DETAILED ACTION
Introduction
This office action is in response to Applicant’s submission filed on February 19, 2025.
Claims 1-20 are pending in the application. As such, claims 1-20 have been examined.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Drawings
The drawings were received on February 19, 2025. These drawings have been accepted and considered by the Examiner.
Claim Rejections - 35 USC § 102
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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-7 and 12-20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Song et al. (US Patent Pub. No. 20210224660 A1), hereinafter Song ‘4660.
Regarding claims 1, 13 and 20, Song ‘4660 teaches a language model updating method, and electronic device, and a non-transitory computer-readable storage medium (Song ‘4660 in [0016] teaches training a student language model, in [0007] teaches using electronic devices, and in [0004] teaches using one or more non-transitory computer-readable media),
[claim 13 only] one or more processors (Song ‘4660 in [0004] teaches using one or more processors);
[claim 13 only] memory storing instructions that, when executed by the one or more processors, cause the electronic device to (Song ‘4660 in [0006] teaches using instructions that, when executed by one or more processors, cause the one or more processors to perform operations, and in [0053] teaches the user computing device includes one or more processors and a memory):
[claim 20 only] non-transitory computer-readable storage medium having instructions recorded thereon, that, when executed by one or more processors, cause the one or more processors to (Song ‘4660 in [0053] teaches the user computing device includes one or more processors and a memory. The one or more processors can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, a FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. The memory can include one or more non-transitory computer-readable storage mediums, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory can store data and instructions which are executed by the processor to cause the user computing device to perform operations):
performed by an electronic device (Song ‘4660 in [0007] teaches using electronic devices),
comprising:
acquiring a first training sequence (Song ‘4660 in [0004] teaches using a first sub-word version of the natural language training input)
comprising at least one of: text used for training or a token sequence corresponding to the text used for training (Song ‘4660 in [0004] teaches using a first sub-word version of the natural language training input);
extracting a first general knowledge representation of the first training sequence using a first language model (Song ‘4660 in [0005] teaches using an output generated by the language model based on the sub-word version of the natural language training input, and in [0004] teaches using a teacher language model);
using a second language model to (Song ‘4660 in [0004] teaches using a student language model):
extract a second general knowledge representation of the first training sequence (Song ‘4660 in [0088] teaches using a second sub-word version of the natural language training input),
and
determine a plurality of prediction results corresponding to the first training sequence (Song ‘4660 in [0081] teaches the outputs can be an output for any number of training tasks such as masked language modeling, two-way sentence sentiment classification, two-way sentence pair classification, next sentence prediction, and/or others);
and
updating the second language model based on (Song ‘4660 in [0004] teaches modifying at least one of the plurality of student sub-word embeddings based at least in part on the loss associated with the teacher output)
the first general knowledge representation (Song ‘4660 in [0005] teaches using an output generated by the language model based on the sub-word version of the natural language training input),
the second general knowledge representation (Song ‘4660 in [0088] teaches using a second sub-word version of the natural language training input),
and
the plurality of prediction results (Song ‘4660 in [0081] teaches the outputs can be an output for any number of training tasks such as masked language modeling, two-way sentence sentiment classification, two-way sentence pair classification, next sentence prediction, and/or others),
wherein each of the first general knowledge representation and the second general knowledge representation comprise at least one of: language patterns or semantic relationships, extracted from the first training sequence (Song ‘4660 in [0027] teaches a language model can perform any number of different language processing tasks. Example language processing tasks include asked language modeling, two-way sentence sentiment classification, two-way sentence pair classification, next sentence prediction, medical record analysis/auto-completion, search query processing, translation, artificial intelligence-based assistant tasks such as question answering, tokenization; sentence boundary detection; shallow parsing; part-of-speech tagging; selection preference; syntax parsing; dependency parsing; constituency parsing; semantics; semantic role labeling; spatial role labeling; opinion mining; word sense disambiguation/induction; and/or others).
Regarding claims 2 and 14, Song ‘4660 teaches the language model updating method and electronic device of claims 1 and 13.
Song ‘4660 further teaches
wherein the updating the second language model comprises:
[claim 14 only] wherein the instructions, when executed by the one or more processors, cause the electronic device to:
determining a first loss based on the first general knowledge representation and the second general knowledge representation (Song ‘4660 in [0006] teaches determining a loss value based on a difference between the teacher parameter values of the teacher intermediate layer and the student parameter values of the student intermediate layer within the shared space);
determining a second loss based on the plurality of prediction results (Song ‘4660 in [0022] teaches a separate loss can be introduced to align the teacher and student language models' trainable variables);
and
updating the second language model based on the first loss and the second loss (Song ‘4660 in [0004] teaches modifying at least one of the plurality of student sub-word embeddings based at least in part on the loss associated with the teacher output [this can be repeated using the first and second losses]).
Regarding claims 3 and 15, Song ‘4660 teaches the language model updating method and electronic device of claims 1 and 13.
Song ‘4660 further teaches
wherein the extracting the first general knowledge representation of the first training sequence comprises:
[claim 15 only] wherein the instructions, when executed by the one or more processors, cause the electronic device to:
determining a first hidden state of the first training sequence using a first encoder in the first language model (Song ‘4660 in [0031] teaches using dimensional embeddings and hidden states, and in [0025] teaches using the Bidirectional Encoder Representations from Transformers (BERT) model, and the BERTBAsE model);
and
determining a first prediction probability for a first plurality of tokens in the first training sequence based on the first hidden state (Song ‘4660 in [0050] teaches using a final loss function which includes, in addition to an optional projection loss, masked language modeling cross-entropy losses for the student as well as the teacher models, since, in some implementations, the teacher model is trained with dual-vocabulary inputs and is not static, and P(y.sub.i=C|θ.sub.s) and P(y.sub.i=c|θ.sub.t) denote the student and teacher model prediction probabilities for class c respectively, and 1 denotes an indicator function, and in [0004] teaches using a first sub-word version of the natural language training input),
wherein the first prediction probability is taken as the first general knowledge representation (Song ‘4660 in [0031] teaches using a general knowledge distillation paradigm of training a smaller student model from a larger teacher model).
Regarding claims 4 and 16, Song ‘4660 teaches the language model updating method and electronic device of claims 1 and 13.
Song ‘4660 further teaches
wherein the using the second language model comprises:
[claim 16 only] wherein the instructions, when executed by the one or more processors, cause the electronic device to:
determining a second hidden state of the first training sequence using a second encoder in the second language model (Song ‘4660 in [0031] teaches using dimensional embeddings and hidden states, and in [0025] teaches using the Bidirectional Encoder Representations from Transformers (BERT) model, and the BERTBAsE model [this can be repeated using the alternate model to get the second versions of the data]);
and
determining a second prediction probability for a first plurality of tokens in the first training sequence based on the second hidden state (Song ‘4660 in [0050] teaches using a final loss function which includes, in addition to an optional projection loss, masked language modeling cross-entropy losses for the student as well as the teacher models, since, in some implementations, the teacher model is trained with dual-vocabulary inputs and is not static, and P(y.sub.i=C|θ.sub.s) and P(y.sub.i=c|θ.sub.t) denote the student and teacher model prediction probabilities for class c respectively, and 1 denotes an indicator function, and in [0004] teaches using a first sub-word version of the natural language training input [this can be repeated using the alternate model to get the second versions of the data]),
wherein the second prediction probability is taken as the second general knowledge representation (Song ‘4660 in [0031] teaches using a general knowledge distillation paradigm of training a smaller student model from a larger teacher model [this can be repeated using the alternate model to get the second versions of the data]).
Regarding claims 5 and 17, Song ‘4660 teaches the language model updating method and electronic device of claims 1 and 13.
Song ‘4660 further teaches
further comprising:
[claim 17 only] wherein the instructions, when executed by the one or more processors, cause the electronic device to:
obtaining a masked sequence for a second training sequence (Song ‘4660 in [0034] teaches using masked language modeling, two-way sentence sentiment classification, two-way sentence pair classification, and/or next sentence prediction can be performed, for example followed by fine-tuning. In the masked language modeling task, words, phrases, and/or entities in context are randomly masked, and the language model needs to predict those words, phrases, or entities given the masked context),
wherein at least a portion of tokens of the second training sequence are masked in the masked sequence (Song ‘4660 in [0034] teaches using masked language modeling, two-way sentence sentiment classification, two-way sentence pair classification, and/or next sentence prediction can be performed, for example followed by fine-tuning. In the masked language modeling task, words, phrases, and/or entities in context are randomly masked, and the language model needs to predict those words, phrases, or entities given the masked context);
and
updating the first language model based on the second training sequence and the masked sequence (Song ‘4660 in [0004] teaches modifying at least one of the plurality of student sub-word embeddings based at least in part on the loss associated with the teacher output, and in [0006] teaches modifying at least one of the teacher parameter values of the teacher intermediate layer or the student parameter values of the student intermediate layer to reduce the loss value).
Regarding claims 6 and 18, Song ‘4660 teaches the language model updating method and electronic device of claims 5 and 17.
Song ‘4660 further teaches
wherein the updating the first language model based on the second training sequence and the masked sequence comprises:
[claim 18 only] wherein the instructions, when executed by the one or more processors, cause the electronic device to:
determining a third hidden state of the masked sequence and a fourth hidden state of the second training sequence using a first encoder in the first language model (Song ‘4660 in [0031] teaches using dimensional embeddings and hidden states, and in [0025] teaches using the Bidirectional Encoder Representations from Transformers (BERT) model, and the BERTBAsE model [this can be repeated using the alternate model to get the third and fourth versions of the data]);
determining a third prediction probability for a second plurality of tokens in the masked sequence based on the third hidden state (Song ‘4660 in [0050] teaches using a final loss function which includes, in addition to an optional projection loss, masked language modeling cross-entropy losses for the student as well as the teacher models, since, in some implementations, the teacher model is trained with dual-vocabulary inputs and is not static, and P(y.sub.i=C|θ.sub.s) and P(y.sub.i=c|θ.sub.t) denote the student and teacher model prediction probabilities for class c respectively, and 1 denotes an indicator function, and in [0004] teaches using a first sub-word version of the natural language training input [this can be repeated using the alternate model to get the third versions of the data and the second plurality of tokens]),
determining a fourth prediction probability for a third plurality of tokens in the second training sequence based on the fourth hidden state (Song ‘4660 in [0050] teaches using a final loss function which includes, in addition to an optional projection loss, masked language modeling cross-entropy losses for the student as well as the teacher models, since, in some implementations, the teacher model is trained with dual-vocabulary inputs and is not static, and P(y.sub.i=C|θ.sub.s) and P(y.sub.i=c|θ.sub.t) denote the student and teacher model prediction probabilities for class c respectively, and 1 denotes an indicator function, and in [0004] teaches using a first sub-word version of the natural language training input [this can be repeated using the alternate model to get the fourth versions of the data and the third plurality of tokens]);
determining a third loss based on the third prediction probability (Song ‘4660 in [0022] teaches a separate loss can be introduced to align the teacher and student language models' trainable variables [this can be repeated to obtain the third loss]);
determining a fourth loss based on the fourth prediction probability (Song ‘4660 in [0022] teaches a separate loss can be introduced to align the teacher and student language models' trainable variables [this can be repeated to obtain the fourth loss]);
and
updating the first language model based on the third loss and the fourth loss (Song ‘4660 in [0004] teaches modifying at least one of the plurality of student sub-word embeddings based at least in part on the loss associated with the teacher output [this can be repeated using the third and fourth losses]).
Regarding claims 7 and 19, Song ‘4660 teaches the language model updating method and electronic device of claims 6 and 18.
Song ‘4660 further teaches
wherein the determining the third loss based on the third prediction probability comprises:
[claim 19 only] wherein the instructions, when executed by the one or more processors, cause the electronic device to:
determining the third loss based on (Song ‘4660 in [0022] teaches a separate loss can be introduced to align the teacher and student language models' trainable variables [this can be repeated to obtain the third loss])
the third prediction probability for a fourth plurality of tokens in a second section (Song ‘4660 in [0050] teaches using a final loss function which includes, in addition to an optional projection loss, masked language modeling cross-entropy losses for the student as well as the teacher models, since, in some implementations, the teacher model is trained with dual-vocabulary inputs and is not static, and P(y.sub.i=C|θ.sub.s) and P(y.sub.i=c|θ.sub.t) denote the student and teacher model prediction probabilities for class c respectively, and 1 denotes an indicator function, and in [0004] teaches using a first sub-word version of the natural language training input [this can be repeated using the alternate model to get the third versions of the data and the fourth plurality of tokens], and in [0034] teaches using masked language modeling where task, words, phrases, and/or entities in context are randomly masked, and the language model needs to predict those words, phrases, or entities given the masked context, and in the next sentence prediction task, given a pair of sentences, the language model predicts whether the pair is consistent [second section is mapped to randomly masked portion])
and
a plurality of real tokens corresponding to the fourth plurality of tokens in the second section (Song ‘4660 in [0050] teaches using a final loss function which includes, in addition to an optional projection loss, masked language modeling cross-entropy losses for the student as well as the teacher models, since, in some implementations, the teacher model is trained with dual-vocabulary inputs and is not static, and P(y.sub.i=C|θ.sub.s) and P(y.sub.i=c|θ.sub.t) denote the student and teacher model prediction probabilities for class c respectively, and 1 denotes an indicator function, and in [0004] teaches using a first sub-word version of the natural language training input [this can be repeated using the alternate model to get the third versions of the data and the fourth plurality of tokens (here the fourth plurality of tokens are the real tokens in the plurality of real tokens)], and in [0034] teaches using masked language modeling where task, words, phrases, and/or entities in context are randomly masked, and the language model needs to predict those words, phrases, or entities given the masked context, and in the next sentence prediction task, given a pair of sentences, the language model predicts whether the pair is consistent [second section is mapped to randomly masked portion]),
wherein the determining the fourth loss based on the fourth prediction probability comprises:
determining the fourth loss based on the fourth prediction probability for a fifth plurality of tokens in a third section and the third prediction probability for the fourth plurality of tokens in the second section (Song ‘4660 in [0050] teaches using a final loss function which includes, in addition to an optional projection loss, masked language modeling cross-entropy losses for the student as well as the teacher models, since, in some implementations, the teacher model is trained with dual-vocabulary inputs and is not static, and P(y.sub.i=C|θ.sub.s) and P(y.sub.i=c|θ.sub.t) denote the student and teacher model prediction probabilities for class c respectively, and 1 denotes an indicator function, and in [0004] teaches using a first sub-word version of the natural language training input [this can be repeated using the alternate model to get the fourth versions of the data and the fifth plurality of tokens and repeated for each corresponding section], and in [0034] teaches using masked language modeling where task, words, phrases, and/or entities in context are randomly masked, and the language model needs to predict those words, phrases, or entities given the masked context, and in the next sentence prediction task, given a pair of sentences, the language model predicts whether the pair is consistent [second section is mapped to randomly masked portion] [third section is mapped to pair of sentences where one sentence is the target]),
and
wherein
the second section indicates a sequence for the at least the portion of the tokens in which the at least the portion of the tokens are masked (Song ‘4660 in [0034] teaches using masked language modeling where task, words, phrases, and/or entities in context are randomly masked, and the language model needs to predict those words, phrases, or entities given the masked context, and in the next sentence prediction task, given a pair of sentences, the language model predicts whether the pair is consistent [second section is mapped to randomly masked portion]),
and
the third section indicates a non-masked sequence for the at least the portion of the tokens in which the at least the portion of the tokens are not masked (Song ‘4660 in [0034] teaches using masked language modeling where task, words, phrases, and/or entities in context are randomly masked, and the language model needs to predict those words, phrases, or entities given the masked context, and in the next sentence prediction task, given a pair of sentences, the language model predicts whether the pair is consistent [third section is mapped to pair of sentences where one sentence is the target]).
Regarding claim 12, Song ‘4660 teaches the language model updating method of claim 1.
Song ‘4660 above teaches the second language model.
Song ‘4660 further teaches
wherein [the second language model] comprises at least one of:
a smart reply model for text call reply, short message reply, email reply, chat reply, and a machine translation model (Song ‘4660 in [0071] teaches using applications which include a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, etc., and in [0081] teaches a teacher output is generated by the teacher language model based on the first sub-word version of the natural language training input).
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.
Claim 8 is rejected under 35 U.S.C. 103 as being unpatentable over Song ‘4660, in view of Song et al. (US Patent Pub. No. 20210374536 A1), hereinafter Song ‘4536.
Regarding claim 8, Song ‘4660 teaches the language model updating method of claim 6.
Song ‘4660 above teaches determining the third hidden state of the masked sequence and the fourth hidden state of the second training sequence using the first encoder in the first language model, the masked sequence and the second training sequence, and determining the third hidden state of the masked sequence and the fourth hidden state of the second training sequence using the first encoder in the first language model.
Song ‘4660 does not teach, however Song ‘4536 teaches
[wherein the determining the third hidden state of the masked sequence and the fourth hidden state of the second training sequence using the first encoder in the first language model comprises:
determining, based on the masked sequence and the second training sequence],
a contextual mask matrix in which values of elements in a first column corresponding to a non-masked token are a first value, and a value of an element on a diagonal is the first value and values of remaining elements are a second value in a second column corresponding to a masked token, wherein the first value is 0, and the second value is greater than a predetermined value (Song ‘4536 in [0087 and Fig. 11] teaches using a first mask matrix generator which may generate a first pre-mask matrix by setting tokens included in the first character string information as rows and columns. Also, the first mask matrix generator may determine neighboring nodes adjacent to each other from the first graph information based on a preset reference distance. Also, the first mask matrix generator may generate the first mask matrix based on the determined neighboring nodes. For example, the first mask matrix generator may generate the first mask matrix by allocating information about nodes for calculating a self-attention matrix to the first pre-mask matrix, and by assigning ‘1’ to neighboring nodes adjacent to the reference node and ‘0’ to the remaining nodes. The generation of the first mask matrix by the first mask matrix generator based on neighboring nodes of the reference node is to train the retrosynthesis prediction model by paying further attention to the neighboring nodes that have a large chemical relationship with the reference node, and Fig. 11 shows corresponding values of elements may be on a diagonal);
PNG
media_image1.png
286
308
media_image1.png
Greyscale
and
[determining the third hidden state of the masked sequence and the fourth hidden state of the second training sequence using the first encoder in the first language model], based on the contextual mask matrix (Song ‘4536 in [0087 and Fig. 11] teaches using a first mask matrix generator which may generate a first pre-mask matrix by setting tokens included in the first character string information as rows and columns. Also, the first mask matrix generator may determine neighboring nodes adjacent to each other from the first graph information based on a preset reference distance. Also, the first mask matrix generator may generate the first mask matrix based on the determined neighboring nodes. For example, the first mask matrix generator may generate the first mask matrix by allocating information about nodes for calculating a self-attention matrix to the first pre-mask matrix, and by assigning ‘1’ to neighboring nodes adjacent to the reference node and ‘0’ to the remaining nodes. The generation of the first mask matrix by the first mask matrix generator based on neighboring nodes of the reference node is to train the retrosynthesis prediction model by paying further attention to the neighboring nodes that have a large chemical relationship with the reference node).
Song ‘4536 is considered to be analogous to the claimed invention because it is in the same field of using a masked matrix. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Song ‘4660 further in view of Song ‘4536 to allow for using a first mask matrix generator which may generate a first pre-mask matrix by setting tokens included in the first character string information as rows and columns. Motivation to do so would allow for accurately predicting reactants from products by using neural networks (Song ‘4536 [0004]).
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Song ‘4660, in view of Krishnaswamy et al. (US Patent Pub. No. 20200174759 A1), hereinafter Krishnaswamy.
Regarding claim 9, Song ‘4660 teaches the language model updating method of claim 1.
Song ‘4660 above teaches the second language model, and the first training sequence.
Song ‘4660 does not teach, however Krishnaswamy teaches
further comprising:
outputting a user interface, via a display of the electronic device, comprising a setup interface for [the second language model], wherein the setup interface comprises at least one of: an interface indicating whether self-updating is on; an interface for authorizing obtaining of [the first training sequence]; an interface for selecting [the first training sequence]; an interface for downloading [the second language model]; and an interface for setting a self-updating frequency (Krishnaswamy in [0006] teaches using a setup interface for a model (MLM), regarding the updating of the model, and regarding the selection of certain data).
Krishnaswamy is considered to be analogous to the claimed invention because it is in the same field of using displays with user interfaces. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Song ‘4660 further in view of Krishnaswamy to allow for using a setup interface for a model (MLM), regarding the updating of the model. Motivation to do so would allow for providing a user with an indication of the degree of compliance with Agile principles thereby enabling the user to further modify the potential location independent Agile delivery model such that the constraints are optimally managed (Krishnaswamy [0045]).
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Song ‘4660, in view of Kohita et al. (US Patent Pub. No. 20200364298 A1), hereinafter Kohita.
Regarding claim 10, Song ‘4660 teaches the language model updating method of claim 1.
Song ‘4660 above teaches the first training sequence.
Song ‘4660 does not teach, however Kohita teaches
wherein [the first training sequence] is related to user activity associated with a user of the electronic device, the user activity comprising at least one of a chat history of the user, an email history of the user, or a text history of the user (Kohita in [0026] teaches the preferred words and non-preferred words can be identified automatically by a computer system, such as a mobile device, a tablet device, and/or cloud-based servers, based on historical user activity (e.g., words used by user in text messages, web search terms and results, etc.) monitored by the computer system and processed by trained machine-learning models).
Kohita is considered to be analogous to the claimed invention because it is in the same field of using user activity. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Song ‘4660 further in view of Kohita to allow for words to be identified automatically by a computer system, such as a mobile device, a tablet device, and/or cloud-based servers, based on historical user activity. Motivation to do so would allow for providing a user with a plurality of words with features similar to a seed word, as the candidate words (Kohita [0023]).
Claim 11 is rejected under 35 U.S.C. 103 as being unpatentable over Song ‘4660, in view of Kristensson et al. (US Patent Pub. No. 20160062458 A1), hereinafter Kristensson.
Regarding claim 11, Song ‘4660 teaches the language model updating method of claim 1.
Song ‘4660 above teaches the first training sequence.
Song ‘4660 does not teach, however Kristensson teaches
wherein [the second language model] is a personalized language model configured to adapt to text associated with a user of the electronic device (Kristensson in [0074] teaches using a language model (i.e., a word likely to be used again), allowing the user to personalize a vocabulary used by the decoder module. In a further embodiment, the system can optionally determine a measurement of manual engagement with the string based on user interactions. Such interactions comprise the number and/or proportion of locked letters indicated while entering the string, and the number and/or proportion of characters that have been edited using an error correction interfaces. The measurement can comprise a determined confidence value based on the interactions, and the confidence value can have a configurable threshold. In a yet further embodiment, a language model can use the confidence value to determine if or how often the language model should adapt to the user's entered text, i.e., if the user frequently corrects the system-selected hypotheses, the system may adapt and select hypotheses that are more similar to the user's entered text. Higher confidence values can result in stronger adaptation, allowing the system to incorporate words that are important to the user).
Kristensson is considered to be analogous to the claimed invention because it is in the same field of personalizing a language model. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Song ‘4660 further in view of Kristensson to allow for adapting a language model to a user's entered text. Motivation to do so would allow for a user to personalize a vocabulary used by the decoder module (Kristensson [0074]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PAUL J. MUELLER whose telephone number is (571)272-1875. The examiner can normally be reached M-F 9:00am-5:00pm (Eastern).
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, Daniel C. Washburn can be reached at 571-272-5551. 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.
PAUL MUELLER
Examiner
Art Unit 2657
/PAUL J. MUELLER/Examiner, Art Unit 2657
Prosecution Insights