DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of Claims
Claims 1-13 have been cancelled in Preliminary Amendment dated 4/25/2024. New claims 14-26 are currently pending examination.
Information Disclosure Statement
The Information Disclosure Statement (IDS) submitted by Applicant on 4/11/2024 has been considered.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefore, subject to the conditions and requirements of this title.
Claims 14-26 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., abstract idea) without significantly more.
Regarding claim 14,
Step 1: Claim 14 is directed to a method.
Step 2A, Prong 1: Claim 14 recites the limitations of:
determining a first set of embeddings of parts of the character string; (i.e., a person can mentally and/or with the aid of pen and paper determine set of embeddings corresponding to a character string – e.g. mapping a character string to a vector of numbers)
determining a second set of embeddings of parts of the character string; (i.e., a person can mentally and/or with the aid of pen and paper determine set of embeddings corresponding to a character string – e.g. mapping a character string to a vector of numbers)
determining the fact for the knowledge base, depending on variables for predicting the fact, wherein, for mutually corresponding embeddings from the first and second sets, one variable of the variables for predicting the fact is determined in each case. (i.e., a person can mentally and/or with the aid of pen and paper determine a fact for a knowledgebase, as claimed)
Hence, the claim recites an abstract idea.
Step 2A, Prong 2: Claim 14 recites the additional elements of “a computer-implemented method for machine learning of a fact for populating a knowledge base, the method comprising the following steps:”. This additional element is recited at a high level of generality such that it amounts to no more than mere instructions to apply the exception using generic computer components. (See MPEP 2106.05(f)). Furthermore, the additional element of “providing a character string;” is considered insignificant extra-solution activity. (See MPEP 2106.05(g)) Hence the claim does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 14 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract idea into a practical application, the additional element of “a computer-implemented method for machine learning of a fact for populating a knowledge base, the method comprising the following steps:” is recited at a high level of generality such that it amounts to no more than mere instructions to apply the exception using generic computer components. (See MPEP 2106.05(f)). Furthermore, the additional element of “providing a character string;” is considered insignificant extra-solution activity consisting of mere data transmission. (See MPEP 2106.05(g)) As such, it is re-evaluated under Step 2B to determine if it is what the courts have considered to be well-understood, routine, and conventional activity in the field. The court decisions cited in MPEP 2106.05(d)(II) have held that mere data transmission over a network (as it is claimed in the present claim) is considered well-understood, routine, and conventional activity in the field supported by Berkheimer. (see also, i. Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information) Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 15,
Step 2A, Prong 1: Claim 15 recites an abstract idea as inherited from claim 14. Claim 15 further recites the following limitation:
wherein the variables are determined independently of an embedding from the first set for which the second set does not include a corresponding embedding. (i.e., a person can mentally and/or with the aid of pen and paper determine a variable independently of an embedding from the first set for which the second set does not include a corresponding embedding)
Hence, the claim further recites an abstract idea.
Step 2A, Prong 2: Claim 15 does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 15 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 16,
Step 2A, Prong 1: Claim 16 recites an abstract idea as inherited from claim 14. Claim 16 further recites the following limitation:
wherein, for an embedding from the first set for which the second set does not include a corresponding embedding, an embedding is provided independently of the character string and one of the variables for predicting the fact is determined depending on the embedding from the first set and the embedding provided independently of the character string. (i.e., a person can mentally and/or with the aid of pen and paper determine on one of the variables for predicting the fact depending on the embedding from the first set and the embedding provided independently of the character string.)
Hence, the claim further recites an abstract idea.
Step 2A, Prong 2: Claim 16 does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 16 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 17,
Step 2A, Prong 1: Claim 17 recites an abstract idea as inherited from claim 14. Claim 17 further recites the following limitation:
wherein, for a first embedding from the first set for which the second set does not include a corresponding embedding a second embedding from the first set for which the second set includes a corresponding embedding is determined, and one of the variables for predicting the fact is determined depending on the first embedding and the second embedding and the embedding from the second set that corresponds to the second embedding (i.e., a person can mentally and/or with the aid of pen and paper determine a second embedding from the first set for which the second set includes a corresponding embedding and also determine one of the variables for predicting the fact depending on the first embedding and the second embedding, as claimed.)
Hence, the claim further recites an abstract idea.
Step 2A, Prong 2: Claim 17 does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 17 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 18,
Step 2A, Prong 1: Claim 18 recites an abstract idea as inherited from claims 14 and 17. Claim 18 further recites the following limitation:
wherein the variable for predicting the fact depending on the first embedding and the second embedding and the embedding corresponding to the second embedding is determined depending on an average of the first embedding and the second embedding. (i.e., a person can mentally and/or with the aid of pen and paper determine the variable for predicting the fact depending on an average of the first embedding and the second embedding)
Hence, the claim further recites an abstract idea.
Step 2A, Prong 2: Claim 18 does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 18 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 19,
Step 2A, Prong 1: Claim 19 recites an abstract idea as inherited from claim 14. Claim 19 further recites the following limitation:
wherein mutually corresponding embeddings of the first and second sets are determined depending on their position in an order of the parts of the character string. (i.e., person can mentally and/or with the aid of pen and paper determine mutually corresponding embeddings of the first and second sets depending on their position in an order of the parts of the character string.)
Hence, the claim further recites an abstract idea.
Step 2A, Prong 2: Claim 19 does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 19 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 20,
Step 2A, Prong 1: Claim 20 recites an abstract idea as inherited from claim 14.
Claim 20 further recites the following limitation:
wherein mutually corresponding embeddings are determined depending on common characters at a beginning of a relevant part of the character string. (i.e., a person can mentally and/or with the aid of pen and paper determine mutually corresponding embeddings depending on common characters at a beginning of a relevant part of the character string)
Hence, the claim further recites an abstract idea.
Step 2A, Prong 2: Claim 20 does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 20 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 21,
Step 2A, Prong 1: Claim 21 recites an abstract idea as inherited from claim 14. Claim 21 further recites the following limitation:
wherein the character string is divided into a first list of parts, wherein the parts in the first list are uniquely assigned to the embeddings from the first set, wherein the character string is divided into a second list of parts, wherein the parts in the second list are uniquely assigned to the embeddings from the second set, wherein mutually corresponding parts in the first and second lists are defined depending on an order in which the parts are arranged in the first and second lists, and wherein the corresponding embeddings are determined depending on the mutually corresponding parts in the first and second lists. (i.e., a person can mentally and/or with the aid of pen and paper divide a character string into a first list of parts and uniquely assign the first list to the embeddings from the fist set. Furthermore, a person can mentally and/or with the aid of pen and paper divide the character string into a second list of parts and uniquely assign the second list to the embeddings from the second set. A person can also mentally and/or with the aid of pen and paper define mutually corresponding parts in the first and second list depending on an order in with the parts are arranged in the first and second lists. A person can also mentally and/or with the aid of pen and paper determine the corresponding embeddings depending on mutually corresponding parts in the first and second lists)
Hence, the claim further recites an abstract idea.
Step 2A, Prong 2: Claim 21 does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 21 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 22,
Step 2A, Prong 1: Claim 22 recites an abstract idea as inherited from claim 14. Claim 22 further recites the following limitation:
wherein the first set of embeddings is determined depending on a first vocabulary, wherein the second set of embeddings is determined depending on a second vocabulary different from the first vocabulary. (i.e., a person can mentally and/or with the aid of pen and paper determine a first set of embeddings depending on a first vocabulary and a second set of embeddings depending on a second vocabulary)
Hence, the claim further recites an abstract idea.
Step 2A, Prong 2: Claim 22 does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 22 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 23,
Step 2A, Prong 1: Claim 23 recites an abstract idea as inherited from claim 14. Claim 23 further recites the following limitation:
wherein the character string is broken down into the parts of the character string by: certain characters of the character string and/or rules for word division and/or syllable division and/or letter division. (i.e., a person can mentally, and/or with the aid of pen and paper break down the character string into parts by certain characters of the character string and/or rules for word division and/or syllable division and/or letter division)
Hence, the claim further recites an abstract idea.
Step 2A, Prong 2: Claim 23 does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 23 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 24,
Step 2A, Prong 1: Claim 24 recites an abstract idea as inherited from claim 14. Claim 24 further recites the following limitation:
wherein mutually corresponding embeddings from the first and second sets are concatenated, or added, or multiplied, to determine the one variable. (i.e., a person can mentally and/or with the aid of pen and paper concatenate, or add, or multiply mutually corresponding embeddings from the first and second sets, to determine the one variable)
Step 2A, Prong 2: Claim 24 does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 24 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 25,
Step 1: Claim 25 is directed to a device.
Step 2A, Prong 1: Claim 25 recites the same and/or analogous limitations as claim 14. Therefore, it is rejected under the same rationale as claim 14.
Step 2A, Prong 2: Claim 25 further recites the additional elements of “a device for machine learning of facts for populating a knowledge base, the device comprising:…” and “at least one processor; and at least one memory, wherein the at least one processor is configured to execute instructions machine learning of a fact for populating a knowledge base, the instructions, when executed by the at least one processor, cause the at least one processor to perform the following steps:…”. These additional elements are recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using generic computer components. (See MPEP 2106.05(f)). Furthermore, the claim recites the additional element of “wherein the at least one memory is configured to store the instructions.” This additional element is considered insignificant extra-solution activity. (See MPEP 2106.05(g)). Hence the claim does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 25 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract idea into a practical application, the additional elements of “a device for machine learning of facts for populating a knowledge base, the device comprising:…” and “at least one processor; and at least one memory, wherein the at least one processor is configured to execute instructions machine learning of a fact for populating a knowledge base, the instructions, when executed by the at least one processor, cause the at least one processor to perform the following steps:…” are recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using generic computer components. (See MPEP 2106.05(f)). Furthermore, the additional element of “wherein the at least one memory is configured to store the instructions.” was determined to be insignificant extra-solution activity consisting of mere data storage in memory. (See MPEP 2106.05(g)) As such, this additional element is further analyzed under Step 2B to see if it is more than what the courts have considered as well-understood, routine, and conventional activity in the field. The court decisions cited in MPEP 2106.05(d)(II) have held that mere data storage in memory (as it is presently claimed) is considered well-understood, routine, and conventional activity in the field supported by Berkheimer. (see also, iv. Storing and retrieving information in memory, Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93;) Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 26,
Step 1: Claim 26 is directed to a non-transitory machine-reachable medium.
Step 2A, Prong 1: Claim 26 recites the same and/or analogous limitations as claim 14. Therefore, it is rejected under the same rationale as claim 14.
Step 2A, Prong 2: Claim 26 recites the additional elements of “a non-transitory machine-reachable medium on which stored a program for machine learning of facts for populating a knowledge base, the program including instructions which, when executed by at least one processor, cause the at least one processor to perform the following steps…”. These additional elements are recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using generic computer components. (See MPEP 2106.05(f)) Hence the claim does not recite additional elements that integrate the judicial exception into a practical application. Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
Step 2B: Claim 26 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the integration of the abstract idea into a practical application, the additional elements of “a non-transitory machine-reachable medium on which stored a program for machine learning of facts for populating a knowledge base, the program including instructions which, when executed by at least one processor, cause the at least one processor to perform the following steps…” are recited at a high-level of generality such that they amount to no more than mere instructions to apply the exception using generic computer components. (See MPEP 2106.05(f)) Hence the claim lacks limitations which amount to significantly more than the judicial exception or an inventive concept, and is rejected. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim(s) 14, 15, 16, 17, 19, 23, 24, 25, and 26 are rejected under 35 U.S.C. 103 as being unpatentable over Lee et al. (US 20180336183 A1, filed May 22, 2017 and published Nov. 22, 2018) in view of Glass et al. (US 20210004672 A1, filed Jul. 3, 2019 and published Jan. 7, 2021)
Regarding claim 14, Lee teaches a computer-implemented method for machine learning of a fact for populating a knowledge base, the method comprising the following steps:
providing a character string (Lee, [0004] teaches the method comprises receiving, by the natural language embedding engine executing on the processor, a document data object of an electronic document, and analyzing, by the natural language embedding engine, a structure of the electronic document to identify one or more structural document elements that have a relationship with the document data object.; Lee, [0037] teaches it should be appreciated that within the present description, the following understanding of various terms is used throughout: [0038] 1. Contiguous dependencies: Given a sequence of words or sentences [ . . . , s(t−1), s(t), s(t+1), . . . ] in a sentence or a document, a set [s(t−d)|d=−p, . . . , 0, . . . , q, and p>0, q>0] comprises a set of words or sentences having contiguous dependencies; [0039] 2. Non-contiguous dependencies: the complement of contiguous dependencies; [0040] 3. Extra-document dependencies: dependencies with content existing outside of the document in which the sentence being embedded exists; [0041] 4. Intra-sentence dependency: those dependencies among words or characters within a sentence; [0042] 5. Extra-sentence dependency: those dependencies among sentences; and [0043] 6. Long distance dependences (or “long dependencies”): dependencies with content and/or structure that includes non-contiguous extra-sentence dependencies and/or extra-document dependencies.)
determining a first set of embeddings of parts of the character string; Lee, Abstract and [0004]-[0005], teaches mechanisms are provided to perform embedding of content of a natural language document. The mechanisms receive a document data object of an electronic document and analyze a structure of the electronic document to identify one or more structural document elements that have a relationship with the document data object. A dependency data structure is generated, representing the electronic document, where edges define relationships between document elements and at least one edge represents at least one relationship between the one or more structural document elements and the document data object. The mechanisms embed the document data object based on the at least one relationship to thereby represent the document data object as a vector data structure. The mechanisms perform natural language processing on the portion of natural language content based on the vector data structure. The one or more structural document elements are non-local non-contiguous with the document data object.; Lee, [0017] teaches The illustrative embodiments provide mechanisms for performing embedding on natural language content based on identified non-contiguous long distanced semantic dependencies and also optionally based on contiguous dependencies.; Lee, [0018] teaches Embedding is a technology that represents words, sentences, or other portions of textual content into a numerical representation, such as a vector representation that machine learning algorithms are able to utilize. Embedding maps words or phrases from a vocabulary to a corresponding vector of real numbers which provides the advantages of improved dimensionality reduction and contextual similarity when compared to other approaches, such as the “bag or words” approach. Lee, [0037] It should be appreciated that within the present description, the following understanding of various terms is used throughout: [0038] 1. Contiguous dependencies: Given a sequence of words or sentences [ . . . , s(t−1), s(t), s(t+1), . . . ] in a sentence or a document, a set [s(t−d)|d=−p, . . . , 0, . . . , q, and p>0, q>0] comprises a set of words or sentences having contiguous dependencies; [0039] 2. Non-contiguous dependencies: the complement of contiguous dependencies; Lee, [0044] The illustrative embodiments provide mechanisms for performing embedding of unstructured data, such as text and images, based on non-contiguous, long distanced, contexts and their dependencies extracted from the document structure and inter-sentence structure. The embedding mechanisms incorporate semantical relevant contexts to produce more accurate semantic representations and improve the performance of natural language processing (NLP) algorithms, such as coreference resolution, paraphrase identification, word analogy test, etc., which may in turn be used by cognitive systems to perform cognitive operations. It should be appreciated that the term “embedding” is utilized in the industry, and herein, to refer to both the method of representing a non-numerical data object as one or more numerical values, such as a vector number representation, and also the results of such a method. Thus, an “embedding” may be the resulting numerical representation of the non-numerical data object generated by an “embedding” method; Lee, [0046] As a further aspect of the illustrative embodiments, the mechanisms of the illustrative embodiments may incorporate external knowledge sources, such as lexicons, knowledge bases, hyperlinks, etc., to assist with providing contextual information for assisting with the embedding process. For example, the mechanisms of the illustrative embodiments may discover the embedding of a natural language text object, e.g., a word, using other words appearing in the definition of the word or lexicon. Similarly, if a knowledge base contains a word or an entity, the mechanisms of the illustrative embodiments may include the entries and relationships in the knowledge base entry in the actual embedding. Hyperlinks, citations, or other references found in a document may be used to extend the context of the natural language text data object being embedded by looking at the natural language content of the target of the hyperlink, i.e. the other document to which the present document is linked by the hyperlink. These knowledge sources provide deeper semantic information about the natural language text data object being embedded, referred to hereafter as the “target object” which can be any portion of natural language text such as a word, a phrase, a sentence, or any other amount of natural language textual content, than merely looking at the contiguous local context of the target object.; Lee, [0068] further teaches As will be discussed in greater detail hereafter, the illustrative embodiments may be integrated in, augment, and extend the functionality of these QA pipeline, or request processing pipeline, mechanisms of a cognitive system with regard to performing aspects of natural language processing in which deep embedding of natural language text data objects, e.g., words, phrases, sentences, or other portions of text, are performed. In some illustrative embodiments, additional embedding of non-textual data objects of electronic documents may also be performed, as discussed hereafter. The natural language processing and deep embedding may be performed when ingesting an electronic document from a corpus for use by the cognitive system when performing its cognitive operations.; Lee [0115] further teaches firstly, for each sentence y (or document element), there is a set of edges {(x.sub.1, label.sub.1, y), . . . , (x.sub.n, label.sub.n, y)}, where x.sub.i is a governing sentence (or document element) and label.sub.i is a dependency label for the i-th dependency edge, forming dependency data structure 143 from dependency annotation logic 142, e.g., (x.sub.i, title-sentence, y) meaning that sentence x.sub.i is the title for the sentence y. Suppose X is a set of dependent sentences {x.sub.1, . . . , x.sub.n} that are dependent on sentence y. The numerical representations, e.g., vectors generated by the embedding logic 144, of the governing sentences X (or document elements) are optimized to predict the incoming edges to y (document elements X generating this document element y), or in some embodiments the outgoing edges, including both edge label (e.g., direct object, title-sentence, etc.), and representation (e.g., vector) of the edges. For example, for a sentence x, the sentence has outgoing edges of the type (x, label, y) where x is the governing sentence and y is the dependent sentence, and has incoming edges of the type (y, label x) where y is the governing sentence, and x becomes the dependent sentence.)
determining a second set of embeddings of parts of the character string; (Lee, Abstract and [0004]-[0005], teaches mechanisms are provided to perform embedding of content of a natural language document. The mechanisms receive a document data object of an electronic document and analyze a structure of the electronic document to identify one or more structural document elements that have a relationship with the document data object. A dependency data structure is generated, representing the electronic document, where edges define relationships between document elements and at least one edge represents at least one relationship between the one or more structural document elements and the document data object. The mechanisms embed the document data object based on the at least one relationship to thereby represent the document data object as a vector data structure. The mechanisms perform natural language processing on the portion of natural language content based on the vector data structure. The one or more structural document elements are non-local non-contiguous with the document data object.; Lee, [0017] teaches The illustrative embodiments provide mechanisms for performing embedding on natural language content based on identified non-contiguous long distanced semantic dependencies and also optionally based on contiguous dependencies.; Lee, [0018] teaches Embedding is a technology that represents words, sentences, or other portions of textual content into a numerical representation, such as a vector representation that machine learning algorithms are able to utilize. Embedding maps words or phrases from a vocabulary to a corresponding vector of real numbers which provides the advantages of improved dimensionality reduction and contextual similarity when compared to other approaches, such as the “bag or words” approach. Lee, [0037] It should be appreciated that within the present description, the following understanding of various terms is used throughout: [0038] 1. Contiguous dependencies: Given a sequence of words or sentences [ . . . , s(t−1), s(t), s(t+1), . . . ] in a sentence or a document, a set [s(t−d)|d=−p, . . . , 0, . . . , q, and p>0, q>0] comprises a set of words or sentences having contiguous dependencies; [0039] 2. Non-contiguous dependencies: the complement of contiguous dependencies; Lee, [0044] The illustrative embodiments provide mechanisms for performing embedding of unstructured data, such as text and images, based on non-contiguous, long distanced, contexts and their dependencies extracted from the document structure and inter-sentence structure. The embedding mechanisms incorporate semantical relevant contexts to produce more accurate semantic representations and improve the performance of natural language processing (NLP) algorithms, such as coreference resolution, paraphrase identification, word analogy test, etc., which may in turn be used by cognitive systems to perform cognitive operations. It should be appreciated that the term “embedding” is utilized in the industry, and herein, to refer to both the method of representing a non-numerical data object as one or more numerical values, such as a vector number representation, and also the results of such a method. Thus, an “embedding” may be the resulting numerical representation of the non-numerical data object generated by an “embedding” method.; Lee, [0046] As a further aspect of the illustrative embodiments, the mechanisms of the illustrative embodiments may incorporate external knowledge sources, such as lexicons, knowledge bases, hyperlinks, etc., to assist with providing contextual information for assisting with the embedding process. For example, the mechanisms of the illustrative embodiments may discover the embedding of a natural language text object, e.g., a word, using other words appearing in the definition of the word or lexicon. Similarly, if a knowledge base contains a word or an entity, the mechanisms of the illustrative embodiments may include the entries and relationships in the knowledge base entry in the actual embedding. Hyperlinks, citations, or other references found in a document may be used to extend the context of the natural language text data object being embedded by looking at the natural language content of the target of the hyperlink, i.e. the other document to which the present document is linked by the hyperlink. These knowledge sources provide deeper semantic information about the natural language text data object being embedded, referred to hereafter as the “target object” which can be any portion of natural language text such as a word, a phrase, a sentence, or any other amount of natural language textual content, than merely looking at the contiguous local context of the target object.; Lee, [0068] further teaches As will be discussed in greater detail hereafter, the illustrative embodiments may be integrated in, augment, and extend the functionality of these QA pipeline, or request processing pipeline, mechanisms of a cognitive system with regard to performing aspects of natural language processing in which deep embedding of natural language text data objects, e.g., words, phrases, sentences, or other portions of text, are performed. In some illustrative embodiments, additional embedding of non-textual data objects of electronic documents may also be performed, as discussed hereafter. The natural language processing and deep embedding may be performed when ingesting an electronic document from a corpus for use by the cognitive system when performing its cognitive operations.; Lee [0115] further teaches firstly, for each sentence y (or document element), there is a set of edges {(x.sub.1, label.sub.1, y), . . . , (x.sub.n, label.sub.n, y)}, where x.sub.i is a governing sentence (or document element) and label.sub.i is a dependency label for the i-th dependency edge, forming dependency data structure 143 from dependency annotation logic 142, e.g., (x.sub.i, title-sentence, y) meaning that sentence x.sub.i is the title for the sentence y. Suppose X is a set of dependent sentences {x.sub.1, . . . , x.sub.n} that are dependent on sentence y. The numerical representations, e.g., vectors generated by the embedding logic 144, of the governing sentences X (or document elements) are optimized to predict the incoming edges to y (document elements X generating this document element y), or in some embodiments the outgoing edges, including both edge label (e.g., direct object, title-sentence, etc.), and representation (e.g., vector) of the edges. For example, for a sentence x, the sentence has outgoing edges of the type (x, label, y) where x is the governing sentence and y is the dependent sentence, and has incoming edges of the type (y, label x) where y is the governing sentence, and x becomes the dependent sentence.))
However, Lee does not distinctly disclose determining the fact for the knowledge base, depending on variables for predicting the fact, wherein, for mutually corresponding embeddings from the first and second sets, one variable of the variables for predicting the fact is determined in each case.
Nevertheless, Glass teaches determining the fact for the knowledge base, depending on variables for predicting the fact, wherein, for mutually corresponding embeddings from the first and second sets, one variable of the variables for predicting the fact is determined in each case (Glass, [0028] teaches to populate the knowledge graph 130 [i.e., understood as to populate the knowledge base] with data from the corpus 140, the processor 120 can identify a set of entities 155 from the corpus 140. The set of entities 155 from the corpus 140 can include or exclude entities that were among the contexts identified for training the model 150. The processor 120 can apply the model 150 on the set of entities 155, and output a predicted unary relation for each one of the entities 155. The predicted unary relations outputted from the model 150 can be in the format (f, R;X), where f denotes a filler argument (e.g., a variable), R denotes a fixed relation (e.g., a constant relation), and X denotes a fixed argument (e.g., a constant). For example, the model 150 can output a predicted unary relation 160 with an associated entity among entities 155, such as (E20, r1:E3), where E20 can be an entity that is absent from the knowledge graph 130. The processor 120 can determine that the entity E20 is absent from the knowledge graph 130 and, in response, can unpack or convert the output 160 (E20, r1:E3) into a triple 170, with the format (f, R, X), such as (E20, r1, E3). The processor 120 can add the triple (E20, r1, E3) into the knowledge graph 130 to provide a new piece of information to the knowledge graph 130. Using a similar approach, the processor 120 can add other new entities and new relations associated with existing entities (in the knowledge graph 130) to the knowledge graph 130.; Glass, [0004] teaches In an example, a knowledge graph can be populated from corpus such as documents and text files by extracting relations between entities with respect to a given schema, such as types of the entities and relations between the entities.; Glass [0016] teaches Knowledge graphs can be used and applied in, for example, natural language processing, data mining, semantic analysis, and machine learning applications. As the amount of data, including nodes and edges (or entities and relations), in the knowledge graph increases, an accuracy of the model can also increase. In an example, an expansion of a knowledge graph using a corpus (e.g., document, text, audio files) can include training an extraction model based on data in the knowledge graph, implementing the extraction model to identify and extract new data (e.g., new concepts, entities, and associated relations) in the corpus, and adding new nodes and edges associated with the new data into the knowledge graph.; Glass, [0020] further teaches The knowledge graph 130 can be stored as a network to represent a knowledge base including structured and unstructured information.; Glass, [0032] teaches the processor 120 can be configured to augment the context 210 to indicate positions of objects, such as words, among the context 210 relative to a focus entity. For example, the processor 120 can be configured to perform lookup functions in a word embedding table 215 to identify a first vector, such as a context vector, where the word embedding table can be based on word embeddings such as word2vec. The word embedding table can be updated during training of the model 150. Further, the processor 120 can be configured to perform lookup functions in a position embedding table 225 to identify a second vector, such as a position vector 220, indicating positions of each object relative to the focus entity of the context 210. The processor 120 can augment the context 210 with the results from the lookup functions. For example, the processor 120 can assign position values to objects among the first vector representing the context 210. In the example shown in FIG. 2, the focus entity E200 is at position 0, and other objects among the context 210 are at different position values relative to the entity E200 at position 0. For example, the entity 202 is being augmented with a position value of 5 to indicate that the entity 202 is five positions away from the entity 200 within the context 210.)
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to have modified the method and system for deep embedding for natural language content, as taught by Lee, with the methods and systems for populating knowledge graphs, as taught by Glass. As the amount of data, including nodes and edges (or entities and relations), in the knowledge graph increases, an accuracy of the model can also increase. (Glass, [0016])
Regarding claim 15, the combination of Lee in view of Glass teaches all of the limitations of claim 14, however, and the combination further teaches wherein the variables are determined independently of an embedding from the first set for which the second set does not include a corresponding embedding (Glass, [0028] teaches to populate the knowledge graph 130 [i.e., understood as to populate the knowledge base] with data from the corpus 140, the processor 120 can identify a set of entities 155 from the corpus 140. The set of entities 155 from the corpus 140 can include or exclude entities that were among the contexts identified for training the model 150. The processor 120 can apply the model 150 on the set of entities 155, and output a predicted unary relation for each one of the entities 155. The predicted unary relations outputted from the model 150 can be in the format (f, R;X), where f denotes a filler argument (e.g., a variable), R denotes a fixed relation (e.g., a constant relation), and X denotes a fixed argument (e.g., a constant). For example, the model 150 can output a predicted unary relation 160 with an associated entity among entities 155, such as (E20, r1:E3), where E20 can be an entity that is absent from the knowledge graph 130. The processor 120 can determine that the entity E20 is absent from the knowledge graph 130 and, in response, can unpack or convert the output 160 (E20, r1:E3) into a triple 170, with the format (f, R, X), such as (E20, r1, E3). The processor 120 can add the triple (E20, r1, E3) into the knowledge graph 130 to provide a new piece of information to the knowledge graph 130. Using a similar approach, the processor 120 can add other new entities and new relations associated with existing entities (in the knowledge graph 130) to the knowledge graph 130).
Motivation to combine same as stated for claim 14.
Regarding claim 16, the combination of Lee in view of Glass teaches all of the limitations of claim 14, and the combination further teaches wherein, for an embedding from the first set for which the second set does not include a corresponding embedding, an embedding is provided independently of the character string, and one of the variables for predicting the fact is determined depending on the embedding from the first set and the embedding provided independently of the character string (Glass, [0028] teaches to populate the knowledge graph 130 [i.e., understood as to populate the knowledge base] with data from the corpus 140, the processor 120 can identify a set of entities 155 from the corpus 140. The set of entities 155 from the corpus 140 can include or exclude entities that were among the contexts identified for training the model 150. The processor 120 can apply the model 150 on the set of entities 155, and output a predicted unary relation for each one of the entities 155. The predicted unary relations outputted from the model 150 can be in the format (f, R;X), where f denotes a filler argument (e.g., a variable), R denotes a fixed relation (e.g., a constant relation), and X denotes a fixed argument (e.g., a constant). For example, the model 150 can output a predicted unary relation 160 with an associated entity among entities 155, such as (E20, r1:E3), where E20 can be an entity that is absent from the knowledge graph 130. The processor 120 can determine that the entity E20 is absent from the knowledge graph 130 and, in response, can unpack or convert the output 160 (E20, r1:E3) into a triple 170, with the format (f, R, X), such as (E20, r1, E3). The processor 120 can add the triple (E20, r1, E3) into the knowledge graph 130 to provide a new piece of information to the knowledge graph 130. Using a similar approach, the processor 120 can add other new entities and new relations associated with existing entities (in the knowledge graph 130) to the knowledge graph 130).
Motivation to combine same as stated for claim 14.
Regarding claim 17, the combination of Lee in view of Glass teaches all of the limitations of claim 14, and the combination further teaches wherein, for a first embedding from the first set for which the second set does not include a corresponding embedding, a second embedding from the first set for which the second set includes a corresponding embedding is determined, and one of the variables for predicting the fact is determined depending on the first embedding and the second embedding and the embedding from the second set that corresponds to the second embedding (Lee, [0122] teaches Also, since the trained neural network 145 can be used to generate a sentence given its dependencies, the application interface logic 146 can use this generation to identify missing document elements or coreference mentions (e.g., it, the program, etc.) in a sentence. For example, the illustrative embodiments may fill a missing document element in a target sentence by feeding those elements, on which the target sentence depends, to the encoders of the neural network 145, and only the known part of the target sentence into the decoder of the neural network 145. In particular, each cell of a recurrent neural network 145 that can be used to implement the decoder, takes the output of the previous cell to consider it. Instead of using the output of the previous cell, in accordance with one illustrative embodiment, the deep embedding engine 140 may provide the known part of the target sentence, if available. One or more additional neural networks(not shown) may be provided to further improve the performance of particular applications by using more application specific features in the embedding with corresponding application specific training data. That is, in addition to, or alternative to, the more general, non-labeled corpus based training of the neural network 145 performed by the above described deep embedding engine 140, for specific applications, application specific training of a neural network may be employed. For example, some tasks, such as sentiment analysis or the like, may be performed using a sentiment labeled corpus. The mechanisms of the illustrative embodiments may be employed to perform embedding using the sentiment labeled corpus and other logic of the ingestion engine 130, via the application interface logic 146, may ingest the embedding and generate a sentiment score, for example. Other applications may utilize other labeled corpora and perform other application specific scoring or operations on the resulting embedded natural language content generated by the deep embedding engine 140.; Lee, [0119] teaches By representing each sentence as a vector representation using the embedding logic 144 in the manner discussed above, more accurate vector representations are generated. Thereafter, with the application interface logic 146, a more accurate natural language processing functionality is made possible, such as a similarity comparison, since the underlying vector representations are more accurate. Such similarity analysis based on vector representations may determine whether the sentences mention the same entity or same type of entity,).
Regarding claim 19, the combination of Lee in view of Glass teaches all of the limitations of claim 14, and the combination further teaches wherein mutually corresponding embeddings of the first and second sets are determined depending on their position in an order of the parts of the character string (Glass, [0032] The processor 120 can be configured to augment the context 210 to indicate positions of objects, such as words, among the context 210 relative to a focus entity. For example, the processor 120 can be configured to perform lookup functions in a word embedding table 215 to identify a first vector, such as a context vector, where the word embedding table can be based on word embeddings such as word2vec. The word embedding table can be updated during training of the model 150. Further, the processor 120 can be configured to perform lookup functions in a position embedding table 225 to identify a second vector, such as a position vector 220, indicating positions of each object relative to the focus entity of the context 210. The processor 120 can augment the context 210 with the results from the lookup functions. For example, the processor 120 can assign position values to objects among the first vector representing the context 210. In the example shown in FIG. 2, the focus entity E200 is at position 0, and other objects among the context 210 are at different position values relative to the entity E200 at position 0. For example, the entity 202 is being augmented with a position value of 5 to indicate that the entity 202 is five positions away from the entity 200 within the context 210.)
Motivation to combine same as stated in claim 14.
Regarding claim 23, the combination of Lee in view of Glass teaches all of the limitations of claim 14, and the combination further teaches wherein the character string is broken down into the parts of the character string by: certain characters of the character string and/or rules for word division and/or syllable division and/or letter division (Lee, [0037] It should be appreciated that within the present description, the following understanding of various terms is used throughout: [0038] 1. Contiguous dependencies: Given a sequence of words or sentences [ . . . , s(t−1), s(t), s(t+1), . . . ] in a sentence or a document, a set [s(t−d)|d=−p, . . . , 0, . . . , q, and p>0, q>0] comprises a set of words or sentences having contiguous dependencies; [0039] 2. Non-contiguous dependencies: the complement of contiguous dependencies; [0040] 3. Extra-document dependencies: dependencies with content existing outside of the document in which the sentence being embedded exists; [0041] 4. Intra-sentence dependency: those dependencies among words or characters within a sentence; [0042] 5. Extra-sentence dependency: those dependencies among sentences; and [0043] 6. Long distance dependences (or “long dependencies”): dependencies with content and/or structure that includes non-contiguous extra-sentence dependencies and/or extra-document dependencies.; Lee, [0045] further teaches with the mechanisms of the illustrative embodiments, both contiguous local context and non-contiguous, long-distanced, contexts providing dependencies are evaluated to perform embedding, i.e. representing a natural language text data object, e.g., a word, sentence, or other portion of text, as a vector representation having a plurality of vector elements that comprise values indicative of context information for the natural language text data object.).
Motivation to combine same as stated for claim 14.
Regarding claim 24, Lee in view of Glass teaches all of the limitations of claim 14, and the combination further teaches wherein mutually corresponding embeddings from the first and second sets are concatenated, or added, or multiplied, to determine the one variable (Glass, [claim 11] teaches wherein the processor is configured to: identify a first vector in a word embedding table, the first vector including a plurality of objects, and the plurality of objects includes a particular entity; identify a second vector in a position embedding table, the second vector including positions of the plurality of objects relative to the particular entity; concatenate the first vector with the second vector to produce an third vector;).
Motivation to combine same as stated for claim 14.
Regarding claim 25,
Claim 25 recites the same and/or analogous limitations as claim 14. Therefore it is rejected under the same rationale and motivation as claim 14.
Glass further teaches at least one processor; and at least one memory, wherein the at least one processor is configured to execute instructions machine learning of a fact for populating a knowledge base, the instructions, when executed by the at least one processor, cause the at least one processor to perform the following steps:
wherein the at least one memory is configured to store the instructions.
(Glass, [0007] teaches in some examples, a system for populating a knowledge graph is generally described. The system can include a memory and a processor comprising hardware, configured to be in communication with each other. The memory can be configured to store a knowledge graph and an unannotated corpus.; Glass, [0003] teaches the present application relates generally to computers and computer applications, and more particularly to machine learning systems that can be implemented to perform distant supervision learning to train a model for predicting unary relations and applying the trained model to populate a knowledge graph.)
Regarding claim 26,
Claim 26 recites the same and/or analogous limitations as claim 14. Therefore, claim 26 is rejected under the same rationale and motivation as claim 14.
Glass further teaches a non-transitory machine-reachable medium on which stored a program for machine learning of facts for populating a knowledge base, (Glass [0008] teaches a computer program product for populating a knowledge graph is generally described. The computer program product may include a computer readable storage medium having program instructions embodied therewith. The program instructions may be executable by a processing element of a device to cause the device to perform one or more methods described herein.; Glass, [0048] further teaches A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.; Glass, [0003] teaches the present application relates generally to computers and computer applications, and more particularly to machine learning systems that can be implemented to perform distant supervision learning to train a model for predicting unary relations and applying the trained model to populate a knowledge graph.)
Claims 18, 20, and 22 are rejected under 35 U.S.C. 103 as being unpatentable over Lee et al. in view of Glass et al., as applied to claim 17, and further in view of Mokeev et al. (US 20210342711 A1, filed May 1, 2020 and published Nov. 4, 2021)
Regarding claim 18, the combination of Lee in view of Glass teaches all of the limitations of claim 17, however the combination does not distinctly disclose wherein the variable for predicting the fact depending on the first embedding and the second embedding and the embedding corresponding to the second embedding is determined depending on an average of the first embedding and the second embedding.
Nevertheless, Mokeev teaches wherein the variable for predicting the fact depending on the first embedding and the second embedding and the embedding corresponding to the second embedding is determined depending on an average of the first embedding and the second embedding (Mokeev teaches [0062] In operation 410, each embedding-consolidating node combines one or more instances of local token embeddings for each token that it handles. For instance, the embedding-consolidating node C1 receives a local token embedding from the embedding-updating node B1 for the word “funny.” It also receives a local token embedding from the embedding-updating node B3 for the same word “funny.” It then averages these two vectors on a dimension-by-dimension basis to produce a consolidated token embedding for the token “embedding.”; Mokeev, [0070] further teaches Although not mentioned in the description of FIGS. 3 and 4, the embedding-updating nodes also calculate local bias parameter values (w.sub.i) for the tokens that they handle, and send them to appropriate embedding-consolidating nodes. Each embedding-consolidating node that receives plural instances of these local bias parameter values for a particular token can then average them to produce a consolidated local bias parameter value. That embedding-consolidating node can then disseminate the consolidated local bias parameter value to selected embedding-updated nodes. For example, assume that two embedding-updating nodes process the word “funny.” They will produce two local versions of the embedding for this word and two local versions of the local bias parameter value for this word. They will send this information to the same embedding-consolidating node. The embedding consolidating node will then average the two local embeddings for “funny” to produce a consolidating embedding for “funny,” and average the two local bias parameter values for “funny” to produce a consolidated local bias parameter value for “funny.” The embedding-consolidating node will then send this consolidated information out to any embedding-updating node that needs this information to perform updating operations in subsequent training cycles.).
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to have modified the method and system for deep embedding for natural language content, as taught by Lee in view of Glass, to further include the method and system for assessing similarity between items using embeddings, as taught by Mokeev. As a first general point, the similarity-assessing engine 120 relies on trained embeddings already produced by the training framework 108. By virtue of this fact, the similarity-assessing engine 120 does not necessarily require the execution of a complex machine-trained model when processing a user's query. And for this reason, the similarity-assessing engine 120 can be expected to consume less computing resources compared to systems that invoke complex machine-trained models to process a query. The computing resources include memory, processing, and data storage resources. The similarity-assessing engine 120 can also be expected to offer better latency-related performance compared to systems that invoke complex machine-trained models to process the user's query. (Mokeev, [0048])
Regarding claim 20, the combination of Lee in view of Glass teaches all of the limitations of claim 14, however the combination does not distinctly disclose wherein mutually corresponding embeddings are determined depending on common characters at a beginning of a relevant part of the character string.
Nevertheless, Mokeev teaches wherein mutually corresponding embeddings are determined depending on common characters at a beginning of a relevant part of the character string ([0060] For example, assume that the prediction generated by the embedding-updating node B1 indicates that item D1 is not a good match for “funny video of cat,” when it is known that it is in fact a good match. In response the embedding-updating node B1 will adjust the embeddings associated with the training example to penalize its inaccurate prediction. If the prediction indicates that the item D1 is a good match, the embedding-updating node B1 will adjust the embeddings associated with the training example to further bolster the prediction it has made. In the terminology used herein, each embedding-updating node produces a set of local token embeddings. For example, the embedding-updating node B1 produces a set of updated embeddings for the tokens “funny,” “video,” “of,” “cat,” and “D1.” The embeddings are referred to as “local” because they reflect insight gleaned from predictions made by just the embedding-updating node B1, and do not yet reflect insight harvested from other embedding-updating nodes (e.g., B2 and B3) that operate in parallel with the embedding-updating node B1. Nor do the predictions made by the embedding-updating node B1 depend on predictions made by other embedding-updating nodes, or require communication with these other embedding-updating nodes. This is true even when different embedding-updating nodes are implemented by the same computing device, and even when different embedding-updating nodes process queries that include at least one common token (e.g., the word “funny”).).
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to have modified the method and system for deep embedding for natural language content, as taught by Lee in view of Glass, to further include the method and system for assessing similarity between items using embeddings, as taught by Mokeev. As a first general point, the similarity-assessing engine 120 relies on trained embeddings already produced by the training framework 108. By virtue of this fact, the similarity-assessing engine 120 does not necessarily require the execution of a complex machine-trained model when processing a user's query. And for this reason, the similarity-assessing engine 120 can be expected to consume less computing resources compared to systems that invoke complex machine-trained models to process a query. The computing resources include memory, processing, and data storage resources. The similarity-assessing engine 120 can also be expected to offer better latency-related performance compared to systems that invoke complex machine-trained models to process the user's query. (Mokeev, [0048])
Regarding claim 22, the combination of Lee in view of Glass teaches all of the limitations of claim 14, and the combination further teaches wherein the first set of embeddings is determined depending on a first vocabulary (Lee [0167] teaches in some illustrative embodiments, the target sentences and aggregated dependencies may be processed using an integralization or on-hot encoding operation by the dependency annotation logic 142, 392. The integralization operation may comprise using a vocabulary obtained from a collection of all of the words in the corpus 510, and representing each word numerically through the building of a one-to-one mapping between numbers and words. For example, if the vocabulary is {cat, dog, child}, then numerical values may be assigned to these words as follows {0, 1, 2}. It is only necessary to assign distinct numbers, so the numerical values could also be {99, 30, 54} or any other numerical representation, depending on the particular implementation. The one-hot encoding turns this numerical representation into a vector where only the dimension specified by the number is 1, and all the other dimensions are 0. For example, using the previous example, the word “cat” will be [1, 0, 0], the word “dog” will be [0, 1, 0], and the word “child” will be [0, 0, 1]. With the illustrative embodiments, the integralized representation may be directly used by the encoders 522 of the neural network 520, or the one-hot representation may be utilized, depending on the desired implementation of the illustrative embodiment.),
However, the combination does not distinctly disclose wherein the second set of embeddings is determined depending on a second vocabulary different from the first vocabulary.
Nevertheless, Mokeev teaches wherein the second set of embeddings is determined depending on a second vocabulary different from the first vocabulary (Mokeev, [0063] teaches In operation 412, the training framework 108 distributes the consolidated token embeddings produced in operation 410 to embedding-updating nodes that will need them to perform subsequent operations. For example, assume that the embedding-updating node B1 will next process another training example having a query that contains the word “funny.” In response, the training framework 108 will send the consolidated token embedding for the word “funny” to this embedding-updating node B1. When performing a next updating operation, the embedding-updating node B1 will rely on the consolidated token embedding for the word “funny,” along with other consolidated token embeddings received from one or more other embedding-consolidating nodes. Operation 412 also implies that the embedding-updating nodes can host different subsets of a complete set of embeddings. For instance, the different embedding-updating nodes can host different embeddings associated with different subsets of words within an entire vocabulary of words.).
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to have modified the method and system for deep embedding for natural language content, as taught by Lee in view of Glass, to further include the method and system for assessing similarity between items using embeddings, as taught by Mokeev. As a first general point, the similarity-assessing engine 120 relies on trained embeddings already produced by the training framework 108. By virtue of this fact, the similarity-assessing engine 120 does not necessarily require the execution of a complex machine-trained model when processing a user's query. And for this reason, the similarity-assessing engine 120 can be expected to consume less computing resources compared to systems that invoke complex machine-trained models to process a query. The computing resources include memory, processing, and data storage resources. The similarity-assessing engine 120 can also be expected to offer better latency-related performance compared to systems that invoke complex machine-trained models to process the user's query. (Mokeev, [0048])
Claim 21 is rejected under 35 U.S.C. 103 as being unpatentable over Lee in view of Glass, as applied to claim 14, and further in view of Zhao et al. (US 20240086731 A1, filed Jan. 13, 2023 and published Mar. 14, 2024)
Regarding claim 21, the combination of Lee in view of Glass teaches all of the limitations of claim 14, however, the combination does not distinctly disclose wherein the character string is divided into a first list of parts, wherein the parts in the first list are uniquely assigned to the embeddings from the first set, wherein the character string is divided into a second list of parts, wherein the parts in the second list are uniquely assigned to the embeddings from the second set, wherein mutually corresponding parts in the first and second lists are defined depending on an order in which the parts are arranged in the first and second lists, and wherein the corresponding embeddings are determined depending on the mutually corresponding parts in the first and second lists
Nevertheless, Zhao teaches wherein the character string is divided into a first list of parts, wherein the parts in the first list are uniquely assigned to the embeddings from the first set, wherein the character string is divided into a second list of parts, wherein the parts in the second list are uniquely assigned to the embeddings from the second set, wherein mutually corresponding parts in the first and second lists are defined depending on an order in which the parts are arranged in the first and second lists, and wherein the corresponding embeddings are determined depending on the mutually corresponding parts in the first and second lists (Zhao, Abstract, teaches the present invention relates to a knowledge-graph extrapolating method and system based on multi-layer perception, the method comprising: using relational graph convolutional network encoders to learn embedding representations, and capturing dynamic evolution of a fact; designing emerging task processing units to construct multiple layers of entity sets, and assigning a matching historical relevance; classifying prediction tasks into different reasoning scenes, and connecting them to the corresponding processing unit for partition of entity sets; and using a multi-class task solving method to acquire predicted probability distributions of target entities, and taking the highest one as a prediction answer, so as to accomplish extrapolation of a temporal knowledge graph, wherein the prediction tasks are classified into different reasoning scenes according to whether it contains any entity or relation that has never appeared historically.; Zhao, [0011] The present invention provides a novel knowledge graph extrapolation model TMP-Net that is based on multi-layer perception, which, for every snapshot, continuously learns representation of entities, relations, and timestamps, thereby being capable of capturing long-term dependency, allowing the model to be more capable of semantic representation. Further, the present invention can divide existing entity sets according to their historical relevance levels, so as to provide entity sets of four different layers of relevance for prediction tasks, and can make prediction more accurately through combination of reviewing known mechanisms and envisioning unknown mechanisms. What makes the present invention special is that entities and relations are considered separately, making it possible to flexibly process four different classes of reasoning scenes including an entity or a relation that have never appeared historically.; Zhao, [0054] teaches At S1, a temporal knowledge graph is divided into sub-graphs custom-character={custom-character.sub.1, custom-character.sub.2, . . . , custom-character.sub.t, . . . } by timestamp, and relational graph convolutional network encoders are used to learn embedding representations of entities, relations, and timestamps and capture dynamic evolution of a fact.).
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to have modified the method and system for deep embedding for natural language content, as taught by Lee in view of Glass, to further include the knowledge-graph extrapolation method and system, as taught by Zhao, in order allow the model to be more capable of semantic representation. (Zhao, [0011])
Conclusion
The following prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Grazoli et al. (US 20250045561 A1) – System and Method for Inductive Learning on Graphs with Knowledge from Language Models - [0055] Although the present invention mainly aims at solving the inductive learning problems, which are particularly challenging when the nodes 112 do not have numerical/categorical features, embodiments of the system 100 may nonetheless support a setting in which nodes 112 do have feature vectors. To this end, the system 100 may include a feature collector 180 that is configured to get the original node features 182, which are then going to be combined with those computed by the LM 150 and used by the graph model 210 for training and prediction, as will be described in detail further below. The entities features 182 can be combined with those extracted by the LM 150 e.g. with a simple concatenation step, but also with more sophisticated techniques. [0057] Entity embeddings computed with the steps described above are then associated to the respective entities 112 in order to construct a graph 190 whose vertexes have numerical embeddings as attributes. If numerical/categorical entity features 182 are provided by the feature collector 180, they can be combined with the entity embeddings, as described above, e.g. simply be concatenating them to the entity embeddings.
Vinod et al. (US 20230177357 A1) - a computerized method for predicting a related field name and an attributes name given an entity name or an attribute name comprising: mining an entity name, a field name, and a datatype information as an extracted data from a specified open source; aggregating the extracted data as a domain knowledge base; implementing an approximate match on an entity name and a field name using a pre-trained word embedding; given the entity name, performing a look up to find one or more closely matching entity names and obtaining a list of potential attributes; and using the domain knowledge base to train a deep learning neural network to predict the attribute name given the entity name or an attribute name.
Cervantes et al. (US 20220179882 A1) - generating a first context-aware vector representation of a first location entity in a first data source and a second context-aware vector representation of a second location entity in a second data source. The approach also comprises processing the first context-aware vector representation and the second context-aware vector representation using a machine learning model to perform a classification of the first location entity as the same as the second location entity. The approach further comprises combining the first data source and the data source into a new database based on the classification and providing the new database as an output. [0007] According to one embodiment, a method comprises creating a knowledge graph of a plurality of location entities of a database. The method also comprises enriching the knowledge graph with multi-modal data. The method further comprises creating a plurality of node embeddings to respectively represent the plurality of location entities based on the enriched knowledge graph. The method further comprises using a machine learning model to predict a relation between any two location entities of the new database based on the plurality of node embeddings. The method further comprises providing the relation as an output. [0045] The new database 115 (combined database) will then contain data records of the multiple data sources 111 such that matched location entities between the multiple data sources 111 are marked as being the same entity (e.g., correlated with a relationship indicator “SAME_AS” or equivalent). The new database 115 can be combined according to any data structure such as, but not limited to, any data structure of a selected one of the multiple data sources 111. For example, if the internal data 101 from an OEM platform 103 is combined with the geographic database 107 of the mapping platform 109, the new internal data 101 can be ingested into the data structure of the geographic database 107 (e.g., the structure as described with respect to FIG. 15 below).
Mullenbach et al. (US 20210319861 A1) - Electronic health records may be organized into problem-oriented medical records. Generating problem-oriented medical record may be based on problem and target relations. Problem and target relations may be determined from a knowledge base. An initial knowledge base may be determined from medical data sets and annotated problem and target relations. The initial knowledge base may be completed to establish new problem target relations using a trained model and/or site embeddings, data statistics, and combined embeddings. The completed knowledge base may be used in generation of problem-oriented medical records. See also Fig. 5.
Abboud et al., “BoxE: A Box Embedding Model for Knowledge Base Completion” (2020)
Li et al., “Position-Aware Relational Transformer for Knowledge Graph Embedding” (April 5, 2023)
Broscheit et al., “Can We Predict New Facts with Open Knowledge Graph Embeddings?, (2020)
Boschin et al. “Combining Embeddings and Rules for Fact Prediction” (International Research School in Artificial Intelligence in Bergen, 2022)
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BEATRIZ RAMIREZ BRAVO whose telephone number is 571-272-2156. The examiner can normally be reached Mon. - Fri. 7:30a.m.-5:00p.m..
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, USMAAN SAEED can be reached at 571-272-4046. 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.
/B.R.B./Examiner, Art Unit 2146
/USMAAN SAEED/Supervisory Patent Examiner, Art Unit 2146
Prosecution Insights