Ex Parte Fritsch et al

Patent Trial and Appeal Board

Jun 30, 2017

12471167 (P.T.A.B. Jun. 30, 2017)

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.
12/471,167 05/22/2009 Juergen Fritsch M0002-1001C2 5334
24208 7590
Robert Plotkin, P.C.
1 Broadway, 14th Floor
Cambridge, MA 02142
07/05/2017 EXAMINER
DESIR, PIERRE LOUIS
ART UNIT PAPER NUMBER
2659
NOTIFICATION DATE DELIVERY MODE
07/05/2017 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):
mail@rplotkin.com
hdas@rplotkin.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
BEFORE THE PATENT TRIAL AND APPEAL BOARD
Ex parte JUERGEN FRITSCH, MICHAEL FINKE, DETLEF KOLL,
MONIKA WOSZCZYNA, and GIRIJA YEGNANARAYANAN
Appeal 2017-004854
Application 12/471,1671
Technology Center 2600
Before LINZY T. McCARTNEY, NATHAN A. ENGELS, and
JAMES W. DEJMEK, Administrative Patent Judges.
ENGELS, Administrative Patent Judge.
DECISION ON APPEAL
STATEMENT OF THE CASE
Appellants appeal under 35 U.S.C. § 134(a) from a Final Rejection of
claims 1—38. We have jurisdiction over the pending claims under 35 U.S.C.
§ 6(b).
We affirm.
1 Appellants identifies Multimodal Technologies, LLC as the real party in
interest. Br. 1.
Appeal 2017-004854
Application 12/471,167
REPRESENTATIVE CLAIM
Claims 1 and 20 are independent claims. Claim 1, reproduced below,
is representative of the claimed subject matter:
1. A computer-implemented method comprising steps of:
(A) identifying a probabilistic language model including a
plurality of probabilistic language models associated with a
plurality of concepts, wherein the plurality of probabilistic
language models is logically organized in a first hierarchy;
(B) using a speech recognition decoder to apply the
probabilistic language model to a spoken audio stream to
produce a document including content organized into a plurality
of sub-structures logically organized in a second hierarchy
having a logical structure defined by a path through the first
hierarchy.
Br. 11 (Claims App’x).
THE REJECTIONS
Claims 1—8, 10-27, and 29-38 stand rejected under
35 U.S.C. § 103(a) as unpatentable over Wang (US 6,785,681 Bl; iss.
Aug. 31, 2004), Alshawi (US 5,870,706; iss. Feb. 9, 1999), and Thelen et al.
(US 6,526,380 Bl; iss. Feb. 25, 2003) (“Thelen”). Final Act. 5-16.
Claims 9 and 28 stand rejected under 35 U.S.C. § 103(a) as
unpatentable over Wang, Alshawi, Thelen, and Ross, Jr. et al. (US
5,823,948; iss. Oct. 20, 1998) (“Ross”). Final Act. 17-18.
ANALYSIS
The Examiner’s Findings
Wang teaches a system that can receive input through various user-
input interfaces, including a speech-capture interface capable of converting
2
Appeal 2017-004854
Application 12/471,167
user speech into various types of outputs. Wang, col. 6,11. 30—33, Fig. 2.
Wang teaches that each user-input interface uses a recognition engine and an
associated language model, where the recognition engines use the language
models to identify “surface semantic structures” to represent the respective
inputs. Wang, col. 6,11. 38—43.
Among other findings, the Examiner finds Wang teaches “identifying
a probabilistic language model including a plurality of probabilistic language
models associated with a plurality of concepts” with Wang’s disclosures of
language models including “any one of a collection of known stochastic
models” such as an N-gram model, a context-free grammar, and
combinations of such models. Final Act. 5 (citing Wang, col. 6,11. 53—58
(“For language-based user input such as speech and handwriting, the
language model used by the recognition engine can be any one of a
collection of known stochastic models. For example the language model can
be an N-gram model that models the probability of a word in a language
given a group of N proceeding works in the input.)); see Wang, col. 6,
11. 58—64 (“The language model can also be a context free grammar that
associates semantic . . . and/or syntactic information with particular words or
phrases. In one embodiment of the present invention, a unified language
model is used that combines an N-gram language model with a context free
grammar.”); cf., e.g., Spec. 149 (“context-free grammars and n-gram
language models 306a-e are both examples of ‘probabilistic language
models’ as that term is used herein”).
Wang also teaches that its language models are capable of generating
“hierarchical surface semantic structures” that use tokens as place values for
words and N-gram probabilities for hypothesized combinations of words and
3
Appeal 2017-004854
Application 12/471,167
tokens. See Final Act. 5—6 (citing Wang, col. 7,11. 1—8); see Wang, col. 6,
11. 59-col. 7,11. 1—8; see also Wang, Fig. 4 (depicting “a graphical
representation of a surface semantic tree structure generated by a recognition
engine for a speech input” (Wang, col. 3,11. 14—16)), col. 9,11. 25—32 (“In an
embodiment that uses a tree structure for the discourse semantic structure,
the semantic tokens appear as nodes on a tree, and the attributes of the token
appear as children of the node. In such embodiments, discourse engine 214
attempts to collapse the discourse tree from the bottom up so that children
nodes are collapsed first and the resolution of the nodes ‘bubbles up.’”),
col. 11,11. 22—58. Further, Wang’s methods generate and use XML pages
and SML pages with the XSLT standard to generate “an appropriate output
page such as an html page, a wml page, or some other output” (Wang
col. 17,11. 32—36), which the Examiner finds teaches “generating a
document” (Final Act. 6 (citing Wang, col. 15,11. 40-41, col. 16,11. 28-44)),
but the Examiner additionally cites Alshawi’s methods of using language
recognition to generate a “transcription” as a further teaching of “generating
a document” as required by claim 1. See Final Act. 6—7 (citing Alshawi,
col. 2,11. 57-60, col. 3,11. 5-8, col. 4,11. 13-22, 34—41, col. 6,11. 30-35); cf.
Spec. H59 (describing use of XML documents to render a structured textual
document on a stylesheet to produce a rendered document; “Techniques for
generating stylesheets and for rendering documents in accordance with
stylesheets are well known to those having ordinary skill in the art.”); 103
(“the structured textual document 310 may be implemented as an XML
document or other document which supports nested structures”). Further,
although Examiner finds Wang and Alshawi both teach use of hierarchical
language models, the Examiner additionally cites Thelen as teaching the
4
Appeal 2017-004854
Application 12/471,167
claimed “plurality of probabilistic language models is logically organized in
a first hierarchy,” as required by claim 1. Final Act. 7 (citing Thelen, col. 8,
11. 53-61, col. 9,11. 10-24).
Appellants’ Arguments
Without rebutting the Examiner’s findings regarding Wang (see Final
Act. 5—6 (citing Wang, col. 4,11. 29-30, col. 6,11. 53—58, col. 7,11. 1—8, 19-
32, col. 15,11. 41—42, col. 16,11. 28-44)), Appellants argue that, because the
Examiner cited Thelen in the Final Office Action, the Examiner “appears to
accept [Appellants’] assertion” that neither Wang nor Alshawi teaches or
suggests “a plurality of language models logically organized in a hierarchy.”
Br. 6 (emphasis omitted). Appellants argue that the Examiner cited Thelen
“[t]o cure this deficiency in Wang and Alshawi” (Br. 6) and, according to
Appellants, even if Thelen teaches a plurality of language models logically
organized in a first hierarchy, the combination of Wang, Alshawi, and
Thelen still does not teach or suggest “using a speech recognition decoder
... to produce a document including content organized ... in a second
hierarchy having a logical structure defined by a path through the first
hierarchy,” as recited in claim 1 (the “disputed limitation”) (Br. 7—9).
Accord Br. 9 (“The mere teaching of a first plurality of probabilistic
language models organized in a first hierarchy, in other words, does not
teach or suggest the express limitation of using a speech recognition decoder
to produce content organized in a second hierarchy having a logical structure
defined by a path through the first hierarchy.”)
But Appellants’ arguments do not substantively address the combined
teachings of the prior art—in particular, Wang’s methods of using
5
Appeal 2017-004854
Application 12/471,167
hierarchical language models and XML, SML, and XSLT for generation of
html pages or other appropriate outputs in combination with the teachings of
Thelen and Alshawi.2 Moreover, the Examiner cites Thelen for the
additional teaching of a plurality of probabilistic language models logically
organized in a first hierarchy, not for the “producing a document” limitation.
See Final Act. 7 (citing Thelen, col. 8,11. 53—61, col. 9,11. 10-24); but see
Spec. 59 (“XML documents . . . typically are rendered in a form that is
more easily readable before being presented to the end user. . . . Techniques
for generating stylesheets and for rendering documents in accordance with
stylesheets are well-known to those having ordinary skill in the art.”), 105
(“Any of a variety of techniques, including techniques well-known to those
of ordinary skill in the art, may be used to search through the language
model hierarchy. . . . The structured document generator 308 selects the
candidate structured document having the highest fitness score as the final
structured textual document 310.”).
Indeed, as noted in the Examiner’s answer, Appellants’ arguments
improperly attack the cited references individually without substantively
addressing what a person of ordinary skill would have understood from the
references’ teachings in combination. See In re Keller, 642 F.2d 413, 426
(CCPA 1981) (“[0]ne cannot show non-obviousness by attacking references
individually where, as here, the rejections are based on combinations of
references.”). In combining Wang, Alshawi, and Thelen, the Examiner
2We note that Wang alone teaches or at least suggests the disputed
limitation. See Spec. 94—102, Figs. 11 A, 12A. See In re Bush, 296 F.2d
491, 496 (CCPA 1961) (sustaining a multiple reference rejection under 35
U.S.C. § 103(a) by relying on less than all of the references); In re Boyer,
363 F.2d 455, 458 n.2 (CCPA 1966).
6
Appeal 2017-004854
Application 12/471,167
articulates a rationale to combine—increasing speech recognition
accuracy—drawn directly from the cited prior art, which Appellants do not
persuasively rebut. See Final Act. 7—8 (citing Alshawi, col. 2, col. 3,11. 3—7;
Thelen, col. 9,11. 10-24). Further, Appellants cite no evidence to show that
applying Thelen’s hierarchically arranged speech recognition model to
Wang’s system (as modified by Alshawi)—to produce a document including
content logically organized in a discourse semantic tree having a logical
structure defined by a path through Thelen’s hierarchically arranged speech
recognition model—would have been uniquely challenging or anything
more than a routine exercise of applying known techniques to achieve
predictable results. See KSRInt’l Co. v. Teleflex Inc., 550 U.S. 398, 416—17
(2007) (explaining as examples of combinations likely to be obvious “[t]he
combination of familiar elements according to known methods . . . when it
does no more than yield predictable results” and “the mere application of a
known technique to a piece of prior art ready for the improvement”);
Leapfrog Enters., Inc. v. Fisher-Price, Inc., 485 F.3d 1157, 1162 (Fed. Cir.
2007) (citing KSR, 550 U.S. at 418).
For the reasons stated above, having considered the Examiner’s
rejection of claim 1 in light of each of Appellants’ arguments and the
evidence of record, we disagree with Appellants and sustain the Examiner’s
rejection of claim 1 under 35 U.S.C. § 103(a), as well as the rejections of
claims 2—38, which are not argued separately with particularity beyond the
arguments advanced for claim 1. See Br. 6 (“Dependent claims 2—19 and
21—38 are not argued separately herein.”), 9 (“The arguments [applied to
claim 1] above are equally applicable to claims 2—38.”). We adopt as our
own the Examiner’s findings, conclusions, and reasons in the rejections of
7
Appeal 2017-004854
Application 12/471,167
claims 1—38 and the Examiner’s answer to the extent consistent with the
above.
DECISION
We affirm the Examiner’s decision rejecting claims 1—38 under
35 U.S.C. § 103(a).
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(1). See 37 C.F.R.
§§ 41.50(f), 41.52(b).
AFFIRMED
8