Apple Inc.

Patent Trials and Appeals Board

Nov 3, 2020

2019004354 (P.T.A.B. Nov. 3, 2020)

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.
15/141,645 04/28/2016 Jerome R. BELLEGARDA P25879US1/77870000211101 6931
150004 7590 11/03/2020
DENTONS US LLP - Apple
4655 Executive Dr
Suite 700
San Diego, CA 92121
EXAMINER
COLUCCI, MICHAEL C
ART UNIT PAPER NUMBER
2656
NOTIFICATION DATE DELIVERY MODE
11/03/2020 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):
dentons_PAIR@firsttofile.com
patent.docket@dentons.com
patents.us@dentons.com
PTOL-90A (Rev. 04/07)

UNITED STATES PATENT AND TRADEMARK OFFICE
____________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
____________

Ex parte JEROME R. BELLEGARDA and JANNES G. DOLFING

Appeal 2019-004354
Application 15/141,645
Technology Center 2600
____________

Before MAHSHID D. SAADAT, JOHNNY A. KUMAR, and
CARL L. SILVERMAN, Administrative Patent Judges.

SAADAT, Administrative Patent Judge.

DECISION ON APPEAL1

Pursuant to 35 U.S.C. § 134(a), Appellant2 appeals from the
Examiner’s decision to reject claims 1 and 3–17, which are all the claims in
this application.3 We have jurisdiction under 35 U.S.C. § 6(b).
We REVERSE.

1 An oral hearing was held for this appeal on October 26, 2020.
2 We use the word “Appellant” to refer to “applicant” as defined in
37 C.F.R. § 1.42(a). Appellant identifies the real party in interest as Apple
Inc. Appeal Br. 4.
3 Claim 2 has been canceled previously.
Appeal 2019-004354
Application 15/141,645

2
STATEMENT OF THE CASE
Appellant’s disclosure is directed to “statistical language modeling,
and more specifically to techniques for speech prediction and recognition.”
Spec. ¶ 2. Claim 1 is illustrative of the invention and reads as follows:
1. A non-transitory computer-readable storage
medium storing one or more programs, the one or more
programs comprising instructions, which when executed by one
or more processors of an electronic device, cause the electronic
device to:
receive a character of a sequence of characters;
determine a current character context based on the
received character of the sequence of characters and a previous
character context;
determine a current word representation based on the
current character context, wherein the current word
representation has a first vector dimensionality;
determine, by the one or more processors, a current word
context based on the current word representation and a previous
word context;
determine, by the one or more processors, a next word
representation based on the current word context, wherein the
next word representation has a second vector dimensionality;
and
provide, by a display of the electronic device, the next
word representation.
REFERENCES AND REJECTIONS
The prior art relied upon by the Examiner is:
Name Reference Date
Medlock US 2012/0223889 A1 Sept. 6, 2012
Boehl US 2014/0230055 Al Aug. 14, 2014
Bangalore US 2016/0004690 Al Jan. 7, 2016
Zhai US 2016/0299685 Al Oct. 13, 2016

Appeal 2019-004354
Application 15/141,645

3
Claims 1, 3–8, and 12–14 stand rejected under 35 U.S.C. § 103(a) as
unpatentable over Medlock and Zhai. Final Act. 4–12.
Claims 9–11 stand rejected under 35 U.S.C. § 103(a) as unpatentable
over Medlock, Zhai, and Boehl. Final Act. 12–14.
Claims 15–17 stand rejected under 35 U.S.C. § 103(a) as unpatentable
over Medlock, Zhai, and Bangalore. Final Act. 14–17.

ISSUE ON APPEAL
Appellant’s arguments in the Appeal Brief present the following
dispositive issue4:
Whether the Examiner erred in finding the combination of Medlock
and Zhai teaches or suggests the following limitation recited in independent
claims 1, 13, and 14:
determine a current word representation based on the current
character context, wherein the current word representation has a
first vector dimensionality;
. . .
determine, by the one or more processors, a next word
representation based on the current word context, wherein the
next word representation has a second vector dimensionality.
Appeal Br. 11.

ANALYSIS
We have reviewed the Examiner’s rejections in light of Appellant’s
contentions in the Appeal Brief and the Reply Brief that the Examiner has

4 We do not address Appellant’s other contentions because this contention is
dispositive of the issue on appeal.
Appeal 2019-004354
Application 15/141,645

4
erred, as well as the Examiner’s response to Appellant’s arguments in the
Appeal Brief. As discussed below, we are persuaded by Appellant’s
contentions of Examiner error.
For the limitation at issue, the Examiner relies on the disclosure in
Medlock of an interface as displayed on a small display of an electronic
watch in Figure 17. Final Act. 5 (citing Medlock ¶¶ 43, 69, 95, 160, 161,
261–293, Fig. 17). The Examiner maps the recited “first vector” to
Medlock’s entries of “See,” “See you,” and “See you later” and explains that
“This also abides by word prediction algorithms where each vector adds to
the dimensionality ... and word context from preceding input and next word
prediction analogous to terms expressly and predictions thereof (0047, 0220,
and fig. 13e).” Final Act. 5–6. The Examiner further adds Zhai “for clarity
and rather than taking official notice or an inherency to teach otherwise
well-known word prediction based on preceding input” and also as
disclosing the next word representation having a second vector
dimensionality, similar to the disclosure of Medlock. Final Act. 6 (citing
Zhai ¶¶ 88, 91, 112; Medlock ¶¶ 28, 66, 94). Finally, the Examiner finds
that
Therefore, it would have been obvious to one of ordinary
skill in the art at the time of the invention to modify the system
of Medlock to incorporate the above claim limitations as taught
by Zhai to allow for well-known fundamental usage of
preceding N number of word prediction in context i.e.
probability of a next term (word) given the previous analogous
to Medlock but in the context of words (present in Medlock)
made explicit by Zhai.
Final Act. 7.
Appellant contends the Examiner erred in reading the claimed vector
on the number of words or characters in Medlock because, “as set forth in
Appeal 2019-004354
Application 15/141,645

5
the Specification, a representation of the current word, such as a vector, is
provided by encoding the current word” and is different from a character or
a word. Appeal Br. 14. According to Appellant, “the Specification makes
clear that the word representations have dimensions corresponding various
factors” that have a certain dimension, but “none even remotely define the
term as a number of characters corresponding to a respective word (the
alleged ‘vector), or alternatively, the number of words in a respective phrase
(also the alleged ‘vector), as asserted by the Examiner.” Appeal Br. 14–15.
Appellant further argues that the Examiner’s interpretation of the claim term
“vector” is inconsistent with the meaning of the word, as used in the cited
prior art or the related publications. See Appeal Br. 16–25.
In response, the Examiner points to paragraph 229 of the Specification
to support the applied claim interpretation that “dimensionality refers to the
number of characters or words such as a user types a word character by
character applicable to multiple words in sequence.” Ans. 4–6. Based on
this interpretation, the Examiner reiterates that the cited portions of Medlock
“demonstrate that the processing of BOTH characters/words AND
probabilities are essential to predicting the next word or character based on
the current word or character and context thereof.” Ans. 7 (citing Medlock
¶¶ 43, 69, 95; Fig. 17). With respect to Zhai, the Examiner adds that “Zhai
analogous to Medlock expressly teaches vector and dimension and
analogous to the claims as well as in 0039-0044” and in “fig. 1, fig. 5, and
0088 of Zhai a word is built based on character sequences as most likely
word candidates form e.g. ‘S’ and ‘A’ to build one of ‘saying’ or ‘satin’.”
Ans. 8.
Appeal 2019-004354
Application 15/141,645

6
To the extent paragraph 229 of the Specification can be characterized
as how “a number of characters” or “a number of words” correspond to the
recited dimensionality, such connection relates to the word representation,
and not the number of characters or words a user inputs. See Spec. ¶ 229.
We find no indication in this paragraph or the related ones describing the
process shown in Appellant’s Figure 11, that “the dimensionality refers to
the number of characters or words such as a user types a word character by
character,” as asserted by the Examiner. See Ans. 4. As stated by
Appellant,
In fact, reading paragraph [0229] in isolation would not even
lead a reader to make the same conclusion as the Examiner.
For example, the sentence in paragraph [0229] describes “a
dimensionality of the character of the sequence of characters
corresponds to a number of characters.” However, it is unclear
how the dimensionality of a single character would somehow
correspond to multiple typed characters. Similarly, it is unclear
how the dimensionality of a single word representation would
correspond to multiple words, based on the description of “a
dimensionality of the next word representation corresponds to
a number of words.”
Reply Br. 9. We also agree with Appellant that “the Application as-filed
makes clear that the dimensionality of the next word representation,
corresponding to ‘a number of words,’ does not refer to an amount of typed
words in a phrase” and instead, “‘a number of words’ corresponds to the
size of a vocabulary to which a predicted next word belongs, which may be
based on the type of encoding utilized.” Id.
In view of the above analysis, we are also persuaded by Appellant’s
contentions that the text prediction process disclosed in Medlock is not
related to the current and next word representation according to a vector
dimensionality. See Appeal Br. 28–30. Similarly, as argued by Appellant,
Appeal 2019-004354
Application 15/141,645

7
Zhai discloses a vector in relation with a set of probabilities for the character
selection in a neural network, which is different from a word representation
having a vector dimensionality. See Appeal Br. 30.
Accordingly, we agree with Appellant that the Examiner does not
provide prima facie support for the rejection. “[T]he examiner bears the
initial burden, on review of the prior art or on any other ground, of
presenting a prima facie case of unpatentability.” In re Oetiker, 977 F.2d
1443, 1445 (Fed. Cir. 1992). Therefore, on the record before us, we are
constrained to conclude the Examiner errs in rejecting claim 1, as well as
independent claims 13 and 14 that recite similar limitations, as obvious.

CONCLUSION
For the reasons stated above, we do not sustain the obviousness
rejection of independent claims 1, 13, and 14, as well as claims 3–8 and 12
dependent therefrom, over Medlock and Zhai. The Examiner has not relied
upon the additional references to teach or suggest the above-identified
limitation in rejecting claims 9–11 and 15–17. See Final Act. 12–17.
Accordingly, for reasons similar to those above for independent claims 1,
13, and 14, we do not sustain the remaining rejections.

DECISION SUMMARY
In summary:
Claims
Rejected
35 U.S.C. § Basis Affirmed Reversed
1, 3–8,
12–14
103(a) Medlock, Zhai 1, 3–8, 12–14
9–11 103(a) Medlock, Zhai,
Boehl
9–11
Appeal 2019-004354
Application 15/141,645

8
Claims
Rejected
35 U.S.C. § Basis Affirmed Reversed
15–17 103(a) Medlock, Zhai,
Bangalore
15–17
Overall
Outcome
1, 3–17

REVERSED