John M. Boyer et al.

Patent Trials and Appeals Board

Aug 2, 2019

14966802 - (D) (P.T.A.B. Aug. 2, 2019)

UNITED STATES PATENT AND TRADEMARK OFFICE
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www.uspto.gov
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
14/966,802 12/11/2015 John M. Boyer AUS920150290US2 7512
50170 7590 08/02/2019
IBM CORP. (WIP)
c/o WALDER INTELLECTUAL PROPERTY LAW, P.C.
1701 N. COLLINS BLVD.
SUITE 2100
RICHARDSON, TX 75080
EXAMINER
RIFKIN, BEN M
ART UNIT PAPER NUMBER
2123
MAIL DATE DELIVERY MODE
08/02/2019 PAPER
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
____________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
____________

Ex parte JOHN M. BOYER, KSHITIJ P. FADNIS,
COLLIN J. MURRAY, and JUSTIN A. ZINIEL
____________

Appeal 2018-008676
Application 14/966,802
Technology Center 2100
____________

Before JOSEPH L. DIXON, JAMES W. DEJMEK, and
STEPHEN E. BELISLE, Administrative Patent Judges.

BELISLE, Administrative Patent Judge.

DECISION ON APPEAL
Appellants1 appeal under 35 U.S.C. § 134(a) from a Final Rejection of
all pending claims, namely, claims 1–22. App. Br. 5. We have jurisdiction
under 35 U.S.C. § 6(b).
We affirm.

1 Appellants identify International Business Machines Corporation as the
real party in interest. App. Br. 2.
Appeal 2018-008676
Application 14/966,802

2
STATEMENT OF THE CASE
The Claimed Invention
Appellants’ invention generally relates to “an improved data
processing apparatus and method and more specifically to mechanisms for
improving a ground truth answer key of a question answering [(“QA”)]
cognitive system using a similar passage cognitive system trained with a
ground truth answer key from the question answering cognitive system.”
Spec. ¶ 1.
According to the Specification, an example of a QA system is the
“IBM Watson™” system, which is available from International Business
Machines (IBM) Corporation, and is “an application of advanced natural
language processing, information retrieval, knowledge representation and
reasoning, and machine learning technologies to the field of question
answering.” Spec. ¶ 3.
Claim 1, reproduced below, is representative of the claimed subject
matter on appeal:
1. A method, in a data processing system comprising at least
one processor and a memory comprising instructions which,
when executed by the at least one processor, causes the at least
one processor to improve ground truth in a question answering
cognitive system, the method comprising:
training, by the data processing system, a similar passage
machine learning model for a similar passage cognitive system
using a question and answer key to form a trained similar passage
machine learning model, wherein the question and answer key
comprises a list of question and answer specification pairs
forming a ground truth for a question answering cognitive
system, wherein each question is a text string and each answer
specification references one or more text passages from a corpus
of information;
Appeal 2018-008676
Application 14/966,802

3
responsive to a search event, sending at least one text input
to the similar passage cognitive system operating in accordance
with the trained similar passage machine learning model and
executing on the at least one processor of the data processing
system, wherein the text input comprises a given text passage
from a given answer specification of the question and answer
key, and receiving from the similar passage cognitive system
configured with the trained similar passage machine learning
model a response list of references to text passages from the
corpus of information that are similar to the given text input;
responsive to an acceptance event for at least one text
passage from the response list, supplementing, by the data
processing system, the question and answer key by adding the at
least one text passage to the given answer specification to form
a supplemented question and answer key; and
training a question answering machine learning model of
the data processing system using the supplemented question and
answer key such that the question answering cognitive system
operates in accordance with the trained question answering
machine learning model and executes on the at least one
processor of the data processing system.
App. Br. 24–25 (Claims Appendix).

References
The Examiner relied on the following references as evidence of
unpatentability of the claims on appeal:
Nakazawa US 2008/0195378 A1 Aug. 14, 2008
Heck US 2009/0162824 A1 June 25, 2009

Appeal 2018-008676
Application 14/966,802

4
Rejection
The Examiner made the following rejection of the claims on appeal:
Claims 1–222 stand rejected under 35 U.S.C. § 103 as being
unpatentable over Heck and Nakazawa. Final Act. 2–7.

ANALYSIS3
Spread over thirty pages of briefing without any clear argument
delineations, Appellants submit what we have categorized as nine arguments
as to the nonobviousness of various sets of claims 1–22 over Heck and
Nakazawa. See App. Br. 5–22; Reply Br. 2–13. We identify and address
each of these arguments below, but find them unpersuasive. We first turn to
certain teachings of Heck and Nakazawa.
Heck, titled “Automated Learning From a Question and Answering
Network of Humans,” generally is directed to:
A QA robot [i.e., automated question and answer system] learns
how to answer questions by observing human interaction over
online social networks. The QA robot observes the way people
ask questions and how other users respond to those questions. In
one embodiment, the QA robot observes which questions are
most helpful and analyzes those questions to identify the
characteristics of those questions that are most helpful. The QA

2 In the Final Action, the Examiner inadvertently captioned this rejection
as applying to claims “1, 3–9, 11–16, and 18–22,” but substantively included
claims 2, 10, and 17 in this rejection. See Final Act. 5–6. We agree with
Appellants that the Examiner intended this rejection to apply to claims 1–22.
See App. Br. 5.
3 Throughout this Decision, we have considered Appellants’ Appeal Brief
filed March 16, 2018 (“App. Br.”); Appellants’ Reply Brief filed August 30,
2018 (“Reply Br.”); the Examiner’s Answer mailed July 2, 2018 (“Ans.”);
the Final Office Action mailed September 20, 2017 (“Final Act.”); and
Appellants’ Specification filed December 11, 2015 (“Spec.”).
Appeal 2018-008676
Application 14/966,802

5
robot then uses those observations to enhance the way questions
are asked and answered in the future.
Heck ¶¶ 1, 15. Heck discloses that “[o]ne way to teach the QA robot how to
answer questions is to boot it into an initial training mode,” where “the QA
robot can then be populated with test questions and answers, archived
questions and answers from a social network, and information from other
sources.” Heck ¶ 31 (emphases added). “The QA robot uses those sources
of information to learn.” Heck ¶ 31. “[T]his training may be supervised by
people to ensure that the answers to a question are correct and that answers
are being stored and indexed properly.” Heck ¶ 31 (emphasis added).
Heck also discloses that “[a] QA robot generates answers to questions,
but there may be times that more than one answer may be applicable to a
question. In one embodiment, the QA robot analyzes candidate answers to
determine the most ‘correct’ answer (e.g., the answer most likely to be
correct answer).” Heck ¶ 38 (emphasis added).
Nakazawa, titled in part “Question and Answer Data Editing Device,”
generally is directed to “a detecting unit that detects a part of the dialogue
content similar to existing question and answer data stored,” and an
“extracting unit that extracts a context in which the dialogue content is made
from dialogue content in the proximity of the similar part detected and
registers the context extracted as new question and answer data or as index
information of the question and answer data.” Nakazawa, Abstract
(emphases added).
Claims 1, 4, 5, 9, 11, 12, 16, 18, and 19
Appellants argue nonobviousness of the independent claims, namely,
claims 1, 9, and 16, along with dependent claims 4, 5, 11, 12, 18, and 19, as
a group. We select independent method claim 1 as the representative claim
Appeal 2018-008676
Application 14/966,802

6
and address Appellants’ arguments thereto, and any claim in this set not
argued separately will stand or fall with our analysis of the rejection of
claim 1. See 37 C.F.R. § 41.37(c)(1)(iv) (2017).
Our reviewing court has held the relevant inquiry in an obviousness
analysis is whether the Examiner has set forth “some articulated reasoning
with some rational underpinning to support the legal conclusion of
obviousness.” In re Kahn, 441 F.3d 977, 988 (Fed. Cir. 2006) (cited with
approval in KSR Int’l Co. v. Teleflex, Inc., 550 U.S. 398, 418 (2007)).
The test for obviousness is not whether the claimed invention is expressly
suggested in any one or all of the references, but whether the claimed subject
matter would have been obvious to those of ordinary skill in the art in light
of the combined teachings of those references. See In re Keller, 642 F.2d
413, 425 (CCPA 1981); In re Burckel, 592 F.2d 1175, 1179 (CCPA 1979)
(“[A] reference must be considered not only for what it expressly teaches,
but also for what it fairly suggests.”). One of ordinary skill can use his or
her ordinary skill, creativity, and common sense to make the necessary
adjustments and further modifications to result in a properly functioning
device. See KSR, 550 U.S. at 418 (“[A] court can take account of the
inferences and creative steps that a person of ordinary skill in the art would
employ.”).
Argument 1: Appellants argue Heck does not teach “training . . . a
similar passage machine learning model for a similar passage cognitive
system using a question and answer key to form a trained similar passage
machine learning model,” as recited in claim 1. See App. Br. 6–7; Reply
Br. 2–6.
Appeal 2018-008676
Application 14/966,802

7
Here, the Examiner finds:
[Heck’s] QA robot learns (i.e. is trained) to generate answers to
questions by observing users, particularly experts, as they
respond to questions and answers on a social network. Here the
questions and answers on the social network represent the
question and answer key, which is used to train the QA robot.
The QA Robot is the “similar passage machine learning model”
which is trained based on questions and answers from a social
network.
Ans. 9. We agree here with the Examiner, and also find, as discussed above,
that Heck expressly discloses training the QA robot using a question and
answer key (i.e., “test questions and answers”) to form a trained QA robot.
See Heck ¶ 31; see also Heck ¶¶ 32–34.
Nevertheless, Appellants argue “[t]he features [of claim 1] being
addressed train a machine learning model for a similar passage cognitive
system that is completely separate from the claimed question answering
cognitive system. Conflating the two cognitive systems is a clear error.”
App. Br. 10; see id. at 9 (“The Examiner makes no distinction between a
similar passage machine learning model and a question answering machine
learning model.”); Reply Br. 4 (“[A] similar passage machine learning
model is not trained to answer questions; it is trained to provide a response
list of references to text passages from a corpus of information that are
similar to a given text input.”); see also Reply Br. 2–4, 9–10. The
Examiner responds that “the QA robot providing [the] answers learns from a
set of question and answers observed overtime,” and “responds to and
provides responses from questions and answers that are similar to the
questions asked,” which “clearly denotes training . . . and provid[ing]
similar passages” and “meets the broadest reasonable interpretation of the
claim . . . .” Ans. 12–13. We again agree with the Examiner.
Appeal 2018-008676
Application 14/966,802

8
Appellants’ Specification states that (1) “a Question Answering
cognitive system (QA system) is an artificial intelligence application
executing on data processing hardware that answers questions pertaining to
a given subject-matter domain presented in natural language” (Spec. ¶ 27
(emphasis added)); and (2) “a Similar Passage cognitive system (SP system)
is an artificial intelligence application executing on data processing
hardware that provides similar passages pertaining to a given subject-matter
domain presented in natural language” (Spec. ¶ 28 (emphasis added)).
We find that, under the broadest reasonable interpretation of claim 1, an
application or system that both (1) answers questions and (2) provides
similar passages as recited in claim 1 would teach or at least fairly suggest to
the skilled artisan separately articulated applications performing (1) and (2).
See KSR, 550 U.S. at 417 (A claim is obvious where it ‘“simply arranges old
elements with each performing the same function it had been known to
perform’ and yields no more than one would expect from such an
arrangement.”); In re Am. Acad. of Sci. Tech Ctr., 367 F.3d 1359, 1364 (Fed.
Cir. 2004) (during prosecution, an application’s claims are given their
broadest reasonable scope consistent with the specification).
Based on the foregoing and our review of the Briefs, we find
Appellants do not show persuasively that the Examiner erred in finding that
Heck at least fairly suggests “training . . . a similar passage machine learning
model for a similar passage cognitive system using a question and answer
key to form a trained similar passage machine learning model,” as recited in
claim 1.
Argument 2: Appellants argue neither Heck nor Nakazawa teaches
“sending at least one text input to the similar passage cognitive system . . . ,
Appeal 2018-008676
Application 14/966,802

9
wherein the text input comprises a given text passage from a given answer
specification of the question and answer key, and receiving . . . a response
list of references to text passages from the corpus of information that are
similar to the given text input,” as recited in claim 1. See App. Br. 11
(“Heck does not teach submitting a text passage of answers in the question
and answer key to receive other text passages in the corpus that are similar
to the text passage.”); id. at 7–11, 13; Reply Br. 2–6, 10. In particular,
Appellants argue “Nakazawa teaches matching text from a first database to a
second database to supplement the second database. Neither database is
used to train a machine learning model.” App. Br. 13–14; see Reply Br. 7–
10.
The Examiner finds:
The Heck reference explicitly denotes allowing the system to
search through its confines for similar questions and answers in
order to answer the user. In fact, Heck specifically states “In one
embodiment, the QA robot stores the questions and their
associated answers directly into its knowledgebase and retrieves
that information when similar questions are subsequently
asked.” This clearly denotes that the system seeks to find
similar answers, and shows that it is not constrained to questions
or answers which are worded exactly the same.
Ans. 11–12 (emphases in original). The Examiner also finds:
[T]he applicant appears to be attempting to view the Heck
reference in a vacuum. The applicant is correct that the Heck
reference does not disclose “sending a given text passage from a
given answer specification of the question and answer key to the
similar passage cognitive system operating in accordance with
the trained similar passage machine learning model.” The
Examiner stated this within the rejection. Heck discloses
receiving questions from the user via a search event, and
processing that search event. However, the remaining portions
of these limitations are not cited in the Heck reference, but are
Appeal 2018-008676
Application 14/966,802

10
met by the Nakazawa reference. The Nakazawa reference clearly
denotes taking in answer information and searching to find
similar answers . . . . When combined with the Heck reference,
this denotes allowing the system to take in answers in order to
find similar answers, which meets the broadest reasonable
interpretation of the claims.
Ans. 13–14; see id. at 15–16, 18 (“The Nakazawa reference has been
brought in . . . to show it is known in the art of question and answer systems
. . . to use both questions and answers to identify new answers for
questions.”). The Examiner also explains:
The Heck reference makes clear that it can take in information,
and search for similar information within its system. The Heck
[r]eference makes use of questions in order to perform this
action. It does not, as the claim requires, make use of answer
information to perform this search, as the examiner has explicitly
stated in the rejection. The Nakazawa reference, however,
clearly shows that it would be obvious to [the skilled artisan] to
make use of answer information to search for new answers to add
to the system . . . . The combined references meet the claimed
limitations.
Ans. 19.
We agree with and adopt as our own the Examiner’s explanation of
obviousness reproduced above, and find Appellants do not persuasively
explain why the combined teachings of Heck and Nakazawa do not at least
fairly suggest to the skilled artisan the broadly claimed features at issue in
representative claim 1. We also agree with the Examiner that Appellants,
despite protestations to the contrary (see, e.g., App. Br. 6; Reply Br. 7),
improperly argue nonobviousness here by attacking Heck and Nakazawa
individually, rather than addressing their combined teachings. See In re
Keller, 642 F.2d at 426 (One cannot show nonobviousness by attacking
Appeal 2018-008676
Application 14/966,802

11
references individually where the rejections are based on combinations of
references.); In re Merck & Co., 800 F.2d 1091, 1097 (Fed. Cir. 1986).
Based on the foregoing and our review of the Briefs, we find
Appellants do not show persuasively that the Examiner erred in finding that
Heck and Nakazawa at least fairly suggest:
sending at least one text input to the similar passage cognitive
system . . . , wherein the text input comprises a given text passage
from a given answer specification of the question and answer
key, and receiving . . . a response list of references to text
passages from the corpus of information that are similar to the
given text input,
as recited in claim 1.
Argument 3: Appellants argue neither Heck nor Nakazawa teaches
“supplementing . . . the question and answer key by adding the at least one
text passage to the given answer specification to form a supplemented
question and answer key,” as recited in claim 1. See App. Br. 7; Reply
Br. 3–4, 10. In particular, Appellants argue, for example:
Heck teaches a specific type of question, which is an opinion
question. There is no correct answer, only a better answer. Heck
gives the example of “where can I buy good Indian food in
Portland, Oreg.?” There are multiple answers, and each answer
is only subjectively better than others. This is not the same as
improving the ground truth by supplementing the question and
answer key.
App. Br. 7 (emphasis added); see Reply Br. 3.
The Examiner finds:
First, the claim at no time requires any particular type of
“question” to be asked. The claim at no time excludes opinion
based questions. Opinion questions can have answers like any
other question can. The claim at no time limits a question to a
single answer. The applicant then goes on to state that “this is
Appeal 2018-008676
Application 14/966,802

12
not the same as improving the ground truth by supplementing the
question and answer key.” The Heck reference clearly discloses
that the system actively learns over time by taking in additional
question and answers over time in order to provide better answers
to the user . . . . This shows that the system is not static, and that
new question and answers are added over time. The examiner
interprets this to be “supplementing” the system with additional
question and answers . . . .
Ans. 10. We agree here with the Examiner, and also find that, contrary to
Appellants’ position, Heck not only teaches “opinion” questions but also
clearly teaches answering “fact” questions. For example, Heck expressly
discloses:
In the case of non-subjective data, a correct answer can be
relatively easy to find, assuming there is an authoritative answer
to the question. For example, a user asks who wrote the book
“Tale of Two Cities”. In one embodiment, the QA robot can find
the information relatively easily in a database that stores
informational data.
Heck ¶ 38 (emphases added). Also, Appellants concede that Heck teaches
accumulating multiple answers to a question and providing such answers in
response to a question. See App. Br. 9 (“Appellants agree that Heck teaches
answering a question with a list of answers.”).
In addition, the Examiner finds:
As [for] the argument that Heck fails to teach supplementing the
question and answer key with the at least one text passage, the
Heck system clearly denotes taking in new question and answer
responses over time. However, the Examiner was very clear,
Heck does not disclose, “by adding the at least one text passage
to the given answer specification.” Once again, the Nakazawa
[reference] meets these limitations, clearly denoting looking for
similar answers, and adding those similar answers to the system
if they meet a matching threshold . . . .
Appeal 2018-008676
Application 14/966,802

13
Ans. 15. We again agree with and adopt as our own the Examiner’s
explanation of obviousness reproduced above, and find Appellants do not
persuasively explain why the combined teachings of Heck and Nakazawa do
not at least fairly suggest to the skilled artisan the broadly claimed features
at issue in representative claim 1.
Based on the foregoing and our review of the Briefs, we find
Appellants do not show persuasively that the Examiner erred in finding that
Heck and Nakazawa at least fairly suggest “supplementing . . . the question
and answer key by adding the at least one text passage to the given answer
specification to form a supplemented question and answer key,” as recited in
claim 1.
Argument 4: Appellants argue neither Heck nor Nakazawa teaches
“training a question answering machine learning model . . . using the
supplemented question and answer key,” as recited in claim 1. See App.
Br. 14; Reply Br. 10. In particular, Appellants argue:
The Final Office Action states that Nakazawa “is never called for
training a machine learning model for a question and answer
system.” Applicants agree. No reference is cited for using
similar passages to train a question and answer machine learning
model. The Examiner only concludes that Heck and Nakazawa
could be combined to result in these features. The Examiner does
not explain how the references can be combined or how such a
combination would result in the claim features.
App. Br. 14–15 (emphases omitted); Reply Br. 9. The Examiner finds:
Each and every limitation is met by the combined references of
Heck and Nakazawa, and each limitation has been clearly cited
in the rejection. Furthermore, the examiner has shown that Heck
and Nakazawa are analogous art, as both involve Q&A systems,
and they have a motivation to combine, to make use of answers
similar to answers you already have in order to improve the Q&A
system.
Appeal 2018-008676
Application 14/966,802

14
Ans. 19–20; see Final Act. 4–5 (“The motivation for [combining Heck and
Nakazawa] would be to ‘detect . . . contents of a dialogue similar to question
and answer data from data of a history of dialogues made in the past . . .’” or
“in the case of Heck, allow the system to also search for similar answers,
now just based on the questions.”). The Examiner repeatedly explains that
Heck’s QA robot “learns” (i.e., is trained) to generate answers to questions,
which is based on “supplementing” as discussed above in Argument 3.
See, e.g., Ans. 9. We again agree with the Examiner, and find Appellants’
arguments unpersuasive.
To the extent that Appellants argue the Examiner did not show how
Nakazawa could be bodily incorporated into Heck (see, e.g., App. Br. 14–
15; Reply Br. 7 (“[T]he combination would not result in the features of
claims 1, 9, and 16”)), obviousness does not require such a showing. It is
well-established that a determination of obviousness based on teachings
from multiple references does not require an actual, physical substitution of
elements. In re Etter, 756 F.2d 852, 859 (Fed. Cir. 1985) (en banc) (“Etter’s
assertions that Azure cannot be incorporated in Ambrosio are basically
irrelevant, the criterion being not whether the references could be physically
combined but whether the claimed inventions are rendered obvious by the
teachings of the prior art as a whole.”); In re Sneed, 710 F.2d 1544, 1550
(Fed. Cir. 1983) (citation omitted) (“[I]t is not necessary that the inventions
of the references be physically combinable to render obvious the invention
under review.”); In re Keller, 642 F.2d at 425 (“The test for obviousness is
not whether the features of a secondary reference may be bodily
incorporated into the structure of the primary reference . . . .”).
Appeal 2018-008676
Application 14/966,802

15
Based on the foregoing and our review of the Briefs, we find
Appellants do not show persuasively that the Examiner erred in finding that
Heck and Nakazawa at least fairly suggest “training a question answering
machine learning model . . . using the supplemented question and answer
key,” as recited in claim 1.
Accordingly, we sustain the Examiner’s rejection of representative
method claim 1 under 35 U.S.C. § 103. As noted above, Appellants do not
separately argue patentability of independent claims 9 and 16 and dependent
claims 4, 5, 11, 12, 18, and 19, but instead rely only on their arguments for
patentability of method claim 1. Accordingly, for the same reasons set forth
above for independent claim 1, we sustain the Examiner’s rejection under
35 U.S.C. § 103 of claims 4, 5, 9, 11, 12, 16, 18, and 19. See 37 C.F.R.
§ 41.37(c)(1)(iv).
Claims 2, 10, and 17
Argument 5: Appellants argue Heck does not teach “prompting a user
to provide a distinguishing question,” as recited, for example, in claim 2.
See App. Br. 15–16; Reply Br. 11. The Examiner finds:
Heck clearly discloses the system encountering questions it feels
are not acceptable and removing them. (See Heck, paragraphs
0032–0034). In fact, the Heck reference clearly denotes making
new question and answer pairs during these actions, as Heck
discloses removing an answer in regards to reviews that a
restaurant is bad, and instead adding that answer to a question of
which restaurants to avoid (see Heck, Paragraph 0034).
Furthermore, the Heck reference denotes asking for additional
information (i.e. distinguishing questions) when the question is
unanswerable (i.e. rejected). (See Heck, paragraph 0055). This
paragraph clearly denotes the Heck reference asking the user to
clarify the question (i.e. provide a distinguishing question).
Appeal 2018-008676
Application 14/966,802

16
Ans. 21. Appellants respond essentially by asserting that the Examiner’s
cited passages in Heck do not equate to the claim language. See Reply
Br. 11. In particular, Appellants argue “Heck teaches prompting the user
that submitted a bad question to clarify the question or provide additional
information. This is not equivalent to prompting a user to provide a
distinguishing question responsive to an answer rejection event for a given
text passage from the response list.” App. Br. 15 (emphasis added); id. at 16
(“[A]sking the user for more information to understand the user’s question is
not equivalent to prompting the user to provide a distinguishing question
. . . .” (Emphasis added)).
But the issue here is not whether Heck discloses the “same” or
“equivalent” features as in claim 2, rather, the issue is whether Heck at least
fairly suggests to the skilled artisan “prompting a user to provide a
distinguishing question” and the other limitations as recited in claim 2. We
agree with and adopt as our own the Examiner’s explanation of obviousness
reproduced above, and find Appellants do not persuasively explain why
Heck does not at least fairly suggest such broad features.
Accordingly, we sustain the Examiner’s rejection of claim 2 under
35 U.S.C. § 103. Appellants do not separately argue patentability of claims
10 and 17, but instead rely only on their arguments for patentability of
claim 2. Accordingly, for the same reasons set forth above for claim 2, we
sustain the Examiner’s rejection under 35 U.S.C. § 103 of claims 10 and 17.
See 37 C.F.R. § 41.37(c)(1)(iv).
Claim 3
Argument 6: Appellants argue Heck does not teach the limitations
recited in claim 3. App. Br. 16–18. In particular, Appellants argue:
Appeal 2018-008676
Application 14/966,802

17
Heck teaches sending questions that cannot be answered by the
QA robot system to human experts to be answered. This portion
of Heck does not teach a question and answer specification pair
in the question and answer key, which forms a ground truth for a
question answering cognitive system, for which the number of
text passages in the answer specification is less than a threshold
value. The cited portion of Heck makes no mention of a question
and answer key, an answer specification, a number of text
passage, or a threshold value. The cited portion of Heck is
concerned with a question from the user, not a question and
answer key, for which an answer cannot be found.
App. Br. 17. The Examiner finds:
[T]he Heck reference denotes a threshold for not having enough
answers to answer a question, in this case, that threshold is not
having an answer at all. (See Heck, Paragraph 0055). In
response to this issue, the Heck reference responds by requesting
that an expert answer the question (i.e. provide at least one
additional text passage reference) and then making use of that
answer within the system. (See Heck, Paragraph 0055). The
applicant at no time required the Threshold to be more than one,
nor do they provide any detail at all as to what this threshold
might encompass. In this case, the threshold is having at least a
single answer. This meets the broadest reasonable interpretation
of the claims . . . .
Ans. 22. In response, Appellants again argue that Heck’s disclosed features
are “not equivalent” to the limitations recited in claim 3, and argue that
claim 3 requires that “every question has one or more answers.” See Reply
Br. 11.
The issue again here is not whether Heck discloses the “same” or
“equivalent” features as in claim 3, rather, the issue is whether Heck at least
fairly suggests to the skilled artisan “determining a question and answer
specification pair for which the number of text passage references in the
answer specification is less than a threshold value,” “prompting a user to
Appeal 2018-008676
Application 14/966,802

18
provide at least one additional text passage reference,” and “amending the
answer specification to include the at least one additional text passage
reference.” We agree with and adopt as our own the Examiner’s explanation
of obviousness reproduced above, and find Appellants do not persuasively
explain why Heck does not at least fairly suggest such broad features.
Accordingly, we sustain the Examiner’s rejection of claim 3 under
35 U.S.C. § 103.
Claims 6, 13, and 20
Argument 7: Appellants argue Heck does not teach “automatically
generating at least one search event upon completion of an ingestion
operation that updates the corpus of information,” as recited, for example, in
claim 6. App. Br. 18–19; Reply Br. 11–12.
The Examiner finds: “The Heck reference clearly denotes a system
that actively learns and seeks to improve by monitoring social networks for
question and answers. . . . Furthermore, the reference goes on to state that it
not only brings in new information, but actively updates and manages
information it already contains . . . .” Ans. 24 (citing Heck ¶¶ 32–34).
In response, Appellants argue merely that “the cited portion of Heck
makes no mention of [the recited claim language]” (App. Br. 18), and that
patentability of claims 6, 13, and 20 derives from their dependency on
independent claims 1, 9, and 16, respectively (see App. Br. 18–19; Reply Br.
11–12), which are not substantive arguments as to the independent
patentability of dependent claims 6, 13, and 20. Thus, we find Appellants’
arguments unpersuasive of Examiner error. See 37 C.F.R. § 41.37(c)(1)(iv)
(Any claim not argued separately will stand or fall with our analysis of the
rejection of the underlying claims.); In re Lovin, 652 F.3d 1349, 1357 (Fed.
Appeal 2018-008676
Application 14/966,802

19
Cir. 2011) (“[T]he Board reasonably interpreted Rule 41.37 to require more
substantive arguments in an appeal brief than a mere recitation of the claim
elements and a naked assertion that the corresponding elements were not
found in the prior art.”); In re Geisler, 116 F.3d 1465, 1470 (Fed. Cir. 1997)
(It is well settled that mere attorney arguments and conclusory statements,
which are unsupported by factual evidence, are entitled to little probative
value.); In re Pearson, 494 F.2d 1399, 1405 (CCPA 1974) (attorney
argument is not evidence).
Accordingly, we sustain the Examiner’s rejection of claims 6, 13,
and 20 under 35 U.S.C. § 103.
Claims 7, 14, and 21
Argument 8: Appellants argue Heck does not teach “invoking at least
one search event for at least one text passage of the question and answer
key,” as recited, for example, in claim 7. App. Br. 19–20; Reply Br. 12.
The Examiner finds:
The Heck reference clearly denotes searching for answers in
response to questions (see Heck, Paragraph 0023). The Heck
reference even clearly denotes that questions can be search
queries (see Heck, Paragraph 0023). There is no constraint on
the user that this information can’t be a question that the system
has already seen before. Furthermore, the system clearly denotes
comparing questions to other questions within its database in
order to find similar ones (see Heck, paragraph 0040). The Heck
reference even explicitly denotes uses using the same questions
over and over with the system (see Heck, Paragraph 0064).
Ans. 25.
Appellants also argue Heck does not teach “presenting a list of
answers that are passages similar to a text passage from an answer
Appeal 2018-008676
Application 14/966,802

20
specification of the question and answer key.” App. Br. 20–21; Reply
Br. 12.
The Examiner finds:
First, this claim does not call for “presenting a list of answers that
are passages similar to a text passage from an answer
specification of the question and answer key.” Claims 7, 14, and
21 call for “populating a user interface with at least one response
list received [from] the at least one search events.” All the
portions of the dependency have been met by the previous
claims, and there is no need for these rejections to be repeated in
these claims. Heck clearly denotes providing a user interface list
of responses to the search event (see Heck, paragraphs 0078-
0079). All the rest of the limitations have been dealt with and
responded to in the claims in which they reside . . . .
Ans. 26. Appellants respond “the references do not teach these features in
either wording.” Reply Br. 12.
Here again, Appellants argue merely that the portions of Heck cited
by the Examiner do not teach Appellants’ particular claim language (without
explaining why that is), and that patentability of claims 7, 14, and 21 derives
from their dependency on independent claims 1, 9, and 16, respectively,
which are not substantive arguments as to the independent patentability of
claims 7, 14, and 21. See App. Br. 19–21; Reply Br. 12. Thus, we find
Appellants’ arguments unpersuasive of Examiner error. See 37 C.F.R.
§ 41.37(c)(1)(iv); In re Lovin, 652 F.3d at 1357; In re Geisler, 116 F.3d at
1470; In re Pearson, 494 F.2d at 1405.
Accordingly, we sustain the Examiner’s rejection of claims 7, 14,
and 21 under 35 U.S.C. § 103.
Claims 8, 15, and 22
Argument 9: Appellants argue “Heck makes no mention of reordering
the response list according to a text analytical proximity measure for each
Appeal 2018-008676
Application 14/966,802

21
text passage in the response list relative to the text input sent to the search
event that returned the response list,” as generally recited, for example, in
claim 8. App. Br. 21–22; Reply Br. 12–13.
The Examiner finds:
The Heck reference clearly denotes judging potential answers
based on a confidence score related to the question asked (see
Heck, paragraph 0076). This confidence threshold is dependent
on the material being asked and can change based upon that
information (see Heck, paragraphs 0073–0074). The examiner
is interpreting this to be a [ ] “text analytical proximity measure”
as it seeks to find the textual answer most likely to answer the
user’s question. The applicant at no time quantifies just what a
“text analytical proximity measure[”] is.
Ans. 27. Appellants respond that “[t]here is no teaching or suggestion that
the confidence score [in Heck] is a text analytical proximity measure,” but
do so without defining a “text analytical proximity measure” and without
explaining why there is no such teaching. Reply Br. 12–13.
Again, Appellants argue merely that the portions of Heck cited by the
Examiner do not teach Appellants’ particular claim language (without
explaining why that is), and that patentability of claims 8, 15, and 22 derives
from their dependency on independent claims 1, 9, and 16, respectively,
which are not substantive arguments as to the independent patentability of
claims 8, 15, and 22. See App. Br. 21–22; Reply Br. 12–13. Thus, we find
Appellants’ arguments unpersuasive of Examiner error. See 37 C.F.R.
§ 41.37(c)(1)(iv); In re Lovin, 652 F.3d at 1357; In re Geisler, 116 F.3d at
1470; In re Pearson, 494 F.2d at 1405.
Accordingly, we sustain the Examiner’s rejection of claims 8, 15,
and 22 under 35 U.S.C. § 103.

Appeal 2018-008676
Application 14/966,802

22
DECISION
We affirm the Examiner’s obviousness rejections of claims 1–22.
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a). See 37 C.F.R.
§ 41.50(f).

AFFIRMED