Opinion
Civil Action 1:18-CV-0549 (BKS/CFH)
09-30-2021
RENSSELAER POLYTECHNIC INSTITUTE and CF DYNAMIC ADVANCES LLC, Plaintiffs, v. AMAZON.COM, INC., Defendant.
FOR PLAINTIFFS: ROBINS KAPLAN LLP FOR DEFENDANT: KNOBBE MARTENS HANCOCK ESTABROOK, LLP BRYAN J. VOGEL, ESQ. DANIELLE ROSENTHAL, ESQ. BENJAMEN C. LINDEN, ESQ. CHRISTOPHER SEIDL, ESQ. MARY PHENG, ESQ. SHUI LI, ESQ. TAYLOR MOORE-WILLIS, ESQ. CHRISTINE YUN SAUER, ESQ. LI ZHU, ESQ. JEREMY A. ANAPOL, ESQ. JON W. GURKA, ESQ. JOSEPH S. CIANFRANI, ESQ. JOSEPH R. RE, ESQ. COLIN HEIDEMAN, ESQ. JOHN G. POWERS, ESQ. JAMES P. YOUNGS, ESQ.
FOR PLAINTIFFS: ROBINS KAPLAN LLP
FOR DEFENDANT: KNOBBE MARTENS HANCOCK ESTABROOK, LLP
BRYAN J. VOGEL, ESQ. DANIELLE ROSENTHAL, ESQ. BENJAMEN C. LINDEN, ESQ. CHRISTOPHER SEIDL, ESQ. MARY PHENG, ESQ. SHUI LI, ESQ. TAYLOR MOORE-WILLIS, ESQ. CHRISTINE YUN SAUER, ESQ. LI ZHU, ESQ.
JEREMY A. ANAPOL, ESQ. JON W. GURKA, ESQ. JOSEPH S. CIANFRANI, ESQ. JOSEPH R. RE, ESQ. COLIN HEIDEMAN, ESQ. JOHN G. POWERS, ESQ. JAMES P. YOUNGS, ESQ.
REPORT AND RECOMMENDATION
DAVID E. PEEBLES, U.S. MAGISTRATE JUDGE
Plaintiffs Rensselaer Polytechnic Institute ("RPI") and CF Dynamic Advances, LLC ("Dynamic Advances") have commenced this action against defendant Amazon.com, Inc. ("Amazon"), alleging patent infringement. The patent at issue discloses a natural language interface ("NLI") consisting of a method for processing natural language inputs provided by a user through a search of language-based databases. Plaintiffs allege that certain of Amazon's products and services infringe upon claims contained within the patent.
Despite their best efforts, the parties have been unable to agree on the proper construction to be attributed to several claim terms contained within RPI's patent, and now seek guidance from the court with regard to eleven such terms. The question of claim construction has been referred to me for the issuance of a report and recommendation to the assigned district judge. Based upon the comprehensive submissions from the the recommendations that follow.
I. BACKGROUND
At the center of this case is United States Patent No. 7, 177, 798 ("'798 Patent"), issued on February 13, 2007, and entitled "Natural Language Interface Using Constrained Intermediate Dictionary of Results." The '798 Patent discloses innovations in the area of NLI that are used to process inputs in the form of natural language queries. The patent lists Dr. Cheng Hsu, a professor at RPI, and Dr. Veera Boonjing, a former doctoral candidate at RPI who now teaches in Thailand, as the inventors, and RPI as the assignee of rights under the patent. Dynamic Advances, a limited liability company headquartered in Texas, is the exclusive licensee under the patent. The '798 Patent was issued based upon an application filed on May 21, 2001, and traces its roots back to a continuation-in-part application filed on or about April 7, 2000, but later abandoned, and a provisional application filed on May 19, 2000.
The '798 Patent can be found at various locations on the docket. See, e.g., Dkt. Nos. 1-1, 142-2, 143-2.
Although the specification identifies the filing date of the priority application as April 17, 2000, Amazon's expert in this action has suggested that the application was filed on April 7, 2000, and that the reference to the 17th is a typographical error. See Dkt. No. 142-14 at 9 n.2. Because the exact date is immaterial to the issues discussed.
The eleven disputed terms currently before the court for construction appear in several of the patent's claims. Claims 1 and 9, which are set forth below, are the patent's only two independent claims and provide a useful backdrop for addressing the terms implicated in the claim construction exercise.
The disputed terms in these two claims are set forth in bold text.
'798 Patent, Claim 1
1. A method for processing a natural language input provided by a user, the method comprising:
providing a natural language query input by the user;
performing, based on the input, without augmentation, a search of one or more language-based databases including at least one metadata database comprising at least one of a group of information types comprising:
case information;
keywords;
information models; and
database values;
providing, through a user interface, a result of the search to the user;
identifying, for one or more language-based databases, a finite number of database objects; and determining a plurality of combinations of the finite number of database objects.
'798 Patent, Claim 9
9. A computer-implemented method for processing a natural language input comprising:
receiving a natural language input;
providing from said natural language input a
plurality of language-based database objects;
identifying a finite number of permutations of the plurality of database objects, the database objects being stored in a metadata database comprising at least one of a group of information comprising
case information,
keywords,
information models, and database values; and
interpreting at least one of the permutations to provide determination of a result of the natural language input.
Figure 2 of the '798 Patent, which depicts "a natural language query processor 201 according to one embodiment of the invention," '798 Patent, 12:52-53, is also useful in placing the disputed claim terms in context:
(Image Omitted)
In this embodiment, the processor 201 receives a natural language inquiry and produces a query result 203. Id. at 12:53-55. This is accomplished by mapping the query 202 to a plurality of database objects 204A using a reference dictionary 208 containing keywords 209, case information 210, information models 211, and database object values 204B. Id. at 12:60-64.
According to the '798 Patent specification, "the natural language query can take several forms including, by way of example, a paragraph, a sentence, sentence fragment, or a plurality of keywords." '798 Patent, 12:55-57.
The query processor 201 includes a reference dictionary object identifier 205 that parses the query 202 and generates one or more objects recognized in the reference dictionary 208. '798 Patent, 13:17-20. The database processor 201 also accepts and processes a number of database objects 204A. Id. at 13:23-24. Significantly, in the reference dictionary 208, keywords 209 may be "learned" from the user or may be provided through a separate keyword administrator interface associated with the query processor 201. Id. at 13:30-34.
The query processor 201 also includes an interpreter and dictionary processor 207 that determines optimal interpretations of the received objects, resolves ambiguities, updates information models 211, and interacts with users to facilitate learning. '798 Patent, 13:35-42. Rules 212 and heuristics 213 are utilized by the processor 207 to resolve ambiguities in determining the optimal interpretation of the query 202. Id. at 13:42-44.
A mapping processor 206 performs a mapping between incoming objects and database objects 204A, and may generate database queries from the objects and interpretations provided by the identifier 205 and the processor 207, respectively. '798 Patent, 13:52-56. The processor 206 may, for example, generate structural query language ("SQL") queries used to locate database objects 204A, and those queries may be executed by an SQL search engine, and the processor 201 may provide a query result 203 to the user through, for example, a graphical user interface. Id. at 13:56-60.
II. PROCEDURAL HISTORY
Plaintiffs commenced this action on May 8, 2018. Dkt. No. 1. In their complaint plaintiffs allege that several of defendant's products including, inter alia, Amazon Echo, Amazon Alexa, and Amazon Fire TV stick, and all Alexa-enabled devices, infringe at least Claim 1 of the '798 Patent. On August 21, 2019, Amazon answered plaintiffs' complaint, generally denying its material allegations and asserting, as defenses, (1) non- infringement, (2) invalidity, (3) prosecution history estoppel, (4) unavailability of injunctive relief, (5) limitation on damages and costs, and (6) inequitable conduct. Dkt. No. 82 at 8-11. Amazon's answer also asserts counterclaims seeking declarations of non-infringement, patent invalidity, and patent unenforceability. Id. at 11. Plaintiff replied in answer to Amazon's counterclaims on September 11, 2019. Dkt. No. 85.
The accused products made, sold, and/or used by Amazon are listed in Appendix A to plaintiffs' complaint. See Dkt. No. 1 at 25.
Additional procedural history not relevant to the issue of claim construction includes Amazon's motions to transfer the action to the Western District of Washington and for dismissal or, in the alternative, a stay of the action. Dkt. Nos. 34, 89. Both of those motions were denied. Dkt. Nos. 81, 108.
In accordance with this court's local patent rules, the parties have requested construction of the following disputed terms:
Term
Claims Implicated
based on the input, without augmentation
1-8
metadata database
1-21
case information
1-21
case
14-17
providing, through a user interface, a result of the search to the user
1-8
determining a plurality of combinations of the finite number of database objects
1-8
resolving ambiguity between the keywords and the plurality of combinations
5, 6
identifying a finite number of permutations of the plurality of database objects
9-21
interpreting at least one of the permutations to provide determination of a result of the natural language input
9-21
provide determination of a result
9-21
result of the natural language input
9-21
In addition, after conferring, the parties have reached agreement with regard to the following eleven claim terms:
Based upon the parties' agreement, and having reviewed the '798 Patent, I recommend that the court adopt the parties' agreed-upon claim term constructions.
Term
Claims Implicated
Construction
natural language query input
1-8
input that is articulated in a human language in a way that a native speaker could understand and use sensibly
natural language query input
20
natural language input
natural language input
1-21
input that is articulated in a human language in a way that a native speaker could understand and use sensibly
language-based databases
1-21
databases that include and store natural language information
database objects
1-21
objects included within a database, including but not limited to data types, data instances, database computational operators, data semantics, data structures, data instances or data values, and data operators
language-based database objects
9-21
database objects in a language-based database
keywords
1-21
natural words and phrases that users use to refer to database objects and information model elements in natural articulation
information models
1-21
webs of concepts for enterprise databases
reference dictionary
4, 14-17
set of database entries or objects and their relations
at least one metadata database comprising at least one of a group of information types comprising: case information; keywords; information models; and database values
1-8
at least one metadata database comprising one or more groups of information types including at least one group containing all four of the information types: case information, keywords,
information models and database values
a metadata database comprising at least one of a group of information comprising case information, key words, information models, and database values
9-21
a metadata database comprising one or more groups of information types including at least one group containing all four of the information types: case information, keywords, information models, and database values
Dkt. No. 129 at 1-2.
Following the submission of extensive briefing and supporting materials regarding the eleven disputed claim terms, I held a claim construction hearing on April 29, 2021. At the conclusion of that hearing, decision was reserved. The question of claim construction has been referred to me for the issuance of a report and recommendation. See Dkt. No. 150.
III. DISCUSSION
A. Claim Construction: The Legal Framework
1. General Principles
Patent claim construction is a mixed issue of law and fact to be decided by the court. Teva Pharmas. USA, Inc. v. Sandoz, Inc., 574 U.S. 318, 331-32 (2015). When the court "reviews . . . evidence intrinsic to the patent (the patent claims and specification, along with the patent's prosecution history), the . . . determination will amount . . . to a determination of law[.]" Teva Pharmas. USA, Inc., 574 U.S. at 331. If it becomes necessary to "look beyond the patent's intrinsic evidence" and consult extrinsic evidence to provide accurate interpretations of disputed claims, the determinations are considered "subsidiary factual findings." Id. Subsidiary factual findings have been described as "the 'evidentiary underpinnings' of claim construction." Id. at 331-32.
When construing a patent, as a general rule, a court assigns disputed terms their ordinary and customary meaning as they would have been understood by a person of ordinary skill in the art ("POSITA"), as of the date of filing on or about April 7, 2000, when read in the context of the patent specification and prosecution history. SpeedTrack, Inc. v. Amazon.com ("SpeedTrack"), 998 F.3d 1373, 1377 (Fed. Cir. 2021); Power Integrations, Inc. v. Fairchild Semiconductor Int'l, Inc., 904 F.3d 965, 971 (Fed. Cir. 2018); Phillips v. AWH Corp., 415 F.3d 1303, 1312-13 (Fed. Cir. 2005). "[T]he ordinary and customary meaning of a claim term is the meaning that the term would have to a [POSITA] in question at the time of the invention[.]" Phillips, 415 F.3d at 1313; accord, Finjan, Inc. v. Cisco Sys., Inc., 837 Fed.Appx. 799, 804 (Fed. Cir. 2020).
The parties did not address the characteristics of a POSITA in their briefs. Plaintiffs' expert, Dr. Steven Shwartz, has opined that a POSITA is "a person with two years of graduate or post-graduate level work in natural language processing and database systems and at least two years of experience in commercial research or development of natural language interfaces for databases." Dkt. No. 142-15 at 8. Dr. Earl Sacerdoti, Amazon's expert, has described a POSITA as a person "hav[ing] a graduate degree in computer science, or a similar field, with course work in artificial intelligence including natural language processing" or, alternatively, as a person who "would have a bachelor's degree in such a field with 2 or more years of industry experience in natural language query processing." Dkt. No. 142-14 at 9. In my view, the distinctions between the proffered characteristics of a POSITA in this case are not material for determining any of the issues at stake. For that reason, when resolving the disputes between the parties, I have considered each expert's definition of a POSITA.
There are two exceptions to this general rule, only one of which is relevant to this action. In particular, a claim term will be afforded a meaning other than its ordinary one when the patentee clearly and unmistakably disavows the full scope of the term either in the patent specification or during prosecution of the patent application. SpeedTrack, Inc., 998 F.3d at 1377-78; Butamax(TM) Advanced Biofuels LLC v. Gevo, Inc., 746 F.3d 1302, 1309 (Fed. Cir. 2014); Thorner v. Sony Computer Entm't Am., LLC, 669 F.3d 1362, 1365 (Fed. Cir. 2012). "A disavowal must be clear, but it need not be explicit." Techtronic Indus. Co. Ltd. v. Int'l Trade Comm'n, 944 F.3d 901, 907 (Fed. Cir. 2019) (internal quotation marks omitted). "Disavowal may be inferred from clear limiting descriptions of the invention in the specification or prosecution history," or "an inventor may . . . disavow claim scope by distinguishing the claimed invention over the prior art." Techtronic Indus. Co. Ltd., 944 F.3d at 907 (internal quotation marks omitted). This exception to the general rule that patent terms should be given their ordinary meaning is both narrow and exacting. Thorner, 669 F.3d at 1366-67.
While the words of a patent claim will generally control, they must not be interpreted in isolation and the POSITA "is deemed to read the claim term not only in the context of the particular claim in which the disputed term appears, but in the context of the entire patent, including the specification." Phillips, 415 F.3d at 1313. A patent's specification often constitutes the "single best guide to the meaning of a disputed term." Vitronics Corp. v. Conceptronic, Inc., 90 F.3d 7576, 1582 (Fed Cir. 1996); accord, Phillips, 415 F.3d at 1314. In this respect, a patent specification, which is like an internal dictionary, must be carefully reviewed to determine whether, for example, the inventor has used a particular term in a manner inconsistent with its ordinary meaning. Vitronics Corp., 90 F.3d at 1582. When resorting to a patent's specification for guidance with respect to disputed claim terms, a court must consider it as a whole and, where possible, all portions should be read in a manner that renders the patent internally consistent. Budde v. Harley-Davidson, Inc., 250 F.3d 1369, 1379-80 (Fed. Cir. 2001).
Although the language of a patent specification can provide important clues regarding the proper construction of a disputed claim term, there are limitations upon its usefulness. "[W]hile it is true that claims are to be interpreted in light of the specification and with a view to ascertaining the invention, it does not follow that limitations from the specification may be read into the claims[.]" Sjolund v. Musland, 847 F.2d 1573, 1581 (Fed. Cir. 1988) (emphasis in original); accord, Kinik Co. v. Int'l Trade Comm'n, 362 F.3d 1359, 1364 (Fed. Cir. 2004). "Nor should particular embodiments in the specification be read into the claims; the general rule is that the claims of a patent are not limited to the preferred embodiment." Cornell Univ. v. Hewlett-Packard Co., 313 F.Supp.2d 114, 126 (N.D.N.Y. 2004) (citing, inter alia, Texas Digital Sys., Inc. v. Telegenix, Inc., 308 F.3d 1193, 1204 (Fed. Cir. 2002)).
In addition to the ordinary meaning of a claim term itself and the patent's specification, the prosecution history related to the patent in issue can help inform the determination of a proper claim construction. Phillips, 415 F.3d at 1317. That history is generally comprised of "the complete record of proceedings before the Patent and Trademark Office [("PTO")], including any express representations made by the applicant regarding the scope of the claims" and an examination of any relevant prior art. Vitronics, 90 F.3d at 1582-83. Such evidence, which typically chronicles the dialogue between the inventor and the PTO leading up to the issuance of a patent, and thus can act as a reliable indicator of any limitations or concessions on the part of the applicant, oftentimes proves highly instructive during claim construction. See Phillips, 415 F.3d at 1317 ("[T]he prosecution history can often inform the meaning of the claim language by demonstrating how the inventor understood the invention and whether the inventor limited the invention in the course of prosecution, making the claim scope narrower than it would otherwise be.").
2. Stare Decisis
On June 12, 2014, I issued a decision and order in Rensselaer Polytechnic Institute v. Apple, Inc. ("Apple"), Civil Action No. 13-CV-0633 (N.D.N.Y., June 12, 2014), construing several disputed claim terms from the '798 Patent. Inasmuch as that case was ultimately settled shortly before trial, the Federal Circuit Court of Appeals did not have occasion to pass upon the claim construction determination from that action. Although the defendant in that case was Apple, Inc., and Amazon was not a party to the litigation, and thus is not bound by res judicata or collateral estoppel with regard to those claim constructions, plaintiffs argue that I should recommend the same constructions of mutually disputed claim terms in the interests of consistency and predictability. Addressing that argument in supplemental submissions, Amazon argues that the Federal Circuit Court of Appeals' recent decision in SpeedTrack undermines the significance of my constructions in Apple.
Specifically, that decision addressed the terms "natural language query input," "based on the input, without augmentation," "metadata database," "case," "case information," "information models," "database objects," "determining a plurality of combinations of the finite number of database objects," "identifying a finite number of permutations of plurality of database objects," and "reference dictionary." Apple, No. 13-CV-0633, Dkt. No. 104 at 44.
Horizontal stare decisis is a discretionary doctrine. Most district courts do not treat other district courts' decisions, including those issued by courts in the same district, as precedential. This is true for district courts presiding over claim construction issues where another district court has ruled on identical claim terms. See, e.g., Verizon Cal., Inc. v. Ronald Katz Tech. Licensing, P.A., 326 F.Supp.2d 1060, 1069 (C.D. Calif. 2003); Texas Instruments, Inc. v. Linear Techs. Corp., 182 F.Supp.2d 580, 589 (E.D. Tex. 2002). This notwithstanding, however, claim construction rulings by other district courts are generally deemed highly relevant and can be regarded as persuasive authority. Depomed, Inc. v. Sun Pharma Global FZE, No. 11-CV-3553, 2012 WL 3201962, at *4 (D. N.J. Aug. 3, 2012); see also KX Indus., L.P. v. PUR Water Purification Prods., Inc., 108 F.Supp.2d 380, 387 (D. Del. 2000) (affording deference to the court's own previous claim construction rulings related to the same patent); cf. Tech. Licensing Corp. v. Blackmagic Design Pty Ltd., No. 13-CV-5184, 2016 WL 8902602, at *5 (N.D. Calif. Nov. 23, 2016) (concluding that stare decisis does not apply to different district courts' interpretations of disputed claims but describing those decisions "as relevant intrinsic evidence" for purposes of its own claim construction analysis).
To the extent stare decisis applies to this court's previous claim construction ruling in the Apple litigation, it would apply only to the following two disputed terms: (1) "based on the input, without augmentation," and (2) "metadata database." Although there are three additional claim terms that were construed by the court in the Apple litigation and are now in dispute in this action, Amazon contends those terms are indefinite - an argument that was not presented by Apple and is separate and apart from the proper interpretation of the disputed terms. Those three claim terms include: (1) "case information," (2) "case," and (3) "determining a plurality of combinations of the finite number of database objects." Other than arguing indefiniteness, Amazon does not offer proposed constructions for these three terms.
The Patent Trial and Appeal Board ("PTAB") has not construed the terms "case" or "determining a plurality of combinations of the finite number of database objections." It did, however, construe the term "case information" in the same way as the court did in Apple on more than one occasion, including in decisions denying two inter partes review ("IPR") petitions filed by both Apple and Amazon, and in denying motions by Amazon for rehearing. See Dkt. Nos. 143-4 at 6-7, 143-5 at 4-7, 14-17, 143-7 at 10-11.
In supplemental briefing the parties have exchanged views concerning the significance of the Federal Circuit's decision in SpeedTrack on the question of stare decisis. See Dkt. Nos. 163, 164. The district court in SpeedTrack issued an initial claim construction decision in which it construed a term ("category description") that paralleled the construction adopted in a prior infringement suit filed by SpeedTrack. 998 F.3d at 1376. In construing the term "category description," the district court applied stare decisis in light of the Federal Circuit's affirmance of its construction of that term in another action involving the patent-in-suit. See SpeedTrack, Inc. v. Amazon.com, Inc., No. 09-CV-4479, 2019 WL 5864630, at *2 (N.D. Cal. Nov. 8, 2019). Based on subsequent disputes between the parties, however, the district court thereafter issued a second claim construction decision that clarified certain points and resolved the parties' specific disputes. Id. at 1377.
Absent from the Federal Circuit's discussion in SpeedTrack are any arguments that the district court's application of stare decisis was unreasonable or erroneous. In any event, the district court in SpeedTrack applied vertical stare decisis, which is distinct from, and subject to different weight than, horizontal stare decisis. Accordingly, I do not read the Federal Circuit's decision in SpeedTrack as eroding the principle that horizontal stare decisis may be a useful tool in claim construction.
B. Indefiniteness
Among Amazon's defenses is the contention that some of the claim terms of the '798 Patent are invalid because they are indefinite. Dkt. No. 82 at 8. As is relevant to the claim construction issue currently before the court, Amazon maintains that the following method claim limitations are indefinite:
(1) providing, through an interface, a result of the search to the user (Claim 1)
(2) determining a plurality of combinations of the finite number of database objects (Claim 1)
(3) resolving ambiguity between the keywords and the plurality of database objects (Claim 5)
(4) identifying a finite number of permutations of the plurality of database objects (Claim 9)
(5) interpreting at least one of the permutations to provide determination of a result of the natural language input (Claim 9).
The first issue that the court must decide is whether the foregoing claim terms are subject to 35 U.S.C. § 112, ¶ 6 (2000). Section 112, ¶ 6 provides as follows:
The Leahy-Smith American Invents Act, ("AIA") which took effect in 2012, resulted in § 112, ¶ 6 being renumbered as § 112(f). I am using the pre-AIA section numbering in this report and recommendation because the '798 Patent was issued prior to the enactment of the AIA amendments.
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.35 U.S.C. § 112, ¶ 6.
In O.I. Corp. v. Tekmar Co., Inc., 115 F.3d 1576 (Fed. Cir. 1997), the Federal Circuit concluded that method claims generally, like the ones at issue in the '798 Patent, can be subject to § 112, ¶ 6. O.I. Corp., 115 F.3d at 1582-83. When claims do not use the words "step for" or "means for," however, there is a presumption that the limitations described are not subject to § 112, ¶ 6. Masco Corp. v. United States, 303 F.3d 1316, 1327 (Fed. Cir. 2002); Watts v. XL Sys., Inc., 232 F.3d 877, 880 (Fed. Cir. 2000) (citing Personalized Media Commc'ns, LLC v. Int'l Trade Comm'n, 161 F.3d 696, 703-04 (Fed. Cir. 1998)). This presumption is "rebutted if the evidence intrinsic to the patent and any relevant extrinsic evidence so warrant." Personalized Media Commc'ns, LLC, 161 F.3d at 704. The presumption is not rebutted "because the claim limitation[s] recite[] or refer[] to terms that are reasonably well understood in the art as names for structure [or acts] and which perform the recited function[s]" referred to. Watts, 232 F.3d at 880. "[T]he focus remains on whether the claim as properly construed recites sufficiently definite structure [or acts] to avoid the ambit of §112, ¶ 6." Personalized Media Commc'ns, 161 F.3d at 704. "[W]here a claim recites a function, but then goes on to elaborate sufficient structure, material, or acts within the claim itself to perform entirely the recited function, the claim is not in means-plus-function format." Sage Prods. v. Devon Indus., Inc., 126 F.3d 1420, 1427-28 (Fed. Cir. 1997); see also Masco Corp., 303 F.3d at 1327 ("Where the claim drafter has not signaled his intent to invoke § 112, paragraph 6 by using the 'step[s] for' language, we are unwilling to resort to that provision to constrain the scope of coverage of a claim limitation without a showing that the limitation contains nothing that can be construed as an act.").
Amazon has attempted to diminish the significance of Masco by arguing that it was overruled by the Federal Circuit's en banc decision in Williamson v. Citrix Online, LLC, 792 F.3d 1339 (Fed. Cir. 2015). See, e.g., Dkt. No. 148 at 9-10. Williamson, however, did not overrule Masco. Instead, Williamson eviscerated the "strong" presumption against applying § 112, ¶ 6 to a claim limitation that does not recite "means for" (or "step for") language. Williamson, 792 F.3d at 1349. While Masco may be the origin of the presumption against applying § 112, ¶ 6 to a claim term in the absence of the required trigger words, it was not until Lighting World, Inc. v. Birchwood Lighting, Inc., 354 F.3d 1354, 1358 (Fed. Cir. 2004), that the Federal Circuit began to describe that presumption as "strong." In Williamson, the Federal Circuit abandoned its characterization of the presumption as a "strong" one and reiterated that § 112, ¶ 6 can (and does) apply when a claim limitation recites a function without also reciting sufficient structure to perform that function. Williamson, 792 F.3d at 1348-49. Accordingly, despite Amazon's attempts to undercut the relevance of Masco, I find that Masco remains good law to the extent that it instructs courts to apply a regular (unweighted) presumption against applying § 112, ¶ 6 to claim terms that do not employ the words "step for" or "means for." See Masco Corp., 303 F.3d at 1327; accord, Neurografix ('360) Patent Litig., 201 F.Supp.3d 206, 214 (D. Mass. 2016).
Judge Rader's concurring opinion in Seal-Flex, Inc. v. Athletic Track & Court Constr., 172 F.3d 836 (Fed. Cir. 1999), addresses the distinction between the "function" and the "act" of a claim element, and is thus helpful in determining whether a claim limitation is subject to § 112, ¶ 6. In his concurrence, Judge Rader explained that "the 'underlying function' of a method claim element corresponds to the what that element ultimately accomplishes in relationship to what the other elements of the claim and the claim as a whole accomplish. 'Acts,' on the other hand, correspond to the how the function is accomplished." Seal-Flex, Inc., 172 F.3d at 849-50 (emphasis in original).
The parties' experts, Dr. Shwartz (for plaintiffs) and Dr. Sacerdoti (for Amazon) have offered competing opinions as to the underlying functions of each of the disputed claim limitations. For their part, plaintiffs contend that the language of the specific claim limitations at issue comprise the "act" (or the "how") of accomplishing the underlying function, whereas the specification describes the "function" or the "what" of the limitation. From plaintiffs' perspective, because the claim limitations describe acts, they fall outside the ambit of § 112, ¶ 6. Amazon, on the other hand, construes the claim limitations as describing functions that are unsupported by any acts within the claim limitations themselves.
In determining whether § 112, ¶ 6 applies, a court must decide whether the claim limitation describes a function without a corresponding act to complete the function. Because of the presumption that applies absent inclusion of the "means for" and "step for" language, and the competing opinions regarding the particular functions of the disputed claim limitations, I am more persuaded by plaintiffs' contentions and, accordingly, recommend the court conclude that § 112, ¶ 6 does not apply to the method claim limitations. Plaintiffs' position is supported by O.I. Corp., 115 F.3d at 1583. In that case, the district court had applied § 112, ¶ 6 to a claim limitation after concluding that the "function" of the claim limitation was found within the patent's preamble (rather than in the claim itself). Id. In reversing the district court, the Federal Circuit observed that, "[p]erforming a series of steps inherently produces a result, in this case the removal of water vapor from the analyte slug, but a statement in a preamble of a result that necessarily follows from performing a series of steps does not convert each of those steps into step-plus-function clauses." Id.
Having set forth the foregoing general principles and foreshadowing my recommendations, I will now turn to analyzing the pertinent claim limitations to which Amazon contends § 112, ¶ 6 applies.
1. Providing, through a user interface, a result of the search to the user
With respect to the first method claim limitation found in Claim 1 at issue - "providing, through a user interface, a result of the search to the user" - plaintiffs contend that the function is "to output a result of processing the natural language input." Dkt. No. 143 at 28; see also Dkt. No. 142-15 at 24. They argue that "this 'providing' step describes how outputting a result of processing the natural language input is accomplished-i.e., providing a result through a user interface." Dkt. No. 143 at 28; Dkt. No. 147 at 12; see also Dkt. No. 162 at 123. Amazon responds that the underlying function of this claim term is "determining the result," not simply outputting the result itself, and that the claim includes no description of any acts necessary to accomplish this function. Dkt. No. 142 at 16; Dkt. No. 148 at 12; Dkt. No. 162 at 140.
As was previewed above, I recommend the court side with plaintiffs and find that the underlying function of the claim term is to output a result of the natural language input. The act necessary to accomplish this function is expressly stated within the claim limitation: provide the result through a user interface. The specification supports this conclusion. '798 Patent, 9:10-12, 9:65-10:16.
Even if the court were to conclude that the function of this claim limitation is, as Amazon argues, "determining the result," according to the language of Claim 1, in order to determine a result, a search must be performed, and that "search" is described in the paragraph immediately preceding the claim term at issue. More specifically, Claim 1 contemplates that the method for processing a natural language input includes "performing, based on the input, without augmentation, a search of one or more language-based databases including at least one metadata database comprising at least one of a group of information types comprising: case information; keywords; information models; and database values." '798 Patent, 36:41-48 (emphasis added).
Ultimately, however, I am not persuaded by Amazon's primary contention that the claim term at issue describes a function because it ignores the purpose of the claim limitation, which is to provide a result of the natural language input through a user interface. Amazon's arguments erroneously focus on how the underlying search is performed, rather than how the result of the search is provided.
Because the claim limitation at issue does not describe a function untethered to an act sufficient to perform that function, I recommend the court conclude that § 112, ¶ 6 does not apply.
2. Determining a plurality of combinations of the finite number of database objects, identifying a finite number of permutations of the plurality of database objects, and interpreting at least one of the permutations to provide determination of a result of the natural language input
With respect to the next method claim limitation in Claim 1 at issue - "determining a plurality of combinations of the finite number of database objects" - plaintiffs argue that the function of the claim is "to determine candidate interpretations of the natural language input." Dkt. No. 143 at 28; Dkt. No. 147 at 13; see also Dkt. No. 142-15 at 25. As to the acts required to accomplish that function, plaintiffs have stated in their briefs that "this 'determining' step describes how interpretations of the natural language input is accomplished-i.e., by determining a plurality of combinations of the finite number of database objects." Dkt. No. 143 at 28-29 (emphasis in original); see also Dkt. No. 147 at 13. As was already noted, plaintiffs clarified their position during oral argument by explaining that the specification describes the function of this claim limitation and that the claim limitation itself is the act needed to accomplish the function. Dkt. No. 162 at 152-53.
Plaintiffs also identify the same function for one of the two method claim limitations at issue in Claim 9, "identifying a finite number of permutations of the plurality of database objects." Plaintiffs then articulate a very similar function for the second method claim term at issue in Claim 9, "interpreting at least one of the permutations to provide determination of a result of the natural language input." More specifically, plaintiffs allege that the underlying function of the first Claim 9 limitation is "to determine candidate interpretations of the natural language input." Dkt. No. 143 at 29; Dkt. No. 147 at 14; see also Dkt. No. 142-15 at 26. They argue that the function of the second claim limitation is "to determine the best candidate interpretation of the natural language input." Dkt. No. 142-15 at 26 (emphasis added).
The parties agree that "permutation" is a reference to potential query interpretations. See Dkt. No. 142 at 17 (citing the experts' declarations).
As was noted above, plaintiffs argue that Claim 9's claim limitations, like Claim 1's claim limitation of "determining . . . combinations," constitute the acts necessary to support the functions of determining candidate interpretations of the natural language input and determining the best candidate interpretation.
For its part, Amazon contends that these claim limitations describe functions - specifically, "determining . . . combinations," "identifying . . . permutations," and "interpreting . . . permutations" - and that the claim limitations do not describe any acts necessary to accomplish the functions. Dkt. No. 142 at 16-17; Dkt. No. 148 at 14-16.
Taking into consideration the entirety of Claims 1 and 9, as well as the written description of the '798 Patent as a whole, I am persuaded by plaintiffs' arguments that the functions underlying the claim limitations at issue are (1) "to determine candidate interpretations of the natural language input" for Claim 1's "determining a plurality" limitation and Claim 9's "identifying . . . permutations" limitation, and (2) "to determine the best candidate interpretation of the natural language input" for Claim 9's "interpreting . . . permutations" limitation. Dr. Shwartz's declaration points out that the '798 Patent specification supports this conclusion insofar as it explains that "[t]he basic idea of search is to enumerate all possible paths (query graphs) and to evaluate (rank) them to determine the best path (interpretation) for a query." '798 Patent, 33:37-39. The claim limitations describe the acts required to accomplish that function. In other words, to determine candidate interpretations of the natural language input, the method must call for determining a plurality of combinations of the finite database objects (as stated in the Claim 1 limitation at issue) and identifying a finite number of permutations of the database objects (as stated in the first Claim 9 limitation at issue). Similarly, to determine the best candidate interpretation of the natural language input, the method must call for interpreting at least one of the permutations to provide determination of a result of the natural language input (as stated in the second Claim 9 limitation at issue).
Because Claims 1 and 9 describe steps associated with the method of processing a natural language input and not functions, I recommend the court conclude that § 112, ¶ 6 does not apply. See O.I. Corp., 115 F.3d at 1583 ("[C]laiming a step by itself, or even a series of steps, does not implicate section 112, ¶ 6. Merely claiming a step without recital of a function is not analogous to a means plus function.").
3. Resolving ambiguity between the keywords and the plurality of combinations
As to the last method claim limitation at issue, which is found in Claim 5 - "resolving ambiguity between the keywords and the plurality of combinations" - plaintiffs contend that the function is "to determine the best candidate interpretation of the natural language input" and the act needed to carry out this function is "by resolving the ambiguity between the keywords and the plurality of combinations." Dkt. No. 143 at 29; Dkt. No. 162 at 160. Amazon argues that the claim limitation "resolving ambiguity" is the function, which is "unsupported by any acts." Dkt. No. 148 at 17; see also Dkt. No. 142 at 17.
After again considering the claim and patent as a whole, I agree with plaintiffs that the claim limitation at issue here describes an act, rather than a function. For that reason, I recommend the court conclude that § 112, ¶ 6 does not apply.
4. Conclusion
Having recommended that the court conclude that § 112, ¶ 6 does not apply to the five method claim limitations, Amazon's contention that the claims associated with those limitations are indefinite must be rejected. In addition, because Amazon has not offered a competing proposed construction of those method claim terms, and, as reflected in the analyses above concerning each term, they appear to be readily understood as stated, I recommend the court decline to construe the method claim limitations. In particular, I recommend the court not construe any of the following claim terms:
(1) providing, through an interface, a result of the search to the user (Claim 1)
(2) determining a plurality of combinations of the finite number of database objects (Claim 1)
(3) resolving ambiguity between the keywords and the plurality of database objects (Claim 5)
(4) identifying a finite number of permutations of the plurality of database objects (Claim 9)
(5) interpreting at least one of the permutations to provide determination of a result of the natural language input (Claim 9).
Finally, it should be noted that my recommended finding that § 112, ¶ 6 does not apply to the five method claim limitations does not prejudice Amazon in moving for summary judgment based on lack of enablement, failure to meet the written description requirements, or indefiniteness under § 112, ¶ 2. Ultimately, each of Amazon's arguments with respect to the five method claim limitations leads back to its primary contention that the '798 Patent does not work. My rejection of Amazon's arguments that the claim limitations are written as step-plus-function still allows Amazon to argue in a summary judgment motion or at trial that the patent's claims and specification fail to disclose a specific way of understanding and interpreting a natural language input.
In addition to arguing, correctly so, that the working protype argument is not legally relevant to the claim construction inquiry, plaintiffs vigorously dispute Amazon's contention that, by their own admission, the patentees were unable to assemble a working prototype practicing the '798 Patent invention. Plaintiffs maintain that the argument is predicated upon incomplete depositions excerpts and mischaracterization of deposition testimony. See, e.g., Dkt. No. 155 at 6-8. At this stage of the litigation, the court need not resolve this dispute between the parties.
C. Construction of Disputed Terms
In addition to their disputes concerning the foregoing method claim terms, the parties also offer competing constructions concerning five other claim terms.
1. Based on the input, without augmentation
Plaintiff's Proposed Construction
Amazon's Proposed Construction
Court's Construction
No construction necessary, or alternatively given plain and ordinary meaning
Result of the natural language input alone, without additional sources of input or constraints on how the natural language is expressed
No construction necessary
This claim term appears in independent Claim 1, and thus also applies to the seven dependent claims that follow. Plaintiffs' proposed definition is consistent with my construction of the term from the Apple litigation. Amazon would like to add a slight twist, proposing that the court add "without additional sources of input or constraints on how the natural language is expressed." The PTAB has not weighed in on the construction of this term.
The parties' disagreement over this claim term centers upon an exchange that occurred during the prosecution of the application that ripened into the '798 Patent concerning whether the issuance of the patent would be invalid as anticipated by U.S. Patent No. 5, 966, 126 ("Szabo"). During prosecution, the applicants distinguished the Szabo system from the claimed invention by noting that Szabo allegedly improved a typical query by "graphically assist[ing] the user in refining the search." Szabo at 6:55-65. The applicants described the input in Szabo as
In its opening and reply claim construction briefs, Amazon proposed a different construction for this term than the one it finally settled on at oral argument. In particular, Amazon originally proposed to construe this term as "based solely on the natural language input without augmentation of any kind (such as additional sources of input or constraints on how the natural language is expressed)." Dkt. No. 143 at 12; Dkt. No. 142 at 25. At oral argument, Amazon streamlined its proposed construction to "based solely on the natural language input without additional sources of input or constraints on how the natural language is expressed." See Dkt. No. 162 at 186.
The Szabo patent is found in the record at Dkt. No. 143-8.
graphically assist[ing] the user in refining the search. That is, his input is not a natural language input in that it is not solely a natural language query input. It is an augmented query which provides structure and content not present in a true natural language input.Dkt. No. 142-5 at 38 (emphasis in original). The inventors argued that "[c]learly, within the context of the disclosed and claimed subject matter, Szabo does not, in fact, provide a natural language query input or perform a search based (only) on that input." Id. Instead, they argued, the invention disclosed in the Szabo patent augments the query by way of a significantly altered derivative version of the natural language input, the purpose of which was to provide structure and content not present through the natural language input. Id. at 37-39. The phrase "without augmentation" was added during prosecution to overcome a rejection by the examiner based upon Szabo. Id. at 33.
Amazon apparently agrees that the court correctly resolved the particular dispute presented in the Apple litigation concerning this term. See Dkt. No. 142 at 25. Amazon goes on, however, by proposing to specifically exclude two potential means of augmentation that are not referenced in Claim 1, including "additional sources of input" and "constraints on how the natural language is expressed." It is true, as Amazon argues, that those two examples represent potential types of augmentation. The addition of these examples of augmentation, however, would render the language of Claim 7, which is dependent upon Claim 1, superfluous. That claim provides as follows:
In Apple, the parties focused on whether "without augmentation" precludes any supplementation of the natural language input or, instead, precludes only supplementation by the user through the addition of information or structure, as plaintiffs in that case argued.
7. The method of claim 1 wherein providing a natural language query input by the user includes providing a natural language query input by the user without constraining said query grammatically or structurally.'798 Patent, 37:8-11. Inclusion of Amazon's proposed qualifier would therefore render that claim superfluous and, accordingly, would violate the doctrine of claim differentiation. See Comark Commc'ns, Inc. v. Harris Corp., 156 F.3d 1182, 1187 (Fed. Cir. 1998) ("There is presumed to be a difference in meaning and scope when different words or phrases are used in separate claims. To the extent that the absence of such difference in meaning and scope would make a claim superfluous, the doctrine of claim differentiation states the presumption that the difference between claims is significant.").
I find, moreover, that the inclusion of the additional language is unnecessary since the claim already makes it clear, by way of the qualifier "without augmentation," that the search being performed must be based solely on the natural language input. I find it unnecessary, and unsupported, to include the additional qualifiers now proposed by Amazon. I therefore recommend that the construction of this term be consistent with that reached in Apple, and that it be construed to mean "based solely on the natural language input without augmentation."
2. Result of the natural language input
In the Apple litigation, Apple urged that the term "of the natural language input," appearing in Claim 9, should be construed the same as the term "based on the input, without augmentation," found in Claim 1. See Apple, No. 13-CV-0633, Dkt. No. 104 at 18 n.7. Because the two claims are independent, I found no reason to construe the two terms identically and rejected Apple's invitation in that regard. Id. In that decision, ultimately there was no construction of the term "of the natural language input" from Claim 9.
Plaintiff's Proposed Construction
Amazon's Proposed Construction
Court's Construction
No construction necessary, or alternatively given plain and ordinary meaning
Result of the natural language input alone, without additional sources of input or constraints on how the natural language is expressed
No construction necessary
Amazon insists that this claim term, which is found in Claim 9, be construed to specifically exclude any other inputs aside from the natural language input. See, e.g., Dkt. No. 142 at 27. As it argued with respect to Claim 1's "without augmentation" claim term, Amazon relies on the patent applicants' representations during prosecution distinguishing Szabo from the '798 Patent. Id. According to Amazon, the applicants argued that the limitation set forth in Claim 9 must "preclude other inputs" beyond the natural language, and that "[o]therwise, logically speaking, the result would not be 'of the natural language input' but of something else." Dkt. No. 142-5 at 38-39.
For the same reason that I did not impose any further limits on this claim term in Apple, I decline Amazon's invitation to do so with regard to Claim 9. Once again, to accept Amazon's position would violate the doctrine of claim differentiation. Moreover, in an effort to distinguish Szabo during patent prosecution, the inventors purposely amended Claim 1 to include "without augmentation" but did not similarly amend Claim 9.
To the extent that Amazon is relying on the patent prosecution history to find disclaimer of any augmentation or constraints, this argument fails. As was discussed above in Part III.A.1., a claim term will be afforded a different meaning other than its ordinary one when the patentee disavows the full scope of the term either in the patent specification or during prosecution of the patent application. See, e.g., SpeedTrack, 998 F.3d at 1377-78. "A disavowal must be clear, but it need not be explicit." Techtronic Indus. Co. Ltd., 944 F.3d at 907 (internal quotation marks omitted). "Disavowal may be inferred from clear limiting descriptions of the invention in the specification or prosecution history," or "an inventor may . . . disavow claim scope by distinguishing the claimed invention over the prior art." Id. (internal quotation marks omitted). This exception to the general rule that patent terms should be given their ordinary meaning is both narrow and exacting. Thorner, 669 F.3d at 1366-67.
In this case, the record does not reflect the applicants' clear intention to cabin the natural language input as strictly as Amazon now proposes. In my view, Amazon's reliance on the patent applicants' statements during patent prosecution is misplaced. The statements made by the applicants were in specific reference to Szabo's prior art - an invention that graphically assists the user in refining the search - and, therefore, the query is augmented in an effort to provide structure and content. Dkt. No. 142-5 at 38. The applicants further explained that Szabo's system "receive[s] an input which is a significantly altered derivative of a natural language input, at best, augmented by a superimposed logic structure. By contrast the present invention [the '798 Patent] allows the natural language queries (input) to be constructed in any way a human user might cho[o]se to do, without re-structuring or augmentation." Id. Because the applicants' statements were in reference to a specific prior art, I do not find they rise to the level of disavowal.
Based upon the foregoing, I recommend that the court reject Amazon's limiting construction of this term and decline to construe the term because it is readily understood by a POSITA.
3. Metadata database
Amazon originally proposed a slightly different construction of this term, see Dkt. No. 129 at 3. By the time it filed its opening construction brief, however, Amazon had modified its proposed definition. See, e.g., Dkt. No. 142 at 28; Dkt. No. 142-14 at 47.
The next claim term in dispute, "metadata database," appears in both independent Claim 1 and independent Claim 9, as well as the remaining nineteen dependent claims.
The parties do not seriously dispute that the two words that comprise this claim term, "metadata" and "database," are readily understood by persons of ordinary skill in the art. One court has observed that metadata is "data about data," and, in the context of addressing the metadata of a document, has described it as
electronically-stored evidence that describes the history, tracking, or management of an electronic document. It includes the hidden text, formatting codes, formulae, and other information associated with an electronic document.
Aguilar v. Immigration & Customs Enforcement Div. of United States Dep't of Homeland Sec., 255 F.R.D. 350, 354 (S.D.N.Y. 2008) (internal quotation marks and citations omitted).
This description of a document's metadata, while not precisely analogous, is similar to that referred to in the '798 Patent specification. For example, the patent describes the metadata available to be searched as part of the natural language interface's natural language processing includes "cases, keywords, information models, and database values." '798 Patent, 8:51-54. To those of ordinary skill in the art, these all represent information about structuring, using, or interpreting data. Plaintiffs' suggested construction, which proposes that a metadata database is simply a database consisting of metadata, which in turn includes information about structuring, using, or interpreting data, is consistent with the '798 Patent specification. '798 Patent, 16:13-49.
Amazon does not challenge the court's construction of "metadata database" rendered in the Apple litigation insofar as it referred to "metadata" as including information about structuring, using, or interpreting data. Amazon correctly notes, however, that I did not construe the term "database" in Apple. See Apple, No. 13-CV-0633, Dkt. No. 269 at 67 n.35. In my summary judgment decision in the Apple litigation, I pointed out that various courts have found the ordinary and customary meaning of database to be an "organized collection of electronic information" or "structured set of data." See Id. (citing Vasudevan Software, Inc. v. Int'l Bus. Mach. Corp., No. 09-CV-5897, 2011 WL 196884, at *1, 2 (N.D. Cal. Jan. 20, 2011) (concluding that "database" is "a structured set of data"); Visicu, Inc. v. iMDsoft, Ltd., No. 07-CV-4562, 2009 WL 1291330, at *12 (E.D. Pa. May 7, 2009) (concluding that a database is "an organized collection of electronic information"); Transcenic, Inc. v. Google, Inc., 7 F.Supp.3d 405, 411, (D. Del. 2013) (finding that a "database" is "a collection of data organized for search and retrieval by a computer")).
Dr. Shwartz, plaintiffs' expert, generally agrees with a definition offered by Dr. Jaime Carbonell in the Apple litigation, defining databases as "organized electronic stores of information." Dkt. No. 142-15 at 11. That definition is consistent with both caselaw and the construction of the term that would have been understood by a POSITA at the time that patent was issued.
One major point of contention over this term appears to be whether, as Amazon argues, the term must refer to a collection of "persistently" stored data. According to Dr. Sacerdoti, the concept of persistency specifically excludes storage of data in volatile random access memory. Dkt. No. 147-3 at 6. As Dr. Shwartz responds, however, there is no support for the narrowing that Amazon now proposes, beyond what a POSITA would understand its plain and ordinary meaning to be. Dkt. No. 142-15 at 22.
According to Dr. Shwartz, at the time of the invention there were several examples of in-memory databases, such as dbVISTA/Raima and DataBlitz, where data could be accessed from a computer memory rather than disk drives. Dkt. No. 142-15 at 22.
While acknowledging that defining "database" could be viewed as somewhat redundant, when considering the claims, and in order to aid the factfinder at trial in understanding the term, I recommend that "metadata database" be construed as "collection of stored data that has a searchable organizational scheme and includes information about structuring, using, or interpreting data."
4. Case information/case
Case Information
Plaintiff's Proposed Construction
Amazon's Proposed Construction
Court's Construction
Information about prior instances of use or the natural language processing method
Indefinite
Information about prior instances of use or the natural language processing method
Case
Plaintiff's Proposed Construction
Amazon's Proposed Construction
Court's Construction
A prior use of the natural language processing method
Indefinite
A prior use of the natural language processing method
The term "case information" appears in both independent Claims 1 and 9 of the '798 Patent. The parties agree that construction of the term "case information" will be determinative of the proper construction of the related term "case," which, in its plural form, is found in Claims 14 and 17. Dkt. No. 129 at 6 n.4. Plaintiffs have proposed the construction the court adopted in the Apple litigation, which was also adopted by the PTAB. Dkt. No. 143-4 at 6-7; Dkt. No. 143-5 at 4-7, 14-17; Dkt. No. 143-7 at 10-11. Aside from arguing indefiniteness, Amazon has not taken a position regarding the proper construction of this term.
With respect to its indefiniteness argument, Amazon argues that plaintiffs' statements before the PTAB in response to Amazon's IPR petition contradict their proposed construction of "case information." Dkt. No. 142 at 22. In particular, according to Amazon, in an effort to distinguish a piece of prior art (Meng), plaintiffs argued that Meng did not disclose the claimed "case information" of the '798 Patent because Meng "only stores information consisting of the users clarifying input." Dkt. No. 142 at 22 (referencing Dkt. No. 142-9 at 49); see also Dkt. No. 153 at 10. Amazon has interpreted this statement to mean that "information consisting of the user's clarifying input" is not case information, a conclusion which is in direct conflict with the '798 Patent specification. See Dkt. No. 142 at 22 ("It is unclear what remains in the scope of 'case information' once users' clarifications are excluded."); Dkt. No. 153 at 10. In my view, Amazon's reading of plaintiffs' explanation is a tortured one that ignores the context in which the explanation was offered.
The Meng prior art consists of a technical report describing an NLI or processing system "as an extension of an existing Hi-level Query Formulator for SQL series." See Dkt. No. 142-10 at 29.
Indeed, the '798 Patent repeatedly explains that information consisting of a user's clarifying input is case information. '798 Patent, 5:30-37, 8:5-19, 17:58-63, 20:34-51, 31:50-56, 34:28-38, 35:62-63-36:1-2, 36:18-25.
Meng describes a "rote-based" learning system that prompts the user for information, and the information is then processed based on the user's response, rather than any prior use of the natural language processing method. Dkt. No. 142-9 at 50. In their IPR response, plaintiffs noted that Meng "only stores information consisting of the user's clarifying input, as opposed to any self-learning or storage of past cases (or use of said case information)." Dkt. No. 143-10 at 49-50. Thus, according to plaintiffs, while Meng stores clarifying information to "update statistics for the next time," the '798 Patent stores information in the form of a case: a problem definition, a solution, and optionally an outcome. Id. at 49; see also Dkt. No. 144 at 17-20. Thus, while case information can include clarifications from users, that is not the full permissible extent of case information.
This is consistent with plaintiffs' expert, Dr. Shwartz, who has uniformly taken the position that user dialogues and the result of the user dialogues can be stored as part of a case. Dkt. No. 147-2 at 120:6-10; Dkt. No. 142-15 at 18-19. Indeed, Dr. Shwartz specifically noted that "information from a user clarification can become part of the case information when stored by the NLP system with the intention of making it available to the natural language query process to help it resolve meaningful term ambiguities in future queries. Not just any information from a user clarification, however, is case information." Dkt. No. 144 at 17. Moreover, the PTAB rejected Amazon's argument that "case information" excludes clarifications from the user. See Dkt. No. 143-5 at 5-6.
Accordingly, for the foregoing reasons, I adopt the claim constructions of these two terms rendered in the Apple litigation, and recommend the term "case information" to be construed as "information about prior instances of use of the natural language method," and that "case" be similarly construed, to mean "a prior use of the natural language processing method."
5. Provide determination of a result
Plaintiff's Proposed Construction
Amazon's Proposed Construction
Court's Construction
No construction necessary, or alternatively given plain and ordinary meaning
Output, to a user, a result
No construction necessary
The last term to be construed appears in Claim 9 of the '798 Patent and was not addressed in the Apple litigation. Plaintiffs urge the court to decline to provide a construction for this term because one is not necessary. Amazon responds by adding a qualifier to the term.
Claim 1 of the '798 Patent calls for a "result of the search [to be provided] to the user." '798 Patent, 36:49-50. On its face, Claim 9 does not similarly specify to whom the resulting determination is provided. Id. at 37:28-29. To add, as Amazon urges, a provision indicating that the result achieved through the process described Claim 9 must be provided to the user would violate the principle of claim differentiation. See Curtiss-Wright Flow Control Corp. v. Velan, Inc., 438 F.3d 1374, 1380-81 (Fed. Cir. 2006) (noting the "presumption that each claim in a patent has a different scope"). Because both Claim 1 and Claim 9 are independent claims, it would be improper to import into Claim 9, as Amazon now urges, the additional restriction concerning to whom the result is to be provided.
Amazon's proposed construction also lacks support from the '798 Patent specification. Addressing Figure 2, the specification notes that the processor depicted 202 "receives a natural language query and a plurality of database objects 204, and produces a query result." '798 Patent, 12:53-55. The specification goes on to provide that "queries may be executed by an SQL search engine, and processor 201 may provide query result 203 to user through, for example, a graphical user interface." Id. at 13:58-60 (emphasis added). Amazon's construction would therefore eliminate a preferred embodiment and require that the result of the search must always be provided to the user. A construction that excludes a preferred embodiment "is rarely, if ever, correct." Hill-Rom Servs. v. Striker Corp., 755 F.3d 1367, 1379 (Fed. Cir. 2014).
In support of its proposed construction, Amazon also relies upon the prosecution history. Claim 9, which was originally Claim 12 but renumbered when the '798 Patent was issued, see Dkt. No. 142-5 at 11, originally required interpretation of permutations "to determine a result." Id. at 69. The patent examiner rejected this claim because it simply "manipulates an abstract idea" and did "not have any post or pre computer process activity." Id. at 25. In response to the rejection, the applicants amended the claim to require that the computer "provide a determination," rather than simply "determine a result." Id. at 16-17. The applicants then argued, based upon the amended language, that "an output is generated which is representative of an interpretation of the natural language input" and that such output is "hardly a mere abstraction" and is instead a "useful, concrete" result. Id. at 20-21. Based upon this exchange, Amazon's expert, Dr. Sacerdoti, contends that a POSITA "would recognize that the only 'useful, concrete' result that arguably occurs 'post computer' in the patented process is that the user receives the output that the computer has generated." Dkt. No. 142-14 at 48 (alteration omitted).
Amazon's reasoning is flawed. As the prosecution history establishes, the inventors argued that Claim 9 was allowable because it included pre-computer processing activity of receiving a natural language input, not post-processing activity of outputting the result to the user. Dkt. No. 142-5 at 20-21. By adding, following a rejection, the limitation "receiving a natural language input" and changing "determine a result" to "provide a determination of a result" the inventors argued that the output "is generated which is representative of an interpretation of the natural language input." Id. at 20. Accordingly, the applicants noted, "there is pre-computer process activity of receiving natural language input." Id. at 21.
Based upon the foregoing I recommend that the court reject the suggestion by Amazon that the result being provided in Claim 9 must always be to the user. I further recommend the court not construe this term as proposed by plaintiffs.
IV. SUMMARY AND ORDER
After carefully considering the '798 Patent as a whole, including its specification, the prosecution history associated with it, extrinsic materials submitted by the parties, and the arguments of counsel, it is hereby
RECOMMENDED that the court adopt the following claim term constructions:
A. Agreed Upon Constructions
Term
Agreed Upon Construction
natural language query input (Claims 1-8)
input that is articulated in a human language in a way that a native speaker could understand and use sensibly
natural language query input (Claim 20)
natural language input
natural language input (Claims 1-21)
input that is articulated in a human language in a way that a native speaker could understand and use sensibly
language-based databases (Claims 1-21)
databases that include and store natural language information
database objects (Claims 1-21)
objects included within a database, including but not limited to data types, data instances, database computational operators, data semantics, data structures, data instances or data values, and data operators
language-based database objects (Claims 9-21)
database objects in a language-based database
Keywords (Claims 1-21)
natural words and phrases that users use to refer to database objects and information model elements in natural articulation
information models (Claims 1-21)
webs of concepts for enterprise databases
reference dictionary (Claims 4, 14-17)
set of database entries or objects and their relations
at least one metadata database comprising at least one of a group of information types comprising: case information; keywords; information models; and database values (Claims 1-8)
at least one metadata database comprising one or more groups of information types including at least one group containing all four of the information types: case information, keywords, information models, and database values
a metadata database comprising at least one of a group of information comprising case information, key words, information models, and database values (Claims 9-21)
a metadata database comprising one or more groups of information types including at least one group containing all four of the information types: case information, keywords, information models, and database values
B. Disputed Claim Term Constructions
based on the input, without augmentation (Claims 1-8)
based solely on the natural language input without augmentation
metadata database (Claims 1-21)
collection of stored data that has a searchable organizational scheme and includes information about structuring, using, or interpreting data
case information (Claims 1-21)
information about prior instances of use or the natural language processing method
case (Claims 14-17)
a prior use of the natural language processing method
providing, through a user interface, a result of the search to the user (Claims 1-8)
no construction necessary
determining a plurality of combinations of the finite number of database objects (Claims 1-8)
no construction necessary
resolving ambiguity between the keywords and the plurality of combinations (Claims 5, 6)
no construction necessary
identifying a finite number of permutations of the plurality of database objects (Claims 9-21)
no construction necessary
interpreting at least one of the permutations to provide determination of a result of the natural language input (Claims 9-21)
no construction necessary
provide determination of a result (Claims 9-21)
no construction necessary
result of the natural language input (Claims 9-21)
no construction necessary
NOTICE: Pursuant to 28 U.S.C. § 636(b)(1), the parties may lodge written objections to the foregoing report. Such objections must be filed with the Clerk of the court within FOURTEEN days of service of this report. FAILURE TO SO OBJECT TO THIS REPORT WILL PRECLUDE APPELLATE REVIEW. 28 U.S.C. § 636(b)(1); Fed.R.Civ.P. 6(a), 6(d), 72; Roldan v. Racette, 984 F.2d 85 (2d Cir. 1993).
It is hereby ORDERED that the Clerk of the court serve a copy of this report and recommendation upon the parties in accordance with this court's local rules.