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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under, including the fee set forth in 37 CFR1.17(e), was filed in this application after final rejection. Since this application is eligiblefor continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e)has been timely paid, the finality of the previous Office action has been withdrawnpursuant to 37 CFR 1.114. Applicant's submission filed on 04/22/2026 has been entered.
Status of the Claims
Claims 1-3, 5-12, and 14-22 are pending.
Response to Applicant’s Arguments
In response to “Independent claims 1, 10, and 19 are directed to a method, an apparatus, and a computer program product for obtaining a manual annotation result and an automatic annotation result to train an inference model” and “As stated in paragraphs 4, 7, and 45 of the application, training Al models requires large quantities of annotated corpuses, but manual annotation by domain experts is time-consuming and expensive. Embodiments of the present disclosure solve this by working at a semantic category granularity having domain experts manually annotate only dozens to hundreds of corpuses per category, then the system automatically annotates the remaining corpuses within that category based on semantic distance calculations. This transforms corpus annotation from a massive manual effort into a manageable task with automated propagation within each semantic category”.
According to the specification US 2024/0020482 A1 at ¶59: “For example, in the case that the corpus is a segment including a plurality of sentences, sentence segmentation may be performed on the corpus, to segment a corpus A into a sentence A1, a sentence A2, and a sentence A3”. Therefore, the corpus being annotated comprises of sentences.
Further according to the specification US 2024/0020482 A1 at ¶79:“ In the user annotation interface provided in this embodiment of this disclosure, a semantic category that needs to be annotated may be selected to be displayed, such as a character semantic category of corpuses or a movie and television semantic category of corpuses. In FIG. 4, the movie and television semantic category of corpuses is used as an example to describe the user annotation interface provided in this embodiment of this disclosure” and at ¶81: “a corpus 1 “Movie B is a humorous television (TV) series directed by director C and director D, and is starred by actors including actor E and actor F” corresponds to four pieces of annotation information. A first piece of annotation information is a subject “Movie B”, a subject type “movie and television work”, a subject-predicate relationship type “directed”, a predicate “director C”, and a predicate type “character”. A second piece of annotation information is the subject “Movie B”, the subject type “movie and television work”, the subject-predicate relationship type “directed”, a predicate “director D”, and the predicate type “character”. A third piece of annotation information is the subject “Movie B”, the subject type “movie and television work”, a subject-predicate relationship type “starred”, a predicate “actor E”, and the predicate type “character”. A fourth piece of annotation information is the subject “Movie B”, the subject type “movie and television work”, the subject-predicate relationship type “starred”, a predicate “actor F”, and the predicate type “character”.
PNG
media_image1.png
788
676
media_image1.png
Greyscale
Therefore, the annotation of corpus corresponds to classifying sentences.
Collecting data, recognizing certain data within the collected data set, and storing that recognized data in a memory are drawn to an abstract idea. TLI, 823 F.3d at 613 (citing Content Extraction and Trans. v. Wells Fargo Bank, 776 F.3d 1343, 1347 (Fed. Cir. 2014)). See also See MPEP 2106.04(a)(2)IIIA (“a claim to “collecting information, analyzing it, and displaying certain results of the collection and analysis," where the data analysis steps are recited at a high level of generality such that they could practically be performed in the human mind”).
In TLI, the claims were drawn to the concept of “classifying an image and storing the image based on its classification” that required components such “telephone unit” and “server”. TLI, 823 F.3d at 611. The claims were not directed to a specific improvement to computer functionality because the “telephone unit” with the addition of “a digital image pick up unit for recording images” are merely a conduit for the abstract idea of classifying an image and storing the image based on its classification. Id. at 612. Similarly, the “server” performs generic computer functions such as storing, receiving, and extracting data without any meaningful limitation. Id. at 612-13.
Therefore, the claims in TLI were not directed to a solution to a “technological problem” and instead were simply directed to the abstract idea of classifying and storing digital images in an organized manner as a well established “basic concept”. Id. at 613. Such “basic concept” is not patentable as they are basic tools of scientific and technological work. Gottschalk v. Benson, 409 U.S. 63, 67 (1972).
Similarly, claim 1 in the instant application recites a method, comprising:
obtaining a corpus set comprising a plurality of semantic categories of first corpuses for annotating;
obtaining, based on the corpus set, a manual annotation corpus and an automatic annotation corpus falling within a target semantic category in the corpus set, wherein the manual annotation corpus comprises a subset of corpuses within the target semantic category, and wherein the automatic annotation corpus comprises remaining corpuses within the target semantic category;
obtaining a manual annotation result of the manual annotation corpus;
calculating a semantic distance between the manual annotation corpus and
the automatic annotation corpus; and
annotating, based on the manual annotation result of the manual annotation corpus, a syntax structure of the manual annotation corpus, and a syntax structure of the automatic annotation corpus, the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the semantic distance satisfies a preset condition, wherein the manual annotation result and the automatic annotation result are configured to train a first inference model.
Like the TLI claims for image classification, sentence annotation / classification is a basic concept. The claim as a whole corresponds to manual annotation or classification of sentences within a subset of corpuses within the target semantic category and a broad recitation for “automatic annotation corpus comprises remaining corpuses within the target semantic category”, the automatic annotation involves annotating the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the semantic distance satisfies a preset condition based on (1) he manual annotation result of the manual annotation corpus, (2) a syntax structure of the manual annotation corpus, and (3) a syntax structure of the automatic annotation corpus.
Here, the term “automatic” does not integrate the claim into a practical application because like the image pickup unit or the server in TLI, “automatic” was recited with such a high level of generality such that computer device implied by the term “automatic” is merely a conduit or tool for the “basic concept” of sentence classification / annotation rather than focusing on a specific improvement to a specifically asserted technology.
Neither did the clause “wherein the manual annotation result and the automatic annotation result are configured to train a first inference model” because a patent may issue for the means or method of producing a certain result, or effect, and not for the result or effect produced. Diamond v. Diehr, 450 U.S. 175, 182 n. 7 (1981).
Here, the claims recited no specifically asserted structure for the first inference model or a particular means or method of training said first inference model for a specifically asserted field of technological use.
In other words, patent eligibility hinged on whether the claim “focus on a specific means or method that improves the relevant technology or are instead directed to a result or effect that itself is the abstract idea and merely invoke generic processes and machinery”. McRO, Inc. v. Bandai Namco Games America, Inc., 837 F.3d 1299, 1314 (Fed. Cir. 2016).
In an exemplary patent eligible automation claim, in McRO, the CAFC noted that prior art method of generating morph weight set with values between “0” and “1” for computer animation of facial expressions are manually determined. Id. at 1304-5. The claimed improvement in McRO allows computers to produce “accurate and realistic lip synchronization and facial expressions in animated characters” that previously could only be produced by human animators through the automated use of rules, rather than artists, to set the morph weights and transitions between phonemes. Id. at 1313.
Specifically, the claims were directed to the incorporation of claimed rules, not the use of the computer that improved existing technological process by allowing automation of further tasks that goes beyond merely organizing existing information into a new form. Id. at 1314-15.
In particular, the claimed process used a combined order of specific rules that renders information into a specific format that is then used and applied to create a sequence of synchronized, animated characters that prevent pre-emption of all processes for achieving automated lip-synchronization of 3-D characters. Id. at 1315. Therefore, the CAFC held that the ordered combination of claimed steps, using unconventional rules that relate sub-sequences of phonemes, timing, and morph weight sets is patent eligible. Id. at 1302-3.
See also MPEP 2106.04(d)I (“an improvement in the functioning of a computer or an improvement to other technology or technical field, as discussed in MPEP 2106.04(d)(1) and 2106.05(a)”).
In other words, simply reciting the term “automatic” for “automatic annotation corpus comprises remaining corpuses within the target semantic category” at a high level of generality is insufficient to focus the claim on a specifically asserted technological improvement.
Rather, patent eligibility requires specific means or method of how to employ a computer to automatically annotate the corpus that cannot be practically performed in a human mind.
Specifically, in the instant claims, there is no specifically asserted automation step on how automatic annotation is performed that amounted to a specifically asserted technological improvement.
The claim broadly requires automation to “annotating, based on (1) the manual annotation result of the manual annotation corpus, (2) a syntax structure of the manual annotation corpus, and (3) a syntax structure of the automatic annotation corpus, the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the semantic distance satisfies a preset condition, wherein the manual annotation result and the automatic annotation result are configured to train a first inference model.
This is akin to computing / calculating the morph weights for accurate and realistic lip synchronization and facial expressions in animated characters without a specifically asserted automation rules to set the morph weights and transitions between phonemes that substitutes artists manually setting the morph weights and transitions between phonemes.
Therefore, claim eligibility requires a specifically asserted automation in annotation that amounted to an improvement in machine annotation techniques.
In response to “Gao states that the present invention uses a baseline decision tree parser, a similarity measure, and a set of support vector machine based classifiers to perform training on a mini-corpus (e.g., a corpus set) and generate a unique set of semantic tags, labels, and connections for each word of the sentences in the mini-corpus. Accordingly, Gao does not obtain, based on its mini-corpus, a manual annotation corpus and an automatic annotation corpus falling within a target semantic category in the mini-corpus. For instance, Gao does not obtain two different corpuses (i.e., a manual annotation corpus and an automatic annotation corpus) from its mini-corpus. Also, Gao does not discuss any target semantic categories. Instead, Gao states that it performs training on its mini-corpus to obtain semantic tags, labels, and connections for each word of the sentences in the mini-corpus. Additionally, Kunnumma does not make up for the shortcomings in Gao. Therefore, the combination of Gao and Kunnumma fails to obtain, based on a corpus set, a manual annotation corpus and an automatic annotation corpus falling within a target semantic category in the corpus set, wherein the manual annotation corpus comprises a subset of corpuses within the target semantic category, and wherein the automatic annotation corpus comprises remaining corpuses within the target semantic category”.
Gao discloses a limited manually annotated corpus / “mini corpus” comprising 10000 to 15000 sentences (Col 4, Rows 63-66) where the mini corpus is further divided into chunks of 1000 or 2000 sentences (Col 4, Rows 66-67).
In particular, a human domain expert uses an annotation tool generates parse tree for the first chunk of the mini-corpus (Col 5, Rows 9-10 and Rows 52-55) for trainers to learn semantic structures of the sentences from the annotated data to build models to predict semantic structure of new sentences (Col 5, Rows 10-16) in a particular domain such as medical domain or a target semantic category (Col 7, Rows 3-5).
The parse trees are fed into a triple annotation engine, where the next chunk of mini-corpus is analyzed (Col 5, Rows 20-25). Specifically, the triple annotation engine includes an engine for a baseline decision tree parser, a similarity measure engine, and a SVM classifier engine that takes the generated models to annotate the next chunk of mini-corpus (Col 5, Rows 20-28 in view of Rows 1-5; see also Col 8, Row 62- Col 9, Row 5).
Therefore, the mini-corpus comprises a subset of first chunk corresponding to manual annotation corpus manually annotated by a human medical domain expert and a subset of subsequent chunks corresponding to automatic annotation corpus to be automatically annotated by the triple annotation engine in increments (Col 5, Rows 39-40).
Claim Rejections - 35 USC § 101
35 U.S.C. §101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-3, 5-12, and 14-22 are rejected under 35 USC 101 as directing toward non-statutory subject matter.
Claim 1 recites a method (“process”). Claim 10 recites an apparatus (“machine”). Claim 19 recites computer program product comprising non-transitory computer readable medium (“manufacture”).
To distinguish ineligible claims that merely recite a judicial exception from eligible claims that require an implementation of judicial exception, the Supreme Court uses a two-step framework: Step One (Step 2A), determine whether the claims at issue are directed to one of those patent-ineligible concepts; and Step Two (Step 2B), if so, ask “what else is there in the claims?” to determine whether the additional elements transform the nature of the claim into a patent eligible application. Alice Corp. Pty. Ltd. v. CLS Bank Int’l., 134 S. Ct. 2347, 2355 (2014).
Step One (Step 2A) is a two prong test that requires the determination of whether the claims at issue are directed to an enumerated patent ineligible concept. See MPEP 2106.04.
Specifically, Step 2A Prong (1) requires the determination of the specific limitations in the claim under examination (individually or in combination) that the examiner believes recites an abstract idea and determining whether the identified limitations falls within the subject matter groupings of abstract ideas enumerated. See MPEP 2106.04(a).
The enumerated patent ineligible concepts comprising:
(a) Mathematical Concepts – mathematical relationships, mathematical formulas or equations, mathematical calculations;
(b) Certain methods of organizing human activity – fundamental economic principles / practices (including hedging, insurance, mitigating risk); commercial or legal interactions (including agreements in the form of contracts; legal obligations; advertising, marketing or sales activities or behaviors; business relations); managing personal behavior or relationships or interactions between people (including social activities, teaching, and following rules / instructions) and
(c) Mental processes – concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a).
If the claim recites an enumerated patent ineligible concept, then Prong (2) of Step One (Step 2A) requires the determination of whether the claim integrates the patent ineligible concept into a practical application. Individually and in combination, identifying whether there are any additional elements recited in the claim beyond the judicial exceptions and evaluating those additional elements to determine whether they integrate the exception into a practical application, using one or more of the considerations laid out by the Supreme Court and the Federal Circuit. See MPEP 2106.04(d).
Under Step Two (Step 2B), if the claim does not integrate the ineligible concept into a practical application and therefore directed to a judicial exception, evaluate whether the claim provides an inventive concept by determining whether there are additional elements, individually and in ordered combination, amount to significantly more than the exception itself. See MPEP 2106.04.
Step 2A Prong (1)
The “directed to” inquiry does not ask whether the claims involve a patent ineligible concept but, considered in light of the specification, whether the claim as a whole is directed to excluded subject matter or directed to an improvement to computer functionality. Enfish L.L.C. v. Microsoft Corp., 822 F.3d 1327, 1335 (Fed. Cir. 2016).
Therefore, Prong (1) of Step 2A requires identifying specific limitations in the claims that recites (“describes” or “set forth”) an abstract idea and determine whether the identified limitations falls within the subject matter groupings of abstract ideas enumerated. See MPEP 2106.04 (“Thus, it is sufficient for this analysis for the examiner to identify that the claimed concept (the specific claim limitation(s) that the examiner believes may recite an exception) aligns with at least one judicial exception”).
In particular, MPEP 2106.04(a)(2) states “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number, e.g., performing an arithmetic operation such as exponentiation”.
Under Prong (1), claim 10 recites an apparatus, comprising:
a memory configured to store instructions; and
one or more processors coupled to the memory and configured to execute the instructions to:
(1) obtain a corpus set comprising a plurality of semantic categories of first corpuses for annotating;
(2) obtaining, based on the corpus set, a manual annotation corpus and an automatic annotation corpus falling within a target semantic category in the corpus set, wherein the manual annotation corpus comprises a subset of corpuses within the target semantic category, and wherein the automatic annotation corpus comprises remaining corpuses within the target semantic category;
(3) obtain a manual annotation result of the manual annotation corpus;
(4) calculate a semantic distance between the manual annotation corpus and the automatic annotation corpus; and
(5) annotate, based on the manual annotation result of the manual annotation corpus, a syntax structure of the manual annotation corpus, and a syntax structure of the automatic annotation corpus, the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the semantic distance satisfies a preset condition, wherein the manual annotation result and the automatic annotation result are configured to train a first inference model.
Claim 1 recites a corresponding method.
Claim 19 recites a computer program product comprising instructions stored on a non-transitory computer-readable medium that, when executed by one or more processors, cause an apparatus to perform the method of claim 1 and functions of claim 10.
Individually, (1) obtains a corpus set, (2) obtains manual annotation corpus and an automatic annotation corpus, and (3) obtain manual annotation result correspond to collecting information. Collecting information, including when limited to particular content, is within the realm of abstract ideas. Electric Power Grp., L.L.C. v. Alstom SA, 830 F.3d 1350, 1353 (Fed. Cir. 2016).
Individually, (4) correspond to mathematical calculations.
According to the specification US 2024/0020482 A1 at ¶88: “the calculation unit 1022-1 may calculate a magnitude of a vector difference between the first vectorized feature and the second vectorized feature (that is, a vector distance between the two vectorized features), and use the magnitude as the semantic distance between the manual annotation corpus and the automatic annotation corpus”.
Therefore, (4) corresponds to analyzing information by mathematical algorithms that are essentially mental processes within the abstract idea category. Id.
Individually, (5) corresponds to corpus data annotation or classification that are text analysis.
According to the specification US 2024/0020482 A1 at ¶76 “The client 101 presents the manual annotation corpus to the user, and obtains the manual annotation result of the user for the manual annotation corpus. In an example, when the corpus annotation method provided in this embodiment of this disclosure is used for performing a tuple annotation on a corpus, tuple information annotated by the user for the manual annotation corpus includes a subject, a predicate, a relationship type between the subject and predicate” and at ¶90: “when the semantic distance between the two corpuses satisfies the preset condition, the annotation unit 1022-2 may automatically annotate the subject “Li(3) Si(4)” in the corpus “Li(3) Si(4) was born in city C” as a character based on the manual annotation result “character” for the subject “Zhang(1) San(1)”, and annotate the object “city C” in the automatic annotation corpus as a location based on the manual annotation result “location” for the object “city B””.
Therefore, (5) corresponds to corpus data annotation or classification by steps people go through in their minds or by mathematical algorithms that are essentially mental processes. Electric Power Grp., 830 F.3d at 1354.
In ordered combination, steps (1)-(5) correspond to receiving corpus set to perform text analysis and text annotation.
Thus, claims 1, 10, and 19 described patent ineligible subject matter enumerated under category (c) Mental processes – concepts performed in the human mind (including an observation, evaluation, judgment, opinion).
Step 2A Prong (2).
Under Prong (2) of Step 2A, the goal is to determine whether the claim is directed to the recited exception by evaluating whether the claim as a whole integrates the recited judicial exception into a practical application of the exception. See MPEP 2106.04II(A).
In particular, evaluating integration into a practical application requires identifying whether there are any additional elements recited in the claim beyond the judicial exception and evaluating those additional elements, individually and in combination, to determine whether they integrate the exception into a practical application, using one or more of the considerations laid out by the Supreme Court and the Federal Circuit (“CAFC”). See MPEP 2106.04(d).
According to the Supreme Court, a patent may issue for the means or method of producing a certain result, or effect, and not for the result or effect produced. Diamond v. Diehr, 450 U.S. 175, 182 n. 7 (1981). Therefore, the focus is on whether the claim “focus on a specific means or method that improves the relevant technology or are instead directed to a result or effect that itself is the abstract idea and merely invoke generic processes and machinery”. McRO, Inc. v. Bandai Namco Games America, Inc., 837 F.3d 1299, 1314 (Fed. Cir. 2016).
In an exemplary patent eligible automation claim, in McRO, the CAFC noted that prior art method of generating morph weight set with values between “0” and “1” for computer animation of facial expressions are manually determined. Id. at 1304-5. The claimed improvement in McRO allows computers to produce “accurate and realistic lip synchronization and facial expressions in animated characters” that previously could only be produced by human animators through the automated use of rules, rather than artists, to set the morph weights and transitions between phonemes. Id. at 1313.
Specifically, the claims were directed to the incorporation of claimed rules, not the use of the computer that improved existing technological process by allowing automation of further tasks that goes beyond merely organizing existing information into a new form. Id. at 1314-15.
In particular, the claimed process used a combined order of specific rules that renders information into a specific format that is then used and applied to create a sequence of synchronized, animated characters that prevent pre-emption of all processes for achieving automated lip-synchronization of 3-D characters. Id. at 1315. Therefore, the CAFC held that the ordered combination of claimed steps, using unconventional rules that relate sub-sequences of phonemes, timing, and morph weight sets is patent eligible. Id. at 1302-3.
See also MPEP 2106.04(d)I (“an improvement in the functioning of a computer or an improvement to other technology or technical field, as discussed in MPEP 2106.04(d)(1) and 2106.05(a)”).
On the other hand, the Supreme Court held that mere recitation of a generic computer cannot transform a patent-ineligible abstract idea into a patent-eligible invention. Alice, 134 S. Ct. at 2358.
For example, collecting data, recognizing certain data within the collected data set, and storing that recognized data in a memory are drawn to an abstract idea. TLI, 823 F.3d at 613 (citing Content Extraction and Trans. v. Wells Fargo Bank, 776 F.3d 1343, 1347 (Fed. Cir. 2014)). See also See MPEP 2106.04(a)(2)IIIA (“a claim to “collecting information, analyzing it, and displaying certain results of the collection and analysis," where the data analysis steps are recited at a high level of generality such that they could practically be performed in the human mind”).
In TLI, the claims were drawn to the concept of “classifying an image and storing the image based on its classification” that required components such “telephone unit” and “server”. TLI, 823 F.3d at 611. The claims were not directed to a specific improvement to computer functionality because the “telephone unit” with the addition of “a digital image pick up unit for recording images” are merely a conduit for the abstract idea of classifying an image and storing the image based on its classification. Id. at 612. Similarly, the “server” performs generic computer functions such as storing, receiving, and extracting data without any meaningful limitation. Id. at 612-13.
Therefore, the claims in TLI were not directed to a solution to a “technological problem” and instead were simply directed to the abstract idea of classifying and storing digital images in an organized manner as a well established “basic concept”. Id. at 613. Such “basic concept” is not patentable as they are basic tools of scientific and technological work. Gottschalk v. Benson, 409 U.S. 63, 67 (1972).
In other words, the Supreme Court and the CAFC distinguished between (1) computer-functionality improvements from the (2) uses of existing computers as tools in aid of processes focused on abstract ideas. Electric Power Grp., L.L.C. v. Alstom SA, 830 F.3d 1350, 1354 (Fed. Cir. 2016) (“…we relied on the distinction made in Alice between, on one hand, computer-functionality improvement and, on the other, uses of existing computers as tools in aid of processes focused on “abstract ideas”…”).
In the instant application, claims 1, 10, and 19 described steps for receiving corpus set to perform text analysis and text annotation that are essentially mental processes within the abstract idea category.
Like the TLI claims for image classification, sentence annotation / classification is a basic concept. The claim as a whole corresponds to manual annotation or classification of sentences within a subset of corpuses within the target semantic category and a broad recitation for “automatic annotation corpus comprises remaining corpuses within the target semantic category”, the automatic annotation involves annotating the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the semantic distance satisfies a preset condition based on (1) he manual annotation result of the manual annotation corpus, (2) a syntax structure of the manual annotation corpus, and (3) a syntax structure of the automatic annotation corpus.
Here, the term “automatic” does not integrate the claim into a practical application because like the image pickup unit or the server in TLI, “automatic” was recited with such a high level of generality such that computer device implied by the term “automatic” is merely a conduit or tool for the “basic concept” of sentence classification / annotation rather than focusing on a specific improvement to a specifically asserted technology.
Therefore, unlike the patent eligible automation claims in McRO that set forth a specific means of automating animation (i.e., using a combined order of specific rules that renders information into a specific format that is then used and applied to create a sequence of synchronized, animated characters to improve a computer process through the automated use of rules), steps (1)-(5) recited no particular means or method of automating the annotation process and the claims recited no particular improvement to an annotation apparatus for automatic annotation.
Rather, the claim broadly requires automation to “annotating, based on (1) the manual annotation result of the manual annotation corpus, (2) a syntax structure of the manual annotation corpus, and (3) a syntax structure of the automatic annotation corpus, the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the semantic distance satisfies a preset condition, wherein the manual annotation result and the automatic annotation result are configured to train a first inference model.
This is akin to computing / calculating the morph weights for accurate and realistic lip synchronization and facial expressions in animated characters without a specifically asserted automation rules to set the morph weights and transitions between phonemes that substitutes artists manually setting the morph weights and transitions between phonemes.
Neither did the clause “wherein the manual annotation result and the automatic annotation result are configured to train a first inference model” integrates the claim into a practical application because a patent may issue for the means or method of producing a certain result, or effect, and not for the result or effect produced. Diamond v. Diehr, 450 U.S. 175, 182 n. 7 (1981). Here, the claims recited no specifically asserted structure for the first inference model or a particular means or method of training said first inference model for a specifically asserted field of technological use.
Further, in Alice, the Supreme Court held that data processing systems with data storage unit and transmission units were purely functional and generic and such recitation of hardware failed to offer any meaningful limitation beyond generally linking the use of a method to a particular technological environment. Id. at 2360. See MPEP 2106.04(d)I (“Generally linking the use of a judicial exception to a particular technological environment or field of use, as discussed in MPEP 2106.05(h)”). Neither stating an abstract idea while adding the words “apply it” nor limiting the use of an abstract idea to a particular technological environment is enough for patent eligibility. Id. at 2350.
Therefore, the recitation of apparatus with memory / non-transitory computer readable medium storing instructions for execution by processor in claims 10 and 19 do not make the claims patent eligible because the use of a generic computer merely invoked generic machinery that did not go beyond merely organizing existing information (syntax structure of annotation corpus) into new form (annotated form).
Applicant argued “The corpus annotation apparatus automatically annotates the automatic annotation corpus based on the manual annotation result of the manual annotation corpus to obtain an automatic annotation result of the automatic annotation corpus. This shortens time consumed for generating the annotation corpus, improves efficiency of generating the annotation corpus, and reduces labor costs”.
However, adding computer functionality to increase the speed or efficiency of the process confer patent eligibility on an otherwise abstract idea. Intellectual Ventures I, 792 F.3d at 1367 (citing Bancorp Servs., LLC v. Sun Life Insurance Co. of Can., 687 F.3d 1266, 1278 (Fed. Cir. 2012) (“The fact that the required calculations could be performed more efficiently via a computer does not materially alter the patent eligibility of the claimed subject matter”)).
In other words, simply adding the word “automatic” to broadly claim an automatic annotation apparatus / computer without specific means or method of how such automatic annotation apparatus automatically annotate the respective annotation corpus failed to specifically assert a technological improvement because claims 1, 10, and 19 did not set forth a specific means or method of producing a certain result, or effect (i.e., “shortens time consumed for generating the annotation corpus, improves efficiency of generating the annotation corpus, and reduces labor costs”) corresponding to automatic corpus annotation.
Here, the broad recitation of computer, memory / non-transitory computer readable medium, processor in claims 1, 10, and 19 offered no meaningful limitation beyond generally linking the application of annotation to computers.
Therefore, claims 1, 10, and 19 are directed to receiving or collecting corpus set to perform text data analysis and text data annotation that are essentially mental processes within the abstract idea category.
Step 2B Inventive Concept.
The Guideline stated that if the additional elements do not integrate the exception into a practical application, then the claim is directed to the recited judicial exception, and requires further analysis under Step 2B where it may still be eligible if it amounts to an “inventive concept”. See MPEP 2106.04IIA and MPEP 2106.05.
Further, an inventive concept can be found in the non-conventional and non-generic arrangement of known conventional pieces. BASCOM Global Internet Servs. v. AT&T Mobility, 827, F3d 1341, 1350 (Fed. Cir. 2016).
In BASCOM, the CAFC held that filtering content is an abstract idea because it is a longstanding, well-known method of organizing human behavior similar to concepts previously found to be abstract. BASCOM, 827 F.3d at 1348. However, the CAFC determined that the claims did not merely recite filtering content along with the requirement to perform it on the internet or on a set of generic computer components, nor did the claims preempt all ways of filtering content on the internet. Id. at 1350.
Rather, the inventive concept described and claimed was the installation of a filtering tool at a specific location, remote from the end-users, with customizable filtering features specific to each end user that gives the filtering tool both the benefits of a filter on a local computer and the benefits of a filter on an internet service provider “ISP” server. Id. By taking a prior art filter solution (one size fits all filter at internet service provider “ISP” server) and making it more dynamic and efficient (providing individualized filtering at the ISP server), the claimed invention improves the performance of the computer system itself. Id. at 1351.
On the other hand, implementation via computers does not offer a meaningful limitation beyond generally linking the use of an abstract idea to a particular technological environment. Alice, 134 S. Ct. at 2360 (“Nearly every computer will include a “communications controller” and “data storage unit” capable of performing the basic calculation, storage, and transmission functions required by the method claims”). Intellectual Ventures I L.L.C. v. Capital One Bank, 792 F.3d 1363, 1370-71 (Fed. Cir. 2015) (“Steps that do nothing more than spell out what it means to “apply it on a computer” cannot confer patent-eligibility).
Similarly, limiting an abstract idea to one field of use do not convert otherwise ineligible concept into an inventive concept. Intellectual Ventures I L.L.C. v. Erie Indem. Co., 850 F.3d 1315, 1328 (Fed. Cir. 2017). Neither does adding computer functionality to increase the speed or efficiency of the process confer patent eligibility on an otherwise abstract idea. Intellectual Ventures I, 792 F.3d at 1367 (citing Bancorp Servs., LLC v. Sun Life Insurance Co. of Can., 687 F.3d 1266, 1278 (Fed. Cir. 2012) (“The fact that the required calculations could be performed more efficiently via a computer does not materially alter the patent eligibility of the claimed subject matter”)).
Individually, in the instant application, claim 1 recites automated annotation, claim 10 recites an apparatus comprising processor and memory storing instructions, while claim 19 recites processor and non-transitory computer readable medium.
Such individual recitation of generic computer components (processor, non-transitory computer readable medium, automated annotation) are purely functional and generic because nearly every computer will include such processor and data storage unit capable of performing basic calculation necessary for analysis and storage.
As an ordered combination, unlike BASCOM that describes an unconventional combination of a conventional ISP server with a customized filter specific to each user that is remote from end-users to provide both the benefits of a filter on a conventional local computer and the benefits of a filter on the conventional ISP server, implementing (1) –(5) on a computer with processor and non-transitory computer readable medium do not involve a unconventional combination of conventional pieces because the combination amounts to “apply it on a computer” or limiting steps (1)-(5) to a computer.
To the extent that implementing (1)-(5) on a processor / computer in the field of computer results in reduction in memory requirement and computational requirement, merely adding computer functionality to increase the speed or efficiency does not confer patent eligibility on an otherwise abstract idea.
Dependent claims also failed to integrated claims 1, 10, and 19 into a practical application under step 2A or an inventive step under step 2B for the following reasons:
Dependent claims 2-3, 11-12, and 20 stated an intent to train an AI inference model but there is no limitation describing a particularly structured AI inference model or how such structured AI inference model is improved according to the training.
Likewise, dependent claims 5 and 15 correspond to mathematically calculations to calculate semantic distance between annotation corpora when performing annotation, which corresponds to analyzing information by mathematical algorithms that are essentially mental processes within the abstract-idea category. Electric Power Grp., 830 F.3d at 1354
Dependent claims 6-9 and 15-18 correspond to further steps for corpus data analysis without limitations describing any particular improvement to a particularly structured AI inference model or a specifically asserted application of the AI inference model to perform machine translation, question answering, or speech recognition.
Dependent claim 21 recites an implementation of the corpus annotation computing apparatus with a communication interface communicating with a client device, one or more processors to perform calculation of the semantic distance, and a broad requirement for the processor of the corpus annotation computing apparatus to perform the annotation of the automatic annotation corpus.
The Supreme Court held that implementation via computers does not offer a meaningful limitation beyond generally linking the use of an abstract idea to a particular technological environment. Alice, 134 S. Ct. at 2360 (“Nearly every computer will include a “communications controller” and “data storage unit” capable of performing the basic calculation, storage, and transmission functions required by the method claims”). Intellectual Ventures I L.L.C. v. Capital One Bank, 792 F.3d 1363, 1370-71 (Fed. Cir. 2015) (“Steps that do nothing more than spell out what it means to “apply it on a computer” cannot confer patent-eligibility).
In other words, the communication interface for communicating with a client device and processor for calculating semantic distance do not offer a meaningful limitation beyond generally linking the use of an abstract idea to a particular technological environment because nearly every computer includes a controller capable of performing basic calculation and transmission function required by claim 21.
Dependent claim 22 requires displaying a user interface on a display device for receiving manual annotations and receiving manual annotations from user inputs on the user interface.
The CAFC held that a process of gathering and analyzing information of a specified content and then displaying the results do not correspond to any particular asserted inventive technology for performing those functions and therefore directed to an abstract idea. Electric Power Grp., 830 F.3 at 1354.
In other words, a display device with a user interface to gathering information of a specified content (i.e., manual annotation) does not correspond to any particular asserted inventive technology for performing (1)-(5) in claim 1 and therefore directed to the abstract idea set forth in Claim 1.
In conclusion, without limitations setting forth a particularly structured AI inference model or, in the alternative, a particular means or method of automating the annotation process, claims 1-3, 5-12, and 14-22 are essentially directed to collecting corpus data and performing corpus text data analysis and corpus text annotation that are essentially mental processes within the abstract idea category.
Therefore, claims 1-3, 5-12, and 14-22 are not eligible for a patent.
Claim Rejections - 35 USC § 112
35 U.S.C. §112(b) reads as follows:
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
Claims 1-3, 5-12, and 14-22 are rejected under 35 USC 112(b) for not particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
Exemplary claim 1 in the instant application recites a method, comprising:
obtaining a corpus set comprising a plurality of semantic categories of first corpuses for annotating;
obtaining, based on the corpus set, a manual annotation corpus and an automatic annotation corpus falling within a target semantic category in the corpus set, wherein the manual annotation corpus comprises a subset of corpuses within the target semantic category, and wherein the automatic annotation corpus comprises remaining corpuses within the target semantic category;
obtaining a manual annotation result of the manual annotation corpus;
calculating a semantic distance between the manual annotation corpus and
the automatic annotation corpus; and
annotating, based on the manual annotation result of the manual annotation corpus, a syntax structure of the manual annotation corpus, and a syntax structure of the automatic annotation corpus, the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the semantic distance satisfies a preset condition, wherein the manual annotation result and the automatic annotation result are configured to train a first inference model.
It is unclear what is being annotated in the last clause of claim 1. Is it:
(a) based on the manual annotation result of the manual annotation corpus, annotating a syntax structure of the manual annotation corpus, and a syntax structure of the automatic annotation corpus, the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the semantic distance satisfies a preset condition, wherein the manual annotation result and the automatic annotation result are configured to train a first inference model; or
(b) based on (1) the manual annotation result of the manual annotation corpus, (2) a syntax structure of the manual annotation corpus, and (3) a syntax structure of the automatic annotation corpus, annotating the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the semantic distance satisfies a preset condition, wherein the manual annotation result and the automatic annotation result are configured to train a first inference model.
Claims 10 and 19 exhibit the same deficiency. Dependent claims 2-3, 5-9, 11-12, 14-18, and 20-22 are dependent on claims 1, 10, and 19 and therefore deficient as well.
For the purpose of claim interpretation, the instant office action corresponds to interpretation (b).
Claim Rejections - 35 USC § 103
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 103 that form the basis for the rejections under this section made in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-3, 5-7, 10-12, 14-16, and 19-22 are rejected under 35 USC 103(a) as being unpatentable over Gao et al. (US 7996211 B2) in view of Kunnumma et al. (US 2020/0320430 A1).
Regarding Claims 1, 10, and 19, Gao discloses an apparatus (Figs. 2-3), comprising:
a memory configured to store instructions (Col 4, Rows 17-21, memory 304 storing program instructions); and
one or more processors coupled to the memory (Col 4, Rows 17-21, processor 302 executes instructions loaded into memory 304) and configured to execute the instructions to:
obtain a corpus set comprising a plurality of semantic categories of first corpuses for annotating (Col 4, Rows 60-67, a mini-corpus comprising 10000 to 15000 sentences for semi-automatic semantic annotation; per Col 5, Rows 5-16 and Col 8, Rows 48-50, a first chunk (first 1000 sentences 602) of the mini-corpus comprising parse trees generated by human using an annotation tool; the parse tree comprises a set of semantic tags, labels, and connections for each word of the sentences; e.g., Col 7, Rows 3-6, word “English” having different semantic categories such as language, person or discipline);
obtain, based on the corpus set, a manual annotation corpus and an automatic annotation corpus falling within a target semantic category in the corpus set (Col 7, Rows 3-5, the mini-corpus being in the medical domain or “target semantic category”), wherein the manual annotation corpus comprises a subset of corpuses within the target semantic category (Col 4, Rows 63-67, given a limited manually annotated corpus or mini-corpus of 10000 to 15000 sentences divided into chunks of 1000 or 2000 sentences; Col 5, Rows 9-10 and Rows 52-55, human annotator / domain expert uses annotation tool to generate parse tree for a first chunk of the mini-corpus), and wherein the automatic annotation corpus comprises remaining corpuses within the target semantic category (Col 5, Rows 1-5, Rows 20-28, and Col 8, Row 62 – Col 9, Row 5, feed the manually generated parse tree into a triple annotation engine to annotate the next chunk of the mini-corpus);
obtain a manual annotation result of the manual annotation corpus (Col 7, Rows 48-60, similarity engine determines the best reference sentence from the training sentences (i.e., the first chunk of 1000 manually annotated sentences); Col 5, Rows 39-40, correctly annotated sentences are used as training data for the next round of incremental annotation; Col 7, Rows 43-52, sentence to be annotated is the candidate sentence (i.e., to be automatically annotated) and find the most relevant segment of the training sentence (i.e., in the manually annotated training data) containing the words to be annotated with similar left and right context size);
calculate a similarity measure between the manual annotation corpus and the automatic annotation corpus (Col 7, Rows 18-21 and 43-47, calculating a fully automatic evaluation metric or respective BLEU scores as respective similarity measure between sentence to be annotated and all sentences in the training data (first 1000 sentences that were manually annotated) containing the word to be annotated as possible reference sentences); and
annotating, based on the manual annotation result of the manual annotation corpus (Col 8, Rows 48-62, feeding the first chunk of 1000 manually annotated sentences into the baseline decision tree statistical parser), a syntax structure of the manual annotation corpus (Col 5, Rows 9-10 and Col 8, Rows 48-50 in view of Col 1, Rows 31-33, human annotator uses annotation tool 606 to generate parse trees for the sentences in the first chunk of the mini-corpus corresponding to grammar based approach capturing syntax and semantics), and a syntax structure of the automatic annotation corpus (Col 5, Rows 25-28 and Col 9, Rows 1-5, send the next 1000 sentences / chunk of the mini-corpus into the baseline decision tree statistical parser to extract meaning and model the syntactic and semantic structure of the second chunk of 1000 sentences; Col 5, Rows 37-40, correctly annotated sentences are used as training data for the next round of incremental annotation; per Col 7, Rows 43-47, sentences to be annotated is treated as the candidate sentence and all the sentences in the training data containing the word to be annotated are possible reference sentences), the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the similarity measure satisfies a preset condition (Col 9, Rows 6-11, annotate the sentences with a unique set of tags, labels, and word connections if (1) the parser engine and (2) the similarity engine agreed on the unique set of tags, labels and word connections; i.e., the similarity engine determined the best reference sentence containing the word to be annotated based on the respective similarity measures / BLEU scores (per Col 8, Rows 54-57, using BLEU score as pre-set condition to determine the best reference sentence), the best reference sentence’s tags, labels, and word connections agree with the tags, labels, and word connection determined by the baseline decision tree statistical parser engine),
wherein the manual annotation result and the automatic annotation result are configured to train a first inference model (Col 5, Rows 13-16, trainers are mechanisms that learn semantic structures of the sentences from the annotated data to build models).
Gao does not teach that the similarity measure is a semantic distance between the manual annotation corpus and the automatic annotation corpus.
Kunnumma teaches an automatic annotation apparatus (¶25) obtains a manual annotation result of a manual annotation corpus (¶52, training data comprising seed data with labels / annotations for a first dataset such as labelled / annotated clauses in a contract), calculates a semantic distance / vector distance in vector space representation between initially labeled sentences / contract clauses (i.e., manual annotation corpus) and unlabeled sentences / contract clauses (i.e., automatic annotation corpus) (¶52, the machine learning classification model may use clauses in the contracts and the labels corresponding to the clauses as initial seed data (first dataset or manual annotation corpus) to evaluate the confidence of the system in identifying an unlabeled clause characterized by the determination of certainty score (second dataset or automatic annotation corpus; per ¶14 and ¶16, generate presented subset of data from the second dataset based on the (1) certainty threshold value and (2) computed candidate vector distance for calibration comprising a user input from to indicate the correctness of the presented subset of data; i.e., ¶52, expert user labels only the dataset where (1) certainty score drops below a predefined certainty threshold and ¶65, (2) those lying at the periphery of the clusters); ¶27, ¶38, and ¶65, in machine learning of similar sentences, candidate vector distance is a pairwise distance metric computed for candidates to determine their similarity wherein the basis for similarity measure is their vector space representation with a cosine distance value 1 indicating they are completely similar and value 0 indicating that they are completely dissimilar.1), and based on the manual annotation result of the manual annotation corpus, annotates the automatic annotation corpus to obtain an automatic annotation result of the automatic annotation corpus when the semantic distance satisfies a preset condition (¶27 and ¶65, using clustering technique to group sentences together. In view of ¶52, when data classifier identified candidate sentences / unlabeled contract clauses as having semantic / vector distances satisfying a preset condition to labeled contract clauses, the candidate sentences are clustered or grouped as similar sentences to the labeled contract clauses / sentences where those lying at the periphery of the clusters are for user to disambiguate / annotate to indicate correctness per ¶16).
It would’ve been obvious to one ordinarily skilled in the art before the effective filing date of the invention to implement a fully automatic similarity measure for annotation (Gao, Col 7, Rows 17-21) base on calculating a semantic distance between the manual annotation corpus and the automatic annotation corpus to annotate the automatic annotation corpus when the semantic distance satisfies a preset condition in order to group similar sentences together for automatic annotation while outliers at the peripheral of the clusters that do not satisfy the preset condition are presented for human annotation (Kunnumma, ¶27; compare Gao, Col 5, Rows 35-40 and Col 9, Rows 21-34, unreliable annotations are comparable to outliers in Kunnumma that requires human annotation).
Further regarding claim 19, Gao discloses a computer program product comprising instructions stored on a non-transitory computer-readable medium that, when executed by one or more processors, cause an apparatus to implement the functions of claims 1 and 10 (Col 4, Rows 17-21, memory 304 storing program instructions for execution by processor 302; see also claim 21 of Gao).
Regarding Claims 2, 11, and 20, Gao discloses wherein the one or more processors are further configured to execute the instructions to train, based on the manual annotation result and the automatic annotation result, the first inference model (Col 5, Rows 13-16, baseline decision tree parser, similarity measure, and the set of support vector machines or trainers, learn semantic structures of the sentences from the annotated data to build models. The resulting models are used in the corresponding engines to predict semantic structure of new sentences).
Regarding Claims 3 and 12, Gao discloses wherein the one or more processors are further configured to execute the instructions to:
receive a selection operation (Col 9, Rows 21-24 and Rows 33-37, sentences tagged as unreliable or low confidence sentences are forwarded to human annotation to be inspected and corrected; i.e., human annotator will select the proper labels for the low confidence sentences); and
train, based on the selection operation, the manual annotation result, and the automatic annotation result, a second inference model (Col 9, Rows 30-37, once the sentences are inspected and corrected by human annotation, these annotated sentences may also be used as training data in addition to the initially annotated sentences (i.e., first chunk) for the next round of annotation; per Col 5, Rows 13-16, trainers learn semantic structures of sentences from the annotated data (i.e., next round of annotation) to build models).
Regarding Claims 5 and 14, Gao as modified by Kunnumma discloses wherein the one or more processors are further configured to execute the instructions to:
calculate the semantic distance comprises
obtain a first vectorized feature of the manual annotation corpus and a second vectorized feature of the automatic annotation corpus (Kunnumma, ¶52, use clauses in the contracts and the labels corresponding to the clauses (manual annotation corpus) to build the data classifier to label / identify an unlabeled clause (automatically labelled second dataset / automatic annotation corpus) and evaluate the confidence in identifying the unlabeled clause characterized by the determination of certainty score; in view of ¶14 and ¶16, generate presented subset of second dataset based on the certainty threshold value and computed candidate vector distance for user calibration (i.e., expert user label only the dataset where the certainty score drops below certainty threshold and those lying at the periphery of clusters per ¶52 and ¶65); ¶38 and ¶65, evaluating the confidence in identifying the unlabeled clause requires computing vector distance in vector space representation as pairwise distance metric for candidates to determine their similarity); and
calculate, based on the first vectorized feature and the second vectorized feature, the semantic distance to annotate the manual annotation corpus and the automatic annotation corpus when the semantic distance satisfies a preset condition (Kunnumma, ¶27 and ¶65, clustering sentence vectors that are similar in vector space representation based on vector distance (i.e., unlabeled sentences that are similar to labeled sentences in vector space are clustered together under the same label based on they are close or are similar per ¶38)).
Regarding Claims 6 and 15, Gao as modified by Kunnumma discloses wherein the one or more processors are further configured to execute the instructions to:
calculate the semantic distance by using an artificial intelligence (AI) model (Gao, Col 7, Rows 16-20 and Rows 41-47, automatically calculate similarity measure for annotation; Kunnumma, ¶38 and ¶65, compute similarity measure in vector space representation);
obtain a manual check result for an annotation result of the automatic annotation corpus (Gao, Col 9, Rows 32-37, sentences tagged as unreliable or low confidence sentences are forwarded to human annotation where they are inspected and corrected by human annotation; Kunnumma, ¶66, outliers lying at the periphery of the clusters are presented to user for labeling or annotating); and
update the AI model by using the automatic annotation corpus and the manual check result when the manual check result indicates that the automatic annotation corpus is incorrectly annotated (Gao, Col 9, Rows 34-37, once they are inspected and corrected, these annotated sentences may also be used as training data for the next round of annotation; i.e., using the annotated data to build models per Col 5, Rows 13-14; Kunnumma, ¶66, the annotated candidates are coupled with the training data before proceeding to a subsequent build stage; i.e., to build the data classifier model per ¶52).
Regarding Claims 7 and 16, Gao as modified by Kunnumma discloses wherein the annotation result comprises a confidence value (Gao, Col 9, Rows 24-25, annotated sentences are generated with confidence 622; Kunnumma, ¶36, data classifier determines the prediction probability or certainty score of a candidate belonging to target classes), and wherein the one or more processors are further configured to execute the instructions to obtain the manual check result when the confidence value is less than a confidence threshold (Gao, Col 9, Rows 32-34, sentences tagged as low confidence sentences are forwarded to human annotation to be inspected; Kunnumma, ¶14, generate presented subset of data from the second dataset (i.e., the set of sentences) based on certainty threshold value and a computed candidate vector distance; ¶37, using certainty threshold to determine if a candidate will be taken up for disambiguation / annotation by the user; e.g., present candidate to human annotation if certainty score is below the certainty threshold).
Regarding Claim 21, Gao discloses wherein the method is implemented by a corpus annotation computing apparatus (Figs. 1-2, server 104 or any of clients 108-112), wherein obtaining the corpus comprises obtaining, from a client device and through a communication interface that communicatively couples the corpus annotation computing apparatus to the client device (Figs 1-3, server 104 implemented by data processing system 200; Col 3, Rows 24-29, server / system 200 with communications links to clients 108-112 through modem 218 and network adapter 220 and clients 108-112 with modem 322 and interface 314 per Col 4, Rows 1-2), the corpus set (Col 5, Rows 9-10 and Col 8, Rows 44-47, the first chunk of the mini-corpus being generated by human using an annotation tool implemented using a graphical user interface (with keyboard and mouse on a client data processing system 300 per Col 3, Rows 48-52 and Col 4, Rows 1-2) with which a user interacts with to perform semantic annotation; i.e., the mini-corpus were obtained from client computer since the first chunk of the corpus were obtained from the client computer where the first chunk was annotated),
wherein obtaining the manual annotation corpus and the automatic annotation corpus comprises obtaining, by one or more processors of the corpus annotation computing apparatus (Col 3, Rows 9-15, processors 202 and 204 for server or processor 302 for clients 108-112), the manual annotation corpus and the automatic annotation corpus (see claim 21 of Gao; see further Col 4, Rows 64-67, Col 5, Rows 9-10 and Rows 37-40, and Col 8, Rows 48-50, for a mini-corpus of 10000 to 15000 sentences divided into chunks of 1000 or 2000 sentences, the first 1000 sentences are manually annotated “manual annotation corpus” and correctly annotated sentences are used as training data for the next round of incremental annotation by triple annotation engine to annotate the next chunk of the mini-corpus per Col 5, Rows 17-25),
wherein calculating the semantic distance comprises calculating, by the one or more processors of the corpus annotation computing apparatus, the semantic distance (Kunnumma, Abstract, computer processors to implement data classifier for computing vector distance in vector space per ¶38; when modifying Gao, the processors of either server 104 or any of clients 108-112 can be modified to implement the computer processor to compute vector distance in vector space), and
wherein annotating the automatic annotation corpus comprises annotating, by the one or more processors of the corpus annotation computing apparatus, the automatic annotation corpus (see claim 21 of Gao, when implemented on either server 104 or any of the clients 108-112).
Regarding Claim 22, Gao discloses displaying, on a display device (Fig. 1, any of the display devices for server 104 and clients 108-112), a user interface for receiving manual annotations (Col 8, Rows 44-47, annotation framework 600 implemented using a graphical user interface with which a user may interact with to perform semantic annotation); and
receiving, from user inputs on the user interface, the manual annotations (Col 5, Rows 7-10, the parse tree itself (i.e., tags, labels, and connections for each word of the sentence) is generated by human using an annotation tool for the first chunk of the mini-corpus).
Claims 8-9 and 17-18 are rejected under 35 USC 103(a) as being unpatentable over Gao et al. (US 7996211 B2) and Kunnumma et al. (US 2020/0320430 A1) as applied to claims 1 and 10, in view of Duta (US 8515736 B1).
Regarding Claims 8-9 and 17-18, Gao does not teach before obtaining the manual annotation corpus and the automatic annotation corpus, provide a semantic category configuration interface.
Duta teaches obtaining manual annotation corpus (Col 9, Rows 37-38, a few manually annotated example utterances) and automatic annotation corpus (Col 9, Rows 38-43, a database of existing utterances) wherein before obtaining the manual annotation corpus and the automatic annotation corpus, provide a semantic category configuration interface (Col 12, Rows 58-65, training manager receives semantic labels and a few sentences for each meaning from customer / developer of a given text classification application);
obtain, in response to a configuration operation on the semantic category configuration interface, second corpuses in the corpus set (Col 9, Rows 38-43, using few example utterances to query a database of existing utterances to extract additional example utterances that are semantically similar to the initial example utterances); and
cluster, based on the semantic categories, the second corpuses (Col 9, Rows 53-60, semantically clustering a set of utterances from a new application to assist developer with the process of labeling (i.e., annotating) utterances);
wherein before clustering the second corpuses, provide a feature configuration interface comprising a plurality of feature candidates (Col 16, Rows 10-14 and Rows 21-23, receive set of manually created semantic classes including a list of categories and subcategories for classifying utterances according to the meanings associated with first utterance (e.g., the few example utterances); per Col 16, Rows 26-28, the set of manually created semantic classes can be created or designed for a specific company, organization, entity, or call center); and
obtain, in response to a selection operation on the feature configuration interface, a target feature for clustering the second corpuses (Col 16, Rows 13-14, list of categories or subcategories for classifying utterances according to meaning; Col 16, Rows 52-55 and Col 17, Rows 1-5, semantically label existing utterances from the database of existing utterances by identifying existing utterances that are semantically similar to the first utterance; per Col 9, Row 53 – Col 10, Row 7, determine semantic distance between existing utterances from the database with few example / annotated utterances and semantically clustering the set of utterances into clusters based on its meaning (i.e., as defined by the list of categories or subcategories)).
It would’ve been obvious to one ordinarily skilled in the art before the effective filing date of the invention to implement a semantic category configuration interface at the annotation tool (Gao, Col 8, Rows 48-50) to receive semantic labels / annotations from customers / developers and a feature configuration interface to receive categories / subcategories from customers / developers to create or design manually created semantic classes for specific company, organization, entity or call center (Duta, Col 16, Rows 26-28).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to examiner Richard Z. Zhu whose telephone number is 571-270-1587 or examiner’s supervisor Hai Phan whose telephone number is 571-272-6338. Examiner Richard Zhu can normally be reached on M-Th, 0730:1700.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/RICHARD Z ZHU/Primary Examiner, Art Unit 2654 07/07/2026
1 Compare specification US 2024/0020482 A1 at ¶88: “the calculation unit 1022-1 may calculate a magnitude of a vector difference between the first vectorized feature and the second vectorized feature (that is, a vector distance between the two vectorized features), and use the magnitude as the semantic distance between the manual annotation corpus and the automatic annotation corpus”.
Prosecution Insights