you like to book your event? [i.e.one or more executable instructions], that specify one or more conditional steps [i.e. one or more next steps]; For example, based on “if” and “else” statements of the script [i.e. one or more executable instructions] that depend on BusinessName variable, the machine conversation dialog flow may prompt “Where would you like to book your event? [i.e. one or more next steps of the conversation]) (302 & 333 - Fig. 3, 401 & 403 - Fig. 4, 501, 502 & 503 - Fig. 5, ¶ 0040, ¶ 0043 – 0044 and ¶ 0047); and
causing a conversational output to be generated through execution of at least one of the one or more executable instructions (i.e. based on “if” and “else” statements of the script [i.e. through execution of at least one of the one or more executable instructions] that depend on BusinessName variable, the machine conversation dialog flow may prompt “Where would you like to book your event? [i.e. causing a conversational output to be generated]) (501, 502 & 503 - Fig. 5, ¶ 0043 – 0044 and ¶ 0047).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the method/system of Kelkar to include the features for wherein each dialogue flow comprising one or more executable instructions that specify one or more next steps of the conversation; and causing a conversational output to be generated through execution of at least one of the one or more executable instructions as taught by Kannan so that the method/system may be implemented with a user interface that permits developers to edit/update the dialog flow in the script syntax (¶ 0047).
Regarding Claim 2, Kelkar and Kannan disclose, in particular Kelkar teaches:
wherein the machine learning model is trained based at least on a plurality of conversations and a second plurality of canonical forms for additional sequences of messages included in the plurality of conversations (i.e. ACAI [i.e. the machine learning model] which is trained based on conversation logs [i.e. a plurality of conversations] may be retrained with new conversation flows and new phrases [i.e. a second plurality of canonical forms] discovered from addition messages/conversations [i.e. additional sequences of messages included in the plurality of conversations]) (¶ 0100 – 0105).
Regarding Claim 9, Kelkar and Kannan disclose, in particular Kelkar teaches:
wherein the machine learning model includes a large language model (LLM) (i.e. The Generative DNN model is a language model described in the NLP literature in many applications. Some applications where a language model is used to generate text include summarization of text, story telling, programming code generation, etc. [i.e. a large language model (LLM)]) (¶ 0029).
Regarding Claim 10, Kelkar and Kannan disclose, in particular Kelkar teaches:
wherein the plurality of canonical forms and the one or more dialogue flows are specified in a formal modeling language (i.e. A conversation flow [i.e. dialogue flows] comprising a sequence of sentence-level auto-intents [i.e. the plurality of canonical forms] and turn-level auto-responses are specified natural language understanding NLU and natural language generation NLG models [i.e. a formal modeling language]) (¶ 0008 - 0009).
Regarding Claim 11, Kelkar discloses:
A processor (i.e. a conversational AI system) (Fig. 1 and ¶ 0016) comprising:
one or more processing units to perform operations (i.e. processor/processing modules) (¶ 0061 and ¶ 0063)) comprising:
inputting a plurality of conversations into a machine learning model (i.e. conversations including sentences, utterances, etc. [i.e. a plurality of conversations], may be inputted into a Generative DNN model [i.e. a machine learning model]) (Fig. 4, ¶ 0025 and ¶ 0029);
generating, based at least on the machine learning model processing the plurality of conversations, a plurality of annotations comprising a plurality of constrained semantic representations for respective messages of sequences of messages included in the plurality of conversations (i.e. method/system, e.g. Annotation module, may generate, via execution of the Generative DNN model [i.e. the machine learning model] processing conversations including sentences, utterances, etc. [i.e. the plurality of conversations], annotations [i.e. a plurality of annotations] comprised of auto-intents, e.g. “cancel subscription”, “cancel reason”, etc. [i.e. constrained semantic representations], corresponding to the user utterances, e.g. “I want to cancel my plan”, “I just don’t need anymore”, etc. [i.e. respective messages of sequences of messages included in the plurality of conversations]) (Fig. 4, Fig. 6, ¶ 0029 –0032);
generating one or more dialogue flows using the plurality of constrained semantic representations (i.e. Once all the conversations are annotated, AutoFlows Generator 209 generates the conversation flows [i.e. one or more dialogue flows] per auto-topic and auto-subtopic in a graph where the root node denotes the start of all conversation flows having the same auto-topic and auto-subtopic. A conversation flow [i.e. dialogue flows] is a sequence of sentence-level auto-intents [i.e. the plurality of constrained semantic representations] and turn-level auto-responses) (¶ 0039 - 0040); and
causing a conversational output to be generated (i.e. the method/system can automatically have dialog with users [i.e. causing a conversational output to be generated] using ACAI system trained based on the conversation flows [i.e. the one or more dialogue flows]) (Abstract, Fig. 7, ¶ 0041 and ¶ 0077 - 0078).
However, Kelkar does not explicitly disclose:
each dialogue flow comprising one or more executable instructions that specify one or more next steps of the conversation; and
On the other hand, in the same field of endeavor, Kannan teaches:
each dialogue flow comprising one or more executable instructions that specify one or more next steps of the conversation (i.e. developer may author/generate a machine conversation dialog flow comprising scripts, e.g. if BusinessName = empty then Prompt “Where would you like to book your event? [i.e.one or more executable instructions], that specify one or more conditional steps [i.e. one or more next steps]; For example, based on “if” and “else” statements of the script [i.e. one or more executable instructions] that depend on BusinessName variable, the machine conversation dialog flow may prompt “Where would you like to book your event? [i.e. one or more next steps of the conversation]) (302 & 333 - Fig. 3, 401 & 403 - Fig. 4, 501, 502 & 503 - Fig. 5, ¶ 0040, ¶ 0043 – 0044 and ¶ 0047); and
causing a conversational output to be generated through execution of at least one of the one or more executable instructions (i.e. based on “if” and “else” statements of the script [i.e. through execution of at least one of the one or more executable instructions] that depend on BusinessName variable, the machine conversation dialog flow may prompt “Where would you like to book your event? [i.e. causing a conversational output to be generated]) (501, 502 & 503 - Fig. 5, ¶ 0043 – 0044 and ¶ 0047).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the method/system of Kelkar to include the features for wherein each dialogue flow comprising one or more executable instructions that specify one or more next steps of the conversation; and causing a conversational output to be generated through execution of at least one of the one or more executable instructions as taught by Kannan so that the method/system may be implemented with a user interface that permits developers to edit/update the dialog flow in the script syntax (¶ 0047).
Regarding Claim 12, Kelkar and Kannan disclose, in particular Kelkar teaches:
wherein the machine learning model, prior to deployment, is fine-tuned based on a second plurality of conversations and a second plurality of constrained semantic representations for additional sequences of messages included in the second plurality of conversations (i.e. ACAI [i.e. the machine learning model] which is trained based on conversation logs [i.e. a plurality of conversations] may be retrained [i.e. fine-tuned] with new conversation flows and new phrases [i.e. a second plurality of constrained semantic representations] discovered from addition messages/conversations [i.e. additional sequences of messages included in the plurality of conversations]) (¶ 0100 – 0105).
Regarding Claim 18, Kelkar and Kannan disclose, in particular Kelkar teaches:
wherein the one or more processors are comprised in at least one of:
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing collaborative content creation for 3D assets;
a system for performing one or more deep learning operations;
a system implemented using an edge device;
a system for generating or presenting at least one of virtual reality content, augmented reality content, or mixed reality content;
a system implemented using a robot; a system for performing one or more conversational AI operations; a system implemented using one or more large language models (LLMs); a system for generating synthetic data; a system for performing one or more generative AI operations;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources) (i.e. The Generative DNN model is a language model described in the NLP literature in many applications. Some applications where a language model is used to generate text include summarization of text, storytelling, programming code generation, etc. [i.e. a system for performing one or more generative AI operations]) (¶ 0029).
Regarding Claim 19, Kelkar discloses:
A system comprising: one or more processing units to generate a dialogue policy using sequences of intents corresponding to a plurality of conversations (i.e. Once all the conversations are annotated, AutoFlows Generator 209 [i.e. A system comprising: one or more processing units] generates the conversation flows [i.e. one or more dialogue policy] per auto-topic and auto-subtopic in a graph where the root node denotes the start of all conversation flows having the same auto-topic and auto-subtopic. A conversation flow [i.e. dialogue flows] is a sequence of sentence-level auto-intents [i.e. using sequences of intents corresponding to a plurality of conversations] and turn-level auto-responses) (¶ 0039 - 0040),
the sequences of intents determined based at least on a large language model (LLM) processing data corresponding to the plurality of conversations and associating intents from the sequences of intents with individual messages includes in the plurality of conversations (i.e. method/system, e.g. Annotation module, may generate, via execution of the Generative DNN model [i.e. a large language model (LLM)], annotations comprised of auto-intents, e.g. “cancel subscription”, “cancel reason”, etc. [i.e. the sequences of intents], corresponding to the user utterances, e.g. “I want to cancel my plan”, “I just don’t need anymore”, etc. [i.e. the plurality of conversations and associating intents from the sequences of intents with individual messages includes in the plurality of conversations]) (Fig. 4, Fig. 6, ¶ 0029 – 0030 and ¶ 0032).
However, Kelkar does not explicitly disclose:
wherein the dialogue policy is used to generate a conversational output through execution of at least one of one or more executable instructions comprised in the dialogue policy.
On the other hand, in the same field of endeavor, Kannan teaches:
wherein the dialogue policy is used to generate a conversational output through execution of at least one of one or more executable instructions comprised in the dialogue policy (i.e. based on “if” and “else” statements of the script [i.e. through execution of at least one of the one or more executable instructions comprised in the dialogue policy] that depend on BusinessName variable, the machine conversation dialog flow [i.e. the dialogue policy] may prompt “Where would you like to book your event? [i.e. generate a conversational output]) (501, 502 & 503 - Fig. 5, ¶ 0043 – 0044 and ¶ 0047).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the method/system of Kelkar to include the features for wherein the dialogue policy is used to generate a conversational output through execution of at least one of one or more executable instructions comprised in the dialogue policy as taught by Kannan so that the method/system may be implemented with a user interface that permits developers to edit/update the dialog flow in the script syntax (¶ 0047).
Regarding Claim 20, Kelkar and Kannan disclose, in particular Kelkar teaches:
wherein the one or more processors are comprised in at least one of:
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing collaborative content creation for 3D assets;
a system for performing one or more deep learning operations;
a system implemented using an edge device;
a system for generating or presenting at least one of virtual reality content, augmented reality content, or mixed reality content;
a system implemented using a robot; a system for performing one or more conversational AI operations; a system implemented using one or more large language models (LLMs); a system for generating synthetic data; a system for performing one or more generative AI operations;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources) (i.e. The Generative DNN model is a language model described in the NLP literature in many applications. Some applications where a language model is used to generate text include summarization of text, storytelling, programming code generation, etc. [i.e. a system for performing one or more generative AI operations]) (¶ 0029).
Claim(s) 3, 4, 5, 6, 8, 13, 14 and 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kelkar in views of Kannan as applied to claims 1 and 11 above, and further in view of Sastre et al. (US PG PUB 20230237994), hereinafter "Sastre".
Regarding Claim 3, Kelkar and Kannan disclose all the features with respect to claim 1 as described above.
In addition, Kelkar further teaches:
wherein the generating the one or more dialogue flows comprises: converting the plurality of canonical forms into a graph (i.e. the AutoFlows Generator generates the conversation flows [i.e. the one or more dialog flows] by converting auto-intents [i.e. the plurality of canonical forms] into nodes of a graph) (¶ 0040 and ¶ 0076).
However, the combination f Kelkar and Kannan does not explicitly disclose:
extracting one or more paths corresponding to the one or more dialogue flows from the graph.
On the other hand, in the same field of endeavor, Sastre teaches:
extracting one or more paths corresponding to the one or more dialogue flows from the graph (i.e. the method/system may identify/extract K paths [i.e. one or more paths] corresponding to a set of dialog flows [i.e. the one or more dialogue flow] from a graph) (Abstract, Fig. 2A, Fig. 5E, ¶ 0031 and ¶ 0065).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the method/system of Kelkar and Kannan to include the features for extracting one or more paths corresponding to the one or more dialogue flows from the graph as taught by Sastre so that the method/system may identify most likely conversation paths (Abstract, Fig. 2C, Fig. 5E, ¶ 0031 and ¶ 0065).
Regarding Claim 4, Kelkar, Kannan and Sastre disclose, in particular Sastre teaches:
wherein the graph comprises a plurality of nodes representing the plurality of canonical forms, a plurality of edges representing orderings of the canonical forms within at least one of the conversation or one or more other conversations processed using the machine learning model (i.e. the graph may include a plurality of nodes representing conversations/utterances [i.e. the plurality of canonical forms], a plurality of paths/transitions [i.e. a plurality of edges] representing ranks [i.e. orderings] of the next step of the conversation [i.e. the canonical forms within at least one of the conversation] processed using dialog modeling system [i.e. the machine learning model]) (Abstract, Fig. 2C, Fig. 5E, ¶ 0031 and ¶ 0065), and
a plurality of weights that are associated with the plurality of edges and represent frequencies of corresponding orderings within at least one of the conversation or the one or more other conversations (i.e. the graph also includes probability/frequency values [i.e. weights] associated with the plurality of paths/transitions [i.e. the plurality of edges], wherein the probability/frequency values represent frequencies, e.g. 18, 31, 103, etc., corresponding to respective ranks [i.e. orderings] associated with the conversations [i.e. the conversation or the one or more other conversations]) (Abstract, Fig. 2C, Fig. 5E, ¶ 0031 and ¶ 0065).
The prior art used in the rejection of the current claim is combined using the same motivations as was applied in claim 3.
Regarding Claim 5, Kelkar, Kannan and Sastre disclose, in particular Sastre teaches:
wherein the extracting the one or more paths comprises extracting a default path corresponding to a most frequent dialogue flow from the graph (i.e. the method/system may extract K best paths including the path [i.e. a default path] correspond to the dialogs that are most likely to occur or occur the most frequently [i.e. a most frequent dialogue flow from the graph]) (Fig. 5E and ¶ 0019).
The prior art used in the rejection of the current claim is combined using the same motivations as was applied in claim 3.
Regarding Claim 6, Kelkar, Kannan and Sastre disclose, in particular Sastre teaches:
wherein the extracting the one or more paths further comprises extracting a branching path corresponding to an alternative dialogue flow that deviates from the most frequent dialogue flow from the graph (i.e. the method/system may extract K best paths including a path [i.e. a branching path] correspond to the dialogs that are different from [i.e. an alternative dialogue flow that deviates] the ones that are most likely to occur or occur the most frequently [i.e. the most frequent dialogue flow from the graph]) (Fig. 5E and ¶ 0019).
The prior art used in the rejection of the current claim is combined using the same motivations as was applied in claim 3.
Regarding Claim 8, Kelkar and Kannan discloses all the features with respect to Claim 1 as described above.
However, the combination of Kelkar and Kannan does not explicitly disclose:
wherein the conversation includes a first set of messages from one or more users and a second set of messages from one or more chatbots.
On the other hand, in the same field of endeavor, Sastre teaches:
wherein the conversation includes a first set of messages from one or more users and a second set of messages from one or more chatbots (i.e. conversation flows [i.e. the conversation] may include a set of messages exchanged [i.e. a first set of messages from one or more users and a second set of messages from one or more chatbots] between user and agent / virtual assistant [i.e. one or more chatbots]) (¶ 0015 and ¶ 0019).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the method/system of Kelkar and Kannan to include the features for wherein the conversation includes a first set of messages from one or more users and a second set of messages from one or more chatbots as taught by Sastre so that the method/system may generate fine-tuned dialog flows from messages exchanged between users and virtual agents/(¶ 0015 and ¶ 0019).
Regarding Claim 13, Kelkar and Kannan discloses all the features with respect to Claim 11 as described above.
In addition, Kelkar teaches:
wherein the generating the one or more dialogue flows comprises: generating a plurality of clustered canonical forms from the plurality of constrained semantic representations (i.e. the method/system may generate auto-topic and auto-subtopic [i.e. a plurality of clustered canonical forms] from auto-intents [i.e. the plurality of constrained semantic representations]) (¶ 0033, ¶ 0039 and ¶ 0077); and
converting the plurality of clustered canonical forms into a graph (i.e. the AutoFlows Generator generates the conversation flows [i.e. the one or more dialog flows] per auto-topic and auto-subtopic in a graph where the root node denotes the start of all conversation flows having the same auto-topic and auto-subtopic) (¶ 0039).
However, the combination of Kelkar and Kannan does not explicitly disclose:
extracting one or more paths corresponding to the one or more dialogue flows from the graph.
On the other hand, in the same field of endeavor, Sastre teaches:
extracting one or more paths corresponding to the one or more dialogue flows from the graph (i.e. the method/system may identify/extract K paths [i.e. one or more paths] corresponding to a set of dialog flows [i.e. the one or more dialogue flow] from a graph) (Abstract, Fig. 2A, Fig. 5E, ¶ 0031 and ¶ 0065).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the method/system of Kelkar and Kannan to include the features for extracting one or more paths corresponding to the one or more dialogue flows from the graph as taught by Sastre so that the method/system may identify most likely conversation paths (Abstract, Fig. 2C, Fig. 5E, ¶ 0031 and ¶ 0065).
Regarding Claim 14, Kelkar, Kannan and Sastre disclose, in particular Sastre teaches:
wherein the graph comprises a plurality of nodes representing the plurality of clustered canonical forms, a plurality of edges representing orderings of the canonical forms within at least one of the conversation or one or more other conversations processed using the machine learning model (i.e. the graph may include a plurality of nodes representing conversations/utterances [i.e. the plurality of clustered canonical forms], a plurality of paths/transitions [i.e. a plurality of edges] representing ranks [i.e. orderings] of the next step of the conversation [i.e. the canonical forms within at least one of the conversation] processed using dialog modeling system [i.e. the machine learning model]) (Abstract, Fig. 2C, Fig. 5E, ¶ 0031 and ¶ 0065), and
a plurality of edges representing orderings of the plurality of clustered canonical forms within the plurality of conversations (i.e. the graph also includes probability/frequency values [i.e. weights] associated with the plurality of paths/transitions [i.e. the plurality of edges], wherein the probability/frequency values represent frequencies, e.g. 18, 31, 103, etc., corresponding to respective ranks [i.e. orderings] associated with the conversations [i.e. the plurality of clustered canonical forms within the plurality of conversations]) (Abstract, Fig. 2C, Fig. 5E, ¶ 0031 and ¶ 0065).
The prior art used in the rejection of the current claim is combined using the same motivations as was applied in claim 13.
Regarding Claim 15, Kelkar, Kannan and Sastre disclose, in particular Sastre teaches:
wherein the extracting the one or more paths comprises extracting a default path corresponding to a most frequent dialogue flow (i.e. the method/system may extract K best paths including the path [i.e. a default path] correspond to the dialogs that are most likely to occur or occur the most frequently [i.e. a most frequent dialogue flow from the graph]) (Fig. 5E and ¶ 0019); and
extracting a branching path corresponding to an alternative dialogue flow that deviates from the most frequent dialogue flow from the graph (i.e. the method/system may extract K best paths including a path [i.e. a branching path] correspond to the dialogs that are different from [i.e. an alternative dialogue flow that deviates] the ones that are most likely to occur or occur the most frequently [i.e. the most frequent dialogue flow from the graph]) (Fig. 5E and ¶ 0019).
The prior art used in the rejection of the current claim is combined using the same motivations as was applied in claim 13.
Claim(s) 7, 16 and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kelkar in views of Kannan as applied to claims 1 and 11 above, and further in view of Koneru et al. (US PG PUB 20240282298), hereinafter "Koneru".
Regarding Claim 7, Kelkar and Kannan disclose all the features with respect to Claim 1 as described above
However, the combination of Kelkar and Kannan does not explicitly disclose:
wherein the causing the conversational output to be generated comprises: generating a prompt that includes the one or more dialogue flows and at least a portion of a current conversation; and processing the prompt using a language model to generate the conversational output.
On the other hand, in the same field of endeavor, Koneru teaches:
wherein the causing the conversational output to be generated comprises: generating a prompt that includes the one or more dialogue flows and at least a portion of a current conversation (i.e. the method/system may generate one or more text prompts that include customer utterance [i.e. at least a portion of a current conversation], transcript of the conversation between the customer at the customer device 110 and the virtual assistant server 150, etc. [i.e. includes the one or more dialogue flows]) (Abstract, Fig. 3D and ¶ 0053); and
processing the prompt using a language model to generate the conversational output (i.e. the method/system may generate a plurality of outputs [i.e. the conversational output] from the selected LLM based on processing of a plurality of prompts provided to the selected LLM [i.e. a large language model]) (Abstract, Fig. 5 and ¶ 0095).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the method/system of Kelkar and Kannan to include the features for wherein the causing the conversational output to be generated comprises: generating a prompt that includes the one or more dialogue flows and at least a portion of a current conversation; and processing the prompt using a language model to generate the conversational output as taught by Koneru so that the method/system may employ a large language model for response generation (Abstract, Fig. 5 and ¶ 0095).
Regarding Claim 16, Kelkar and Kannan disclose all the features with respect to Claim 1 as described above
However, the combination of Kelkar and Kannan does not explicitly disclose:
wherein the causing the conversational output to be generated comprises: generating an embedding of a canonical form associated with the one or more dialogue flows; determining one or more canonical forms based at least on the embedding and one or more embeddings of one or more predefined canonical forms; generating a prompt that includes the one or more canonical forms, one or more example outputs associated with the one or more canonical forms, and at least a portion of a current conversation; and inputting the prompt into a language model to generate the conversational output.
On the other hand, in the same field of endeavor, Koneru teaches:
wherein the causing the conversational output to be generated comprises: generating an embedding of a canonical form associated with the one or more dialogue flows (i.e. method/system may generate one or more nodes comprising intent, entity, service, messages, etc. [i.e. an embedding of a canonical form associated with the one or more dialogue flows]) (¶ 0045);
determining one or more canonical forms based at least on the embedding and one or more embeddings of one or more predefined canonical forms (i.e. nodes of the dialog flow may include various types of interactions, such as, for example, questions, prompts, confirmations, and messages, and are configured to gather information from the customer, provide information to the customer, or perform a specific action [i.e. determining one or more canonical forms based at least on the embedding and one or more embeddings of one or more predefined canonical forms]) (¶ 0045);
generating a prompt that includes the one or more canonical forms, one or more example outputs associated with the one or more canonical forms, and at least a portion of a current conversation (i.e. the method/system may generate one or more text prompts that include customer utterance [i.e. at least a portion of a current conversation], a few-shot sample conversations [i.e. one or more example outputs associated with the one or more canonical forms], and transcript of the conversation between the customer at the customer device 110 and the virtual assistant server 150, etc. [i.e. includes the one or more dialogue flows]) (Abstract, Fig. 3D and ¶ 0053); and
inputting the prompt into a language model to generate the conversational output (i.e. the method/system may generate a plurality of outputs [i.e. the conversational output] from the selected LLM based on processing of a plurality of prompts provided to the selected LLM [i.e. a large language model]) (Abstract, Fig. 5 and ¶ 0095).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the method/system of Kelkar and Kannan to include the features for wherein the causing the conversational output to be generated comprises: generating an embedding of a canonical form associated with the one or more dialogue flows; determining one or more canonical forms based at least on the embedding and one or more embeddings of one or more predefined canonical forms; generating a prompt that includes the one or more canonical forms, one or more example outputs associated with the one or more canonical forms, and at least a portion of a current conversation; and inputting the prompt into a language model to generate the conversational output as taught by Koneru so that the method/system may employ a large language model for response generation (Abstract, Fig. 5 and ¶ 0095).
Regarding Claim 17, Kelkar, Kannan and Koneru disclose, in particular Kelkar teaches:
wherein the machine learning model comprises at least one of a generative model or a large language model (LLM) (i.e. The Generative DNN model is a language model described in the NLP literature in many applications. Some applications where a language model is used to generate text include summarization of text, story telling, programming code generation, etc. [i.e. a large language model (LLM)]) (¶ 0029).
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SOE MIN HLAING whose telephone number is (303)297-4282. The examiner can normally be reached Monday-Friday 9AM - 5PM.
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, Christopher Parry can be reached at 571-272-8328. 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.
/Soe Hlaing/ Primary Examiner, Art Unit 2451
Prosecution Insights