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 .
Status of Claims
This action is in reply to the claims filed on 27 November 2024.
Claim 16 has been cancelled.
Claims 1-15 and 17-21 are currently pending and have been examined.
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 therefore, subject to the conditions and requirements of this title.
Claim(s) 17 is/are rejected under 35 USC § 101.
Claim(s) 17 is/are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claim(s) does/do not fall within at least one of the four categories of patent eligible subject matter because claim 17 encompasses a transitory medium given the claim's broadest reasonable interpretation in light of paragraph [0129] of the specification. Such media have been held to be ineligible subject matter under 35 USC 101. See In re Nuijten, 500 F.3d 1346, 1356-57 (Fed. Cir. 2007).
The Examiner notes claim(s) 17 has/have been interpreted as not being directed to a statutory category. The following rejection below has been provided as if claim 17 has/have been interpreted as being directed to a statutory category.
Claims 1-15 and 17-21 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., an abstract idea) without significantly more.
Independent claims 1, 17, and 18 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claims regard a process that, as drafted under its broadest reasonable interpretation, covers performance of the limitations as a mental process, but for the recitation of generic computer hardware (e.g., a storage medium using one or more programs, the one or more programs comprising computer instructions, which, when executed by a processor, cause the processor to perform the method (claim 17) and a terminal, comprising a memory and a processor, wherein the memory stores one or more programs, the one or more programs comprising computer instructions, which, when executed by the processor, cause the processor to perform the method (claim 18)).
In regard to the processing of independent claims 1, 17, and 18, the claimed functionality could be practiced as a mental process in the following manner:
acquiring a current utterance entered by a user; (a human can mentally listen to another user)
reading a preset topic transfer graph and a preset target topic, wherein the topic transfer graph comprises a plurality of nodes and connecting lines between the nodes, the plurality of nodes correspond to topics in one-to-one correspondence, each of the connecting lines points from a first node to a second node, a weight of the connecting line indicates probability of transferring from a topic corresponding to the first node to a topic corresponding to the second node, and the topic transfer graph comprises a node corresponding to the target topic; (a human can read a graph on paper as described by the limitation)
determining a topic of reply content of the current utterance at least based on the current utterance, the topic transfer graph and the target topic, and recording the topic of the reply content of the current utterance as a reply topic; (a human can make a determination based on what they have heard and what they see on paper and store a determination in their mind or on paper) and
generating the reply content of the current utterance at least based on the reply topic. (a human can mentally generate a response to an utterance that was spoken to them)
This judicial exception is not integrated into a practical application. Outside of the identified abstract idea, the claimed invention only includes a storage medium using one or more programs, the one or more programs comprising computer instructions, which, when executed by a processor, cause the processor to perform the method (claim 17) and a terminal, comprising a memory and a processor, wherein the memory stores one or more programs, the one or more programs comprising computer instructions, which, when executed by the processor, cause the processor to perform the method (claim 18), which amounts to no more than mere instructions to implement an otherwise abstract idea using generic components. Note that the computing components here are being used for their ordinary purpose of executing a program to carry out a process (i.e., being used as a tool) instead of being improved as a tool.
Independent claims 1, 17, and 18 do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the above additional element(s) merely use a computer as a tool to perform an abstract idea, which does not render a claim as being significantly more than the judicial exception.
Therefore, claim 1, 17, and 18 are not eligible subject matter under 35 USC 101.
The remaining dependent claims fail to add patent eligible subject matter to their respective parent claims:
Claim 2 and claim 19 further recites acquiring data in a way that a human can by mentally listening or reading and generating a graph in a way that a human can using pen and paper. Claim 2 further recites a computer. This/these additional element(s) alone or in ordered combination does no more than merely use a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)), which does not integrate the claim(s) into a practical application, nor does it render a claim as being significantly more than the abstract idea.
Claim 3 and claim 20 further recite determining a probability and making a determination based on that probability in a way that can be mentally understood and performed by a human.
Claim 4 and claim 21 further regard making a determination in a way that can be mentally understood by a human.
Claim 5 further regards getting information and creating vectors out of information to calculate a probability in a way that can be mentally understood by a human. Claim 5 further regards a pre-trained topic planning model comprising a language representation network, an attention network, a first feature calculation network, and a first classifier. This/these additional element(s) alone or in ordered combination does no more than merely use a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)), which does not integrate the claim(s) into a practical aapplication, nor does it render a claim as being significantly more than the abstract idea. The above identified additional generic computer components are no more than mere instructions to apply the exception using generic computer components that are well-known, routine, and conventional as is evidenced by Paragraph [0075] of the instant specification that shows the language representation network being a popular BERT network, Paragraph [0055] of Chae (US 20220366890 A1), which teaches attention mechanisms being known, Paragraph [0036] of Takase (US 20170046428 A1), which teaches methods known in the art for calculating feature vectors, and Paragraph [0067] of Zheng (US 20150112918 A1), which teaches known classifiers for classifying topics.
Claim 6 further regards manipulating data and vectors to calculate a probability in a way that can be mentally understood by a human.
Claim 7 further regards a mathematical equation falling under the abstract idea of math.
Claim 8 further regards manipulating data and vectors to calculate a probability in a way that can be mentally understood by a human. Claim 8 further regards a second feature calculation network and a second classifier. This/these additional element(s) alone or in ordered combination does no more than merely use a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)), which does not integrate the claim(s) into a practical application, nor does it render a claim as being significantly more than the abstract idea. The above identified additional generic computer components are no more than mere instructions to apply the exception using generic computer components that are well-known, routine, and conventional as is evidenced by Paragraph [0036] of Takase (US 20170046428 A1), which teaches methods known in the art for calculating feature vectors, and Paragraph [0067] of Zheng (US 20150112918 A1), which teaches known classifiers for classifying topics.
Claim 9 further regards manipulating data and vectors to calculate a probability in a way that can be mentally understood by a human.
Claim 10 further regards a mathematical equation falling under the abstract idea of math.
Claim 11 further regards making a determination from a graph that can be mentally read by a human. The determination can be mentally understood by a human.
Claim 12 further regards calculating a similarity between topics and using knowledge graphs in a way that can be mentally understood by a human given all the proper information verbally or on pen and paper.
Claim 13 further regards manipulating data and vectors to calculate a probability in a way that can be mentally understood by a human. Claim 13 further regards a pre-trained reply generation model comprising an encoder, a knowledge selector, and a decoder. This/these additional element(s) alone or in ordered combination does no more than merely use a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)), which does not integrate the claim(s) into a practical application, nor does it render a claim as being significantly more than the abstract idea. The above identified additional generic computer components are no more than mere instructions to apply the exception using generic computer components that are well-known, routine, and conventional as is evidenced by Paragraph [0024] of Balatsos (US 20070274245 A1), which teaches encoder and decoders being well understood and Moon (see attached NPL teaches common knowledge augmented dialog systems in in section 5 on page 852.
Claim 14 further regards manipulating data and vectors to calculate a probability in a way that can be mentally understood by a human.
Claim 15 further regards manipulating data and vectors to calculate a probability in a way that can be mentally understood by a human.
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)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 1, 3, 4, 17, 18, 20, and 21 is/are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Liu (US 20230088445 A1).
Regarding claim 1, Liu teaches a conversation content generation method, comprising:
acquiring a current utterance [historical conversation information] entered by a user; (see at least Paragraph [0038] “In operation S210, a historical conversation information is acquired.”; Paragraph [0041] “the historical conversation information may be text information or voice information.”; Paragraph [0051] “the historical conversation information may be a conversation information in a duration from a time instant 0 to a time instant t in the human-computer interaction.”; Fig. 2)
reading a preset topic transfer graph [conversation target graph] and a preset target topic [target conversation guiding information], (see at least Paragraph [0049] “for operation S220, determining the target conversation object to be generated, from the conversation target graph based on the historical conversation information includes: determining, based on the historical conversation information and a target conversation guiding information, the target conversation object from the conversation target graph.”; Paragraph [0054] “a type of the target conversation object includes at least one of [...], a conversation topic”; Fig. 2)
wherein the topic transfer graph comprises a plurality of nodes and connecting lines between the nodes, the plurality of nodes correspond to topics in one-to-one correspondence, each of the connecting lines points from a first node to a second node, a weight of the connecting line indicates probability of transferring from a topic corresponding to the first node to a topic corresponding to the second node, and the topic transfer graph comprises a node corresponding to the target topic; (Paragraph [0043] “the conversation target graph may include a plurality of object nodes. The plurality of object nodes may have a connection edge with each other. The object node may be configured to represent a conversation object. The connection edge may be configured to represent an association relationship between two connected object nodes.”; Paragraph [0054] “a type of the target conversation object includes at least one of [...], a conversation topic”; Paragraph [0057] “An element in the cost parameter is used to represent a transition probability between two adjacent candidate object nodes.”; Paragraph [0041] “historical conversation information generated during the conversation in a process of human-computer interaction.”; Fig. 2; Fig. 3A)
determining a topic of reply content [target conversation object] of the current utterance at least based on the current utterance [historical conversion information], the topic transfer graph [conversion target graph] and the target topic [target conversation guiding information], (see at least Paragraph [0049] “for operation S220, determining the target conversation object to be generated, from the conversation target graph based on the historical conversation information includes: determining, based on the historical conversation information and a target conversation guiding information, the target conversation object from the conversation target graph.”; Fig. 2)
and recording the topic of the reply content of the current utterance as a reply topic; (see at least Paragraph [0101] “The target conversation object may be target conversation object at a current moment, e.g. a time instant t. The sequence of historical target conversation objects may include a sequence of historical target conversation objects generated during a historical time period, e.g., a time period from a time instant 0 to the time instant t, in the conversation.”)
generating the reply content of the current utterance at least based on the reply topic. (see at least Paragraph [0040] “In operation S230, a target conversation information for recommendation is generated based on the target conversation object.”; Fig. 2)
Regarding claim 3, Liu teaches the method according to claim 1. Liu further teaches:
wherein said determining the topic of the reply content of the current utterance at least based on the current utterance, the topic transfer graph and the target topic comprises:
calculating a transfer probability of a current topic, wherein the current topic is a last reply topic or is determined based on the current utterance; (see at least Paragraph [0071] “a probability 370 of switching the target conversation object node is determined based on the cost parameter of the candidate object node”; Paragraph [0074] “a historical target conversation object at a time instant t in the sequence of historical target conversation objects may be used as a target conversation object at a time instant t+1.”; Fig. 3A; Examiner notes the sequence of historical target conversation objects tracks how the conversation moves from one topic to another.)and
determining whether the transfer probability is greater than a first preset threshold, determining the reply topic from the topic transfer graph in response to the transfer probability being higher than the first preset threshold, (see at least Paragraph [0071] “In a case that the probability of switching is determined to be greater than or equal to a predetermined switching threshold, the target conversation object is determined from the candidate object node(s) based on the cost parameter(s) of the candidate object node(s).”; Fig. 3A)
and using the current topic as the reply topic in response to the transfer probability being lower than or equal to the first preset threshold. (see at least Paragraph [0074] “in the case that the probability of switching is determined to be smaller than the predetermined switching threshold, a historical target conversation object at a time instant t in the sequence of historical target conversation objects may be used as a target conversation object at a time instant t+1.”; Fig. 3A)
Regarding claim 4, Liu teaches the method according to claim 3. Liu further teaches:
wherein said determining whether the transfer probability is greater than the first preset threshold comprises:
in response to similarities between the current topic and topics corresponding to each node in the topic transfer graph being less than or equal to a second preset threshold, determining that the transfer probability is lower than the first preset threshold. (see at least Paragraph [0062] “a gated recurrent unit (GRU) of a recurrent neural network (RNN) may be used to process a representation 320 of the sequence of historical target conversation objects and the transition matrix 310 for the candidate object node(s), to determine a first initial cost parameter 330 of the candidate object node(s).”; Paragraph [0074] “in the case that the probability of switching is determined to be smaller than the predetermined switching threshold” Fig. 3A; Liu sees if topics are similar based on the costs and if the costs are too high meaning the topics are not similar the topic stays the same)
Claim 17:
Claim(s) 17 is/are directed to a storage medium. Claim(s) 17 recite limitations parallel in nature as those addressed above for claim(s) 1, which are directed towards a method. Claim(s) 17 is/are therefore rejected for the same reasons as set above for claim(s) 1. Claim 17 further teaches a storage medium storing one or more programs, the one or more programs comprising computer instructions, which, when executed by a processor, cause the processor to perform the method (see Paragraph [0167] of Liu).
Claim 18:
Claim(s) 18 is/are directed to a system. Claim(s) 18 recite limitations parallel in nature as those addressed above for claim(s) 1, which are directed towards a method. Claim(s) 18 is/are therefore rejected for the same reasons as set above for claim(s) 1. Claim 18 further teaches a terminal, comprising a memory and a processor, wherein the memory stores one or more programs, the one or more programs comprising computer instructions which, when executed by the processor, cause the processor to perform the method (see at least Paragraph [0034] and [0166] of Liu).
Claim 20:
Claim(s) 20 is/are directed to a system. Claim(s) 20 recite limitations parallel in nature as those addressed above for claim(s) 3, which are directed towards a method. Claim(s) 20 is/are therefore rejected for the same reasons as set above for claim(s) 3. Claim 20 further teaches a terminal, comprising a memory and a processor, wherein the memory stores one or more programs, the one or more programs comprising computer instructions which, when executed by the processor, cause the processor to perform the method (see at least Paragraph [0034] and [0166] of Liu).
Claim 21:
Claim(s) 21 is/are directed to a system. Claim(s) 21 recite limitations parallel in nature as those addressed above for claim(s) 4, which are directed towards a method. Claim(s) 21 is/are therefore rejected for the same reasons as set above for claim(s) 4. Claim 21 further teaches a terminal, comprising a memory and a processor, wherein the memory stores one or more programs, the one or more programs comprising computer instructions which, when executed by the processor, cause the processor to perform the method (see at least Paragraph [0034] and [0166] of Liu).
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim(s) 2 and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liu (US 20230088445 A1) in further view of Kidd (US 20040268366 A1).
Regarding claim 2, Liu teaches the method according to claim 1. Liu further teaches:
wherein a method for constructing the topic transfer graph comprises:
acquiring a plurality of dialogue corpora, wherein each of the plurality of dialogue corpora comprises a plurality of rounds of human-computer dialogue samples, (see at least Paragraph [0051] “the historical conversation information may be a conversation information in a duration from a time instant 0 to a time instant t in the human-computer interaction.”) each of the plurality of rounds of human-computer dialogue samples has pre-labeled first label and second label, the first label indicates a topic of the human-computer dialogue sample of the round, and the second label indicates whether the topic of the human-computer dialogue sample of the round is the same as a topic of a next round of human-computer dialogue sample; (see at least Paragraph [0121] “the training sample may include a sample conversation information and a label corresponding to the sample conversation information, and the label includes a sample conversation object. A type of the label includes at least one of the conversation type of recommendation, the conversation topic and the topic attribute.”; Fig. 5) and
generating the topic transfer graph based on the plurality of dialogue corpora. (see at least Paragraph [0119] “a conversational recommendation model including the initial conversation target graph is trained by using the training sample, to obtain the conversation target graph in the trained conversational recommendation model.”; Paragraph [0121] “the training sample may include a sample conversation information”; Fig. 5)
Liu does not teach:
the second label indicates whether the topic of the human-computer dialogue sample of the round is the same as a topic of a next round of human-computer dialogue sample.
However, Kidd teaches:
a label (i.e., a flag) that represents a change in state (i.e., a topic) (see at least Paragraph [0028] “the flags such as state change flags and moderation flags have been described as a flag having a set value of ‘1’ and a reset value of ‘0’”)
This technique of Kidd is applicable to the method of Liu as they both share characteristics and capabilities, namely, they are directed to managing the changes of states in a computer environment. The sole difference between Liu and the claimed invention is the specifics of the second label. The secondary reference shows that state change flags were well known in the art at the time of the claimed invention. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have substituted the second label of Liu to incorporate the state change flags as taught by Kidd. Since each individual element and its function are shown in the prior art, albeit shown in separate references, the difference between the claimed subject matter and the prior art rests not on any individual element or function but in the very combination itself- that is in the substitution of the state change flag of the secondary reference(s) for the second label of the primary reference. Thus, the simple substitution of one known element for another producing a predictable result renders the claim obvious. Furthermore, one of ordinary skill in the art before the effective filling date of the claimed invention would have been motivated to modify Liu in order to provide a significant improvement may be made in memory usage (see paragraph [0012] of Kidd).
Claim 19:
Claim(s) 19 is/are directed to a system. Claim(s) 19 recite limitations parallel in nature as those addressed above for claim(s) 2, which are directed towards a method. Claim(s) 19 is/are therefore rejected for the same reasons as set above for claim(s) 2. Claim 19 further teaches a terminal, comprising a memory and a processor, wherein the memory stores one or more programs, the one or more programs comprising computer instructions which, when executed by the processor, cause the processor to perform the method (see at least Paragraph [0034] and [0166] of Liu).
Claim(s) 5-6, 8-9, 11, and 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liu (US 20230088445 A1) in further view of Moon (see attached NPL).
Regarding claim 5, Liu teaches the method according to claim 3. Liu further teaches:
wherein the transfer probability of the current topic is calculated by a pre-trained topic planning model, the topic planning model comprises a language representation network, an attention network, a first feature calculation network and a first classifier, (see at least Paragraph [0034] which recites an embedding layer representing a language representation network; Paragraph [0061] and [0095] which describes using an attention mechanism; Paragraph [0062] which discusses the GRU which represents the first feature calculation network; Paragraph [0072] explains the SoftMax function which is a classifier; Fig. 3A shows a method for calculating a probability of switching topics; Fig. 5 recites the training process.)
and said calculating the transfer probability of the current topic comprises:
extracting semantic information of the current utterance and a target utterance using the language representation network to acquire a semantic feature vector; (see at least Paragraph [0034] “the representation of the sequence of historical target conversation objects may be obtained by processing the sequence of historical target conversation objects by using an input layer (e.g., an embedding layer).”)
calculating an attention vector [the aggregated data from the attention model] using the attention network (see at least Paragraph [0095] “the object node representation of the conversation target sub-graph at a next level are aggregated by using attention mechanism”)
wherein the topic transfer matrix [transition matrix] is acquired by vectorizing the topic transfer graph using a graph embedding algorithm [GCN]; (see at least Paragraph [0061] “As shown in FIG. 3A, a graph convolutional network (GCN) may be used to process the candidate object node(s) in the conversation target graph to obtain a transition matrix 310 for the candidate object node(s).”; Fig. 3A shows how the conversation target graph is processed by the GCN algorithm to create a transition matrix 310.)
calculating a topic evaluation vector [cost parameters] using the first feature calculation network [GRU] based on the attention vector and the semantic feature vector; (see at least Fig. 3A shows how the GRU [first feature calculation network] uses the output of the attention network and GCN [attention vector] and the representation of the sequence of historical conversation objects 320 to create a topic evaluation vector [cost parameters].) and
calculating the transfer probability using the first classifier based on the topic evaluation vector. (see at least Paragraph [0081] “the cost parameter of the candidate object node may be processed by using a third fully connected layer, a fourth fully connected layer and a second activation function (e.g., the 2FC+softmax activation function) which are cascaded, in order to determine the probability of generating the target conversation object node.”; Paragraph [0068] “in FIG. 3A, a multi-layer perceptron (MLP) may be used to process the first initial cost parameter 330 and the second initial cost parameter 350 to obtain a multi-classification result.”)
Liu does not teach:
calculating an attention vector using the attention network based on the semantic feature vector and a topic transfer matrix, wherein the topic transfer matrix is acquired by vectorizing the topic transfer graph using a graph embedding algorithm.
However, Moon teaches:
calculating an attention vector using the attention network based on the semantic feature vector and a topic transfer matrix. (see Page 848 of Moon which recites:
PNG
media_image1.png
76
263
media_image1.png
Greyscale
This equation shows an attention vector αt-. This attention vector is calculated using ht-1 which includes a semantic representation of the dialogue history. Then αt is used with rk∈RKG, which is the topic transfer matrix to make another attention vector as claimed in the claimed invention.)
This step of Moon is applicable to the method of Liu as they both share characteristics and capabilities, namely, they are directed to naturally transversing between topics using a knowledge graph. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have modified the method of Liu to incorporate the attention vector as taught by Moon. One of ordinary skill in the art before the effective filling date of the claimed invention would have been motivated to modify Liu in order to create natural knowledge paths among entities in dialog contexts (see Introduction of Moon).
Regarding claim 6, Liu in view of Moon teaches the method according to claim 5. Liu does not teach:
wherein said calculating the transfer probability based on the topic evaluation vector comprises:
fusing the topic evaluation vector and the attention vector to acquire a first fusion vector; and
calculating the transfer probability based on the first fusion vector.
However, Moon teaches:
wherein said calculating the transfer probability based on the topic evaluation vector comprises:
fusing the topic evaluation vector and the attention vector to acquire a first fusion vector; and
calculating the transfer probability based on the first fusion vector. (Page 848, Equation 7 of Moon teaches combining ht-1, which includes data that would be considered a topic evaluation vector, with the attention vector created with RKG and αt to create zt.)
The motivation for making this modification to the teachings of Liu is the same as that set forth above, in the rejection of claim 5.
Regarding claim 8, Liu in view of Moon teaches the method according to claim 5. Liu further teaches:
wherein the topic planning model further comprises a second feature calculation network and a second classifier, (Fig. 3B of Liu shows multiple GRU calculation for a first and second cost parameter and multiple SoftMax classifiers) and said selecting the reply topic from the topic transfer graph comprises:
calculating a topic guidance vector using the second feature calculation network based on the topic evaluation vector and the attention vector; (see at least Fig. 3B shows how the GRU [second feature calculation network] uses the output of the attention network and GCN [attention vector] and the a topic evaluation vector [cost parameters] to create a vector of the cost parameters to input into the SoftMax under the BRI of topic guidance vector.) and
determining the reply topic using the second classifier based on the topic guidance vector. (Fig. 3B shows how the SoftMax function is used as a classifier to determine target conversation objects)
Regarding claim 9, Liu in view of Moon teaches the method according to claim 8. Liu does not teach:
wherein said determining the reply topic based on the topic guidance vector comprises:
fusing the topic guidance vector and the attention vector to acquire a second fusion vector; and determining the reply topic based on the second fusion vector.
However, Moon teaches:
wherein said determining the reply topic based on the topic guidance vector comprises:
fusing the topic guidance vector and the attention vector to acquire a second fusion vector; and determining the reply topic based on the second fusion vector. (Page 848, Equation 7 of Moon teaches combining ht-1, which includes data that would be considered a topic evaluation vector, with the attention vector created with RKG and αt to create zt. This zt is used to score and rank candidate knowledge graph entities.)
The motivation for making this modification to the teachings of Liu is the same as that set forth above, in the rejection of claim 5.
Regarding claim 11, Liu teaches the method according to claim 1. Liu further teaches:
wherein said generating the reply content of the current utterance at least based on the reply topic (see at least Paragraph [0040] “In operation S230, a target conversation information for recommendation is generated based on the target conversation object.”; Fig. 2) comprises:
reading a preset knowledge graph, (see at least Paragraph [0113] “An initial conversation target graph 650 is generated based on the sequence 630 of conversation objects and a knowledge graph 640.”)
determining target knowledge from the knowledge graph based on the reply topic; (see at least Paragraph [0039] “In operation S220, a target conversation object to be generated is determined from a conversation target graph based on the historical conversation information”; Fig. 2) and
generating the reply content based on the target knowledge and the reply topic. (see at least Paragraph [0040] “In operation S230, a target conversation information for recommendation is generated based on the target conversation object.”; Fig. 2)
Liu does not teach:
reading a preset knowledge graph, wherein the knowledge graph comprises common sense knowledge and/or specific knowledge, the specific knowledge refers to knowledge in a specific field, and the specific field is determined by the target topic.
However, Moon teaches:
reading a preset knowledge graph, wherein the knowledge graph comprises common sense knowledge and/or specific knowledge, the specific knowledge refers to knowledge in a specific field, and the specific field is determined by the target topic. (See at least the large-scale common-fact KG [knowledge graph] on Page 845-846 of Moon)
This step of Moon is applicable to the method of Liu as they both share characteristics and capabilities, namely, they are directed to naturally transversing between topics using a knowledge graph. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have modified the knowledge graph of Liu to incorporate common sense knowledge as taught by Moon. One of ordinary skill in the art before the effective filling date of the claimed invention would have been motivated to modify Liu in order to create natural knowledge paths among entities in dialog contexts (see Introduction of Moon).
Regarding claim 13, Liu in view of Moon teaches the method according to claim 11. Liu further teaches:
wherein the reply content is calculated by a pre-trained reply generation model, the reply generation model comprises an encoder, a knowledge selector and a decoder, (see at least Paragraph [0104] “a bidirectional GRU and a hierarchical gated fusion unit (HGFU) may be respectively used as an encoder and a decoder, […] thereby generating the target conversation information for reply.”) and
said generating the reply content of the current utterance at least based on the reply topic (see at least Paragraph [0040] “In operation S230, a target conversation information for recommendation is generated based on the target conversation object.”; Fig. 2) comprises:
generating the reply content using the decoder at least based on the fusion encoding vector; (see at least Paragraph [0104] “a hierarchical gated fusion unit (HGFU) may be respectively used as an encoder and a decoder, […] thereby generating the target conversation information for reply”)
and/or a dialogue encoding vector, (see at least Paragraph [0104] “an encoder and a decoder, to process the target conversation object, the historical conversation information”) the topic encoding vector is acquired by encoding the reply topic using the encoder, (see at least Paragraph [0104] “an encoder and a decoder, to process the target conversation object,”) and the dialogue encoding vector is acquired by encoding a dialogue history with the user using the encoder. (see at least Paragraph [0104] “an encoder […] to process […] the historical conversation information”)
Liu does not teach:
calculating a target knowledge encoding vector using the knowledge selector based on an initial knowledge encoding vector and a content encoding vector;
fusing the target knowledge encoding vector and the content encoding vector to acquire a fusion encoding vector; and
wherein the initial knowledge encoding vector is acquired by encoding the knowledge graph using the encoder.
However, Moon teaches:
calculating a target knowledge encoding vector using the knowledge selector based on an initial knowledge encoding vector and a content encoding vector; (Examiner notes equation 7 on Page 848 of Moon shows that a target knowledge encoding vector (z_t) generated using a n attention mechanism based on knowledge graph relation embeddings (r_k) and aggregated input (see x In equation 3 on page 847 of Moon)).
fusing the target knowledge encoding vector and the content encoding vector to acquire a fusion encoding vector; (see at least equation 6 on page 848 of Moon
PNG
media_image2.png
20
132
media_image2.png
Greyscale
which shows how the aggregated input is fused with the target knowledge encoding vector z_t) and
wherein the initial knowledge encoding vector is acquired by encoding the knowledge graph using the encoder. (see at least section 2.2 on page 847 of Moon. The Entity representation section recites “We construct KG [Knowledge graph] embeddings to encode each entity mention”)
The motivation for making this modification to the teachings of Liu is the same as that set forth above, in the rejection of claim 11.
Claim(s) 7 and 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liu (US 20230088445 A1) in further view of Moon (see attached NPL) in further view of Arevalo (see attached NPL)
Regarding claim 7, Liu in view of Moon teaches the method according to claim 5. Liu does not teach:
wherein prior to calculating the transfer probability based on the topic evaluation vector, the method further comprises:
calculating a product of the topic evaluation vector and a first enhancement factor, and updating the topic evaluation vector based on the product;
wherein the first enhancement factor is calculated using the following formula:
fe = tanh (we * ae + ve * ap + be)
where fe is the first enhancement factor, we is a preset first weight matrix, ae is the topic evaluation vector, ve is a preset second weight matrix, ap is the attention vector, and be is a preset first bias vector for characterizing disturbance.
However, Moon teaches:
wherein prior to calculating the transfer probability based on the topic evaluation vector, (Examiner notes that Page 848 of Moon show the steps to pruning a tree prior to doing the final ranking calculation) the method further comprises:
calculating a product of(Equation 6 on 848 of Moon explains an equation that gets the tensor product of it and the first enhancement factor (
PNG
media_image3.png
20
156
media_image3.png
Greyscale
). This factor affects the eventual topic evaluation vector ht.)
wherein the first enhancement factor is calculated using the following formula:
fe = tanh (we * ae + ve * ap + be)
where fe is the first enhancement factor, we is a preset first weight matrix, ae is the topic evaluation vector, ve is a preset second weight matrix, ap is the attention vector, and be is a preset first bias vector for characterizing disturbance. (Page 848 of Moon teaches
PNG
media_image3.png
20
156
media_image3.png
Greyscale
. Both equations use the tanh function. Both equations call upon weights (we and ve in the claimed inventio and Wz and Whc in Moon). Both equations have an attention vector (zt of Moon). Both equations have a topic evaluation vectors (ht-1 of Moon). Page 848 of Moon also states that bias terms for gate were omitted for simplicity of notation (be). Therefore, the equation is taught by Moon.)
The motivation for making this modification to the teachings of Liu is the same as that set forth above, in the rejection of claim 5.
Liu in view of Moon does not teach:
calculating a product of the topic evaluation vector and a first enhancement factor, and updating the topic evaluation vector based on the product.
However, Arevalo teaches:
calculating a product of the topic evaluation vector and a first enhancement factor, and updating the topic evaluation vector based on the product. (see at least Page 5 of Arevalo, which recites:
PNG
media_image4.png
35
149
media_image4.png
Greyscale
Where z represents the first enhancement factor (gating factor) and hv represents the topic evaluation vector (representation modified by the gating factor).)
This step of Arevalo is applicable to the method of Liu in view of Moon as they both share characteristics and capabilities, namely, they are directed to classifying topics (genres) of media such as text, speech and video. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have modified the method of Liu in view of Moon to incorporate calculating a product of the topic evaluation vector and first enhancement feature as taught by Arevalo. One of ordinary skill in the art before the effective filling date of the claimed invention would have been motivated to modify Liu in view of Moon in order to determine how each modality contributes to the output of hidden units (see Introduction on page 2 of Arevalo).
Regarding claim 10, Liu in view of Moon teaches the method according to claim 8. Liu does not teach:
wherein prior to determining the reply topic based on the topic guidance vector, the method further comprises:
calculating a product of the topic guidance vector and a second enhancement factor, and updating the topic guidance vector based on the product;
wherein the second enhancement factor is calculated using the following formula:
fc = sigmoid (wc*he+vc*ap+bc)
where fc is the second enhancement factor, wc is a preset third weight matrix, he is the topic guidance vector, vc is a preset fourth weight matrix, ap is the attention vector, and bc is a preset second bias vector for characterizing disturbance.
However, Moon teaches:
wherein prior to determining the reply topic based on the topic guidance vector, (Examiner notes that Page 848 of Moon show the steps to pruning a tree prior to doing the final ranking calculation) the method further comprises:
calculating a product (Equation 6 on 848 of Moon explains an equation that gets the tensor product of it and the first enhancement factor (
PNG
media_image3.png
20
156
media_image3.png
Greyscale
). This factor affects the eventual topic evaluation vector ht.)
wherein the second enhancement factor is calculated using the following formula:
fc = function (wc*he+vc*ap+bc)
where fc is the second enhancement factor, wc is a preset third weight matrix, he is the topic guidance vector, vc is a preset fourth weight matrix, ap is the attention vector, and bc is a preset second bias vector for characterizing disturbance. (Page 848 of Moon teaches
PNG
media_image3.png
20
156
media_image3.png
Greyscale
. Both equations use a function (tanh for Moon and sigmoid for the claimed invention). Both equations call upon weights (wc and vc in the claimed inventio and Wz and Whc in Moon). Both equations have an attention vector (zt of Moon). Both equations have a topic guidance vectors (ht-1 of Moon). Page 848 of Moon also states that bias terms for gate were omitted for simplicity of notation (be). Therefore, the equation is taught by Moon.)
Liu in view of Moon does not teach:
calculating a product of the topic guidance vector and a second enhancement factor, and updating the topic guidance vector based on the product;
and the function being a sigmoid function.
However, Arevalo teaches:
calculating a product of the topic evaluation vector and a first enhancement factor, and updating the topic evaluation vector based on the product. (see at least Page 5 of Arevalo, which recites:
PNG
media_image4.png
35
149
media_image4.png
Greyscale
Where z represents the first enhancement factor (gating factor) and hv represents the topic evaluation vector (representation modified by the gating factor).)
and the function being a sigmoid function. (see Page 5 of Arevalo which shows the sigmoid σ symbol being used on a function similar to the claimed invention (i.e., with weights and biases being applied to x values).
This step of Arevalo is applicable to the method of Liu in view of Moon as they both share characteristics and capabilities, namely, they are directed to classifying topics (genres) of media such as text, speech and video. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have modified the method of Liu in view of Moon to incorporate calculating a product of the topic evaluation vector and first enhancement feature as taught by Arevalo. One of ordinary skill in the art before the effective filling date of the claimed invention would have been motivated to modify Liu in view of Moon in order to determine how each modality contributes to the output of hidden units (see Introduction on page 2 of Arevalo).
Claim(s) 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liu (US 20230088445 A1) in further view of Moon (see attached NPL) in further view of Tang (see attached NPL).
Regarding claim 12, Liu in view of Moon teaches the method according to claim 11. Liu does not teach:
wherein the knowledge graph comprises the common sense knowledge and the specific knowledge, the target knowledge comprises target common sense knowledge and target specific knowledge, and said determining the target knowledge from the knowledge graph based on the reply topic comprises:
calculating a similarity between the reply topic and the target topic; and
selecting the common sense knowledge and the specific knowledge based on the similarity to acquire the target knowledge, wherein the higher the similarity, the greater the proportion of the target specific knowledge in the target knowledge, and the smaller the proportion of the target common sense knowledge in the target knowledge.
However, Moon teaches:
wherein the knowledge graph comprises the common sense knowledge and the specific knowledge, the target knowledge comprises target common sense knowledge and target specific knowledge, (see at least Section 1 on pages 845-846 of Moon, which shows a common-sense knowledge graphs being used in combination with domain specific knowledge “we completely ground dialogs in a large-scale common fact KG, allowing for domain-agnostic conversational reasoning in open-ended conversations across various domains and tasks (e.g. chit-chat, recommendations, etc.)”. Furthermore Section 3 on Page 849 recites “To ensure sufficient separation of the dialog content, we used entities related to movies (titles, actors, directors) and books (titles, authors) for the recommendation task, and entities related to sports (athletes, teams) and music (singers) for the chit-chat task (Table 1).”)
The motivation for making this modification to the teachings of Liu is the same as that set forth above, in the rejection of claim 11.
Liu in view of Moon does not teach:
said determining the target knowledge from the knowledge graph based on the reply topic comprises:
calculating a similarity between the reply topic and the target topic; and
selecting the common sense knowledge and the specific knowledge based on the similarity to acquire the target knowledge, wherein the higher the similarity, the greater the proportion of the target specific knowledge in the target knowledge, and the smaller the proportion of the target common sense knowledge in the target knowledge.
However, Tang teaches:
calculating a similarity between the reply topic and the target topic; (see at least Section 4.2 on Page 5628 of Tang “Given the key word Basketball of the current turn and its closeness score (0.47) to the target Dance, […] we use cosine similarity between normalized word embeddings as the measure of keyword closeness.”) and
selecting the common sense knowledge and the specific knowledge based on the similarity to acquire the target knowledge, wherein the higher the similarity, the greater the proportion of the target specific knowledge in the target knowledge, and the smaller the proportion of the target common sense knowledge in the target knowledge. (section 4.2. on page 5628 of Tang teaches “We constrain that the keyword of each turn must move strictly closer to the end target compared to those of preceding turns. Figure 2, right part, illustrates the rule at a particular step. Given the key word Basketball of the current turn and its closeness score (0.47) to the target Dance, the only valid candidate keywords for the next turn are those with higher target closeness, such as Party with a closeness score of 0.62. On the other hand, transitioning from Basketball to Sport is not allowed in the context as it does not move towards the target. More concretely, we use cosine similarity between normalized word embeddings as the
measure of keyword closeness.”)
This step of Tang is applicable to the method of Liu in view of Moon as they both share characteristics and capabilities, namely, they are directed to tracking a conversation to move towards a particular topic. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have modified the method of Liu in view of Moon to incorporate calculating a similarity between the reply topic and the target topic and selecting the common sense knowledge and the specific knowledge based on the similarity to acquire the target knowledge as taught by Tang. One of ordinary skill in the art before the effective filling date of the claimed invention would have been motivated to modify Liu in view of Moon in order to generate meaningful and effective conversations with a decent success rate of reaching the target topics (see Introduction on Page 5625 of Tang).
Claim(s) 14 and 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liu (US 20230088445 A1) in further view of Moon (see attached NPL) in further view of Yao (see attached NPL).
Regarding claim 14, Liu in view of Moon teaches the method according to claim 13. Liu further teaches:
wherein said generating the reply content at least based on the fusion encoding vector comprises:
decoding the fusion encoding vector to acquire a first latent vector of an i-th word in the reply content, wherein i is a positive integer; (see at least Paragraph [0104] “a hierarchical gated fusion unit (HGFU) may be respectively used as […] a decoder”; Examiner notes a HGFU decoder processes input at time steps to produce an internal hidden state.)
decoding the dialogue encoding vector to acquire a second latent vector of the i-th word in the reply content; (see at least Paragraph [0104] “a hierarchical gated fusion unit (HGFU) may be respectively used as […] a decoder, to process […] the historical conversation information”; Examiner notes historical conversation information is dialogue that is decoded by the HGFU in a way to produce a second internal hidden state (vector))
fusing the first latent vector and the second latent vector to acquire a fusion latent vector of the i-th word in the reply content; (see at least Paragraph [0104] “a hierarchical gated fusion unit (HGFU) may be respectively used as […] a decoder”; Examiner notes the main function of a HGFU is to fuse the vectors.) and
generating the i-th word in the reply content based on the fusion latent vector. (see at least Paragraph [0104] “a hierarchical gated fusion unit (HGFU) may be respectively used as […] a decoder thereby generating the target conversation information for reply.”)
While Liu uses HGFU, Liu in view of Moon does not explicitly teach how the HGFU works.
However, Yao teaches:
decoding the fusion encoding vector to acquire a first latent vector of an i-th word in the reply content, wherein i is a positive integer; (see at least Page 2193-2194 “the new hidden state of the auxiliary decoding h_w is computed by the following equation” and equation 9 on page 2194 of Yao)
decoding the dialogue encoding vector to acquire a second latent vector of the i-th word in the reply content; (see at least Page 2193 of Yao which recites “C_t be the current attention-based context. The current hidden state of the general decoding, h_y, is defined” and the equation 8 on page 2193 of Yao which shows how the dialogue represented by C_t is used to calculate h_y.)
fusing the first latent vector and the second latent vector to acquire a fusion latent vector of the i-th word in the reply content; (see at least section 3.3.3. on Page 2194 of Yao which shows how the fusion unit combines h_y and h_w). and
generating the i-th word in the reply content based on the fusion latent vector. (see at least section 3.3. of Yao on Page 2193 which recites “the fusion unit combines the hidden states of both GRUs to predict the next word y_t.”)
This step of Yao is applicable to the method of Liu in view of Moon as they both share characteristics and capabilities, namely, they are directed to human computer conversation systems. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to have modified the HGFU of Liu in view of Yu to incorporate the functionality as taught by Yao. One of ordinary skill in the art before the effective filling date of the claimed invention would have been motivated to modify Liu in view of Moon in order to incorporate cue words using proposed hierarchical gated fusion unit in a flexible way (see Introduction on page 2191 of Yao).
Regarding claim 15, Liu in view of Moon in further view of Yao teaches the method according to claim 14. Liu further teaches:
wherein prior to fusing the first latent vector and the second latent vector, the method further comprises:
inputting a fusion latent vector of an (i-1)th word in the reply content into the decoder, to make the decoder decode the fusion encoding vector based on the (i-1)th word. (see at least Paragraph [0104] “a hierarchical gated fusion unit (HGFU) may be respectively used as […] a decoder”; Examiner notes fundamental feature of HGFU is to put the calculated hidden state from the previous time step into the network to influence the generation of the current time step.)
Liu does not explicitly state how the HGFU decoder functions.
However, Moon discloses on Page 848, equation 6 how the fusion latent vector of the i-1th word can be used for decoding of the current work (
PNG
media_image5.png
42
238
media_image5.png
Greyscale
). See how h_t-1 is being used in the calculation of h_t.
The motivation for making this modification to the teachings of Liu is the same as that set forth above, in the rejection of claim 11.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DANIELLE ELIZABETH ZEVITZ whose telephone number is (703)756-1070. The examiner can normally be reached Mo-Th 10am-6pm.
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, Andrew Flanders can be reached at (571) 272-7516. 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.
/DANIELLE ELIZABETH ZEVITZ/Examiner, Art Unit 2655
/ANDREW C FLANDERS/Supervisory Patent Examiner, Art Unit 2655
Prosecution Insights