Amazon Technologies, Inc.

Patent Trials and Appeals Board

Mar 15, 2021

2019005632 (P.T.A.B. Mar. 15, 2021)

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.
14/828,176 08/17/2015 Simon Overell SEMSP005C1 7058
136609 7590 03/15/2021
Weaver Austin Villeneuve & Sampson LLP - AMZ
P.O. BOX 70250
OAKLAND, CA 94612-0250
EXAMINER
RIFKIN, BEN M
ART UNIT PAPER NUMBER
2198
NOTIFICATION DATE DELIVERY MODE
03/15/2021 ELECTRONIC
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.
Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the
following e-mail address(es):
USPTO@wavsip.com
eofficeaction@appcoll.com
PTOL-90A (Rev. 04/07)

UNITED STATES PATENT AND TRADEMARK OFFICE
____________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
____________

Ex parte SIMON OVERELL and WILLIAM TUNSTALL-PEDOE
____________

Appeal 2019-005632
Application 14/828,176
Technology Center 2100
____________

Before ROBERT E. NAPPI, JASON J. CHUNG, and
STEVEN M. AMUNDSON, Administrative Patent Judges.

AMUNDSON, Administrative Patent Judge.

DECISION ON APPEAL
Appellant1 seeks our review under 35 U.S.C. § 134(a) from a final
rejection of claims 1, 3–10, and 12–22, i.e., all pending claims. We have
jurisdiction under 35 U.S.C. § 6(b).
We REVERSE.

1 We use the word “Appellant” to refer to “applicant” as defined in
37 C.F.R. § 1.42 (2018). Appellant identifies the real party in interest
as Amazon Technologies, Inc. (a subsidiary of Amazon.com, Inc.).
Appeal Br. 5.
Appeal 2019-005632
Application 14/828,176

2
STATEMENT OF THE CASE
The Invention
According to the Specification, the invention relates to “repositories
of general knowledge, stored in structured form,” and “extracting structured
knowledge from unstructured text for use in a knowledge representation
system.” Spec. 6:9–12, 6:22–24, 227:2–6.2 The Specification states that the
“knowledge representation system includes a knowledge base that represents
knowledge using a structured, machine-readable format” that “includes fact
triples.” Id. at 6:12–15, 6:24–26. The Specification explains that a “fact
triple includes two entity objects and a relationship object expressing a
relationship between the two entity objects.” Id. at 6:15–16, 6:26–28; see id.
at 29:16–21.
According to the Specification, a question may require “an entity to
fill an unknown gap of knowledge, e.g. ‘Who directed Star Wars’ would be
translated to ‘[unknown object] [directed] [star wars]’.” Spec. 219:21–23.
In that situation, “the unknown entity is extracted from the answer.” Id.
at 219:23–24.
Exemplary Claims
Independent claims 1 and 10 exemplify the claims at issue and read
as follows:

2 This decision uses the following abbreviations: “Spec.” for the
Specification, filed August 17, 2015; “Final Act.” for the Final Office
Action, mailed August 29, 2018; “Appeal Br.” for the Appeal Brief, filed
January 28, 2019; “Ans.” for the Examiner’s Answer, mailed May 22, 2019;
and “Reply Br.” for the Reply Brief, filed July 15, 2019.
Appeal 2019-005632
Application 14/828,176

3
1. A computer-implemented method for adding facts to
a knowledge base, comprising:
identifying a question and a corresponding answer in
unstructured text using one or more computing devices;
using the one or more computing devices, identifying
from the question a known entity, an unknown entity, and a
relationship between the known entity and the unknown entity;
using the one or more computing devices, identifying in
the knowledge base a first known entity object representing the
known entity and a relation object representing the relationship;
generating a first fact triple using the one or more
computing devices, the first fact triple including the first known
entity, the relation object, and an unknown entity object
representing the unknown entity;
identifying information in the answer corresponding to
the unknown entity using the one or more computing devices;
identifying a second known entity object in the
knowledge base using the one or more computing devices, the
second known entity object corresponding to the information
identified in the answer;
replacing the unknown entity object of the first fact triple
with the second known entity object to generate a second fact
triple using the one or more computing devices;
adding the second fact triple to the knowledge base using
the one or more computing devices.
10. A system for adding facts to a knowledge base,
comprising one or more computing devices configured to:
identify a question and a corresponding answer in
unstructured text;
determine that the question is a yes-or-no question;
identify from the question a first entity, a second entity,
and a relationship between the first entity and the second entity;
Appeal 2019-005632
Application 14/828,176

4
identify in the knowledge base a first known entity object
representing the first entity, a second known entity object
representing the second entity, and a relation object
representing the relationship;
generate a fact triple from the question, the fact triple
including the first entity object, the second entity object, and
the relation object;
determine a polarity of the fact triple from the answer;
and
add the fact triple to the knowledge base.
Appeal Br. 61–62 (Claims App.).
The Prior Art Supporting the Rejections on Appeal
As evidence of unpatentability under 35 U.S.C. § 103(a),3 the
Examiner relies on the following prior art:

3 The Leahy-Smith America Invents Act (“AIA”), Pub. L. No. 112-29,
125 Stat. 284 (2011), amended 35 U.S.C. § 103 effective March 16, 2013.
Because the effective filing date for Application 14/828,176 predates the
AIA’s amendment to § 103, this decision refers to the pre-AIA version of
§ 103.
Name Reference Date
Naito et al. (“Naito”) US 2002/0065845 A1 May 30, 2002
Cardie et al. (“Cardie”) US 2009/0112892 A1 Apr. 30, 2009
Cong et al. (“Cong”) US 2010/0076978 A1 Mar. 25, 2010
Lee at al. (“Lee”) US 2010/0153094 A1 June 17, 2010
L. Hirschman & R. Gaizauskas, Natural Language Question Answering:
The View from Here, 7 Natural Language Engineering 275–300 (2001)
(“Hirschman”)
Appeal 2019-005632
Application 14/828,176

5
The Rejections on Appeal
Claims 1, 3–6, 8, and 21 stand rejected under 35 U.S.C. § 103(a) as
unpatentable over Cong and Naito. Final Act. 4–10.
Claim 7 stands rejected under 35 U.S.C. § 103(a) as unpatentable over
Cong, Naito, and Lee. Final Act. 11.
Claim 9 stands rejected under 35 U.S.C. § 103(a) as unpatentable over
Cong, Naito, and Cardie. Final Act. 12.
Claims 10, 12–14, 16–20, and 22 stand rejected under 35 U.S.C.
§ 103(a) as unpatentable over Cong, Naito, Hirschman, and Cardie. Final
Act. 13–21, 23–27.
Claim 15 stands rejected under 35 U.S.C. § 103(a) as unpatentable
over Cong, Naito, Hirschman, Cardie, and Lee. Final Act. 22.
ANALYSIS
We have reviewed the § 103(a) rejections in light of Appellant’s
arguments that the Examiner erred. Based on the record before us and for
the reasons explained below, we agree with Appellant’s arguments that the
Examiner erred in finding that the cited portions of the references teach or
suggest the claimed subject matter. We add the following to address and
emphasize specific findings and arguments.
The § 103(a) Rejections of Claims 1, 3–9, and 21
INDEPENDENT CLAIM 1
As noted above, the § 103(a) rejection of independent claim 1 rests on
Cong and Naito. See Final Act. 4–7. Appellant argues that the Examiner
erred in rejecting claim 1 because Cong’s question-context-answer triple
differs from the claimed “fact triple.” See Appeal Br. 40–43; Reply Br. 2–3.
In particular, Appellant contends that Cong’s question-context-answer triple
Appeal 2019-005632
Application 14/828,176

6
“has the elements of [context] [question] [answer], rather than a fact triple
having the elements of [unknown object] [relationship] [known object].”
Appeal Br. 42; Reply Br. 2. According to Appellant, claim 1
“contemplate[s] using, for example, the question ‘Who directed Star Wars,’
to generate the first fact triple [unknown object] [directed] [star wars], which
includes a known entity [Star Wars], an unknown entity [unknown object],
and the relation object between the known entity and the unknown entity
[directed].” Appeal Br. 42; see Reply Br. 2, 9. Appellant also contends that
Cong’s question-context-answer triple and the claimed “fact triple” represent
triples at “different layers of abstraction” because each element of the
claimed “fact triple” is derivable “entirely from a single element (i.e., the
question element) of the question-context-answer triple.” Appeal Br. 42;
Reply Br. 2; see Appeal Br. 40.
Appellant also argues that the Examiner erred in rejecting claim 1
because Cong does not teach or suggest the following limitation in claim 1:
“identifying from the question a known entity, an unknown entity, and a
relationship between the known entity and the unknown entity.” See Appeal
Br. 45–48; Reply Br. 3–9. In particular, Appellant contends that “Cong
identifies from a plurality of sentences whether a particular sentence has a
relationship to another sentence, where the relationship can be a question-
answer relationship, question-question relationship, or question-context
relationship.” Appeal Br. 46; Reply Br. 9. Appellant also contends that
Cong does not disclose “identifying from the question a relationship
between entities identified in the question, such as the relationship of
[directed] between the known entity of [Star Wars] and the [unknown
Appeal 2019-005632
Application 14/828,176

7
object] ‘Who’ in the first fact triple as derived from the question ‘Who
directed Star Wars’.” Appeal Br. 46; Reply Br. 9.
Further, Appellant asserts that the Examiner interprets the limitation
“identifying from the question a known entity, an unknown entity, and a
relationship between the known entity and the unknown entity” too broadly
to improperly encompass identifying a question-answer relationship. See
Appeal Br. 47–48; Reply Br. 3–9. Appellant also asserts that the Examiner’s
interpretation of that limitation renders superfluous the previously recited
limitation “identifying a question and a corresponding answer in
unstructured text.” See Appeal Br. 47; Reply Br. 5–6. According to
Appellant, the Examiner wrongly “equates ‘known entity’ with the question”
and wrongly “equates ‘unknown entity’ with the answer to the question.”
Reply Br. 3, 8.
In addition, Appellant contends that “Naito does not cure the
deficiency of Cong.” Appeal Br. 48; see Reply Br. 10. Appellant also
contends that “Naito falls short of the claimed subject matter, providing only
a possible starting point (i.e., identification of an answer and/or similar
answers that corresponds to a question and/or similar questions) for the
type of fact addition to a knowledge base.” Appeal Br. 48.
In response, the Examiner determines that Cong “denotes having a
question (a known entity), and looking for an answer (an unknown entity),
with the relationship being context, all of which is disclosed.” Ans. 4 (citing
Cong ¶ 33). The Examiner explains that “[t]he known entity is the question
itself, the unknown entity is the answer to the question, and the relationship
is the context between the question and answer.” Id. at 8. The Examiner
reasons that Cong’s question-context-answer triples “are made by searching
Appeal 2019-005632
Application 14/828,176

8
forum posts, finding questions (known entities), and looking for answers
(unknown entities) using the context (the relation).” Id. at 4–5. In addition,
the Examiner determines that “since Cong clearly meets these limitations,”
there is “no need for Naito to meet these limitations.” Id. at 10.
Based on the record before us, we agree with Appellant that the
Examiner has not adequately explained how the cited portions of Cong teach
or suggest (1) the claimed “fact triple” and (2) the limitation “identifying
from the question a known entity, an unknown entity, and a relationship
between the known entity and the unknown entity.” See Appeal Br. 40–43,
45–48; Reply Br. 2–9.
Regarding the claimed “fact triple,” Cong forms a question-context-
answer triple by analyzing textual sentences and classifying sentence
relationships, e.g., as question-question, question-answer, or question-
context. See Cong ¶¶ 12, 17–19, 33–34, Fig. 1. To form a question-context-
answer triple, Cong “first identif[ies] questions in a thread, and then
identif[ies] the context and answer of every question within a uniform
framework.” Id. ¶ 19; see id. ¶ 17. When Cong forms a question-context-
answer triple, the answer corresponds to a known entity, not an unknown
entity. See id. ¶¶ 12, 17–19, 33, Fig. 1. As Appellant contends, Cong’s
question-context-answer triple and the claimed “fact triple” represent triples
at “different layers of abstraction.” See Appeal Br. 42; Reply Br. 2.
Further, we agree with Appellant that interpreting the limitation
“identifying from the question a known entity, an unknown entity, and
a relationship between the known entity and the unknown entity” to
encompass identifying a question-answer relationship renders superfluous
the previously recited limitation “identifying a question and a corresponding
Appeal 2019-005632
Application 14/828,176

9
answer in unstructured text.” See Appeal Br. 47; Reply Br. 5–6. Preferably,
a claim interpretation does not render superfluous any claim language. See
Power Mosfet Techs., L.L.C. v. Siemens AG, 378 F.3d 1396, 1410 (Fed. Cir.
2004).
The Examiner identifies no reason to depart from this principle. Final
Act. 4–7, 28; Ans. 8–10. Instead, the Examiner interprets the limitation
“identifying from the question a known entity, an unknown entity, and a
relationship between the known entity and the unknown entity” too broadly.
See Final Act. 4–7; Ans. 8–10; Spec. 218:21–220:22. As properly
interpreted, this limitation does not encompass simply identifying a
question-answer relationship. Thus, the Examiner wrongly determines that
identifying a question-answer relationship satisfies this limitation. Final
Act. 4–7; Ans. 8–10. Due to that error, the Examiner does not adequately
explain how the cited portions of Cong teach or suggest this limitation. See
Final Act. 4–7; Ans. 8–10.
Hence, for the reasons discussed above, we do not sustain the § 103(a)
rejection of claim 1.
DEPENDENT CLAIMS 3–6, 8, AND 21
Claims 3–6, 8, and 21 depend directly or indirectly from claim 1. For
the reasons discussed for claim 1, we do not sustain the § 103(a) rejection of
these dependent claims.
DEPENDENT CLAIMS 7 AND 9
Claims 7 and 9 depend directly from claim 1. On this record, the
Examiner has not shown how the additionally cited Lee and Cardie
references overcome the deficiencies in Cong discussed above for claim 1.
Appeal 2019-005632
Application 14/828,176

10
Hence, for the reasons discussed for claim 1, we do not sustain the § 103(a)
rejections of claims 7 and 9.
The § 103(a) Rejections of Claims 10, 12–20, and 22
INDEPENDENT CLAIMS 10 AND 20
As noted above, the § 103(a) rejection of independent claims 10
and 20 rests on the combination of Cong, Naito, Hirschman, and Cardie. See
Final Act. 13–16, 23–27. Appellant argues that the Examiner erred in
rejecting claims 10 and 20 for the same reasons that the Examiner erred
in rejecting claim 1. See Appeal Br. 49–50.
Claim 10 recites “generate a fact triple from the question, the fact
triple including the first entity object, the second entity object, and the
relation object.” Appeal Br. 62. Claim 20 recites (1) “generate a first fact
triple, the first fact triple including the first known entity object, the relation
object, and an unknown entity object representing the unknown entity,” and
(2) “generate a third fact triple from the second question, the third fact triple
including the third entity object, the fourth entity object, and the second
relation object representing the second relationship.” Id. at 64. The
Examiner cites Cong as teaching or suggesting a “fact triple” according
to each of claims 10 and 20. See Final Act. 13–14, 23–24.
For the reasons discussed above for a “fact triple” according to
claim 1, Examiner has not adequately explained how the cited portions of
Cong teach or suggest a “fact triple” according to each of claims 10 and 20.
See Final Act. 13–14, 23–24. Hence, we do not sustain the § 103(a)
rejection of claims 10 and 20.
Appeal 2019-005632
Application 14/828,176

11
DEPENDENT CLAIMS 12–14, 16–19, AND 22
Claims 12–14, 16–19, and 22 depend directly or indirectly from
claim 10. For the reasons discussed for claim 10, we do not sustain the
§ 103(a) rejection of these dependent claims.
DEPENDENT CLAIM 15
Claim 15 depends directly from claim 10. On this record, the
Examiner has not shown how the additionally cited Lee reference overcomes
the deficiencies in Cong discussed above for claim 10. Hence, for the
reasons discussed for claim 10, we do not sustain the § 103(a) rejection
of claim 15.
Because the preceding determinations resolve the § 103(a) rejections
for all pending claims, we need not address Appellant’s other arguments
regarding Examiner error. See, e.g., Beloit Corp. v. Valmet Oy, 742 F.2d
1421, 1423 (Fed. Cir. 1984) (explaining that an administrative agency may
render a decision based on “a single dispositive issue”).
CONCLUSION
We reverse the Examiner’s decision to reject claims 1, 3–10, and
12–22.
In summary:
Claim(s)
Rejected 35 U.S.C. § Reference(s)/Basis Affirmed Reversed
1, 3–6,
8, 21 103(a) Cong, Naito
1, 3–6,
8, 21
7 103(a) Cong, Naito, Lee 7
9 103(a) Cong, Naito, Cardie 9
Appeal 2019-005632
Application 14/828,176

12
Claim(s)
Rejected 35 U.S.C. § Reference(s)/Basis Affirmed Reversed
10, 12–14,
16–20, 22 103(a)
Cong, Naito,
Hirschman, Cardie
10, 12–14,
16–20, 22
15 103(a)
Cong, Naito,
Hirschman,
Cardie, Lee
15
Overall
Outcome
1, 3–10,
12–22
REVERSED