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 .
DETAILED ACTION
Claims 1-3, 5-12, and 14-22 are pending. Claims 1, 11, and 20 are independent.
This Application was published as US 20240386200.
Apparent priority is 15 May 2023.
The instant Application is directed to a method of determining topics and parent topics.
Applicant’s amendments and arguments are considered but are either unpersuasive or moot in view of the new grounds of rejection that, if presented, were necessitated by the amendments to the Claims.
This action is Final.
Response to Arguments
35 USC 101
Applicant's arguments with respect to 30 USC 101 have been fully considered and are persuasive. The rejection under 35 USC 101 is withdrawn.
35 USC 102
Applicant’s arguments with respect to 35 USC 102 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Claim Rejections - 35 USC § 103
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.
Claim(s) 1-3, 5, 7-12, 14, and 16-22 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ni et al. (US 20110258229 A1) in view of Wang et al. (US 20220171798 A1) and Amini et al. (US 20110098999 A1).
Regarding claim 1, Ni discloses: 1. A method for natural language processing of a corpus of documents, the method comprising: evaluating the corpus of documents using a machine learning model to choose a plurality of topics, ("[0014] The disclosure describes a technique for organizing, classifying, and locating written content in different languages, and for performing other cross-lingual tasks. The technique utilizes a multi-language, hyperlinked document corpus, in which documents are organized or aligned by concept. ... Topic modeling techniques are used to determine topics of the documents, with the constraint that all the documents corresponding to a particular concept, regardless of language, are assumed to share a common topic distribution. ..." ; see also "[0032] … inferring 320 can be accomplished by performing a modified LDA analysis…” – LDA reads on a machine learning model.)
wherein the machine learning model generates a distribution over a predetermined list of topics for each document in the corpus; ("[0018] ...Each topic is defined as plurality of topic/word (T/W) distributions corresponding respectively to languages L1 through Lj. Each topic/word distribution indicates the probability of any particular word occurring in the corresponding topic." )
using the plurality of topics to generate a topic of topics, ("[0014]... Each universal topic is defined by a plurality of topic/word distributions, corresponding to the different languages of the multi-language document corpus, indicating the words likely to appear in relation to that universal topic. …" )
wherein the topic of topics comprises a finite set of groupings of the plurality of topics; (not explicitly disclosed)
assessing the topic of topics to determine a quality of the natural language processing of the corpus; ("[0069]...Taking the 31st universal-topic as an example, it is nicely represented by English words like football, team and cup, and also by Chinese words ... This indicates that ML-LDA has maintained the advantage of traditional topic modeling methods: it is able to well capture the semantic relationships between words. At the same time, it can also capture such semantic patterns between words in different languages. Another observation is that the extracted universal-topics are quite diverse with respect to their meanings. This is consistent with an assumption that the document-aligned Wikipedia corpus has a broad coverage of knowledge." )
comparing the quality against a training set quality; and modifying at least one parameter of the machine learning model when the quality is less than the training set quality to automatically tune the natural language processing performed using the machine learning model. (Ni discloses in 0084 that traditional model training and classification process is followed after multilingual topic determination, but does not explicitly disclose the steps of training)
Ni does not explicitly disclose a predetermined list of topics, or wherein the topic of topics comprises a finite set of groupings of the plurality of topics. Ni also does not explicitly disclose the details of training the machine learning model.
Wang discloses: wherein the machine learning model generates a distribution over a predetermined list of topics for each document in the corpus; ("[0100] In some embodiments, computing device 110 may use a topic model (e.g., an LDA model) to determine the topic distribution of text topic 125 with respect to set of predetermined topics 1210. That is, computing device 110 may determine set of probabilities 1215 corresponding to set of predetermined topics 1210 based on the topic model for determining text topic 125. In order to make the topic model more suitable for technical scenarios of the embodiments of the present disclosure, the topic model for determining text topic 125 may be trained using the set of text associated with multiple predetermined objects 150. In this way, the topic model for determining text topic 125 can learn the topic and word distributions involved by multiple predetermined objects 150, thereby improving the effectiveness of analysis and statistics on text topic 125 of text 120 by the topic model.”)
And wherein the topic of topics comprises a finite set of groupings of the plurality of topics; (using a predetermined list of topics means that there is a finite number of combinations of those topics.)
Ni and Wang are considered analogous art to the claimed invention because they disclose topic modeling. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Ni with predetermined topics as disclosed by Wang. Doing so would have been beneficial in order improve the effectiveness of analysis (Wang [0100]).
Wang does not explicitly disclose the details of training the machine learning model.
Amini discloses: comparing the quality against a training set quality; (“[0085] The minimum may be calculated on the basis of the difference between the two languages.”)
and modifying at least one parameter of the machine learning model when the quality is less than the training set quality to automatically tune the natural language processing performed using the machine learning model. ("[0080] In an embodiment, after the first classifier performs its second iteration and stores the updated first classification in memory, the second classifier performs a second iteration to update the second classification in view of the updated first classification. The difference between the two classifications is reduced after each iteration as well as the classification errors associated with classifier until the training cost reaches a minimum. Several iterations may be performed by each classifier until the training cost between the two classifications reaches the minimum." )
Ni, Wang and Amini are considered analogous art to the claimed invention because they disclose topic modeling. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination with two classifiers which are trained in iterations as disclosed by Amini. Doing so would have been beneficial in order enhance accuracy (Amini [0076]).
Regarding claim 2, Ni discloses: 2. The method of claim 1, further comprising assigning a probability to each topic from the plurality of topics. ("[0018] ...Each topic is defined as plurality of topic/word (T/W) distributions corresponding respectively to languages L1 through Lj. Each topic/word distribution indicates the probability of any particular word occurring in the corresponding topic." )
Regarding claim 3, Ni discloses: 3. The method of claim 2, wherein a sum of probabilities from the assigning adds to one. ("
Σ
w
∈
W
l
p
w
θ
=
1
" [0043])
Regarding claim 5, Ni discloses: 5. The method of claim 1, wherein the quality comprises a topic coherence metric used to determine a degree of semantic similarity between words in each topic. ("[0036]...The philosophy behind topic modeling is to utilize word co-occurrence information for extracting document semantics. For example, if “ipod” and “iTunes” frequently co-occur in a set of documents, both words may have high scores in one topic." )
Regarding claim 7, Ni discloses: 7. The method of claim 1, wherein the quality comprises alignment between multiple languages in the corpus. ("[0080] An action 706 comprises obtaining topic distribution of documents of an unclassified document corpus. This can be accomplished by comparing an unclassified document in a given language to the topic word distributions corresponding to the given language to estimate a topic distribution of the unclassified document. In some cases, it also comprises comparing new documents of a different language to identify one or more groups of the unclassified documents sharing common topic distributions." see also: "[0069]...Taking the 31st universal-topic as an example, it is nicely represented by English words like football, team and cup, and also by Chinese words ... This indicates that ML-LDA has maintained the advantage of traditional topic modeling methods: it is able to well capture the semantic relationships between words." )
Regarding claim 8, Ni does not explicitly disclose iterative training.
Amini discloses: 8. The method of claim 1, further comprising: repeating the evaluating, using and assessing after modifying the at least one parameter of the machine learning model. ("[0080] In an embodiment, after the first classifier performs its second iteration and stores the updated first classification in memory, the second classifier performs a second iteration to update the second classification in view of the updated first classification. The difference between the two classifications is reduced after each iteration as well as the classification errors associated with classifier until the training cost reaches a minimum. Several iterations may be performed by each classifier until the training cost between the two classifications reaches the minimum." )
See claim 1 for motivation statement.
Regarding claim 9, Ni discloses: 9. The method of claim 1, wherein the machine learning model comprises a Latent Dirichlet Allocation (LDA) model used to choose the plurality of topics. ("[0032] An action 320 comprises inferring the plurality of universal topics from the documents 308 of the concept units 306, based on the generative model 312. As will be described, inferring 320 can be accomplished by performing a modified LDA analysis. Other Bayesian and statistical techniques might alternatively be used to infer universal topics 314 and topic distributions 316. [0033] The following exemplary multilingual modeling uses Wikipedia as an exemplary multi-language document corpus and a modified Latent Dirichlet Allocation (LDA) technique as an exemplary multilingual topic modeling algorithm to build a multilingual model and to infer topic distributions and topic/word distributions. The resulting technique will be referred to as multi-lingual LDA (ML-LDA)." )
Regarding claim 10, Ni discloses: 10. The method of claim 9, wherein the Latent Dirichlet Allocation (LDA) model is used to generate the topic of topics. ("[0032] An action 320 comprises inferring the plurality of universal topics from the documents 308 of the concept units 306, based on the generative model 312. As will be described, inferring 320 can be accomplished by performing a modified LDA analysis. Other Bayesian and statistical techniques might alternatively be used to infer universal topics 314 and topic distributions 316. [0033] The following exemplary multilingual modeling uses Wikipedia as an exemplary multi-language document corpus and a modified Latent Dirichlet Allocation (LDA) technique as an exemplary multilingual topic modeling algorithm to build a multilingual model and to infer topic distributions and topic/word distributions. The resulting technique will be referred to as multi-lingual LDA (ML-LDA)." )
Claim 11 is a device claim with limitations corresponding to the limitations of Claim 1 and is rejected under similar rationale. Additionally, “a processor; and memory” of the Claim are taught by Ni. (“Processor 904”; “Memory 906” Fig. 9)
Claim 12 is a device claim with limitations corresponding to the limitations of Claim 2 and is rejected under similar rationale.
Claim 14 is a device claim with limitations corresponding to the limitations of Claim 5 and is rejected under similar rationale.
Claim 16 is a device claim with limitations corresponding to the limitations of Claim 7 and is rejected under similar rationale.
Claim 17 is a device claim with limitations corresponding to the limitations of Claim 8 and is rejected under similar rationale.
Claim 18 is a device claim with limitations corresponding to the limitations of Claim 9 and is rejected under similar rationale.
Claim 19 is a device claim with limitations corresponding to the limitations of Claim 10 and is rejected under similar rationale.
Claim 20 is a medium claim with limitations corresponding to the limitations of Claim 1 and is rejected under similar rationale. Additionally, “A non-transitory computer readable medium” of the Claim are taught by Ni (“Memory 906” Fig. 9)
Claim 21 is a medium claim with limitations corresponding to the limitations of Claim 9 and is rejected under similar rationale.
Claim 22 is a medium claim with limitations corresponding to the limitations of Claim 10 and is rejected under similar rationale.
Claim(s) 6 and 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ni in view of Wang and Amini as applied in claim 1 above, further in view of Skupin et al. (US 20170228654 A1).
Regarding claim 6, Ni discloses: 6. The method of claim 1, wherein the quality comprises a perplexity score when the natural language processing is applied to a new document. ("[0014]... Once defined in this manner, new documents can be compared to the universal topic space for various multi-lingual applications such as document comparison, document recommendation, document classification using a pre-classified corpus, etc." )
Ni does not explicitly disclose a perplexity score. Neither do Wang or Amini.
Skupin discloses: 6. The method of claim 1, wherein the quality comprises a perplexity score when the natural language processing is applied to a new document. ("[0047] When filtering the text corpus data, the number of topics for topic modeling, can be determined. Perplexity is widely used in natural language processing to evaluate the performance of language processing models. The perplexity of topic models with different input topic numbers can be computed to evaluate them. [0048] The data set input into the topic model can be split into a training set and a test held-out set. The training dataset is trained with a different number of topics and is evaluated against all the test held-out datasets to get the log likelihood of each document. Perplexity for the LDA topic model may then be computed based on the length of each document to produce a perplexity graph (FIG. 3). As shown in FIG. 3, four hundred to six hundred topics would be a good range for the example model, because that is where the model seems to only incrementally change and models with a lower perplexity score, in general, have a better performance." )
Ni, Wang, Amini, and Skupin are considered analogous art to the claimed invention because they disclose topic modeling. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination with a perplexity score as disclosed by Skupin. Doing so would have been beneficial in order to evaluate the performance of the model and find a model with better performance (Skupin [0047]). This combination falls under combining prior art elements according to known methods to yield predictable results or use of known technique to improve similar devices (methods, or products) in the same way. See MPEP 2141, KSR, 550 U.S. at 418, 82 USPQ2d at 1396.
Claim 15 is a device claim with limitations corresponding to the limitations of Claim 6 and is rejected under similar rationale.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Mishra et al. (US 20200302011 A1). Mishra discloses topic modeling, including LDA, for predefined topics (ref [0030]).
Qin et al. (US 20170185601 A1). Qin discloses LDA for predetermined topics (ref [0058]).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JON C MEIS whose telephone number is (703)756-1566. The examiner can normally be reached Monday - Thursday, 8:30 am - 5:30 pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Hai Phan can be reached at 571-272-6338. 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.
/JON CHRISTOPHER MEIS/Examiner, Art Unit 2654
/HAI PHAN/Supervisory Patent Examiner, Art Unit 2654
Prosecution Insights