Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
101 Subject Matter Eligibility Analysis
Step 1: Claims 1-20 are within the four statutory categories (a process, machine, manufacture or composition of matter).
Step 2A Prong One, Step 2A Prong Two, and Step 2B Analysis:
Step 2A Prong One asks if the claim recites a judicial exception (abstract idea, law of nature, or natural phenomenon). If the claim recites a judicial exception, analysis proceeds to Step 2A Prong Two, which asks if the claim recites additional elements that integrate the abstract idea into a practical application. If the claim does not integrate the judicial exception, analysis proceeds to Step 2B, which asks if the claim amounts to significantly more than the judicial exception. If the claim does not amount to significantly more than the judicial exception, the claim is not eligible subject matter under 35 U.S.C. 101.
None of the claims represent an improvement to technology.
1-10 are directed to a method consisting of a series of steps, meaning that it is directed to the statutory category of process. Claims 11-20 are directed a memory device and a processing device which are machines.
Regarding claim 1, the following claim elements are abstract ideas:
creating a plurality of processed events by filtering content of the plurality of event signals using a unified schema comprising a plurality of categories (This is an abstract idea of a mental process. It involves observing actions performed by a user during an online session, determining which predefined categories correspond to the observed actions, and organizing the actions according to these categories. A person could observe a user’s online activities, such as viewed content, selected items, or other interactions, apply judgement to determine which category corresponds to each activity, and create a record of the categorized actions. This type of observation, evaluation, and categorization can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.);
creating a unified action stream by aggregating the plurality of processed events (This is an abstract idea of a mental process. It involves collecting and organizing a user’s prior actions to determine the user’s interests and preferences. A person could observe a user’s prior activities, organize the activities based on observation and judgement, and use the organized activities to determine which content, item, or activity would likely be of interest to the user. This type of observation, evaluation, and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.)
generating a plurality of features using the unified action stream (This is an abstract idea of a mental process. It involves evaluating a user’s prior activities and deriving characteristics associated with a user’s interests, preferences, or behavior from those activities. A person could observe a user’s prior actions, analyze the actions based on observation and judgement, and identify characteristics that describe the user’s interests or preferences. This type of observation, evaluation, and characterization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.);
generating an action sequence using the unified action stream (This is an abstract idea of a mental process. It involves reviewing a user’s prior activities and organizing the activities into an ordered sequence. A person could observe a user’s prior actions, determine the order in which the actions occurred, and arrange actions into a sequence based on observation and judgement. This type of observation and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.);
generating input data for a trained machine learning model, the input data comprising the plurality of features and the action sequence (This is an abstract idea of a mental process. It involves determining what information associated with a user’s activities is relevant to a decision or recommendation and organizing the information for consideration. A person could observe a user’s prior actions, identify characteristics and sequence that are relevant to determining a recommendation based on observation, judgement, and decision-making. This type of observation, evaluation, selection, and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.); and
The following claim elements are additional elements which, taken alone or in combination with the other elements, do not integrate the judicial exception into a practical application nor amount to significantly more than the judicial exception:
receiving a plurality of event signals from a plurality of verticals for an ongoing session of a user of an online system (The step of “receiving” information merely gathers data for use in the abstract idea and amounts to well-understood, routine, and conventional activity.);
generating an output of the trained machine learning model by applying the trained machine learning model to the input data (This step of “generating an output” merely obtains a result based on the abstract idea and therefore amounts to insignificant extra-solution activity.).
Regarding claim 2, the rejection of claim 1 is incorporated herein. Further, claim 2 recites the following abstract ideas:
generating the plurality of features using the retrieved data and the processed event (This is a abstract idea of a mental process. It involves reviewing information associated with a user’s activities and deriving characteristics from the information and activities. A person could observe a user’s actions, review information associated with those actions, and identify characteristics that describe the user’s interests, preferences, or behavior based on observation and judgement. This type of observation, evaluation, and characterization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.).
The following claim elements are additional elements which, taken alone or in combination with the other elements, do not integrate the judicial exception into a practical application nor amount to significantly more than the judicial exception:
retrieving data associated with a processed event of the unified action stream using a category of the plurality of categories (This step of “retrieving data” is merely a generic data operation that amounts to storing and retrieving information in memory, which has been recognized by the courts as well-understood, routine, and conventional computer activity.);
Regarding claim 3, the rejection of claim 1 is incorporated herein. Further, claim 3 recites the following abstract ideas:
generating a user embedding for the user using the action sequence (This is an abstract idea of a mental process. It involves reviewing a user’s prior activities and forming a profile of the user based on the activities. A person could observe a user’s prior actions, identify patterns associated with the actions, and develop a characterization of the user’s interests, preferences, or behavior based on observation and judgement. This type of observation, evaluation, and characterization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.),
The following claim elements are additional elements which, taken alone or in combination with the other elements, do not integrate the judicial exception into a practical application nor amount to significantly more than the judicial exception:
wherein the input data comprises the plurality of features and the user embedding (This limitation merely recites data to be used by the machine learning model and therefore amounts to insignificant extra-solution activity.).
Regarding claim 4, the rejection of claim 3 is incorporated herein. Further, claim 4 recites the following additional elements, which taken alone or in combination with other elements, do not integrate the judicial exception into a practical application nor amount to significantly more than the judicial exception:
causing the recommendation to be presented to the user in the ongoing session (Displaying information to a user merely outputs the result of the abstract idea and amounts to insignificant extra-solution activity.).
Regarding claim 5, the rejection of claim 1 is incorporated herein. Further, claim 5 recites the following abstract ideas:
consolidating two or more event signals of the plurality of event signals into a consolidated event; and filtering content of the consolidated event using the unified schema (This is an abstract idea of a mental process. It involves reviewing multiple user activities, determining that the activities should be considered together, removing information that is not relevant to a particular category, and categorizing the remaining information according to predefined categories. A person could observe multiple actions performed by a user, combine the actions into a single activity based on observation and judgement, disregard information that is not relevant to a particular category, and classify the remaining activity into one or more categories. This type of observation, evaluating, filtering, and categorization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.).
Regarding claim 6, the rejection of claim 5 is incorporated herein. Further, claim 6 recites the following abstract ideas:
wherein consolidating the two or more event signals is in response to determining that each of the two or more event signals do not satisfy a category threshold and that the two or more event signals together do satisfy the category threshold (This is an abstract idea of a mental process. It involves evaluating multiple user activities, determining that the activities individually do not provide sufficient information to satisfy a category, and determining that the activities collectively provide sufficient information to satisfy the category. A person could observe multiple actions performed by a user, determine that each action alone is insufficient to indicate a particular interest or preference, and determine that the actions together are sufficient to indicate the interest or preference based on observation and judgement. This type of observation, evaluation, and decision-making can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.).
Regarding claim 7, the rejection of claim 1 is incorporated herein. Further, claim 7 recites the following abstract ideas:
extracting an action sequence from the unified action stream (This is an abstract idea of a mental process. It involves reviewing a user’s activities and identifying an ordered sequence of the activities. A person could observe a user’s prior actions, determine the order in which the actions occurred, and select the ordered actions from a collection of activities based on observation and judgement. This type of observation, selection, and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.)
The following claim elements are additional elements which, taken alone or in combination with the other elements, do not integrate the judicial exception into a practical application nor amount to significantly more than the judicial exception:
using a sliding time window (This limitation constitutes mere instructions to apply the abstract idea and insignificant extra-solution activity. See MPEP 2106.05(f) and 2106.05(g).).
Regarding claim 8, the rejection of claim 1 is incorporated herein. Further, claim 8 recites the following abstract ideas:
detecting a trigger to generate the action sequence, wherein generating the action sequence is in response to detecting the trigger (This is an abstract idea of a mental process. It involves reviewing a user’s activities, determining whether the activities indicate a particular interest, objective, or intent, and deciding to organize the activities into an action sequence based on that determination. A person could observe a user’s prior actions, such as websites visited or activities performed, and determine based on observation and judgement that the actions indicate a particular interest or objective. For example, a person reviewing a user’s browsing activity may determine that the user is searching for a new job, planning a vacation, or preparing to place a sports wager, and in response organize the user’s activities into a sequence reflecting that interest or objective. This type of observation, evaluation, and decision-making can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.).
Regarding claim 9, the rejection of claim 8 is incorporated herein. Further, claim 9 recites the following abstract ideas:
detecting a subsequent event for the ongoing session (This is an abstract idea of a mental process. It involves observing a user’s activities during an ongoing session and identifying activities that occur after the initial activity. A person could observe a user’s actions, determine the order in which the actions occur, and identify subsequent actions based on observation and judgement. This type of observation, evaluation, and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.)
Regarding claim 10, the rejection of claim 1 is incorporated herein. Further, claim 10 recites the following abstract ideas:
generating a short-term user embedding for the ongoing session using the plurality of features and the action sequence (This is an abstract idea of a mental process. It involves reviewing information associated with a user’s activities during an ongoing session and forming a profile of the user based on the information. A person could observe a user’s activities, evaluate characteristics associated with activities, identify patterns reflected by the activities, and develop a characterization of the user’s interests, preferences, or behavior based on observation and judgement. This type of observation, evaluation, and characterization of information can practically be performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.)
The following claim elements are additional elements which, taken alone or in combination with the other elements, do not integrate the judicial exception into a practical application nor amount to significantly more than the judicial exception:
wherein the input data further comprises the short-term user embedding (This limitation merely identifies additional data for use in conjunction with the abstract idea and therefore amounts to insignificant extra-solution activity.).
Regarding claim 11, the following claim elements are abstract ideas:
create a plurality of processed events by filtering content of the plurality of event signals using a unified schema comprising a plurality of categories (This is an abstract idea of a mental process. It involves observing actions performed by a user during an online session, determining which predefined categories correspond to the observed actions, and organizing the actions according to these categories. A person could observe a user’s online activities, such as viewed content, selected items, or other interactions, apply judgement to determine which category corresponds to each activity, and create a record of the categorized actions. This type of observation, evaluation, and categorization can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.);
create a unified action stream by aggregating the plurality of processed events (This is an abstract idea of a mental process. It involves collecting and organizing a user’s prior actions to determine the user’s interests and preferences. A person could observe a user’s prior activities, organize the activities based on observation and judgement, and use the organized activities to determine which content, item, or activity would likely be of interest to the user. This type of observation, evaluation, and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.)
generate a plurality of features using the unified action stream (This is an abstract idea of a mental process. It involves evaluating a user’s prior activities and deriving characteristics associated with a user’s interests, preferences, or behavior from those activities. A person could observe a user’s prior actions, analyze the actions based on observation and judgement, and identify characteristics that describe the user’s interests or preferences. This type of observation, evaluation, and characterization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.);
generate an action sequence using the unified action stream (This is an abstract idea of a mental process. It involves reviewing a user’s prior activities and organizing the activities into an ordered sequence. A person could observe a user’s prior actions, determine the order in which the actions occurred, and arrange actions into a sequence based on observation and judgement. This type of observation and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.);
generate input data for a trained machine learning model, the input data comprising the plurality of features and the action sequence (This is an abstract idea of a mental process. It involves determining what information associated with a user’s activities is relevant to a decision or recommendation and organizing the information for consideration. A person could observe a user’s prior actions, identify characteristics and sequence that are relevant to determining a recommendation based on observation, judgement, and decision-making. This type of observation, evaluation, selection, and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.); and
The following claim elements are additional elements which, taken alone or in combination with the other elements, do not integrate the judicial exception into a practical application nor amount to significantly more than the judicial exception:
at least one memory device (This a high-level recitation of generic computer components for performing the abstract idea. See MPEP 2106.05(f).);
a processing device (This a high-level recitation of generic computer components for performing the abstract idea. See MPEP 2106.05(f).),
receive a plurality of event signals from a plurality of verticals for an ongoing session of a user of an online system (The step of “receiving” information merely gathers data for use in the abstract idea and amounts to well-understood, routine, and conventional activity.);
generat3 an output of the trained machine learning model by applying the trained machine learning model to the input data (This step of “generating an output” merely obtains a result based on the abstract idea and therefore amounts to insignificant extra-solution activity.).
Regarding claim 12, the rejection of claim 11 is incorporated herein. The claim recites similar limitations corresponding to claim 2. Therefore, the same subject matter analysis that was utilized for claim 2, as described above, is equally applicable to claim 12.
Therefore, claim 12 is ineligible.
Regarding claim 13, the rejection of claim 11 is incorporated herein. The claim recites similar limitations corresponding to claim 3. Therefore, the same subject matter analysis that was utilized for claim 3, as described above, is equally applicable to claim 13.
Therefore, claim 13 is ineligible.
Regarding claim 14, the rejection of claim 11 is incorporated herein. The claim recites similar limitations corresponding to claim 5. Therefore, the same subject matter analysis that was utilized for claim 5, as described above, is equally applicable to claim 14.
Therefore, claim 14 is ineligible.
Regarding claim 15, the rejection of claim 14 is incorporated herein. The claim recites similar limitations corresponding to claim 6. Therefore, the same subject matter analysis that was utilized for claim 6, as described above, is equally applicable to claim 15.
Therefore, claim 15 is ineligible.
Regarding claim 16, the rejection of claim 11 is incorporated herein. The claim recites similar limitations corresponding to claim 7. Therefore, the same subject matter analysis that was utilized for claim 7, as described above, is equally applicable to claim 16.
Therefore, claim 16 is ineligible.
Regarding claim 17, the rejection of claim 11 is incorporated herein. The claim recites similar limitations corresponding to claim 8. Therefore, the same subject matter analysis that was utilized for claim 8, as described above, is equally applicable to claim 17.
Therefore, claim 17 is ineligible.
Regarding claim 18, the rejection of claim 17 is incorporated herein. The claim recites similar limitations corresponding to claim 9. Therefore, the same subject matter analysis that was utilized for claim 9, as described above, is equally applicable to claim 18.
Therefore, claim 18 is ineligible.
Regarding claim 19, the rejection of claim 17 is incorporated herein. The claim recites similar limitations corresponding to claim 10. Therefore, the same subject matter analysis that was utilized for claim 10, as described above, is equally applicable to claim 19.
Therefore, claim 19 is ineligible.
Regarding claim 20, the following claim elements are abstract ideas:
create a plurality of processed events by filtering content of the plurality of event signals using a unified schema comprising a plurality of categories (This is an abstract idea of a mental process. It involves observing actions performed by a user during an online session, determining which predefined categories correspond to the observed actions, and organizing the actions according to these categories. A person could observe a user’s online activities, such as viewed content, selected items, or other interactions, apply judgement to determine which category corresponds to each activity, and create a record of the categorized actions. This type of observation, evaluation, and categorization can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.);
create a unified action stream by aggregating the plurality of processed events (This is an abstract idea of a mental process. It involves collecting and organizing a user’s prior actions to determine the user’s interests and preferences. A person could observe a user’s prior activities, organize the activities based on observation and judgement, and use the organized activities to determine which content, item, or activity would likely be of interest to the user. This type of observation, evaluation, and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.)
generate a plurality of features using the unified action stream (This is an abstract idea of a mental process. It involves evaluating a user’s prior activities and deriving characteristics associated with a user’s interests, preferences, or behavior from those activities. A person could observe a user’s prior actions, analyze the actions based on observation and judgement, and identify characteristics that describe the user’s interests or preferences. This type of observation, evaluation, and characterization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.);
generate an action sequence using the unified action stream (This is an abstract idea of a mental process. It involves reviewing a user’s prior activities and organizing the activities into an ordered sequence. A person could observe a user’s prior actions, determine the order in which the actions occurred, and arrange actions into a sequence based on observation and judgement. This type of observation and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.);
generate input data for a trained recommendation model, the input data comprising the plurality of features and the action sequence (This is an abstract idea of a mental process. It involves determining what information associated with a user’s activities is relevant to a decision or recommendation and organizing the information for consideration. A person could observe a user’s prior actions, identify characteristics and sequence that are relevant to determining a recommendation based on observation, judgement, and decision-making. This type of observation, evaluation, selection, and organization of information can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.); and
generate a recommendation (This is an abstract idea of a mental process. It involves evaluating information associated with a user and determining a recommendation based on the information. A person could review information regarding the user’s activities, interests, and preferences, or behavior, evaluate the information using observation and judgement, and select a recommendation for the user based on the evaluation. This type of observation, evaluation, and decision-making can be practically performed in the human mind or with pen and paper and therefore falls within the mental process grouping of abstract ideas.)
The following claim elements are additional elements which, taken alone or in combination with the other elements, do not integrate the judicial exception into a practical application nor amount to significantly more than the judicial exception:
at least one memory device (This a high-level recitation of generic computer components for performing the abstract idea. See MPEP 2106.05(f).);
a processing device (This a high-level recitation of generic computer components for performing the abstract idea. See MPEP 2106.05(f).),
recommendation model (This a high-level recitation of generic computer components for performing the abstract idea. See MPEP 2106.05(f).),
receive a plurality of event signals from a plurality of verticals for an ongoing session of a user of an online system (The step of “receiving” information merely gathers data for use in the abstract idea and amounts to well-understood, routine, and conventional activity.);
from the trained recommendation model by applying the trained recommendation model to the input data (This limitation merely instructs that the abstract idea be performed using a recommendation model and therefore amounts to mere instructions to apply the exception using a generic computer component. Additionally, the limitation amounts to insignificant extra-solution activity.);
cause the recommendation to be presented to the user in the ongoing session (Displaying information to a user merely outputs the result of the abstract idea and amounts to insignificant extra-solution activity.).
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.
Claims 1-6, 10-15, and 19-20 are rejected under the 35 U.S.C. 103 as being unpatentable over Yao et al., (NPL: “USER: A Unified Information Search and Recommendation Model based on Integrated Behavior Sequence” (Published: 2021) in view of Jiang et al., (Pub. No.: US 20190065606 A1 (Filed: 2017)).
Regarding claim 1, Yao teaches the following limitations:
receiving a plurality of event signals from a plurality of verticals for an ongoing session of a user of an online system (Yao, [Introduction] “When a user browses the article list generated by the recommendation system, she is attracted by the article titled “New energy vehicle:Weilai ...?”. After reading this article, she switches to the search engine and issues a query to seek more knowledge about “New energy vehicle”. Then, she browses the search results and articles recommended along with the clicked document to know more. Such an information-seeking pattern which mixes behaviors made in proactive searches and passive recommendations is common in our surfing process.” – Yao teaches an online information content platform having multiple services including a recommendation system and a search engine. Under the broadest reasonable interpretation, the recommendation system and search engine correspond to the claimed plurality of verticals. Yao further teaches user interactions including browsing recommended articles, issuing search queries, browsing search results, and selecting documents. Under BRI, the browsing activities, search queries, search result interactions, and document selections correspond to the claimed plurality of event signals. The disclosed information-seeking pattern includes user interactions occurring across the recommendation system and search engine during a continuous user activity. Accordingly, Yao teaches receiving a plurality of event signals from a plurality of verticals for an ongoing session of a user of an online system.);
creating a unified action stream by aggregating the plurality of processed events (Yao, [section 3] “On an information content service platform with both search engine and recommendation engine, the user 𝑢 could browse articles 𝐵 in the recommendation system, issue queries 𝑄 to seek for information and click satisfied documents 𝐷 in the search engine. All these behaviors are sequential, so we integrate them into a heterogeneous behavior sequence in chronological order.” – Yao teaches user interactions occurring across multiple services including a recommendation system and a search engine. Yao further teaches integrating browsing activities, search queries, and document selection activities into a heterogeneous behavior sequence in chronological order. Under BRI, the browsing activities, search queries, and document-selection activities correspond to the claimed plurality of processed events, while the heterogeneous behavior sequence corresponds to claimed unified action stream. The integration of multiple behavior type into a heterogeneous behavior sequence corresponds to aggregating the plurality of processed events to create a unified stream.);
generating a plurality of features using the unified action stream; generating an action sequence using the unified action stream (Yao, [section 3] “the user 𝑢 could browse articles 𝐵 in the recommendation system, issue queries 𝑄 to seek for information and click satisfied documents 𝐷 in the search engine. All these behaviors are sequential, so we integrate them into a heterogeneous behavior sequence in chronological order.” [section 4.2] “The two vectors are concatenated to generate the representation of a historical search behavior
r
S
through an MLP layer…For a browsing behavior made in recommendation, it corresponds to only a browsed article 𝐵. Thus, its representation is just the article representation
r
B
calculated by the text encoder. With the representation of all past behaviors in the current session calculated,
H
S
=
{
r
B
1
,
r
S
2
,
.
.
.
}
, we could capture the relationships between the search and browsing behaviors, and fuse the session context into the user’s current intention. We combine
H
S
with the target intention
I
t
and pass them through a session-level transformer for interaction…The action type includes search (S) and browsing (B). Finally, the output of the last position
I
t
s
represents the user’s current intention fusing the session context” – Yao teaches user actions including browsing articles in a recommendation system, issuing search queries, and selecting documents in a search engine. Yao further teaches integrating user actions into a heterogeneous behavior sequence in chronological order. Under BRI, the browsing activities, search activities, and document-selection activities correspond to actions represented within the claimed unified action stream, while the heterogeneous behavior sequence corresponds to the claimed action sequence generated using the unified action stream. Yao additionally teaches generating representations for historical search behaviors (
r
S
), browsing behaviors (
r
B
), session behaviors (
H
S
), and a current user intention representation (
I
t
s
) from the heterogeneous behavior sequence. Yao further teaches identifying action types including search and browsing when processing the sequence. Under BRI, the generated search-behavior representations, browsing-behavior representations, session-behavior representations, and user-intention representations correspond the claimed plurality of features generated using the unified action stream. Accordingly, Yao teaches generating a plurality of features using the unified action stream and generating an action sequence using the unified action stream.);
generating input data for a trained machine learning model, the input data comprising the plurality of features and the action sequence; and generating an output of the trained machine learning model by applying the trained machine learning model to the input data (Yao, [section 4.4] “We represent the user’s current intention as
I
t
that is initialized with the issued query
Q
t
for search or the user embedding
E
m
b
u
for recommendation. The unified problem is to rank the candidate document
D
t
based on the personalized relevance that is calculated with the current intention
I
t
, the query
Q
t
(empty for recommendation) and the user history 𝐻…Through the text encoder, session encoder and history encoder, we get the representations of the user’s current intention and candidate document.. Besides, following [13, 20], we also extract several relevance-based features
F
q
,
d
for personalized search…Finally, the score for the candidate document is calculated by aggregating all these scores and features with an MLP layer as:
p
u
n
i
f
i
e
d
(
D
t
|
I
t
,
Q
t
,
H
)
=
Φ
(
f
u
n
i
f
i
e
d
)
. Φ() represents an MLP layer without an activation function. Whether for the search or recommendation task, we generate personalized document list by calculating relevance scores in this way.” – Yao teaches representing a user’s current intention using a query or user embedding and calculating personalized relevance using the current intention, query information, and user history. Yao further teaches generating representations of a user’s current intention and candidate documents and extract relevance-based features. Under BRI, the user history corresponds to the previously generated action sequence, while the generated representations and extracted relevance-based features corresponds to the previously generated plurality of features. Accordingly, Yao teaches generating input data for a trained machine learning model, the input data comprising a plurality of features and the action sequence. Yao further teaches calculating a candidate-document score by aggregating the scores and features with an MLP layer and generating relevance values
p
u
n
i
f
i
e
d
(
D
t
|
I
t
,
Q
t
,
H
)
. Yao additionally teaches generating a personalized document list based on the calculated relevance scores. Under BRI, the MLP layer corresponds to the claimed training machine learning model, while the personalized relevance values and personalized document list correspond to the claimed output generated by applying the trained machine learning model to the input data.).
However, Yao does not teach but Yao in view of Jiang teaches the following limitations:
creating a plurality of processed events by filtering content of the plurality of event signals using a unified schema comprising a plurality of categories (Yao, “When a user browses the article list generated by the recommendation system, she is attracted by the article titled “New energy vehicle:Weilai ...?”. After reading this article, she switches to the search engine and issues a query to seek more knowledge about “New energy vehicle”. Then, she browses the search results and articles recommended along with the clicked document to know more. Such an information-seeking pattern which mixes behaviors made in proactive searches and passive recommendations is common in our surfing process.” Jiang, paragraph [0086] “When a user takes an action within the social networking system 630, the action is recorded in the activity log 642…When an action is taken within the social networking system 630, an entry for the action is added to the activity log 642. The activity log 642 may be referred to as an action log.” [0088] “Other examples of actions where a user interacts with an external system 620 include a user expressing an interest in an external system 620 or another entity, a user posting a comment to the social networking system 630 that discusses an external system 620 or a web page 622a within the external system 620, a user posting to the social networking system 630 a Uniform Resource Locator (URL) or other identifier associated with an external system 620…Thus, the activity log 642 may include actions describing engagements between a user of the social networking system 630 and an external system 620 that is separate from the social networking system 630.” [0045] “The category filtering module 304 can be configured to filter categories based on one or more filtering criteria….Any categories in the set of specific categories that do not have a confidence score that satisfies the first confidence score threshold (e.g., that do not meet or exceed the first confidence score threshold) can be removed from consideration as a category recommendation….Any categories in the set of general categories that do not have a confidence score that satisfies the second confidence score threshold (e.g., that do not meet or exceed the second confidence score threshold) can be removed from consideration as a category recommendation.” [0046] “filtering of the set of specific categories may result in a filtered set of specific categories, and filtering of the set of general categories may result in a filtered set of general categories.” – Yao teaches user interaction events including browsing recommended articles, issuing search queries, browsing search results, and selecting documents. Jiang further teaches recording user actions as entries in an activity log, including expressing interest in an entity, posting comments, posting identifiers, and attending events. Under BRI, the user interaction events and record action entries correspond to the claimed plurality of events. Jiang additionally teaches multiple categories and category-based filtering in which categories that fail to satisfy confidence score thresholds are removed from consideration, resulting in filtered category sets. Under BRI, the categories correspond to the claimed plurality of categories of the unified schema, while the threshold-based category filtering corresponds to filtering content of the plurality of event signals using the unified schema. The resulting categorized and filtered action information corresponds to the claimed plurality of processed events. Therefore, the combination of Yao and Jiang teaches the claimed limitation.);
Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, having a combination of Yao and Jiang before them, to incorporate the category-based filtering techniques of Jiang into the unified search and recommendation framework of Yao. One would have been motivated to make such a combination in order to classify and filter heterogeneous user interaction information according to predefined categories before generating user behavior representations and recommendations. This would allow interaction information collected from multiple services to be processed in a more structured and consistent manner when generating user behavior representations and recommendations.
Regarding claim 2, Yao in view of Jiang teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. Yao in view of Jiang further teaches:
retrieving data associated with a processed event of the unified action stream using a category of the plurality of categories; and generating the plurality of features using the retrieved data and the processed event (Jiang, paragraph [0045] “ Any categories in the set of specific categories that do not have a confidence score that satisfies the first confidence score threshold (e.g., that do not meet or exceed the first confidence score threshold) can be removed from consideration as a category recommendation.” [0046] “The filtered set of specific categories, the filtered set of general categories, and the set of popular categories can represent a corpus of categories for potential selection as category recommendations.” Yao, [section 4.2] ““The two vectors are concatenated to generate the representation of a historical search behavior
r
S
through an MLP layer…For a browsing behavior made in recommendation, it corresponds to only a browsed article 𝐵. Thus, its representation is just the article representation
r
B
calculated by the text encoder. With the representation of all past behaviors in the current session calculated,
H
S
=
{
r
B
1
,
r
S
2
,
.
.
.
}
, we could capture the relationships between the search and browsing behaviors, and fuse the session context into the user’s current intention.” – Yao teaches user interaction data including search behaviors, browsing behaviors, and document selection behaviors that are integrated into a heterogeneous behavior sequence. As modified by Jiang, Yao’s user interaction data would be categorized and filtered using categories satisfying confidence score thresholds. Under BRI, the category-filtered user interaction data corresponds to data retrieved from the processed events using a category of the plurality of categories. Yao further teaches generating representations of search behaviors, browsing behaviors, and session behaviors form the user interaction data. Under BRI, these generated representations correspond to the claimed plurality of features generated using the retrieved data and the processed event. Accordingly, the combination of Yao and Jiang teaches the claimed limitation.).
Regarding claim 3, Yao in view of Jiang teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. Yao in view of Jiang further teaches:
generating a user embedding for the user using the action sequence, wherein the input data comprises the plurality of features and the user embedding (Yao, [section 4.2] “With the representation of all past behaviors in the current session calculated,
H
S
=
{
r
B
1
,
r
S
2
,
.
.
.
}
, we could capture the relationships between the search and browsing behaviors, and fuse the session context into the user’s current intention. We combine
H
S
with the target intention
I
t
and pass them through a session-level transformer for interaction… Finally, the output of the last position
I
t
s
represents the user’s current intention fusing the session context…
[
H
s
,
I
t
]
P
,
[
H
S
,
I
t
]
T
are the position embedding and type embedding.” – Yao teaches a behavior sequence
H
S
comprising user search and browsing behaviors. Yao further teaches combining the behavior sequence with target intention and processing the combined information through a session-level transformer. The output
I
t
S
represents the user’s current intention fusing the session context. Under BRI, the behavior sequence corresponds to the claimed action sequence, while the resulting user-intention representation corresponds to the claimed user embedding generated using the action sequence. Accordingly, Yao teaches generating a user embedding for the user using the action sequence, wherein the input data comprises the plurality of features and the user embedding.).
Regarding claim 4, Yao in view of Jiang teaches all the elements of claim 3, therefore is rejected for the same reasons as those presented for claim 3. Yao in view of Jiang further teaches:
wherein the trained machine learning model is a recommendation model and the output is a recommendation, the method further comprising: causing the recommendation to be presented to the user in the ongoing session (Yao, [section 4.2] “With the representation of all past behaviors in the current session calculated,
H
S
=
{
r
B
1
,
r
S
2
,
.
.
.
}
, we could capture the relationships between the search and browsing behaviors, and fuse the session context into the user’s current intention.” [section 4.4] “Whether for the search or recommendation task, we generate personalized document list by calculating relevance scores in this way.” [section 4.5] “We adopt a pairwise manner to train our USER model. For both personalized search and recommendation tasks… In the unified scenario, we have access to both search and recommendation data. Thus, we can train one USER model with data from the two tasks and apply the trained model to both of them.” Jiang, paragraph [0067] “ the browser application 612 displays the identified content using the format or presentation described by the markup language document 614…the markup language document 614 includes instructions for generating and displaying a web page having multiple frames that include text and/or image data” – Yao teaches a trained USER model that is applied to recommendation tasks and further teaches generating a personalized document list for the recommendation task using behaviors occurring within a current session. Under BRI, the trained USER model corresponds to the claimed recommendation model, the personalized document list corresponds to the claimed recommendation, and the current session corresponds to the claimed ongoing session. Jiang teaches displaying identified content and generating and displaying a web page to a user through a browser application. Under BRI, displaying the identified content and generated web page correspond to presenting the recommendation to the user.).
Regarding claim 5, Yao in view of Jiang teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. Yao in view of Jiang further teaches:
consolidating two or more event signals of the plurality of event signals into a consolidated event; and filtering content of the consolidated event using the unified schema (Yao, [section 3] “On an information content service platform with both search engine and recommendation engine, the user 𝑢 could browse articles 𝐵 in the recommendation system, issue queries 𝑄 to seek for information and click satisfied documents 𝐷 in the search engine. All these behaviors are sequential, so we integrate them into a heterogeneous behavior sequence in chronological order.” Jiang, paragraph [0045] “The category filtering module 304 can be configured to filter categories based on one or more filtering criteria… Any categories in the set of specific categories that do not have a confidence score that satisfies the first confidence score threshold (e.g., that do not meet or exceed the first confidence score threshold) can be removed from consideration as a category recommendation.” – Yao teaches multiple user interaction events including browsing articles, issuing search queries, and selecting documents. Yao further teaches integrating the multiple user interaction events into a heterogeneous behavior sequence. Under BRI, integrating multiple user interaction events into a single heterogeneous behavior sequence corresponds to consolidating two or more event signals into a consolidated event. Jiang teaches filtering information according to categories and confidence score based filtering criteria. In combination, Jiang’s category-based filtering is applied to Yao’s user interaction events before integration into a heterogeneous behavior sequence. Under BRI, the category filtering corresponds to filtering content of the consolidated event using the previously mapped unified schema comprising a plurality of categories. Accordingly, the combination of Yao and Jiang teaches consolidating two or more event signals of the plurality of event signals into a consolidated event and filtering content of the consolidated event using the unified schema.).
Regarding claim 6, Yao in view of Jiang teaches all the elements of claim 5, therefore is rejected for the same reasons as those presented for claim 5. Yao in view of Jiang further teaches:
wherein consolidating the two or more event signals is in response to determining that each of the two or more event signals do not satisfy a category threshold and that the two or more event signals together do satisfy the category threshold (Yao, [section 3] “On an information content service platform with both search engine and recommendation engine, the user 𝑢 could browse articles 𝐵 in the recommendation system, issue queries 𝑄 to seek for information and click satisfied documents 𝐷 in the search engine. All these behaviors are sequential, so we integrate them into a heterogeneous behavior sequence in chronological order.” Jiang, paragraph [0045] “The category filtering module 304 can be configured to filter categories based on one or more filtering criteria… Any categories in the set of specific categories that do not have a confidence score that satisfies the first confidence score threshold (e.g., that do not meet or exceed the first confidence score threshold) can be removed from consideration as a category recommendation.” – Yao teaches combining multiple user interaction events into a single heterogeneous behavior sequence. Jiang teaches determining whether information satisfies confidence-score thresholds and retaining or removing information based on threshold satisfaction. It would have been obvious to a person of ordinary skill in the art to apply Jiang’s threshold-based determination criteria to the combined user interaction events of Yao and to retain consolidated event information when the combined information satisfies a threshold, even when individual events do not independently satisfy the threshold. Doing so advantageously preserves relevant user interaction information while filtering insignificant events.).
Regarding claim 10, Yao in view of Jiang teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. Yao in view of Jiang further teaches:
generating a short-term user embedding for the ongoing session using the plurality of features and the action sequence, wherein the input data further comprises the short-term user embedding (Yao, [section 3] “Past behaviors in the current session are viewed as the short-term history.” [section 4.2] “With the representation of all past behaviors in the current session calculated,
H
S
=
{
r
B
1
,
r
S
2
,
.
.
.
}
, we could capture the relationships between the search and browsing behaviors, and fuse the session context into the user’s current intention. We combine
H
S
with the target intention
I
t
and pass them through a session-level transformer for interaction…Finally, the output of the last position
I
t
s
represents the user’s current intention fusing the session context.” [section 4.4] “Through the text encoder, session encoder and history encoder, we get the representations of the user’s current intention and candidate document, i.e.
I
t
s
,
I
t
l
,
r
D
t
a
n
d
r
l
D
t
…Finally, the score for the candidate document is calculated by aggregating all these scores and features with an MLP layer…” – Yao teaches representing past behaviors occurring within a session as short-term history and generating a current intention representation, the from the session behavior session behavior sequence. Under BRI, the session behavior sequence corresponds to the previously mapped plurality of features and action sequence, while the current-iteration representation corresponds to the claimed short-term user embedding generated for the ongoing session. Yao further teaches including
I
t
s
among the features used for the subsequent model processing. Accordingly, Yao teaches generating a short-term user embedding for the ongoing session using the plurality of features and the action sequence, wherein the input data further comprises the short-term embedding.).
Regarding claim 11, Yao teaches the following limitations:
receive a plurality of event signals from a plurality of verticals for an ongoing session of a user of an online system (Yao, [Introduction] “When a user browses the article list generated by the recommendation system, she is attracted by the article titled “New energy vehicle:Weilai ...?”. After reading this article, she switches to the search engine and issues a query to seek more knowledge about “New energy vehicle”. Then, she browses the search results and articles recommended along with the clicked document to know more. Such an information-seeking pattern which mixes behaviors made in proactive searches and passive recommendations is common in our surfing process.” – Yao teaches an online information content platform having multiple services including a recommendation system and a search engine. Under the broadest reasonable interpretation, the recommendation system and search engine correspond to the claimed plurality of verticals. Yao further teaches user interactions including browsing recommended articles, issuing search queries, browsing search results, and selecting documents. Under BRI, the browsing activities, search queries, search result interactions, and document selections correspond to the claimed plurality of event signals. The disclosed information-seeking pattern includes user interactions occurring across the recommendation system and search engine during a continuous user activity. Accordingly, Yao teaches receiving a plurality of event signals from a plurality of verticals for an ongoing session of a user of an online system.);
create a unified action stream by aggregating the plurality of processed events (Yao, [section 3] “On an information content service platform with both search engine and recommendation engine, the user 𝑢 could browse articles 𝐵 in the recommendation system, issue queries 𝑄 to seek for information and click satisfied documents 𝐷 in the search engine. All these behaviors are sequential, so we integrate them into a heterogeneous behavior sequence in chronological order.” – Yao teaches user interactions occurring across multiple services including a recommendation system and a search engine. Yao further teaches integrating browsing activities, search queries, and document selection activities into a heterogeneous behavior sequence in chronological order. Under BRI, the browsing activities, search queries, and document-selection activities correspond to the claimed plurality of processed events, while the heterogeneous behavior sequence corresponds to claimed unified action stream. The integration of multiple behavior type into a heterogeneous behavior sequence corresponds to aggregating the plurality of processed events to create a unified stream.);
generate a plurality of features using the unified action stream; generating an action sequence using the unified action stream (Yao, [section 3] “the user 𝑢 could browse articles 𝐵 in the recommendation system, issue queries 𝑄 to seek for information and click satisfied documents 𝐷 in the search engine. All these behaviors are sequential, so we integrate them into a heterogeneous behavior sequence in chronological order.” [section 4.2] “The two vectors are concatenated to generate the representation of a historical search behavior
r
S
through an MLP layer…For a browsing behavior made in recommendation, it corresponds to only a browsed article 𝐵. Thus, its representation is just the article representation
r
B
calculated by the text encoder. With the representation of all past behaviors in the current session calculated,
H
S
=
{
r
B
1
,
r
S
2
,
.
.
.
}
, we could capture the relationships between the search and browsing behaviors, and fuse the session context into the user’s current intention. We combine
H
S
with the target intention
I
t
and pass them through a session-level transformer for interaction…The action type includes search (S) and browsing (B). Finally, the output of the last position
I
t
s
represents the user’s current intention fusing the session context” – Yao teaches user actions including browsing articles in a recommendation system, issuing search queries, and selecting documents in a search engine. Yao further teaches integrating user actions into a heterogeneous behavior sequence in chronological order. Under BRI, the browsing activities, search activities, and document-selection activities correspond to actions represented within the claimed unified action stream, while the heterogeneous behavior sequence corresponds to the claimed action sequence generated using the unified action stream. Yao additionally teaches generating representations for historical search behaviors (
r
S
), browsing behaviors (
r
B
), session behaviors (
H
S
), and a current user intention representation (
I
t
s
) from the heterogeneous behavior sequence. Yao further teaches identifying action types including search and browsing when processing the sequence. Under BRI, the generated search-behavior representations, browsing-behavior representations, session-behavior representations, and user-intention representations correspond the claimed plurality of features generated using the unified action stream. Accordingly, Yao teaches generating a plurality of features using the unified action stream and generating an action sequence using the unified action stream.);
generate input data for a trained machine learning model, the input data comprising the plurality of features and the action sequence; and generate an output of the trained machine learning model by applying the trained machine learning model to the input data (Yao, [section 4.4] “We represent the user’s current intention as
I
t
that is initialized with the issued query
Q
t
for search or the user embedding
E
m
b
u
for recommendation. The unified problem is to rank the candidate document
D
t
based on the personalized relevance that is calculated with the current intention
I
t
, the query
Q
t
(empty for recommendation) and the user history 𝐻…Through the text encoder, session encoder and history encoder, we get the representations of the user’s current intention and candidate document.. Besides, following [13, 20], we also extract several relevance-based features
F
q
,
d
for personalized search…Finally, the score for the candidate document is calculated by aggregating all these scores and features with an MLP layer as:
p
u
n
i
f
i
e
d
(
D
t
|
I
t
,
Q
t
,
H
)
=
Φ
(
f
u
n
i
f
i
e
d
)
. Φ() represents an MLP layer without an activation function. Whether for the search or recommendation task, we generate personalized document list by calculating relevance scores in this way.” – Yao teaches representing a user’s current intention using a query or user embedding and calculating personalized relevance using the current intention, query information, and user history. Yao further teaches generating representations of a user’s current intention and candidate documents and extract relevance-based features. Under BRI, the user history corresponds to the previously generated action sequence, while the generated representations and extracted relevance-based features corresponds to the previously generated plurality of features. Accordingly, Yao teaches generating input data for a trained machine learning model, the input data comprising a plurality of features and the action sequence. Yao further teaches calculating a candidate-document score by aggregating the scores and features with an MLP layer and generating relevance values
p
u
n
i
f
i
e
d
(
D
t
|
I
t
,
Q
t
,
H
)
. Yao additionally teaches generating a personalized document list based on the calculated relevance scores. Under BRI, the MLP layer corresponds to the claimed training machine learning model, while the personalized relevance values and personalized document list correspond to the claimed output generated by applying the trained machine learning model to the input data.).
However, Yao does not teach but Yao in view of Jiang teaches the following limitations:
A system comprising: at least one memory device; and a processing device, operatively coupled with the at least one memory device, to (Jiang, paragraph [0094] “The computer system 700 includes a processor 702, a cache 704, and one or more executable modules and drivers, stored on a computer-readable medium, directed to the processes and features described herein. Additionally, the computer system 700 includes a high performance input/output (I/O) bus 706 and a standard I/O bus 708. A host bridge 710 couples processor 702 to high performance I/O bus 706, whereas I/O bus bridge 712 couples the two buses 706 and 708 to each other. A system memory 714 and one or more network interfaces 716 couple to high performance I/O bus 706”):
create a plurality of processed events by filtering content of the plurality of event signals using a unified schema comprising a plurality of categories (Yao, “When a user browses the article list generated by the recommendation system, she is attracted by the article titled “New energy vehicle:Weilai ...?”. After reading this article, she switches to the search engine and issues a query to seek more knowledge about “New energy vehicle”. Then, she browses the search results and articles recommended along with the clicked document to know more. Such an information-seeking pattern which mixes behaviors made in proactive searches and passive recommendations is common in our surfing process.” Jiang, paragraph [0086] “When a user takes an action within the social networking system 630, the action is recorded in the activity log 642…When an action is taken within the social networking system 630, an entry for the action is added to the activity log 642. The activity log 642 may be referred to as an action log.” [0088] “Other examples of actions where a user interacts with an external system 620 include a user expressing an interest in an external system 620 or another entity, a user posting a comment to the social networking system 630 that discusses an external system 620 or a web page 622a within the external system 620, a user posting to the social networking system 630 a Uniform Resource Locator (URL) or other identifier associated with an external system 620…Thus, the activity log 642 may include actions describing engagements between a user of the social networking system 630 and an external system 620 that is separate from the social networking system 630.” [0045] “The category filtering module 304 can be configured to filter categories based on one or more filtering criteria….Any categories in the set of specific categories that do not have a confidence score that satisfies the first confidence score threshold (e.g., that do not meet or exceed the first confidence score threshold) can be removed from consideration as a category recommendation….Any categories in the set of general categories that do not have a confidence score that satisfies the second confidence score threshold (e.g., that do not meet or exceed the second confidence score threshold) can be removed from consideration as a category recommendation.” [0046] “filtering of the set of specific categories may result in a filtered set of specific categories, and filtering of the set of general categories may result in a filtered set of general categories.” – Yao teaches user interaction events including browsing recommended articles, issuing search queries, browsing search results, and selecting documents. Jiang further teaches recording user actions as entries in an activity log, including expressing interest in an entity, posting comments, posting identifiers, and attending events. Under BRI, the user interaction events and record action entries correspond to the claimed plurality of events. Jiang additionally teaches multiple categories and category-based filtering in which categories that fail to satisfy confidence score thresholds are removed from consideration, resulting in filtered category sets. Under BRI, the categories correspond to the claimed plurality of categories of the unified schema, while the threshold-based category filtering corresponds to filtering content of the plurality of event signals using the unified schema. The resulting categorized and filtered action information corresponds to the claimed plurality of processed events. Therefore, the combination of Yao and Jiang teaches the claimed limitation.);
Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, having a combination of Yao and Jiang before them, to incorporate the category-based filtering techniques of Jiang into the unified search and recommendation framework of Yao. One would have been motivated to make such a combination in order to classify and filter heterogeneous user interaction information according to predefined categories before generating user behavior representations and recommendations. This would allow interaction information collected from multiple services to be processed in a more structured and consistent manner when generating user behavior representations and recommendations.
Regarding claim 12, Yao in view of Jiang teaches all the elements of claim 11, therefore is rejected for the same reasons as those presented for claim 11. The claim recites similar limitations corresponding to claim 2 and is rejected for similar reasons as claim 2 using similar teachings and rationale.
Regarding claim 13, Yao in view of Jiang teaches all the elements of claim 11, therefore is rejected for the same reasons as those presented for claim 11. The claim recites similar limitations corresponding to claim 3 and is rejected for similar reasons as claim 3 using similar teachings and rationale.
Regarding claim 14, Yao in view of Jiang teaches all the elements of claim 11, therefore is rejected for the same reasons as those presented for claim 11. The claim recites similar limitations corresponding to claim 5 and is rejected for similar reasons as claim 5 using similar teachings and rationale.
Regarding claim 15, Yao in view of Jiang teaches all the elements of claim 14, therefore is rejected for the same reasons as those presented for claim 14. The claim recites similar limitations corresponding to claim 6 and is rejected for similar reasons as claim 6 using similar teachings and rationale.
Regarding claim 19, Yao in view of Jiang teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. Yao in view of Jiang further teaches:
generate a short-term user embedding for the ongoing session using the plurality of features and the action sequence, wherein the input data further comprises the short-term user embedding (Yao, [section 3] “Past behaviors in the current session are viewed as the short-term history.” [section 4.2] “With the representation of all past behaviors in the current session calculated,
H
S
=
{
r
B
1
,
r
S
2
,
.
.
.
}
, we could capture the relationships between the search and browsing behaviors, and fuse the session context into the user’s current intention. We combine
H
S
with the target intention
I
t
and pass them through a session-level transformer for interaction…Finally, the output of the last position
I
t
s
represents the user’s current intention fusing the session context.” [section 4.4] “Through the text encoder, session encoder and history encoder, we get the representations of the user’s current intention and candidate document, i.e.
I
t
s
,
I
t
l
,
r
D
t
a
n
d
r
l
D
t
…Finally, the score for the candidate document is calculated by aggregating all these scores and features with an MLP layer…” – Yao teaches representing past behaviors occurring within a session as short-term history and generating a current intention representation, the from the session behavior session behavior sequence. Under BRI, the session behavior sequence corresponds to the previously mapped plurality of features and action sequence, while the current-iteration representation corresponds to the claimed short-term user embedding generated for the ongoing session. Yao further teaches including
I
t
s
among the features used for the subsequent model processing. Accordingly, Yao teaches generating a short-term user embedding for the ongoing session using the plurality of features and the action sequence, wherein the input data further comprises the short-term embedding.).
Regarding claim 20, Yao teaches the following limitations:
receive a plurality of event signals from a plurality of verticals for an ongoing session of a user of an online system (Yao, [Introduction] “When a user browses the article list generated by the recommendation system, she is attracted by the article titled “New energy vehicle:Weilai ...?”. After reading this article, she switches to the search engine and issues a query to seek more knowledge about “New energy vehicle”. Then, she browses the search results and articles recommended along with the clicked document to know more. Such an information-seeking pattern which mixes behaviors made in proactive searches and passive recommendations is common in our surfing process.” – Yao teaches an online information content platform having multiple services including a recommendation system and a search engine. Under the broadest reasonable interpretation, the recommendation system and search engine correspond to the claimed plurality of verticals. Yao further teaches user interactions including browsing recommended articles, issuing search queries, browsing search results, and selecting documents. Under BRI, the browsing activities, search queries, search result interactions, and document selections correspond to the claimed plurality of event signals. The disclosed information-seeking pattern includes user interactions occurring across the recommendation system and search engine during a continuous user activity. Accordingly, Yao teaches receiving a plurality of event signals from a plurality of verticals for an ongoing session of a user of an online system.);
create a unified action stream by aggregating the plurality of processed events (Yao, [section 3] “On an information content service platform with both search engine and recommendation engine, the user 𝑢 could browse articles 𝐵 in the recommendation system, issue queries 𝑄 to seek for information and click satisfied documents 𝐷 in the search engine. All these behaviors are sequential, so we integrate them into a heterogeneous behavior sequence in chronological order.” – Yao teaches user interactions occurring across multiple services including a recommendation system and a search engine. Yao further teaches integrating browsing activities, search queries, and document selection activities into a heterogeneous behavior sequence in chronological order. Under BRI, the browsing activities, search queries, and document-selection activities correspond to the claimed plurality of processed events, while the heterogeneous behavior sequence corresponds to claimed unified action stream. The integration of multiple behavior type into a heterogeneous behavior sequence corresponds to aggregating the plurality of processed events to create a unified stream.);
generate a plurality of features using the unified action stream; generating an action sequence using the unified action stream (Yao, [section 3] “the user 𝑢 could browse articles 𝐵 in the recommendation system, issue queries 𝑄 to seek for information and click satisfied documents 𝐷 in the search engine. All these behaviors are sequential, so we integrate them into a heterogeneous behavior sequence in chronological order.” [section 4.2] “The two vectors are concatenated to generate the representation of a historical search behavior
r
S
through an MLP layer…For a browsing behavior made in recommendation, it corresponds to only a browsed article 𝐵. Thus, its representation is just the article representation
r
B
calculated by the text encoder. With the representation of all past behaviors in the current session calculated,
H
S
=
{
r
B
1
,
r
S
2
,
.
.
.
}
, we could capture the relationships between the search and browsing behaviors, and fuse the session context into the user’s current intention. We combine
H
S
with the target intention
I
t
and pass them through a session-level transformer for interaction…The action type includes search (S) and browsing (B). Finally, the output of the last position
I
t
s
represents the user’s current intention fusing the session context” – Yao teaches user actions including browsing articles in a recommendation system, issuing search queries, and selecting documents in a search engine. Yao further teaches integrating user actions into a heterogeneous behavior sequence in chronological order. Under BRI, the browsing activities, search activities, and document-selection activities correspond to actions represented within the claimed unified action stream, while the heterogeneous behavior sequence corresponds to the claimed action sequence generated using the unified action stream. Yao additionally teaches generating representations for historical search behaviors (
r
S
), browsing behaviors (
r
B
), session behaviors (
H
S
), and a current user intention representation (
I
t
s
) from the heterogeneous behavior sequence. Yao further teaches identifying action types including search and browsing when processing the sequence. Under BRI, the generated search-behavior representations, browsing-behavior representations, session-behavior representations, and user-intention representations correspond the claimed plurality of features generated using the unified action stream. Accordingly, Yao teaches generating a plurality of features using the unified action stream and generating an action sequence using the unified action stream.);
generate a user embedding for the user using the action sequence (Yao, [section 4.2] “With the representation of all past behaviors in the current session calculated,
H
S
=
{
r
B
1
,
r
S
2
,
.
.
.
}
, we could capture the relationships between the search and browsing behaviors, and fuse the session context into the user’s current intention. We combine
H
S
with the target intention
I
t
and pass them through a session-level transformer for interaction… Finally, the output of the last position
I
t
s
represents the user’s current intention fusing the session context…
[
H
s
,
I
t
]
P
,
[
H
S
,
I
t
]
T
are the position embedding and type embedding.” – Yao teaches a behavior sequence
H
S
comprising user search and browsing behavior. Yao further teaches combining the behavior sequence with target intention and processing the combined information through a session-level transformer. The outputs
I
t
s
represents the user’s current intention fusing the session context. Under BRI, the behavior sequence corresponds to the claimed action sequence, while the resulting user-intention representation corresponds to the claimed user embedding generated using the action sequence. Accordingly, Yao teaches generating a user embedding for the user using the action sequence.);
generate input data for a trained recommendation model, the input data comprising the plurality of features and the user embedding (Yao, [section 4.4] “We represent the user’s current intention as
I
t
that is initialized with the issued query
Q
t
for search or the user embedding
E
m
b
u
for recommendation…we also extract several relevance-based features 𝐹𝑞,𝑑 for personalized search…Finally, the score for the candidate document is calculated by aggregating all these scores and features with an MLP layer” [section 4.5] “We adopt a pairwise manner to train our USER model. For both personalized search and recommendation tasks…In the unified scenario, we have access to both search and recommendation data. Thus, we can train one USER model with data from the two tasks and apply the trained model to both of them…Then, we make a copy for each task and finetune it with the corresponding task data to fit the individual data distribution.” – Yao teaches generating user embeddings and relevant-based features that are provided to an MLP layer for recommendation scoring. Yao further teaches training and fine-tuning the USER model using the recommendation data. Under BRI, the trained USER corresponds the claimed trained recommendation model, while the user embedding (
E
m
b
u
) and relevance-based features correspond the claimed input data comprising the plurality of features and the user embedding. Accordingly, Yao teaches generating an input data for a trained recommendation model, the input data comprising a plurality of features and user embeddings.);
However, Yao does not teach but Yao in view of Jiang teaches the following limitations:
A system comprising: at least one memory device; and a processing device, operatively coupled with the at least one memory device, to (Jiang, paragraph [0094] “The computer system 700 includes a processor 702, a cache 704, and one or more executable modules and drivers, stored on a computer-readable medium, directed to the processes and features described herein. Additionally, the computer system 700 includes a high performance input/output (I/O) bus 706 and a standard I/O bus 708. A host bridge 710 couples processor 702 to high performance I/O bus 706, whereas I/O bus bridge 712 couples the two buses 706 and 708 to each other. A system memory 714 and one or more network interfaces 716 couple to high performance I/O bus 706”):
create a plurality of processed events by filtering content of the plurality of event signals using a unified schema comprising a plurality of categories (Yao, “When a user browses the article list generated by the recommendation system, she is attracted by the article titled “New energy vehicle:Weilai ...?”. After reading this article, she switches to the search engine and issues a query to seek more knowledge about “New energy vehicle”. Then, she browses the search results and articles recommended along with the clicked document to know more. Such an information-seeking pattern which mixes behaviors made in proactive searches and passive recommendations is common in our surfing process.” Jiang, paragraph [0086] “When a user takes an action within the social networking system 630, the action is recorded in the activity log 642…When an action is taken within the social networking system 630, an entry for the action is added to the activity log 642. The activity log 642 may be referred to as an action log.” [0088] “Other examples of actions where a user interacts with an external system 620 include a user expressing an interest in an external system 620 or another entity, a user posting a comment to the social networking system 630 that discusses an external system 620 or a web page 622a within the external system 620, a user posting to the social networking system 630 a Uniform Resource Locator (URL) or other identifier associated with an external system 620…Thus, the activity log 642 may include actions describing engagements between a user of the social networking system 630 and an external system 620 that is separate from the social networking system 630.” [0045] “The category filtering module 304 can be configured to filter categories based on one or more filtering criteria….Any categories in the set of specific categories that do not have a confidence score that satisfies the first confidence score threshold (e.g., that do not meet or exceed the first confidence score threshold) can be removed from consideration as a category recommendation….Any categories in the set of general categories that do not have a confidence score that satisfies the second confidence score threshold (e.g., that do not meet or exceed the second confidence score threshold) can be removed from consideration as a category recommendation.” [0046] “filtering of the set of specific categories may result in a filtered set of specific categories, and filtering of the set of general categories may result in a filtered set of general categories.” – Yao teaches user interaction events including browsing recommended articles, issuing search queries, browsing search results, and selecting documents. Jiang further teaches recording user actions as entries in an activity log, including expressing interest in an entity, posting comments, posting identifiers, and attending events. Under BRI, the user interaction events and record action entries correspond to the claimed plurality of events. Jiang additionally teaches multiple categories and category-based filtering in which categories that fail to satisfy confidence score thresholds are removed from consideration, resulting in filtered category sets. Under BRI, the categories correspond to the claimed plurality of categories of the unified schema, while the threshold-based category filtering corresponds to filtering content of the plurality of event signals using the unified schema. The resulting categorized and filtered action information corresponds to the claimed plurality of processed events. Therefore, the combination of Yao and Jiang teaches the claimed limitation.);
generate a recommendation from the trained recommendation model by applying the trained recommendation model to the input data; and cause the recommendation to be presented to the user in the ongoing session (Yao, [section 4.4] “empty. Finally, the score for the candidate document is calculated by aggregating all these scores and features with an MLP layer… Whether for the search or recommendation task, we generate personalized document list by calculating relevance scores in this way.” [section 4.5] “We adopt a pairwise manner to train our USER model. For both personalized search and recommendation tasks…Thus, we can train one USER model with data from the two tasks and apply the trained model to both of them.” Jiang, paragraph [0067] “ the browser application 612 displays the identified content using the format or presentation described by the markup language document 614” – Yao teaches applying the trained USER model to input data to calculate relevance scores and generate a personalized document list for a recommendation task. Under BRI, the personalized document list corresponds to the claimed recommendation. Jiang teaches causing identified content to be displayed to the user through a browser application on the user device.).
Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, having a combination of Yao and Jiang before them, to incorporate the category-based filtering techniques of Jiang into the unified search and recommendation framework of Yao. One would have been motivated to make such a combination in order to classify and filter heterogeneous user interaction information according to predefined categories before generating user behavior representations and recommendations. This would allow interaction information collected from multiple services to be processed in a more structured and consistent manner when generating user behavior representations and recommendations.
Claims 7, 8, 9, 16, 17, and 18 are rejected under the 35 U.S.C. 103 as being unpatentable over Yao et al., (NPL: “USER: A Unified Information Search and Recommendation Model based on Integrated Behavior Sequence” (Published: 2021) in view of Jiang et al., (Pub. No.: US 20190065606 A1 (Filed: 2017)) further in view of Ramamurthy et al., (Pub. No.: US 20220405619 A1 Filed: (2021)).
Regarding claim 7, Yao in view of Jiang teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. However, Yao in view of Jiang does not teach but Yao in view of Jiang further in view of Ramamurthy teaches:
extracting an action sequence from the unified action stream using a sliding time window (Yao, [section 3] “All these behaviors are sequential, so we integrate them into a heterogeneous behavior sequence in chronological order. Referring to existing session segmentation methods [13, 20], we divide the user’s whole behavior sequence into several sessions with 30 minutes of inactivity as the interval.” Ramamurthy, paragraph [0031] “Streaming data is typically processed sequentially and incrementally on a record-by-record (e.g., tuple-by-tuple) basis or over one or more sliding time and/or tumbling windows. Such windows are data objects that move across continuous streaming data, splitting the data into finite sets…In a sliding window, for instance, tuples are grouped within a window that slides across the data stream according to a specified interval.” – Yao teaches generating heterogeneous behavior sequence from integrated user interaction events and further teaches dividing the behavior sequence into sessions based on a temporal interval. As discussed above, the heterogeneous behavior sequence corresponds to claimed action sequence generated from the unified action steam. Under BRI, dividing the behavior sequence into sessions corresponds to extracting portions of the action sequence from the unified action stream. Ramamurthy teaches using a sliding window that moves across streaming data and groups data according to a specified interval. Accordingly, the combination of Yao and Ramamurthy teaches extracting an action sequence from the unified action stream using a sliding time window.).
Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, having Yao, Jiang, and Ramamurthy before them, to incorporate Ramamurthy’s sliding time window processing into Yao’s behavior sequence generation. One would have been motivated to make such a combination in order to extract temporally relevant action sequences from a continuous stream of categorized user interaction events for subsequent machine learning processing. This would allow the generated action sequences to reflect recent user activity while reducing the influence of stale event data.
Regarding claim 8, Yao in view of Jiang teaches all the elements of claim 1, therefore is rejected for the same reasons as those presented for claim 1. However, Yao in view of Jiang does not teach but Yao in view of Jiang further in view of Ramamurthy teaches:
detecting a trigger to generate the action sequence, wherein generating the action sequence is in response to detecting the trigger (Ramamurthy, paragraph [0034] “The triggering event detector 104 is generally responsible for detecting a triggering event in data (e.g., the streaming data filtered by the stream filtering component 102)…” [0035] “A “triggering event” is any event that triggers or causes an estimate or prediction to be generated by one or more machine learning models and/or other intermediate steps…the triggering event is a synchronous request for an insight.” – As discussed above, Yao and Jiang teach filtering user interaction events according to categories and generating a unified action stream from the resulting processed events. Ramamurthy teaches detecting a triggering event in filtered streaming data and further teaches that the triggering event causes subsequent machine learning processing to occur. Under BRI, the triggering event corresponds to the claimed trigger, while initiating subsequent processing of the filtered stream data in response to the triggering event corresponds to generating the action sequence in response to detecting the trigger. Accordingly, the combination of Yao, Jiang, and Ramamurthy teaches detecting a trigger to generate the action sequence, wherein generating the action sequence is in response to detecting the trigger.).
Regarding claim 9, Yao in view of Jiang further in view of Ramamurthy teaches all the elements of claim 8, therefore is rejected for the same reasons as those presented for claim 8. Yao in view of Jiang further in view of Ramamurthy teaches:
wherein the trigger comprises: detecting a subsequent event for the ongoing session (Yao, [section 3] “All these behaviors are sequential, so we integrate them into a heterogeneous behavior sequence in chronological order… Past behaviors in the current session are viewed as the short-term history… The horizontal edges indicate the sequential relationship between two consecutive actions.” Ramamurthy, paragraph [0034] “The triggering event detector 104 is generally responsible for detecting a triggering event in data…” – Yao teaches a sequence of user behaviors occurring within a current session and further teaches sequential relationships between consecutive actions in the session. Under BRI, a consecutive action occurring after a prior action within the current session corresponds to the claimed subsequent event for the ongoing session. Ramamurthy teaches detecting triggering events in data using a triggering event detector. Accordingly, the combination of Yao, Jiang, and Ramamurthy teaches wherein the trigger comprises detecting a subsequent event for the ongoing session.).
Regarding claim 16, Yao in view of Jiang teaches all the elements of claim 11, therefore is rejected for the same reasons as those presented for claim 11. The claim recites similar limitations corresponding to claim 7 and is rejected for similar reasons as claim 7 using similar teachings and rationale.
Regarding claim 17, Yao in view of Jiang teaches all the elements of claim 11, therefore is rejected for the same reasons as those presented for claim 11. The claim recites similar limitations corresponding to claim 8 and is rejected for similar reasons as claim 8 using similar teachings and rationale.
Regarding claim 18, Yao in view of Jiang further in view of Ramamurthy teaches all the elements of claim 17, therefore is rejected for the same reasons as those presented for claim 17. The claim recites similar limitations corresponding to claim 9 and is rejected for similar reasons as claim 9 using similar teachings and rationale.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Daravanh Phakousonh whose telephone number is (571)272-6324. The examiner can normally be reached Mon - Thurs 7 AM - 5 PM, Every other Friday 7 AM - 4PM.
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, Li B Zhen can be reached at 571-272-3768. 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.
/Daravanh Phakousonh/Examiner, Art Unit 2121
/Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights