DETAILED ACTION
This Office Action is in response to the RCE filed on 12/17/2025.
Claims 8, 12, 14, and 15 currently amended.
Claim 16 newly added.
Claims 8-16 are currently pending in this application and have been examined.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
In reference to Applicant’s arguments on page(s) 5-8 regarding rejections made under 35 U.S.C. 103:
Independent claims 8 and 15 stand rejected under 35 U.S.C. § 103 as allegedly
unpatentable over Estes and Liu. However, neither Ester nor Liu teach everything they are purported to teach and therefore the proposed combination is deficient.
Estes does not teach updating a classifier dictionary as required by the instant claims. The Office Action at p. 6 alleges that this feature is disclosed in Estes at [0090] and [0093], where Estes describes specific steps of personal search agent system (PSA) equipped with user agent (UA). See also Estes at [0069], [0096]. Specifically, Estes at [0090] describes that PSA updates its understanding of a user based on user interactions. Estes at [0093] describes that PSA again updates its understanding of a user (e.g., preference, intent) based on user survey data. The user understanding is not a classifier dictionary nor does updating the user understanding correspond to updating a classifier dictionary based on data sources. Instead, the user understanding is used for tasks such as filtering search results. Id. at [0069], [0095]. Moreover, even aside from the UA of the PSA, Estes does not use a classification dictionary as claimed at all; instead it uses a natural language processing technique of tokenization and subsequent association of tokens to form a concept map like network representation. Id. at [0061], [0065], [0103]. Therefore, Estes does not teach all that it is purported to teach.
Furthermore, the Office Action at p. 7 concedes that Estes does not disclose the feature of collecting feature data "based on a classifier dictionary comprising one or more predetermined classifiers, wherein the one or more predetermined classifiers comprise one or more hashtags," implying that a classifier dictionary is not used for collecting feature data from data sources in Estes as required by the instant claims. It is unclear how Estes could purportedly update a classifier dictionary based on data sources if, as admitted, Estes does not use a classifier dictionary with the data sources.
Regarding independent claim 15, the Office Action at pp. 10-11 implies that a hybrid structure of Dynamic Molecular Language (DML) in Estes at [0054] corresponds to a classifier dictionary as claimed because it purports that the six element types of the hybrid data structure are used as classifiers to classify data from a plurality of data sources for feature extraction. However, these elements are not classifiers but rather are used in an "extended version of standard Belief, Desire, Intention (BDI) models of agents." Id. at [0054]-[0055]. Such models are used to guide and coordinate agent behavior for improved workflow, not to classify anything, let alone as claimed in independent claims 8 or 15. See id. at [0055] (discussing using the hybrid data structures to guide "patterns of action within the system" within "codes of conduct" using a "binary transformation space for all possible agent actions"). Furthermore, Estes does not provide any connection between the hybrid structure of DML and the UA of the PSA that is allegedly used to update the hybrid structure. Hence, Estes does not teach updating a classifier dictionary as claimed contrary to the assertions of the Office Action.
While the Office Action acknowledges that Estes does not teach a classifier dictionary comprising one or more predetermined classifiers comprising one or more hashtags, it purports that Liu does teach this limitation and would be used to modify Estes. Office Action at p. 6. However, Liu does not teach "one or more predetermined classifiers compris[ing] one or more hashtags" as claimed. Instead, Liu searches hashtags in social media posts for presence of certain terms for the purpose of generating text labels for images. That is, at best arguendo, the hashtags in Liu are part of the data being classified, not the classifiers used for the classification, Indeed, Liu at Section 3 talks about identifying nouns "in hashtags" (emphasis added) but not using hashtags as classifiers as required by the instant claims. Liu does not teach using hashtags in any other way. Hence Liu does not teach a classifier dictionary as claimed contrary to the assertions of the Office Action.
Moreover, the purported motivation to combine Estes and Liu is deficient. The Office Action at pp. 7-8 alleges that "the techniques for automatic tag extraction from social media" from Liu would have been combined with "the systems and techniques for knowledge discovery" from Estes "to use the system of Liu to generate a dictionary containing all relevant tags, labels, etc. for searching." However, Estes and Liu are for completely different technologies. Namely, Estes describes learning conceptual associations between text for improving search results, while Liu is directed to labeling images from social media posts based on associated text from the posts. There is no reason to add the labeling of images in social media posts from Liu to the knowledge mapping via conceptual associations between text from Estes as these are unrelated functions. In any case, adding "techniques for automatic tag extraction" to tag images, even if added to Estes arguendo does not result in Estes, which is acknowledged to not use a classifier dictionary, having a classifier dictionary, let alone one that is, and is used, as claimed. Furthermore, the suggestion that one of ordinary skill in the art would have been motivated "to use the system of Liu to generate a dictionary" is based on impermissible hindsight, as evidenced by the fact that Estes does not use a classifier dictionary at all and Liu at least does not teach "generat[ing]" a classifier dictionary as claimed (instead it uses the fixed Stanford Named Entity Recognition labels) and therefore the motivation arises not from anything taught in the references but only by using the Applicant's disclosure as a blueprint. Interconnect Planning Corp. v. Feil, 774 F.2d 1132, 1143 (Fed. Cir. 1985). Hence, a person of ordinary skill in the art would not be motivated to combine Estes with Liu. For at least these reasons, instant independent claims 8 and 15 should therefore be allowed, along with their dependent claims. Such allowance is respectfully requested.
Examiner’s response:
Applicant’s arguments have been fully considered but are found to be not persuasive in light of the amendments made on the claims.
Applicant argues that Estes does not teach updating a classifier dictionary as required by the independent claims. Examiner disagrees. The instant application’s claims were examined under BRI and Estes was found to teach the limitations as filed. The content of Estes was found to be in line with what the Applicant has considered to be a classifier dictionary, and as such the excerpts of Estes are cited.
Applicant argues that Estes does not teach collecting feature data "based on a classifier dictionary comprising one or more predetermined classifiers, wherein the one or more predetermined classifiers comprise one or more hashtags” even though Estes was previously cited as teaching updating a classifier dictionary. Examiner disagrees. Estes does teach the action of updating a classifier dictionary, however the main concept of the aforementioned limitation is that the one or more predetermined classifiers comprise one or more hashtags. While Estes teaches the updating of a classifier dictionary, it does not teach the portion of the limitation regarding the one or more hashtags, which is why Liu was brought in to remedy that deficiency.
Applicant argues that the motivation to combine Estes and Liu is deficient since they are “unrelated technologies”. Examiner disagrees. Estes learns conceptual associations between text for improving search results while Liu extracts labels from social media posts. The primary purpose of Liu is to eventually label images from the extracted data, but the motivation to combine references would be to use the text/label extraction of Liu and combine that with the conceptual associations learning of Estes. In this manner, the extracted labels/text from Liu are then used with the dictionary of Estes to increase the amount of learning data that is available in the dictionary to better improve search results.
In light of the arguments presented and the amendments made on the claims, the rejections made under 35 U.S.C. 103 are maintained.
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.
Claim(s) 8-15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Estes (US 20160140236 A1) in view of Liu et al (S. Liu and T. Forss, "Automatic tag extraction from social media for visual labeling," 2015 7th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K), Lisbon, Portugal, 2015, pp. 504-510., hereinafter Liu).
Regarding Claim 8:
Estes teaches
A method for use in searching data sources for specific criteria and automatically updating machine learning training sets and classifier, the method comprising: storing a machine learning algorithm and the machine learning training sets, wherein the machine learning algorithm is configured to reference the machine learning training sets (Estes [0157]: “In addition to the passive personalization that the KDAS provides through observing the searching behavior of the user, the KDAS system supports user-defined pattern discovery. The end user is able to create a conceptual graph of a pattern or relationship and label the entire graph and its components. The KDAS system can then create an “artificial” pathway in concept space for that graph and trigger an alert whenever that particular pathway is activated”; (EN): the user defined graph is stored and can be updated at a later time and is analogous to a training set.);
a processor effectuating a recognition agent over one or more data sources (Estes [0015]: ” a conversational inference module for translating requests from the chatbot to the computer, the storage means, and the data agent or agents; and, one or more data extractors, connected to an existing third-party information system and operable for interfacing with this system in response to requests processed by the chatbot and the conversational inference module”);
a processor effectuating an extractor that collects feature data from the one or more data sources (Estes [0016]: “One or more network data agents may also be provided in the system, for connecting at least one preselected physician information system to at least one health care provider network, and for extracting data from the preselected physician information system and providing the data to the health care provider network”);
an input device receiving the specific criteria (Estes [0014]: “The system may further comprise means for collecting data input generated by interaction between the system and a human user, and means for learning from the input”);
a processor effectuating an updater that dynamically updates the machine learning training sets and the classifier dictionary based on the collected feature data from the one or more data sources and the specific criteria, enabling the machine learning algorithm to identify at least one feature data including the specific criteria from the collected feature data (Este [0090]: “The system updates the user understanding based on the interactions report”; [0093]: “The UA reports the survey data to the system to update the system's understanding about the user's preferences and intent”);
an output device transmitting the at least one feature data including the identified specific criteria (Estes [0132]: “The output layer would be the response from the system and is dealt with in the Agent Inference Module (AIM) detailed in the next section. The AIM also is what dynamically calculates the weights along the nodes based on the context”);
wherein the machine learning algorithm, the machine learning training sets, and the classifier dictionary, are continuously updated using the updater, based on a plurality of received specific criteria and corresponding feature data collected from the extractor to improve performance of the machine learning algorithm over time (Estes [0090]: “The system updates the user understanding based on the interactions report”; [0093]: “The UA reports the survey data to the system to update the system's understanding about the user's preferences and intent”; [0103]: “Thus, the PSA extracts key concept words e.g. nouns, adjectives, combinations, proper names, etc., that co-occur within a sentence, which are taken to be a semantic unit. The concepts and their concept links, i.e. concepts that co-occur with them in the same sentence, are catalogued and sorted into a concept network that represents that structure”).
Estes does not distinctly disclose
based on a classifier dictionary comprising one or more predetermined classifiers, wherein the one or more predetermined classifiers comprise one or more hashtags;
However, Liu teaches
based on a classifier dictionary comprising one or more predetermined classifiers, wherein the one or more predetermined classifiers comprise one or more hashtags (Liu [Section 3.1, p. 3]: “For named entity recognition we applied Stanford Named Entity Recognizer, which identifies names of people, places, organizations quite satisfactorily. Other types of proper nouns, e.g. names of products, books, magazines, movies, sports, other events and activities can often be identified in the hashtags or key phrase list”; (EN): according to Applicant’s specification the dictionary is simply a list; [0026]: “In an example embodiment, the dictionary is a list of the names of entities in which the user is interested, for example, a list of the names of companies or people or products that the user may be interested in analyzing”);
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, and having the teachings of Estes and Liu before him or her, to modify the systems and techniques for knowledge discovery of Estes to include the techniques for automatic tag extraction from social media as shown in Liu. The motivation for doing so would have been to use the system of Liu to generate a dictionary containing all relevant tags, labels, etc. for searching (Liu [Section 3.1, p. 3]: “For named entity recognition we applied Stanford Named Entity Recognizer, which identifies names of people, places, organizations quite satisfactorily. Other types of proper nouns, e.g. names of products, books, magazines, movies, sports, other events and activities can often be identified in the hashtags or key phrase list”).
Regarding Claim 9:
Estes teaches
The method of claim 8, wherein the effectuating over the one or more data sources comprises the recognition agent denoting analytics in the data to aggregate feature data (Estes [0017]: “The system also may be utilized in government and security applications such as processing defense intelligence information, as by clustering and visualizing data which can be provided to a human analyst of defense intelligence as a supplemental tool”).
Regarding Claim 10:
Estes teaches
The method of claim 9, further comprising utilizing the machine learning algorithm to search the aggregated feature data for the at least one feature data that is associated with the predetermined classifiers and the specific criteria (Estes [0114]: “as sufficient features are detected, whole concepts are activated and other “invisible” features are inferred while at the same time, counteracting this rapid expansion in state space, relationships “target” the expansion and modifiers limit the range within particular feature-space dimensions”).
Regarding Claim 11:
Estes teaches
The method of claim 8, wherein the predetermined classifiers are associated with specific criteria types (Estes [0143]: “Within a given configuration, individual agents can execute on dedicated servers optimized for their particular agents”).
Regarding Claim 12:
Estes teaches
The method of claim 11, wherein the specific criteria types are at least one of medical, scientific, legal or news related information, and at least one feature is at least one of an image, text, video, and audio (Estes [0189]: “In addition, numerous data extractors are used to interface with proprietary medical systems”; [0061]: “First, there is the pattern input (the feature detection level of the system) where the stimuli enter into the system through a process of “discretization.” Discretization takes a given perceptual chuck—which at the data level could be a sentence in a text document (unstructured source) or a table in a database (structured source)”).
Regarding Claim 13:
Estes teaches
The method of claim 11, wherein the predetermined classifiers comprise named entities (Estes [0054]: “DML is essentially a hybrid data structure containing six primitive element types: Beliefs 10, Desires 12, Intentions 14, Methods 16, Values 18, and History 20”).
Regarding Claim 14:
Estes teaches
The method of claim 8, wherein the one or more data sources are at least one of social media post feeds, text documents, tabular content or image repositories and the machine learning algorithm is a deep learning algorithm (Estes [0061]: “First, there is the pattern input (the feature detection level of the system) where the stimuli enter into the system through a process of “discretization.” Discretization takes a given perceptual chuck—which at the data level could be a sentence in a text document (unstructured source) or a table in a database (structured source)”; [0065]: “A dynamic conceptual network (DCN) is not just learned, but obtains its residual structure from the second type of feature map discussed earlier”).
Regarding Claim 15:
Estes teaches
A method for searching for information related to posts in a plurality of data sources, the method comprises: extracting at least one feature associated with one or more classifiers from the plurality of data sources (Estes [0061]: “First, there is the pattern input (the feature detection level of the system) where the stimuli enter into the system through a process of “discretization.” Discretization takes a given perceptual chuck—which at the data level could be a sentence in a text document (unstructured source) or a table in a database (structured source)”; [0054]: “DML is essentially a hybrid data structure containing six primitive element types: Beliefs 10, Desires 12, Intentions 14, Methods 16, Values 18, and History 20”);
dynamically updating training sets for a machine learning algorithm and identifying specific criteria from the at least one feature utilizing the machine learning algorithm and the training sets (Estes [0090]: “The system updates the user understanding based on the interactions report”; [0093]: “The UA reports the survey data to the system to update the system's understanding about the user's preferences and intent”);
aggregating the information from data sources of the plurality of data sources containing the identified specific criteria, providing the aggregated information for display in a searchable format, and updating, periodically, the training sets, the classifier dictionary, and the machine learning algorithm by repeatedly performing (i) the step of extracting to extract a plurality of features and (ii) the step of dynamically updating the training sets based on the plurality of features (Estes [0022]: “providing at least one network data agent, operable for connecting at least one preselected physician information system to at least one health care provider network, and further operable for extracting data from the at least one preselected physician information system and providing the data to the at least one health care provider network”; [0090]: “The system updates the user understanding based on the interactions report”; [0093]: “The UA reports the survey data to the system to update the system's understanding about the user's preferences and intent”)
Estes does not distinctly disclose
the one or more classifiers being identified from a classifier dictionary, wherein the one or more classifiers comprise one or more hashtags
However, Liu teaches
the one or more classifiers being identified from a classifier dictionary, wherein the one or more classifiers comprise one or more hashtags (Liu [Section 3.1, p. 3]: “For named entity recognition we applied Stanford Named Entity Recognizer, which identifies names of people, places, organizations quite satisfactorily. Other types of proper nouns, e.g. names of products, books, magazines, movies, sports, other events and activities can often be identified in the hashtags or key phrase list”; (EN): according to Applicant’s specification the dictionary is simply a list; [0026]: “In an example embodiment, the dictionary is a list of the names of entities in which the user is interested, for example, a list of the names of companies or people or products that the user may be interested in analyzing”);
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, and having the teachings of Estes and Liu before him or her, to modify the systems and techniques for knowledge discovery of Estes to include the techniques for automatic tag extraction from social media as shown in Liu. The motivation for doing so would have been to use the system of Liu to generate a dictionary containing all relevant tags, labels, etc. for searching (Liu [Section 3.1, p. 3]: “For named entity recognition we applied Stanford Named Entity Recognizer, which identifies names of people, places, organizations quite satisfactorily. Other types of proper nouns, e.g. names of products, books, magazines, movies, sports, other events and activities can often be identified in the hashtags or key phrase list”).
Regarding Claim 16:
Estes does not distinctly disclose
The method of claim 15, wherein the plurality of data sources comprise posts comprising hashtags.
However, Liu teaches
The method of claim 15, wherein the plurality of data sources comprise posts comprising hashtags (Liu [Section 3.1, p. 3]: “For named entity recognition we applied Stanford Named Entity Recognizer, which identifies names of people, places, organizations quite satisfactorily. Other types of proper nouns, e.g. names of products, books, magazines, movies, sports, other events and activities can often be identified in the hashtags or key phrase list”);
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, and having the teachings of Estes and Liu before him or her, to modify the systems and techniques for knowledge discovery of Estes to include the techniques for automatic tag extraction from social media as shown in Liu. The motivation for doing so would have been to use the system of Liu to generate a dictionary containing all relevant tags, labels, etc. for searching (Liu [Section 3.1, p. 3]: “For named entity recognition we applied Stanford Named Entity Recognizer, which identifies names of people, places, organizations quite satisfactorily. Other types of proper nouns, e.g. names of products, books, magazines, movies, sports, other events and activities can often be identified in the hashtags or key phrase list”).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 20180260680 A1 – Intelligent assistant devices and methods for interacting with a user
US 20170061294 A1 – Predicting Labels Using A Deep-Learning Model
US 20130332460 A1 – Information is obtained from the Internet is combined with processed and rated information from social networking services by a structured and social data aggregator providing highly relevant search results
Any inquiry concerning this communication or earlier communications from the examiner should be directed to COREY M SACKALOSKY whose telephone number is (703)756-1590. The examiner can normally be reached M-F 7:30am-3:30pm EST.
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, Omar Fernandez Rivas can be reached at (571) 272-2589. 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.
/COREY M SACKALOSKY/Examiner, Art Unit 2128
/OMAR F FERNANDEZ RIVAS/Supervisory Patent Examiner, Art Unit 2128