DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Amendments
This Office Action is in response to the amendment filed on 6/9/2026.
No claims have been amended.
Claims 1, 3, 6, and 7 have been cancelled.
Claims 9 and 10 have been added.
The objections and rejections from the prior correspondence that are not restated herein are withdrawn.
Response to Arguments
Applicant's arguments filed on 6/9/2026 have been fully considered.
Applicant's arguments regarding the 35 U.S.C. 101 rejections of the previous office action have been fully considered but are not persuasive. In pages 4-9 of the remarks, applicant argues that claim 9 is not abstract but unconventional and non-routine because it recites a unique ordered combination of technical elements that utilizes unique hardware, solves a technical problem in data science, integrates into a practical application through a particular link between embedding vectors and portfolio optimization, and amounts to significantly more than a judicial exception because the combination of weighting texts and computing a similarity matrix using the same embedding vectors is not well-understood, routine, or conventional.
Examiner respectfully disagrees. The steps of computing inner products, applying a softmax function, computing cosine similarities, and minimizing a risk defined by a similarity matrix are mathematical concepts under Step 2A Prong 1. Further, characterizing the combination as unconventional does not confer eligibility under 101. Using the same embedding vectors to compute weights for texts and a similarity matrix does not change the analysis because that relationship merely defines how the data is generated and reused in a pipeline for determining a portfolio. Additionally, claim 9 recites a method without any indication of the hardware being used to perform the method, and analogous independent non-transitory storage medium claim 10 recites only generic computer components. Neither of these claims recites a particular machine (i.e., unique hardware) or technological improvement that would constitute integration into a practical application. The alleged technological improvement in the form of a processing pipeline that acquires embedding vectors from texts and uses those vectors for portfolio optimization only provides an improvement of the financial output for improving the portfolio construction decision-making process. According to MPEP § 2106.05(a), the judicial exception alone cannot provide the improvement, and the improvement must be provided by one or more additional elements. Considering the claim as a whole, meaning, analyzing if the abstract idea is integrated into a practical application or provides an improvement to technology under Step 2A Prong 2 and analyzing if the abstract idea amounts to significantly more under Step 2B, the claimed invention does not integrate into a practical application, provide an improvement to technology, or amount to significantly more than the abstract idea. Further, the recited ordered combination of known mathematical steps applied to financial data does not amount to significantly more than the abstract idea under Step 2B, as shown in the detailed 101 rejections below.
Applicant's arguments regarding the 35 U.S.C. 103 rejections of the previous office action have been fully considered but are not persuasive. Applicant argues:
“However, claims 1, 3, 6, and 7 have been canceled, and new independent claim 9, and new analogous independent claim 10 have been amended herein with features that are neither taught, suggested nor disclosed in the primary references of Hu ("Listening to Chaotic Whispers: A Deep Learning Framework for News-oriented Stock Trend Prediction") in view of Guosheng ("Deep Stock Representation Learning: From Candlestick Charts To Investment Decisions"), Qi ("Semantic Enhancement and Multi-level Label Embedding for Chinese News Headline Classification"), Miller ("Key-Value Memory Networks for Directly Reading Documents"), and Markowitz ("Portfolio Selection"), hereafter Hu, Guosheng, Qi, Miller, and Markowitz respectively. The differences between new independent claims 9 and 10, and Hu, Guosheng, Qi, Miller, and Markowitz are explained below.”
Examiner respectfully disagrees. Amended claims 9 and 10 are taught by the combination of HU, Qi, MILLER, GUOSHENG, and ZHENG, as shown in the detailed 103 rejection below.
Applicant further argues:
“The Office Action relies on Guosheng for the similarity matrix. However, Guosheng derives similarity based on price-related information and historical price series. In contrast, the present claims require computing a similarity matrix based on similarities between embedding vectors corresponding to targets. These embedding vectors are used in computing weights for texts and reflect relationships between texts and targets. Accordingly, the similarity in the present claims is directed to a different type of similarity from the similarity in Guosheng.”
Examiner respectfully disagrees. GUOSHENG teaches learned Convolutional autoencoder (CAE) feature vectors corresponding to stocks and computes cosine similarity between those learned vectors to determine stock similarity. GUOSHENG further states that the learned vectors capture semantic information about stocks, as shown by the t-SNE visualization in GUOSHENG [pg. 2709, Fig. 3.], where stocks are grouped by industrial sector. These learned features capture both key quantitative (e.g., historical time-series price-related data) and semantic information (see GUOSHENG [pg. 2707, section 2 Methodology]).
Applicant further argues:
“The Office Action relies on Miller for inner-product-based computation. Miller computes addressing probabilities between a query and memory entries in a question-answering framework. In contrast, the present claims compute weights for texts for each target based on an embedding vector corresponding to the target. Thus, the claimed computation is target-dependent and produces different weights for different targets, which is not disclosed in Miller.
“Even if the individual elements were known, the cited references do not suggest using embedding vectors both:
(i) for computing weights for texts for each target, and
(ii) for computing a similarity matrix used in portfolio determination.
The cited references do not appear to provide a suggestion to combine the elements in the manner recited in the claims.
The Examiner's proposed combination of Hu, Guosheng, Qi, Miller, and Markowitz therefore still depends on impermissible hindsight, since none of the cited references-alone or in combination-discloses or suggests these distinctive features of the present invention.”
Examiner respectfully disagrees. In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981 ); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986). It should be noted that the combination of HU, QI, and MILLER teaches the claimed limitation as outlined in the 103 rejections below. More specifically, and as similarly stated in the previous Office Action, the MILLER reference teaches a method for answering questions from text by using a knowledge base. Further, one of ordinary skill having the teachings of HU, which uses natural language processing methods to process news for predicting trends and constructing a portfolio, would see the benefit of using QI’s finance news article quantitative and semantic information feature learning, and using MILLER’s assigning of relevance probabilities and weighted sums for identifying and weighting relevant text information in HU’s stock price trend prediction framework for constructing the portfolio.
Applicant further argues:
“The Office Action relies on Markowitz for portfolio optimization. Markowitz uses a covariance matrix of returns. In contrast, the present claims use a similarity matrix based on embedding vectors. Thus, the claimed risk is not based on covariance of returns but on similarities between embedding vectors.”
Examiner respectfully disagrees. MARKOWITZ is not relied upon for teaching any limitations in this Office Action.
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 9 and 10 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: Claim 9 is directed to a process. Claim 10 is directed to a machine or an article of manufacture.
With respect to claim(s) 9 and 10:
2A Prong 1: The claim(s) recite(s) an abstract idea. Specifically:
(Claim 9) A method for determining a portfolio for a plurality of targets (Mental process – A person can mentally determine a portfolio for a plurality of targets – see MPEP § 2106.04(a)(2)(III))
(Claim 10) […] determining a portfolio for a plurality of targets […] (Mental process – A person can mentally determine a portfolio for a plurality of targets – see MPEP § 2106.04(a)(2)(III))
extracting/extract, for each text, a first feature vector representing a word-level feature, and a second feature vector representing a context-level feature; (Mathematical concept – extracting a first feature vector representing a word-level feature and a second feature vector representing a context-level feature vector involves mathematical calculations (see paragraph [0041]) – see MPEP § 2106.04(a)(2)(I))
computing, for each target and for each text, a score based on an inner product between: the first feature vector and an embedding vector corresponding to the target; (Mathematical concept – determining a score based on an inner product between vectors involves mathematical calculations – see MPEP § 2106.04(a)(2)(I))
computing, for each target, weights for the texts using a softmax function applied to the scores; (Mathematical concepts – Computing weights using a softmax function involves mathematical calculations – see MPEP § 2106.04(a)(2)(I))
generating, for each target, a status vector by computing a weighted sum of the second feature vectors using the weights; (Mathematical concepts – Generating a status vector by computing a weighted sum involves mathematical calculations (see paragraph [0050]) – see MPEP § 2106.04(a)(2)(I))
computing, after the training, a similarity matrix based on cosine similarities between the embedding vectors, and (Mathematical concept – Computing a similarity matrix based on cosine similarities involves mathematical calculations (see paragraphs [0081]) – see MPEP § 2106.04(a)(2)(I))
determining a portfolio vector that minimizes a risk defined using the similarity matrix. (Mathematical concepts – This limitation involves mathematical calculations (see paragraphs [0082-0083]) – see MPEP § 2106.04(a)(2)(I))
If claim limitations, under their broadest reasonable interpretation, cover performance of the limitations as a mental process, but for the recitation of generic computer components, then the claim limitations fall within the mathematical or mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
2A Prong 2: The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination.
Additional elements:
(Claim 10) A non-transitory information recording medium recording a program for […] the program causing a computer to (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
receiving/receive texts released at a plurality of dates and times; (Mere data gathering – Adding insignificant extra-solution activity of mere data gathering to the judicial exception – see § MPEP2106.05(g).)
training/train a model, including learning the embedding vectors, to output classifications indicating whether prices of the targets have increased or decreased based on the status vectors; (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
2B: The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
(Claim 10) A non-transitory information recording medium recording a program for […] the program causing a computer to (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
receiving/receive texts released at a plurality of dates and times; (Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception (WURC)- see MPEP § 2106.05(d)(ll)(i) - Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information).)
training/train a model, including learning the embedding vectors, to output classifications indicating whether prices of the targets have increased or decreased based on the status vectors; (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Claim Rejections - 35 USC § 103
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.
Claims 9 and 10 are rejected under 35 U.S.C. 103 as being unpatentable over HU ("Listening to Chaotic Whispers: A Deep Learning Framework for News-oriented Stock Trend Prediction") in view of QI ("Semantic Enhancement and Multi-level Label Embedding for Chinese News Headline Classification"), MILLER ("Key-Value Memory Networks for Directly Reading Documents”), GUOSHENG ("Deep Stock Representation Learning: From Candlestick Charts To Investment Decisions"), and ZHENG (“DIVERSITY AND SPARSITY: A NEW PERSPECTIVE ON INDEX TRACKING”), hereafter HU, QI, MILLER, GUOSHENG, and ZHENG respectively.
Regarding Claim 9:
HU teaches:
A method for determining a portfolio for a plurality of targets, comprising: (HU [pg. 268, section 5.4] teaches: "Based on these scores, a straightforward portfolio construction strategy called top-K selects K stocks with the highest scores to construct a new portfolio for the next trading day.")
receiving texts released at a plurality of dates and times; (HU [pg. 264, section I. Introduction] teaches: “The stock trend prediction task can be formulated as follows: given the length of a time sequence
N
, the stock
s
and date
t
, the goal is to use the news corpus sequence (i.e., receiving texts) from time
t
-
N
to
t
-
1
, denoted as
[
C
t
-
N
,
C
t
-
N
+
1
,
.
.
.
,
C
t
-
1
]
,
(i.e., released at a plurality of dates and times) to predict the class of
R
i
s
e
_
P
e
r
c
e
n
t
(
t
)
, i.e. DOWN, UP, or PRESERVE. Note that each news corpus
C
i
contains a set of news with the size of
L
,
C
i
=
n
i
1
,
n
i
2
,
…
,
n
i
L
]
, denoting
L
related news on date
i
.” Examiner’s note: Specification paragraph [0024] of the instant application states: “As a unit of date and time, an appropriate unit, such as 1 day, 12 hours, 1 hour, and 30 minutes, can be employed, and dates and times are denoted by integers in chronological order. Therefore, the date and time immediately before a date and time t is denoted by t-1.” Therefore, under BRI, a plurality of dates and times can be interpreted as a news corpus sequence from time
t
-
N
to
t
-
1
.)
extracting, for each text, a first feature vector representing a word-level feature, […]; (HU [pg. 264, section 4.2 Hybrid Attention Networks] teaches: “For each
i
t
h
news (i.e., for each text) in news corpus
C
t
of date
t
, we use a word embedding layer to calculate the embedded vector for each word (i.e., extracting […] a word level feature vector representing a word-level feature) and then average all the words’ vectors to construct a news vector
n
t
i
.”)
training a model […] to output classifications indicating whether prices of the targets have increased or decreased based on the status vectors; (HU [pg. 265, Temporal Attention] teaches: “we use β to calculate the weighted sum
V
, so that it can incorporate the sequential news context information with temporal attention, and will be used for classification.” HU [pg. 265, section 4.2 Hybrid Attention Networks] teaches: "Trend Prediction: The final discriminative network is a standard Multi-layer Perceptron (MLP), which takes
V
as input (i.e., based on the status vectors) and produces the three-class classification (i.e., training a model […] to output classifications) of the future stock trend." HU [pg. 262, section I. Introduction] teaches: “The stock trend prediction task can be formulated as follows: given the length of a time sequence
N
, the stock
s
and date
t
, the goal is to use the news corpus sequence from time
t
-
N
to
t
-
1
, denoted as
[
C
t
-
N
,
C
t
-
N
+
1
,
.
.
.
,
C
t
-
1
]
,
to predict the class of
R
i
s
e
_
P
e
r
c
e
n
t
(
t
)
, i.e. DOWN, UP, or PRESERVE (i.e., classifications indicating whether prices of the targets have increased or decreased).)
HU is not relied upon for teaching:
extracting, for each text, […] a second feature vector representing a context-level feature;
computing, for each target and for each text, a score based on an inner product between: the first feature vector and an embedding vector corresponding to the target;
computing, for each target, weights for the texts using a softmax function applied to the scores;
generating, for each target, a status vector by computing a weighted sum of the second feature vectors using the weights;
training a model, including learning the embedding vectors, […]
computing, after the training, a similarity matrix based on cosine similarities between the embedding vectors, and
determining a portfolio vector that minimizes a risk defined using the similarity matrix.
However, QI teaches: extracting, for each text, […] a second feature vector representing a context-level feature; (QI [pg. 2, section B. Label Embedding] teaches: "In this paper, we attempt to construct a multi-level label embedding strategy to represent as a word-level vector and sentence-level text for news classification.” QI [pg. 4, section B. Multi-dimensional Feature Fusion Network] teaches: “In order to represent more text features and make the sentences have both sequence features and local features, we aggregate the representation results of Bi-GRU and multi-scale CNN together to form the final sentence representation (i.e., a second feature vector representing a context-level feature) […]". QI [pg. 8, section V. Conclusion] teaches: "Moreover, a joint model of the bidirectional GRU and multiscale CNN is designed as the extractor of sentence (i.e., extracting) features to expand sentence semantics from multi-dimensions." QI [pg. 3, Fig. 1.] teaches text being input for feature extraction (i.e., extracting, for each text). Examiner’s note: Under broadest reasonable interpretation, a context-level feature vector can be interpreted as the final sentence representation.)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of HU and QI before them, to include QI’s sentence-level feature extraction in HU’s news-oriented stock prediction framework. One would have been motivated to make such a combination in order to represent more text features and make the sentences have both sequence features and local features and expand sentence semantics from multi-dimensions (QI [pg. 4, section B. Multi-dimensional Feature Fusion Network] and [pg. 7, section V. Conclusion]).
HU in view of QI is not relied upon for teaching:
computing, for each target and for each text, a score based on an inner product between: the first feature vector and an embedding vector corresponding to the target;
computing, for each target, weights for the texts using a softmax function applied to the scores;
generating, for each target, a status vector by computing a weighted sum of the second feature vectors using the weights;
training a model, including learning the embedding vectors, […]
computing, after the training, a similarity matrix based on cosine similarities between the embedding vectors, and
determining a portfolio vector that minimizes a risk defined using the similarity matrix.
However, MILLER teaches: computing, for each target and for each text, a score based on an inner product between: the first feature vector and an embedding vector corresponding to the target; (MILLER [pg. 3, section 3.1 Model Description] teaches: "Key Addressing: during addressing, each candidate memory (i.e., for each text) is assigned a relevance probability by comparing the question (i.e., for each target) to each key:
p
h
i
=
S
o
f
t
m
a
x
A
Φ
X
x
⋅
A
Φ
K
K
h
i
Examiner’s note: Under BRI, an embedding vector corresponding to the target can be interpreted as
A
Φ
X
x
and the first feature vector can be interpreted as
A
Φ
K
K
h
i
. Further, computing, for each target and for each text, a score based on an inner product can be interpreted as the inner product between
A
Φ
X
x
⋅
A
Φ
K
K
h
i
.)
computing, for each target, weights for the texts using a softmax function applied to the scores; (MILLER [pg. 3, section 3.1 Model Description] teaches: "Key Addressing: during addressing, each candidate memory is assigned a relevance probability (i.e., weights for the texts) by comparing the question (i.e., for each target) to each key:
p
h
i
=
S
o
f
t
m
a
x
A
Φ
X
x
⋅
A
Φ
K
K
h
i
Examiner’s note: Under BRI, computing, for each target, weights for the texts using a softmax function applied to the scores can be interpreted as computing the equation above, where a softmax function is applied to the inner product (i.e., scores) of the candidate memory (i.e., for each text) and the question (i.e., for each target).)
generating, for each target, a status vector by computing a weighted sum of the second feature vectors using the weights; (MILLER [pg. 3, section 3.1 Model Description] teaches: "Value Reading: in the final reading step, the values of the memories are read by taking their weighted sum using the addressing probabilities, and the vector
o
is returned:
o
=
∑
i
p
h
i
A
Φ
V
(
v
h
i
)
Examiner’s note: under BRI, generating, for each target, a status vector can be interpreted as computing the summation resulting in
o
by using
p
h
i
(i.e., the weights) and
A
Φ
V
(
v
h
i
)
(i.e., the second feature vectors).)
training a model, including learning the embedding vectors, […] (MILLER [pg. 4, section 3.1 Model Description] teaches: “The whole network is trained end-to-end (i.e., training a model), and the model learns to perform the iterative accesses to output the desired target a by minimizing a standard cross-entropy loss between
a
^
and the correct answer
a
. Backpropagation and stochastic gradient descent are thus used to learn the matrices
A
,
B
(i.e., including learning the embedding vectors) and
R
1
,
…
,
R
H
.”)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of HU, QI, and MILLER before them, to include MILLER’s softmax function for determining relevance probability and weighted sum using the probabilities in HU and QI’s news-oriented stock prediction framework. One would have been motivated to make such a combination in order to improve information extraction from documents to answer questions related to the documents (e.g., stock price prediction), (MILLER [pg. 1, Abstract]).
HU in view of QI and MILLER is not relied upon for teaching:
computing, after the training, a similarity matrix based on cosine similarities between the embedding vectors, and
determining a portfolio vector that minimizes a risk defined using the similarity matrix.
However, GUOSHENG teaches: computing, after the training, a similarity matrix based on cosine similarities between the embedding vectors, and (GUOSHENG [pg. 2708, section 2.2 Clustering] teaches: "We next aim to provide a clustering method for diversified – and hence low risk – portfolio selection. [...] To solve these problems, we introduce the network modularity method [18] to find the cluster structure of the stocks, where each stock is set as one node and the link between each pair of stocks (i.e., a similarity matrix) is set as the cosine similarity (i.e., based on cosine similarities) calculated (i.e., computing) by our learned (i.e., after the training) CAE features (i.e., between the embedding vectors).” GUOSHENG [pg. 2708, section 3.2 Quantitative Results] teaches: “This illustrates the efficacy of our CAE and learned feature for capturing semantic information about stocks.”)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of HU, QI, MILLER, and GUOSHENG before them, to include GUOSHENG’s similarity calculation in HU, QI, and MILLER’s news-oriented stock prediction framework. One would have been motivated to make such a combination in order to improve portfolio construction based on stock similarity features (GUOSHENG [pg. 2709, section 4 Conclusions).
HU in view of QI, MILLER, and GUOSHENG is not relied upon for teaching, but ZHENG teaches: determining a portfolio vector that minimizes a risk defined using the similarity matrix. (ZHENG [pg. 2, section 2.1. Problem Setting] teaches: “
w
∈
R
N
is the weight of each asset to hold (i.e., a portfolio vector) in order to approximate the index
Y
. […] Therefore, the objective function becomes,
min
w
≥
0
,
∑
i
w
i
=
1
X
w
-
Y
2
2
(
2
)
Eq. 2 is known as a non-negative regression problem with sum-to-one constraint, which can easily be solved by quadratic programming (QP) (i.e., determining a portfolio vector).” ZHENG [pg. 2, section 2.2. Diversity] teaches: “we propose to use,
w
T
A
w
(
3
)
where
A
i
j
is a similarity measure (i.e., using a similarity matrix) between assets
i
and
j
, where 0 means most dissimilar and 1 means most similar.” ZHENG [pg. 2, section 2.2.1. Choice of A] teaches: “
A
can be further decomposed as,
A
=
Z
T
Z
.” ZHENG [pg. 3, section 2.3. Sparsity] teaches: “our full objective function can be written as,
min
w
X
w
-
Y
2
2
+
λ
1
Z
w
2
2
+
λ
2
1
T
Z
Z
T
-
1
Z
w
S
u
b
j
e
c
t
t
o
:
w
≥
0
a
n
d
∑
i
w
i
=
1
(
i
.
e
.
,
m
i
n
i
m
i
z
e
s
a
r
i
s
k
)
ZHENG [pg. 2, section 2.2. Diversity] teaches: “In modern portfolio theory, the term
w
T
∑
w
represents the risk (variance) of portfolio, and in our work,
w
T
A
w
serves the similar purpose of reducing the risk of several highly correlated assets plummeting simultaneously.” Examiner’s note: Under BRI, a risk defined using a similarity matrix can be interpreted as the ZHENG’s term
w
T
A
w
, which is the risk defined using the similarity measure
A
. Further, because
A
=
Z
T
Z
, the term
λ
1
Z
w
2
2
in the objective function minimized over
w
is equal to
λ
1
w
T
A
w
. ZHENG states that minimizing the term
w
T
A
w
serves a similar purpose in reducing risk in modern portfolio theory as
w
T
∑
w
.)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of HU, QI, MILLER, GUOSHENG, and ZHENG before them, to include ZHENG’s similarity measure for reducing portfolio risk in HU, QI, MILLER, GUOSHENG, and ZHENG’s news-oriented stock prediction framework. One would have been motivated to make such a combination in order to reduce the risk of several highly correlated assets plummeting simultaneously (ZHENG [pg. 2, section 2.2. Diversity]).
Regarding Claim 10:
The claim recites similar limitations as corresponding claim 9 and is rejected for similar reasons as claim 9 using similar teachings and rationale. Additionally, HU teaches:
[…] determining a portfolio for a plurality of targets […] (HU [pg. 268, section 5.4] teaches: "Based on these scores, a straightforward portfolio construction strategy called top-K selects K stocks with the highest scores to construct a new portfolio for the next trading day.")
HU is not relied upon for teaching, but QI teaches: A non-transitory information recording medium recording a program for determining a […] for a plurality of targets, the program causing a computer to (QI [pg. 5, section C. Implementation Details] teaches: “all experiments in the paper are completed on the open source framework PyTorch on Linux CentOS 7.6.1810 system with NVIDIA TITAN Xp GPU (12G graphics memory).” QI [pg. 5, Table I], [pg. 6, Fig. 2], and [pg. 3, Table III] teach learning deep characteristics of Chinese news texts, such as in finance, to improve the performance of news headline classifications. Therefore, QI teaches determining a classification for finance news headlines (i.e., for a plurality of targets).)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of HU and QI before them to include QI’s GPU system, memory, and classification of finance news headlines in HU’s news-oriented stock prediction framework. HU teaches a portfolio construction strategy based on predicting stock trends using a sequence of recent related news (HU [pg. 262, section 1 Introduction]). Further, one of ordinary skill could use QI’s classification of finance news headlines and QI’s system with NVIDIA TITAN Xp GPU to implement and aid in HU’s portfolio construction strategy. One would have been motivated to make such a combination in order to improve the performance of news headline (and text content) classification (QI [pg. 7, section V. Conclusion]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Alvaro S Laham Bauzo whose telephone number is (571)272-5650. The examiner can normally be reached Mon-Fri 7:30 AM - 11:00 AM | 1:00 PM - 5:30 PM ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Usmaan Saeed can be reached on (571) 272-4046. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/A.S.L./Examiner, Art Unit 2146
/USMAAN SAEED/Supervisory Patent Examiner, Art Unit 2146
Prosecution Insights