Paper No. 58
`Trials@uspto.gov
`571-272-7822
`Date Entered: January 30, 2014
`
`
`
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`____________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`____________
`
`INTERTHINX, INC.
`Petitioner1
`
`v.
`
`CORELOGIC SOLUTIONS, LLC
`Patent Owner
`____________
`
`Case CBM2012-00007
`Patent 5,361,201
`____________
`
`
`
`
`Before, MICHAEL P. TIERNEY, JONI Y. CHANG,
`and BRIAN J. McNAMARA, Administrative Patent Judges.
`
`McNAMARA, Administrative Patent Judge.
`
`
`
`FINAL WRITTEN DECISION
`35 U.S.C. § 328(a) and 37 C.F.R. § 42.73
`
`
`
`
`
`
`
`1 On November 12, 2013, the Board terminated Petitioner’s involvement without
`terminating the proceeding under 37 C.F.R. § 327(a).
`
`
`
`IBG 1082
`IBG v. TT
`CBM2016-00054
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`
`
`BACKGROUND
`
`In its Petition for covered business method patent review of US 5,361,201
`
`(the ’201 Patent), Interthinx, Inc. (“Petitioner”) asserted that claims 1, 5, 6, 9, and
`
`10 were unpatentable under 35 U.S.C. §§ 102 and 103, and recited unpatentable
`
`subject matter under 35 U.S.C. § 101. Pet. 13-80. CoreLogic Information
`
`Solutions, LLC (“Patent Owner”) later disclaimed claim 5. Prelim. Resp. 11, 13.
`
`The Board instituted a trial on January 31, 2013. Decision to Institute, Paper 16.
`
`Petitioner’s involvement terminated late in this proceeding, pursuant to a
`
`settlement with the Patent Owner. Termination of Petitioner Pursuant To
`
`Settlement, Paper 47. The Board retained jurisdiction to issue this Final Written
`
`Decision. 35. U.S.C. § 317(a).
`
`The ’201 Patent, which is expired, is the subject of a jury verdict rendered
`
`on September 28, 2012, and a judgment entered in CoreLogic Information
`
`Solutions, Inc. v. Fiserv, Inc., No. 2:10-CV-132-RSP (E.D. Tex. Oct. 2, 2012).
`
`Among other things, the District Court entered judgment of non-infringement in
`
`favor of Petitioner and in favor of Patent Owner, rejecting Petitioner’s assertion
`
`that patent claims 1 and 10 of the ’201 Patent are invalid as anticipated or obvious.
`
`Ex. 2006. Several days earlier, on September 23, 2012, the District Court denied
`
`Defendant’s Motion for Summary Judgment that the Patent-In-Suit [the ’201
`
`Patent] is Invalid under 35 U.S.C. § 101. Ex. 2003. Post-trial motions filed in the
`
`District Court included Patent Owner’s Motion for Judgment as a Matter of Law
`
`that Petitioner infringed the ’201 Patent, Petitioner’s Motion for Judgment as a
`
`Matter of Law that claims 1 and 10 of the ’201 Patent are invalid under 35 U.S.C.
`
`§ 102 and/or § 103, and Petitioner’s Motion for Judgment as a Matter of Law that
`
`the ’201 Patent is invalid under 35 U.S.C. § 101. The District Court denied all
`
`post-trial motions on September 30, 2013. Ex. 2039; Ex. 2040. On October 25,
`
`
`
`2
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`2013, the parties moved to terminate this covered business method patent review.
`
`Paper 44. On November 12, 2013, the Board terminated Petitioner’s involvement
`
`without terminating the proceeding. 37 C.F.R. § 327(a). Patent Owner presented
`
`arguments at an oral hearing conducted on December 2, 2013.
`
`THE ’201 PATENT (EXHIBIT 1001)
`
`All of the challenged claims are drawn to “[a] computer implemented
`
`method for appraising a real estate property.” Noting that traditional statistical
`
`techniques, such as multiple linear regression and logistical regression, have been
`
`tried in the past, the ’201 Patent identifies uncertainty as to the optimal temporal
`
`and geographical sample size among the difficulties of applying a regression model
`
`to the appraisal problem. Ex. 1001, co1. 1, l. 56 - col. 2, l. 16. The ’201 Patent
`
`addresses these problems with a model development component and a property
`
`valuation component. Ex. 1001, col. 6, ll. 4-6. Using predictive modeling
`
`techniques, such as neural networks and regression modeling, the model
`
`development component uses training data describing a number of real estate
`
`properties, characteristics, and prices to build models containing information
`
`representing learned relationships among a number of variables and to develop
`
`error models, which are typically regression models, to estimate error in predicted
`
`sales prices. Ex. 1001, col. 6, ll. 3-22. The property valuation component feeds
`
`input data describing the subject property and its geographic area to the neural
`
`network models and error models to generate price estimates, error ranges, and
`
`other codes to be output to a display device, printer, or database for future access.
`
`Ex. 1001, col. 6, ll. 23-30.
`
`In our Decision to Institute, we adopted the constructions applied by the
`
`District Court. Paper 16 at 15-16. With the construction of the construed terms
`
`indicated by italics, claim 1 recites:
`
`
`
`3
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`
`A computer implemented method (which does not require a general
`purpose computer and does not exclude human interaction or input) for
`appraising a real estate property, comprising the steps of:
`collecting training data (data which is available regarding real estate
`properties);
`developing a predictive model (which is not limited to a neural
`network and does not exclude a regression model) from the training data
`(data which is available regarding real estate properties);
`storing the predictive model (which is not limited to a neural network
`and does not exclude a regression model);
`obtaining individual property data for the real estate property;
`developing an error model (a model that estimates error in the
`predicted sales price of the subject property generated by the predictive
`model) from the training data (data which is available regarding real estate
`properties);
`storing the error model (a model that estimates error in the predicted
`sales price of the subject property generated by the predictive model) ; and
`generating a signal indicative of an error range for the appraised value
`responsive to the application of the individual property data to the stored
`error model (a model that estimates error in the predicted sales price of the
`subject property generated by the predictive model).
`
`Claim 6, which depends from disclaimed independent claim 5 and
`
`incorporates all the limitations of claim 5, differs from claim 1 in several ways.
`
`Claim 6 limits the training data to individual property training data describing past
`
`real estate sales which is aggregated into area training data sets describing a
`
`plurality of sales within a geographic area. The aggregating step is repeated using
`
`successively larger geographic areas until the number of sales within the
`
`geographic area over a predetermined time period exceeds a predetermined
`
`number. Another important difference between claims 1 and 6 is that claim 6 does
`
`not recite an error model.
`
`Claim 9 differs from claim 1 by reciting the selection of a geographic area
`
`surrounding the real estate property and obtaining area data for the geographic
`
`area. Claim 9 also does not recite an error model.
`
`
`
`4
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`
`Claim 10 depends from claim 9 and recites the same steps of developing an
`
`error model and generating a signal indicative of an error range that are recited in
`
`claim 1.
`
`RES JUDICATA AND COLLATERAL ESTOPPEL
`
`As noted above, in the co-pending litigation, the District Court denied all
`
`post-trial motions for judgment as a matter of law, and the parties subsequently
`
`settled. There has been no appeal of the District Court judgment.
`
`Citing The Restatement of the Law Judgments 2d, Patent Owner argues that
`
`res judicata bars Petitioner’s §101 case because the District Court entered a final
`
`summary judgment on the merits of that claim. PO Resp. 70-71. Citing the
`
`Supreme Court’s decision in Microsoft Corp. v. i4i, 131 S.Ct. 2238, (2011) , and
`
`Cybersource Corp. v. Retail Decisions, Inc., 654 F.3d 1366, 1369 (Fed. Cir. 2011),
`
`Patent Owner argues that collateral estoppel also applies to Petitioner’s challenge
`
`under 35 U.S.C. § 101 because the question is purely one of law, rather than fact,
`
`to which the clear and convincing standard is not applicable. PO Resp. 73-74.
`
`Thus, Patent Owner argues that in this case, the patent is expired and cannot be
`
`amended, the Board adopted the Court’s claim construction, and that for questions
`
`of law, district courts and the Board apply the same standard. Id. at 74-75.
`
`Patent Owner’s underlying assumption that subject matter eligibility
`
`determinations are pure questions of law, not subject to the clear and convincing
`
`evidence standard, is not supported by the Federal Circuit. “[T]he analysis under
`
`[35 U.S.C.] § 101, while ultimately a legal determination, is rife with underlying
`
`factual issues.” Ultramercial, Inc. v. Hulu, LLC, 722 F.3d 1335, 1339 (Fed. Cir.
`
`2013) (citing, e.g., CLS Bank Int'l v. Alice Corp., 717 F.3d 1269, 1304–05 (Fed.
`
`Cir. 2013) (en banc)) (Chief Judge Rader, and Judges Linn, Moore, and O'Malley,
`
`concluding that “any attack on an issued patent based on a challenge to the
`
`
`
`5
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`eligibility of the subject matter must be proven by clear and convincing evidence,”
`
`and Judges Lourie, Dyk, Prost, Reyna, and Wallach, concluding that a statutory
`
`presumption of validity applies when § 101 is raised as a basis for invalidity in
`
`district court proceedings.)). In denying Petitioner’s motion for judgment as a
`
`matter of law under 35 U.S.C. § 101, the District Court stated only that Petitioner
`
`failed to show that claims 1 and 10 of the ’201 Patent do not cover a “new and
`
`useful process, machine, manufacture, or composition of matter, or any new and
`
`useful improvement thereof.” Ex. 2039, 2. The District Court’s decision applied
`
`the clear and convincing evidence standard to the facts underlying its
`
`determination of law when denying Petitioner’s motion for summary judgment and
`
`motion for judgment as a matter of law that the claims of the ’201 Patent do not
`
`recite eligible subject matter. In contrast, the Board reviews the patentability of a
`
`claim, rather than its validity, and applies a preponderance of the evidence standard
`
`to the underlying factual determinations, e.g., whether the claims contain
`
`limitations that narrow or tie them to specific applications of an otherwise abstract
`
`concept. 35 U.S.C. § 326(e); Ultramercial, 722 F.3d at 1339. Because the Board
`
`applies to the underlying facts an evidentiary standard that is different from the
`
`standard applied by the courts, the issue decided by the Board is not identical to the
`
`one decided or litigated in the first action and could not have been essential to the
`
`final judgment in the first action. See In re Freeman, 30 F.3d, 1459, 1465 (Fed.
`
`Cir. 1994). Petitioner did not have an opportunity to litigate the issue in the first
`
`action. Id. Therefore, we conclude that Petitioner’s challenge under 35 U.S.C. §
`
`101 is not barred by res judicata or collateral estoppel.
`
`In denying Petitioner’s motion for judgment as a matter of law under 35
`
`U.S.C. § 102 and/or 35 U.S.C. § 103 the Court stated only that it found substantial
`
`evidence to support the jury’s finding of validity. Ex. 2039, 2. Patent Owner
`
`
`
`6
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`argues that collateral estoppel applies to the prior art ground of unpatentability
`
`because the issues before the District Court and the Board are identical, the issue
`
`was litigated at the District Court, the issue of whether the prior art invalidated the
`
`patent was essential to the final judgment of validity and the Petitioner had a full
`
`and fair opportunity to litigate the issue. However, the jury’s finding that
`
`Petitioner had not proved invalid any claim of the ’201 Patent under the clear and
`
`convincing evidence standard, is not binding on the Board, which evaluates claim
`
`patentability and applies a preponderance of the evidence standard. 35 U.S.C.
`
`326(e); see, In re Swanson, 540 F.3d 1368, 1377 (Fed. Cir. 2008) (stating a court’s
`
`holding that a patent is not invalid “is not binding on subsequent litigation or PTO
`
`reexaminations”).
`
`Patent Owner argues that, because the ’201 Patent is expired and cannot be
`
`amended, the Board should apply the higher clear and convincing evidence
`
`standard applied by the District Court. PO Resp. 79. However, the statute does
`
`not provide an exception for expired patents. 35 U.S.C. § 326(e).
`
`Finally, to apply issue preclusion, the party against whom the estoppel is
`
`being asserted must have been accorded a full and fair opportunity to litigate in the
`
`prior court proceeding the very issue he now seeks to relitigate. In re Freeman, 30
`
`F.3d at 1467. Petitioner is no longer a party to this proceeding and Patent Owner
`
`has failed to demonstrate that the Office, which is proceeding to a final written
`
`decision under 35 U.S.C. § 327(a), had a full and fair opportunity to litigate the
`
`patentability issues in the prior court proceeding.
`
` Therefore, we conclude that res judicata and collateral estoppel do not limit
`
`the Board’s ability to decide the challenges at issue in this proceeding.
`
`
`
`CLAIMS 1, 6, 9 AND 10 DO NOT RECITE PATENT ELIGIBLE SUBJECT
`MATTER UNDER 35 U.S.C. § 101
`
`
`
`7
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`
`
`
`A persistent theme in Patent Owner’s arguments is that the claims recite
`
`patentable subject matter because a computer is central to all the claims and the
`
`processes cannot be done manually. PO Resp. 59-62. The fact that a claim relies
`
`on a method that is implemented on a computer is not a per-se indicator of
`
`patentability. Rather, a challenged claim, properly construed, must incorporate
`
`enough meaningful limitations to ensure that what is claimed is more than just an
`
`abstract idea and is not a mere “drafting effort designed to monopolize [an abstract
`
`idea] itself.” Mayo Collaborative Servs. v. Prometheus Labs, Inc., 132 S. Ct. 1289,
`
`1297 (2012). In order for a machine to impose a meaningful limitation on the
`
`scope of a method claim, it must play a significant part in permitting the claimed
`
`method to be performed, rather than function solely as an obvious mechanism for
`
`permitting a solution to be achieved more quickly. SiRF Tech., Inc. v. Int'l Trade
`
`Comm'n, 601 F.3d 1319, 1333 (Fed. Cir. 2010). Claims that recite a method of
`
`doing business on a computer, and do no more than merely recite the use of the
`
`computer for its ordinary function of performing repetitive calculations, are not
`
`patent eligible. Bancorp Servs., L.L.C. v. Sun Life Assurance Co., 687 F.3d 1266,
`
`1278-79 (Fed. Cir. 2012) (computer used for its most basic function, the
`
`performance of repetitive calculation, does not impose a meaningful claim
`
`limitation).
`
`Patent Owner argues that under the machine or transformation test, the
`
`computer plays a necessary and vital role to the development and storage of the
`
`predictive and error models. PO Resp. 63. During the oral hearing, Patent Owner
`
`argued that the word “model” in the claims requires a specialized computer, not a
`
`mere algorithm, a mere function, or equations. Tr. 10. Patent Owner also argued
`
`that all of the claim limitations require the computer. Tr. 16. Although the
`
`preamble recites a computer implemented process, none of the claim elements,
`
`
`
`8
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`with the possible exception of the “storing” limitations, specifically recites a
`
`relationship to the computer.
`
`According to Patent Owner, the claims pass the Federal Circuit’s “mental
`
`process test” because they recite an automated, predictive model based system for
`
`real estate appraisals using sophisticated error modeling techniques, including
`
`statistical techniques, to overcome human bias and cannot be performed entirely
`
`manually or in the human mind. Id. at 66-70, Tr. 19-21, 23. During the oral
`
`hearing, Patent Owner argued extensively that the claims require implementation
`
`on a computer. Tr. 15-18, 24-26. However, the claims recite collecting training
`
`data, developing the predictive model, and developing the error model in the
`
`abstract, and do not tie necessarily these steps to a computer or a particular
`
`application. Even the claim limitations that recite generating the signal responsive
`
`to application of individual property data to the model do not require any specific
`
`volume of data (as claimed, the individual property data could merely be the
`
`address of the property). For example, Dr. Lipscomb testified that claim 1 does not
`
`provide any threshold for training data. Ex. 1024, 99, 114. Dr. Lipscomb further
`
`testified that a predictive model could be developed manually using a limited
`
`number of observations. Id. at 101.
`
`As previously discussed, the inquiry under § 101 requires a search for
`
`limitations in the claims that narrow or tie the claims to a specific application of an
`
`otherwise abstract concept. Ultramercial, 722 F.3d at 1339. We consider the
`
`claim as a whole. Id. at 1344. Patent Owner argues that the claims satisfy the
`
`“abstract idea” test for patentable subject matter because, rather than being tied
`
`preemptively to a field of use, they are narrowly tied to a specific application, i.e.,
`
`previously unknown computer-based modeling for real estate appraisal by
`
`developing and storing predictive and error models. PO Resp. 64-66.
`
`
`
`9
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`
`A claim is not patent eligible if, instead of claiming an application of an
`
`abstract idea, the claim instead is drawn to the abstract idea itself. Ultramercial,
`
`722 F.3d at 1343. Inventor, Dr. Jost considers his invention to be the process for
`
`predicting the selling price of a home (a point estimate) using statistical analysis to
`
`develop the point estimate, and a co-inventor’s error model, which was also
`
`developed using statistical analysis. Ex. 1022, 72-73, 81-82. Dr. Jost’s testimony
`
`is consistent with claims 1 and 10, which recite appraising real estate by
`
`developing and storing a predictive model and generating a signal indicative of the
`
`appraised (predicted) value, and developing and storing an error model and
`
`generating a signal indicative of the error range for the appraised (predicted) value.
`
` Patent Owner’s expert Dr. Lipscomb characterizes the invention as “a way
`
`to determine the sales price of a property that may not have a recent sales
`
`transaction” and that predictive models “have been around a long time.” Ex. 1024,
`
`42-43. Determining a price has been found to be abstract as a method of
`
`calculating. See, SAP America, Inc. v. Versata Dev. Group, Inc., CBM2012-
`
`00001, 2013 WL 3167735, at *16 (PTAB, June 11 2013). Dr. Jost testified that it
`
`was common in statistics to provide a confidence interval around estimates. Ex.
`
`1022, 102. The testimony of Dr. Jost and Dr. Lipscomb indicates that determining
`
`an error can be accomplished using the same well known techniques as those
`
`applied to the predictive model and that bounding a price estimate by an error
`
`range is an abstract concept. Therefore, we find the claimed development of a
`
`model to predict a value and an error model to assess the error range around the
`
`predicted value, as recited in claims 1 and 10, is an abstract concept. Similarly,
`
`claims 6 and 9 recite limitations that bound the geographic area from which data is
`
`obtained to develop the model. The limitations in claims 6 and 9 recite the abstract
`
`and well known concept of examining the geographic area around the subject
`
`
`
`10
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`property so that relevant samples can be obtained to predict the price of the
`
`property.
`
` A claim is not patent-eligible where it merely recites a law of nature and
`
`adds additional steps that merely reflect routine, conventional activity of those who
`
`work in the field. Mayo, 132 S. Ct. at 1298. As discussed above, we find that
`
`claims 1, 6, 9 and 10 recite abstract concepts and do not transform these ideas into
`
`patent eligible applications of these abstractions. Therefore, we conclude that
`
`claims 1, 6, 9 and 10 of the ’201 Patent recite non-patentable subject matter.
`
`ANALYSIS OF PRIOR ART CHALLENGES
`
`COMPUTER-ASSISTED MASS APPRAISAL SYSTEMS
`
`One theme underlying Patent Owner’s arguments against the challenges to
`
`the claims of the ’201 Patent is that the ’201 Patent concerns an automated
`
`valuation model (AVM), whereas the prior art supporting the challenges concerns
`
`computer-assisted mass appraisal (CAMA) systems. PO Resp. 1-2. Patent Owner
`
`contends that the prior art stands in contrast to the claims of the ’201 Patent
`
`because CAMA systems are used by tax assessors to assess all properties in a
`
`jurisdiction, while the ’201 Patent describes generating a property-specific value to
`
`provide meaningful information to an underwriter. Id. at 2.
`
`Although Patent Owner argues that appraising all properties in a jurisdiction
`
`is “the exact opposite of what the [’201 Patent] describes and claims,” id., Patent
`
`Owner admits that “CAMA systems generate a specific property value that is then
`
`used as part of the property tax assessment calculation.” Id. at 15-16. This
`
`generation of a specific property value highlights the applicability of the CAMA
`
`references. Whether a valuation is desired for purposes of assessing taxes on many
`
`properties, making a purchase offer on a single property, or some other purpose,
`
`the objective of the ’201 Patent claims and the CAMA prior art is the generation of
`
`
`
`11
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`a specific property value. Thus, we do not agree with Patent Owner’s underlying
`
`premise that the CAMA art is inapplicable to the claimed subject matter.
`
`The Patent Owner Response also argues that there is no evidence that the
`
`references, some of which have more than one date, were publicly accessible. PO
`
`Resp. 13-14. However, Patent Owner did not move to exclude the references from
`
`this proceeding, as provided for in 37 C.F.R. § 42.64(c) and the Scheduling Order.
`
`Therefore, the Board does not exclude the CAMA evidence.
`
`
`CLAIMS 1, 6, 9 AND 10 ARE ANTICIPATED BY OR OBVIOUS OVER
`JENSEN -1 (EXHIBIT 1014)
`
`The Board instituted a trial based on Petitioner’s challenge that Jensen-1
`
`(Ex. 1014)2 anticipates claims 1, 6, 9, and 10 of the ’201 Patent. As construed,
`
`claims 1, 6, 9 and 10 recite developing and storing a predictive model, such as a
`
`regression model, from training data, i.e., data which is available regarding real
`
`estate properties. Claims 1 and 10 also recite developing and storing an error
`
`model, i.e., a model that estimates error in the predicted sales price of the subject
`
`property generated by the predictive model, from the same training data used to
`
`develop the predictive model. The claims recite no other limitations on the error
`
`model. Claims 1 and 10 further recite generating a signal indicative of the error
`
`range responsive to application of the individual property data to the error model.
`
`The claims do not limit how the individual property data is applied to the error
`
`model.
`
`Patent Owner argues that Jensen-1 teaches using the training data to value
`
`all the properties and does not disclose obtaining individual property data and then
`
`valuing the individual property, as recited in all the challenged claims. PO Resp.
`
`
`2 Jensen-1 (Ex. 1014) indicates that it was published in The Property Tax Journal
`in September 1987.
`
`
`
`12
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`54-55. Patent Owner’s argument is similar to its underlying theme that CAMA
`
`prior art does not apply to the claimed AVM system. As discussed above, we do
`
`not find this argument persuasive because the ultimate objective of each system is
`
`to determine a value for each individual parcel. The same sentence in Jensen-1
`
`that refers to a model estimating the sales price of all the properties also refers to
`
`adjustments necessary to correct the actual sales prices of selected recent
`
`comparable sales for the property characteristic differences between the subject
`
`property being valued and each comparable sale, in order to ascertain what each
`
`comparable property would have sold for had it been identical to the subject
`
`property in a physical descriptive sense on the valuation date. Ex. 1014, p. 194.
`
`(Emphasis added). Thus, as recited in claims 1, 6, 9 and 10, Jensen-1 obtains
`
`individual property data and recognizes a relationship between individual property
`
`data and the training data.
`
`Jensen-1 also discloses using the training data to predict the individual
`
`property value, as recited in claims 1, 6, 9 and 10. Jensen-1 discusses a model in
`
`which the value of a property is based on the land value qualified by a set of
`
`adjustments. Ex. 1014, p. 198. The adjustments reflect specific property
`
`characteristics. For example, up to a point, the value of a property can be expected
`
`to increase with increasing frontage. Id. Thus, specific properties with different
`
`frontage are valued differently.
`
`Jensen-1 discloses other modeling approaches that determine various per
`
`unit rates within a market model using regression analysis, Longini and Carbone’s
`
`adaptive estimation, and Carlson’s iterative correlative estimation procedure. Id. at
`
`200. Jensen-1’s discussion of interactive correlative estimation specifically states
`
`that, within each iteration, “the model estimate and then the residual error (the
`
`actual sale price minus the model estimate) are computed on a per parcel basis.”
`
`
`
`13
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`Ex. 1014, p. 220. Thus, Jensen-1 discloses generating an appraised value
`
`responsive to the application of individual property data to the predictive error
`
`model. Jensen-1 specifically states that “[o]nce meaningful market models have
`
`been developed from the available sales, they can be used to estimate the fair
`
`market values of all of the parcels in the jurisdiction whether recently sold or not.”
`
`Ex. 2014, p. 224. As discussed above, the reference to all parcels in the
`
`jurisdiction does not imply that all the parcels are evaluated the same. Jensen
`
`discloses that the market models take into account the various characteristics of
`
`each parcel and determine a value for each parcel.
`
`Jensen-1 notes that the model estimate, i.e., the fair market value of each
`
`parcel as determined by the market models, is an abstract number and its reliability
`
`as an approximation of true market value depends on the quality of the individual
`
`estimates. Ex. 1014, p. 224. Jensen-1 specifically notes that when an actual sale
`
`price falls below the value estimate, it may or may not indicate a problem with the
`
`appraisal. Id. Thus, Jensen-1 applies to individual appraisals and attempts to assess
`
`an error in the appraisal of each parcel.
`
`Patent Owner argues that Jensen-1 does not disclose or suggest an error
`
`model, as recited in claims 1 and 10. PO Resp. 51-54. Consistent with its position
`
`that CAMA systems seek to achieve a specific point estimate and do not employ
`
`error models, Patent Owner argues that error statistics disclosed in Jensen-1 are
`
`different from error models. Id. at 52. Patent Owner contends that error statistics
`
`are applicable to all properties in the training data, whereas an error model
`
`generates a signal unique to the individual property being valued. Id. at 16.
`
`Patent Owner further argues that reliability factors, mean square error and
`
`asymptotic confidence bands in Jensen-1 describe observations used in the training
`
`data and do not relate to any individual property. Id. at 53. However, claims 1 and
`
`
`
`14
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`10 both recite that the error model is developed from the training data, rather than
`
`the individual property data. Thus, the use of training data to develop the error
`
`models and confidence bands in Jensen is consistent with the error model
`
`limitations of claims 1 and 10.
`
`During the oral hearing, Patent Owner characterized Petitioner’s arguments
`
`as assuming that, if there’s an error range, an error model exists. Tr. 40. Patent
`
`Owner argued that the claimed error range is generated separately from the
`
`development of both the predictive and error models. Tr. 39. We note, however,
`
`that claims 1 and 10 only recite generating a signal indicative of the error range in
`
`response to the application of the individual property data to the stored model.
`
`(emphasis added). The limitation does not recite generating the error range itself
`
`in response to application of the individual property data. Although a particular
`
`application of the individual property data to the error model may create a separate
`
`signal for a particular property, the claims do not require that the error or the error
`
`range for each property be different. There is no limitation requiring a “signal
`
`unique to the individual property,” as argued by Patent Owner. PO Resp. 52.
`
`Thus, claims 1 and 10 do not preclude confidence bands disclosed in Jensen-1
`
`from the claimed error model, as Patent Owner contends. PO Resp. 52-53.
`
`In consideration of the above, we conclude that claim 1 of the ’201 Patent is
`
`anticipated by or obvious over Jensen-1.
`
`Claim 6 depends from disclaimed claim 5, which does not include an error
`
`model. Claim 6 recites aggregating the training data into training data sets based
`
`on successively larger geographic areas until the number of sales within a
`
`geographic area over a predetermined time exceeds a predetermined number.
`
`Claim 6 does not specifically recite how the aggregated training data sets are
`
`integrated into the predictive model. Patent Owner argues that in contrast to
`
`
`
`15
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`aggregation, i.e., assembling data based on larger geographic area, CAMA models
`
`(and Jensen-1 in particular) disclose “disaggregation,” i.e., moving from a larger
`
`geographic area to a smaller one. PO Resp. 24-25, 57. According to Patent
`
`Owner, Jensen-1 describes disaggregation because it discloses a “global” model
`
`estimated from a common database that is tailored down to small towns and rural
`
`areas. PO Resp. 57.
`
`Jensen-1 recommends that for small sample environments, e.g., small towns,
`
`rural residential areas, farms, and certain types of commercial and non-residential
`
`properties, that individual models be generated from countywide or statewide
`
`models tailored to each town or village via a model update technique, such as
`
`weighted Bayesian regression. Ex. 1014, p. 236. Using this approach, all of the
`
`essential property descriptors will be included in the model with an adequate
`
`number of sales to support them. Id. Thus, instead of disaggregating a “global”
`
`model as Patent Owner suggests, PO Resp. 24-25, 57, Jensen-1 discloses
`
`aggregating information from larger geographic models, such as a statewide model
`
`or a countywide model, to create a model that can be used to predict the value of
`
`one or more properties in a geographic environment where the number of samples
`
`is insufficient for accurate modeling. Tailoring the models to each town or village
`
`using a statewide or countywide model to obtain a sufficient number of samples,
`
`inherently discloses using successively larger geographic areas until a
`
`predetermined number of sales in a predetermined time exceeds a predetermined
`
`number, as recited in claim 6. Even if Jensen-1 is not considered to disclose
`
`explicitly or inherently using successively larger geographic areas, its disclosure of
`
`using statewide or countywide data to model areas lacking an adequate number of
`
`samples suggests aggregating training data into sets based on larger geographic
`
`areas until one obtains a meaningful or predetermined number of sales over a time
`
`
`
`16
`
`

`

`Case CBM2012-00007
`Patent 5,361,201
`
`
`frame. Thus we conclude that the features of claim 6 would have been obvious
`
`under 35 U.S.C. § 103 over Jensen-1. Therefore, we conclude that claim 6 is
`
`anticipated by or at least