`_______________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`_______________________
`
`Ocean Tomo, LLC,
`Petitioner
`
`v.
`
`Patent Ratings, LLC,
`Patent Owner
`______________________
`
`Patent No. 9,075,849
`Filing Date: July 22, 2014
`Issue Date: July 7, 2015
`Title: METHOD AND SYSTEM FOR PROBABILISTICALLY QUANTIFYING
`AND VISUALIZING RELEVANCE BETWEEN TWO OR MORE
`CITATIONALLY OR CONTEXTUALLY RELATED DATA OBJECTS
`_______________________
`
`Case CBM: Unassigned
`_______________________
`
`Declaration of Patrick Thomas, Ph.D. Under 37 C.F.R. § 1.132
`
`
`CHICAGO/#2726795.1
`
`1
`
`
`
`Ocean Tomo Ex. 1003-001
`
`
`
`I.
`
`INTRODUCTION
`
`
`
`1.
`
`This Declaration provides an analysis of US Patent #9,075,849 (referred to
`
`hereafter as the ‘849 patent). This patent was issued on July 7, 2015, and is assigned to
`
`PatentRatings, LLC.
`
`2.
`
`The main objective of this Declaration is to address the question of whether the
`
`‘849 patent is directed to abstract ideas under current patent law. Specifically, it provides an
`
`opinion as to whether the ‘849 patent qualifies as eligible or ineligible following the recent Alice
`
`and Bilski cases decided by the US Supreme Court.
`
`II.
`
`PROFESSIONAL BACKGROUND
`
`3.
`
`Work Experience - I am a science and technology analyst, and my main expertise
`
`is in data mining and intellectual property analytics. I have worked with patent metrics, citation
`
`data and statistical models for over two decades.
`
`4.
`
`I am currently a partner in 1790 Analytics LLC, which I co-founded in 2004.
`
`1790 is a consulting firm focused on developing quantitative intellectual property metrics, and
`
`employing these metrics to answer a broad range of questions. I have consulted with many large
`
`corporations and financial institutions, helping them to identify and capitalize upon technological
`
`and investment opportunities. I have also examined a broad range of science and technology
`
`policy issues for various government agencies. My government clients include the US
`
`Department of Defense (DOD); US Department of Energy (DOE); the Intelligence Advanced
`
`Research Projects Activity (IARPA); National Institute of Standards & Technology (NIST); and
`
`the Small Business Administration (SBA). My commercial and investment clients are not
`
`revealed here for confidentiality reasons.
`
`
`CHICAGO/#2726795.1
`
`2
`
`
`
`Ocean Tomo Ex. 1003-002
`
`
`
`5.
`
`Before my work at 1790 Analytics, I was a Senior Analyst at CHI Research Inc.,
`
`one of the pioneering companies in intellectual property metrics. While at CHI, I consulted with
`
`large corporations and government agencies, and developed subscription products aimed at the
`
`investment community. This work was again based extensively on quantitative patent metrics. I
`
`was employed at CHI from 1998 until 2004, which is when I co-founded 1790. Prior to CHI, I
`
`was an Assistant Professor in Quantitative Methods at Southampton Business School (UK).
`
`6.
`
`Education - I was educated in the United Kingdom, earning a B.S. (First Class) in
`
`Management Science from the University of Manchester in 1991; an M.S. in Computer Science
`
`from the University of Birmingham in 1993; and a Ph.D. in Management Science and Statistics
`
`from Nottingham Trent University in 1998.
`
`7.
`
`My Ph.D. thesis made extensive use of literature citation data, and was my first
`
`exposure to this type of information. Specifically, I examined the scholarly literature associated
`
`with management science theories, and designed models that forecast which theories would have
`
`lasting impact, and which theories would become fashions with only fleeting influence. These
`
`models made extensive use of multivariate statistical analysis, in order to identify characteristics
`
`that would differentiate between long-lasting ideas and fads. My thesis was published in 1999 as
`
`a book entitled “Fashions in Management Research: An Empirical Analysis.”
`
`8.
`
`Publication History - I have published numerous articles in peer-reviewed
`
`journals, covering various subjects including technology assessment, science policy, company
`
`valuation, and investment analysis. While these articles cover a variety of topics, they share the
`
`same analytical core, namely the use of quantitative metrics and statistical models in assessing
`
`published literature and patents.
`
`
`CHICAGO/#2726795.1
`
`3
`
`
`
`Ocean Tomo Ex. 1003-003
`
`
`
`9.
`
`The first of these papers, published when I was still a Ph.D. student, proposed a
`
`quantitative citation-based method for determining the quality of academic journals. After
`
`joining CHI, and since co-founding 1790, my research has focused mainly on metrics related to
`
`patents, rather than literature. I have published a series of papers demonstrating the application of
`
`quantitative patent metrics in numerous contexts. These include forecasting patent renewal
`
`decisions; valuing merger and acquisition candidates; identifying undervalued stocks for
`
`investment purposes; tracing the historical development of technologies; and locating emerging
`
`technologies early in their lifetime. All of these papers have a strong quantitative base, and a
`
`number of them employ multivariate statistical analysis, a subject which is examined in more
`
`detail in this Declaration.
`
`10.
`
`Patenting History - In addition to publishing articles in peer-reviewed journals, I
`
`also invented a US patent (US #7,832,211) based on my research on company valuation. This
`
`patent describes an algorithm that places a valuation on a company based on the strength of its
`
`patent portfolio, as measured via a variety of quantitative patent metrics. The patent-based
`
`valuation, which is derived via multiple regression analysis, is then compared to the current
`
`valuation of the company in the stock market. If the patent-based valuation is higher than the
`
`current valuation, then the company is rated as undervalued, and is thus a target for possible
`
`investment.
`
`III. OVERVIEW OF TOOLS AND TECHNIQUES RELEVANT TO THE ‘849
`PATENT
`
`11.
`
`The ‘849 patent describes a system and method for identifying documents
`
`(especially patent documents) that are closely related to each other, based on their connections in
`
`a citation network. This citation network consists of documents (the nodes in the network)
`
`connected via citations that form the links between these nodes. These citations may be, for
`
`
`CHICAGO/#2726795.1
`
`4
`
`
`
`Ocean Tomo Ex. 1003-004
`
`
`
`example, prior art references in the case of patents, or reference lists in the case of scientific
`
`papers.
`
`12.
`
`The ‘849 patent is directed to two basic tools of research – bibliometrics and
`
`statistical analysis. It is thus instructive to examine the development of these two research areas,
`
`in order to provide context for the subsequent analysis of the ‘849 patent.
`
`IV.
`
`BIBLIOMETRIC TOOLS
`
`13.
`
`Bibliometrics is a basic tool of research used in the social sciences. It can be
`
`defined as the process of extracting measurable data through the statistical analysis of document
`
`contents, plus information about how the texts are being accessed and used by subsequent
`
`researchers. The use of bibliometric tools has a long history, and there are numerous journals that
`
`publish bibliometric research extensively, including Scientometrics, Journal of Information
`
`Science, and Journal of the Association for Information Science and Technology.
`
`14.
`
`Citation analysis is one of the key constituent parts of bibliometrics. The usage of
`
`citation analysis on a significant scale can be traced back to the 1950s, although a few small
`
`studies predate this. In 1955, Eugene Garfield published an article in Science, outlining the basic
`
`concept of a citation index for scientific documents (Garfield, 1955). Garfield subsequently set
`
`up a company named the Institute of Scientific Information (ISI), which published the first
`
`Science Citation Index (SCI) in 1963. ISI is now part of ThomsonReuters, and the SCI is a major
`
`component of the widely-used Web of Science.
`
`15.
`
`In his original paper, Garfield acknowledged that the idea of a science citation
`
`index was inspired in part by the well-established Shepard’s Citations in legal research.
`
`Shepard’s, which dates back to the 19th century, provides a list of all the authorities citing a
`
`particular legal case, statute, or authority, and can thus be used to trace their judicial history and
`
`find other relevant cases directed to a legal issue.
`
`
`CHICAGO/#2726795.1
`
`5
`
`
`
`Ocean Tomo Ex. 1003-005
`
`
`
`16.
`
`Just as the idea of using citation analysis to assess science was inspired by the
`
`legal field, the idea of patent citation metrics was in turn inspired by the increasing use of such
`
`metrics to evaluate scientific literature. Computerized citation data for patents were first made
`
`available in 1975. This enabled companies such as Computer Horizons Inc. (subsequently CHI
`
`Research) to develop patent citation databases. These databases contained citation links between
`
`generations of patents, just as links had been made previously between generations of scientific
`
`papers. There are now numerous companies with patent citation databases, including my own
`
`company 1790 Analytics.
`
`17.
`
`In his Science paper, Garfield outlined two basic uses of scientific citations that
`
`have inspired decades of subsequent bibliometric research using both patent documents and
`
`scientific articles.
`
`18.
`
`Using citations to measure relatedness - the first use of citations is to connect
`
`scientific articles that may be related in some way, even if they are not in the same subject field.
`
`To quote Garfield:
`
`“… this paper considers the possible utility of a citation index that offers a
`19.
`new approach to subject control of the literature of science. By virtue of its
`different construction, it tends to bring together material that would never be
`collated by using the usual subject indexing. It is best described as an
`association-of-ideas index.” (Garfield, 1955, p.108)
`
`20.
`
`Citation indexing is thus a tool to help overcome a major problem facing
`
`researchers, namely that restricting their reading to previous research from their own field may
`
`be insufficient, since there may be useful and relevant documents that are classified elsewhere.
`
`Analogously, in patents, there may be relevant prior art that lies beyond the immediate
`
`technology area, and may not be located via a classification or keyword search. Citations can
`
`thus be used as a complement to standard subject and word searches. As noted in a 2010 essay
`
`on the history of citation indexing:
`
`
`CHICAGO/#2726795.1
`
`6
`
`
`
`Ocean Tomo Ex. 1003-006
`
`
`
`Garfield's achievement lay in establishing the utility and objectivity of a
`21.
`citation index in pulling up related papers in published literature that at first
`glance might not have seemed pertinent to the researcher's inquiry.
`
`Thomson Reuters http://wokinfo.com/essays/history-of-citation-indexing/
`
`22.
`
`The use of citations in this way is thus based on the idea that two documents that
`
`are linked via a citation are likely to be related in some way. These are often known as direct
`
`citation links. Beyond these direct citation links, there has also been extensive research into the
`
`use of indirect citation links. Such links occur where two documents are linked via citations
`
`through a third, intermediate document.
`
`23.
`
`To demonstrate these indirect citation links, consider a simple universe consisting
`
`of three documents, A, B and C. In total, there are four ways in which A and B can be connected
`
`via C:
`
`A cites C, C cites B
`
`B cites C, C cites A
`
`A cites C, B cites C
`
`C cites A, C cites B
`
`24.
`
`The first two of these possibilities are examples of indirect sequential citations.
`
`Such citations are widely used in longitudinal evaluation studies, particularly to trace the impact
`
`of public funding on technological developments. In my own work, I have used indirect
`
`sequential citations in a series of reports for the Department of Energy (DOE). These reports
`
`examine the contribution of DOE funded research to subsequent developments in renewable
`
`energy technology – see for example Ruegg and Thomas (2009).
`
`25.
`
`The third possibility occurs where both A and B reference the same earlier
`
`document (C). This is known as bibliographic coupling. It is based on the idea that, if two pieces
`
`of research reference the same earlier document, they may share the same technical or theoretical
`
`
`CHICAGO/#2726795.1
`
`7
`
`
`
`Ocean Tomo Ex. 1003-007
`
`
`
`foundation to some degree. Bibliographic coupling is often extended to examine the number of
`
`references two documents have in common. The more references two documents have in
`
`common, the greater their assumed similarity. Bibliographic coupling has been in use for more
`
`than 50 years, having been introduced in 1963 by M. M. Kessler. It is thus one of the oldest
`
`forms of citation analysis for linking scientific and technical documents.
`
`26.
`
`The fourth possibility occurs where one document (C) references both A and B.
`
`This is known as co-citation. It is based on the idea that, if two documents are referenced in the
`
`same later document, they may cover similar technical or theoretical information. As in the case
`
`of bibliographic coupling, co-citation is often extended to examine the number of documents that
`
`cite two particular documents. The more documents that cite both of the selected documents, the
`
`greater assumed similarity between these two documents.
`
`27.
`
`Co-citation was introduced by Henry Small of ISI in 1973. It has become very
`
`widely used in evaluation studies, and has generally replaced bibliographic coupling as the main
`
`method for calculating the similarity of documents based on overlaps in their citing or cited
`
`documents. The use of co-citation clustering is particularly widespread. This is a method for
`
`clustering documents based on the extent of their co-citation.
`
`28.
`
`These various indirect citation links, plus the direct citation links referred to
`
`above, form the basic building blocks of citation networks, and are well established in the
`
`research community. From these building blocks, it is a simple programming task to construct
`
`citation networks covering any number of generations, and including any number of documents.
`
`Ultimately, almost all documents in the network are likely to be connected (similar to the ‘six
`
`degrees of separation’ that supposedly link all individuals across the world).
`
`
`CHICAGO/#2726795.1
`
`8
`
`
`
`Ocean Tomo Ex. 1003-008
`
`
`
`29.
`
`Indeed, the comparison to individuals is instructive, since citation networks are
`
`really just one instance of a relationship network. Such networks could be constructed between
`
`individuals based on personal contacts, organizations based on business relationships, keywords
`
`based on co-occurrence, and so on. The basic underlying components – in terms of nodes
`
`consisting of entities, and links consisting of relationships between these nodes – are similar
`
`across all such networks. The types of tools used to analyze such networks are also largely
`
`standard, especially in terms of methods for connecting entities based on the similarity of their
`
`links in the network.
`
`30.
`
`One practical example of a citation network among documents is the ‘neighbor
`
`searching’ algorithm developed by CHI Research in the 1980s. The neighbor searching
`
`algorithm works by counting the number of direct and indirect citation links between documents.
`
`These links are then weighted based on generation, with direct (first-generation) links weighted
`
`higher than second-generation indirect links, which are weighted higher than third-generation
`
`indirect links, and so on. The neighbor searching algorithm is described in US Patent #7,433,884,
`
`which was filed in 2004 and granted in 2008. It has since expired due to failure to pay the
`
`required maintenance fee. A version of neighbor searching is currently used by my company,
`
`under the alternative name ‘N-Degree’.
`
`31.
`
`Using citations to measure impact - The second application of citation indexes
`
`suggested by Garfield is reflected in this statement from his paper in Science:
`
` “In effect, the system would provide a complete listing, for the
`32.
`publications covered, of all the original articles that had referred to the article in
`question. This would clearly be particularly useful in historical research, when
`one is trying to evaluate the significance of a particular work and its impact on
`the literature and thinking of the period. Such an “impact factor” may be much
`more indicative than an absolute count of the number of a scientist's
`publications” (Garfield, 1955, p.109)
`
`
`CHICAGO/#2726795.1
`
`9
`
`
`
`Ocean Tomo Ex. 1003-009
`
`
`
`33.
`
`In other words, citations are suggested as a proxy for the impact of a document
`
`upon subsequent research. That is, if Document B references Document A (as prior art in the
`
`case of patents, or as past research in the case of papers), then the content of Document A is seen
`
`as having influenced the research described in Document B in some way. This idea can then be
`
`extended to measure the impact of a particular document, based on how many citations it
`
`receives from subsequent documents. For example, if Document A is cited by not only
`
`Document B, but also 100 other documents, then Document A is regarded as more influential
`
`than if it is cited by Document B alone.
`
`34. When evaluating patent documents, the basic idea behind patent citation analysis
`
`is that highly cited patents (i.e. patents cited as prior art by many later patents) tend to contain
`
`technological information of particular importance. As such, they form the basis for many new
`
`innovations, and so are cited frequently by later patents. This does not mean that every highly
`
`cited patent is important, or that patents cited infrequently are necessarily trivial. However,
`
`numerous validation studies have revealed the existence of a strong positive relationship between
`
`patent citations and measures of technological importance and commercial value, although there
`
`have been dissenting voices (Wang, 2007). Useful overviews of such validation studies can be
`
`found in Breitzman and Mogee (2002), Sampat and Ziedonis (2004), and Hsieh (2011).
`
`35.
`
`There are thus two well-established abstract concepts that underpin most citation
`
`analysis, irrespective of whether the focus of the analysis is on literature or patents. The first is
`
`that two documents connected by a citation link (whether direct or indirect) are related in some
`
`way. The second is that the citing document in the linked pair has been influenced to some
`
`degree by the cited document. From these basic concepts, one can apply a wide array of
`
`bibliometric and statistical analyses. For example, the ‘849 patent describes an application of the
`
`
`CHICAGO/#2726795.1
`
`10
`
`
`
`Ocean Tomo Ex. 1003-010
`
`
`
`first abstract concept – i.e. that of document relatedness (as discussed in more detail later in this
`
`Declaration).
`
`36.
`
`Conclusions regarding bibliometric tools - the discussion above provides a brief
`
`overview of some of the techniques that form the key foundations of bibliometrics, especially
`
`citation analysis. These techniques are in widespread usage within the field. Indeed, in 1987,
`
`Dorothy Hertzel wrote in the Encyclopedia of Library and Information Science that:
`
`Since the 1973 paper of Small, there does not seem to have been any
`37.
`basically new idea presented; the many publications since then are variations,
`applications and/or extensions of the original hypotheses, laws, or techniques.
`(Hertzel, 1987, p. 168)
`
`38. While this quotation is now over 25 years old, it still largely holds true today.
`
`Techniques for analyzing citation networks have advanced, as improved computing power has
`
`allowed for the development of more complex algorithms. However, these algorithms (including
`
`our own at 1790 Analytics) are, in the main, simply different ways of counting links within
`
`citation networks. In many ways, this is inevitable, since there are only so many ways one can
`
`analyze a network consisting of nodes and links, whether this is a network of documents, people,
`
`organizations etc.
`
`V.
`
`STATISTICAL TOOLS
`
`39.
`
`The ‘849 patent makes extensive use of fundamental multivariate statistical tools
`
`widely used in the social sciences. A brief overview of such tools is thus provided in this section
`
`of the Declaration.
`
`40.
`
`In simple terms, multivariate statistical tools are mathematical formulas used to
`
`model relationships between multiple independent (‘predictor’) variables and on one or more
`
`dependent (‘outcome’) variable. Perhaps the most well-known such tool is multiple regression,
`
`which measures the relationship between multiple independent variables and a single dependent
`
`
`CHICAGO/#2726795.1
`
`11
`
`
`
`Ocean Tomo Ex. 1003-011
`
`
`
`variable. For example, multiple regression could be used to examine the relationship between
`
`independent variables such as age, gender, qualifications, geographical location, profession etc.
`
`and the dependent variable annual income. This would provide insights into which of these
`
`independent variables are related particularly strongly with annual income. The parameters of the
`
`model may also be used to output a forecast for the dependent variable given certain independent
`
`variable values. In this example, the model would output an expected annual income for an
`
`individual with particular characteristics in terms of age, gender, qualifications, location,
`
`profession etc. This output is a numerical scalar value (i.e. a dollar amount).
`
`41. Multiple regression relies on a number of assumptions. In the context of the
`
`current discussion, perhaps the most important of these assumptions is that the dependent
`
`variable should be scalar (i.e. it is a numerical value that can be measured on a scale, such as
`
`annual income in the example above). In the methods described in the ‘849 patent, this
`
`assumption is often violated, since the dependent variable is typically binary (i.e. whether two
`
`patents will be linked directly via a citation). Hence, standard multiple regression is not
`
`applicable.
`
`42.
`
`As noted in its specification, the ‘849 patent instead employs widely-known
`
`specialized types of regression analysis, known as probit regression (or probit model) and
`
`logistic regression (or logit model). These two types of regression are specifically designed for
`
`circumstances such as this, where there are multiple independent variables, and a single binary
`
`variable (which is typically coded as 0 or 1). The purpose of the models is to estimate the
`
`probability that an entity with particular characteristics will fall into one of the two binary
`
`categories, rather than the other. Note that the logit model can also be applied to situations where
`
`there are more than two categories for the dependent variable, in which case it is known as
`
`
`CHICAGO/#2726795.1
`
`12
`
`
`
`Ocean Tomo Ex. 1003-012
`
`
`
`multinomial logistical regression (or ordered logistic regression if the categories are ordered).
`
`Meanwhile the probit model generally applies to situations where the dependent variable is
`
`binary.
`
`43.
`
`To continue the example used above, given the same set of independent variables
`
`(age, gender, qualifications, location, profession etc), one could use a logit or probit model to
`
`determine the likelihood of an individual voting Republican or Democrat in an upcoming
`
`presidential election (assuming no third-party candidates). If Republican is coded as 0, and
`
`Democrat as 1, the model would output a value somewhere between these two endpoints for each
`
`individual. This value represents the likelihood of an individual voting Republican or Democrat
`
`based on the values of the independent variables for that individual. If the value is greater than
`
`0.5, the individual is forecast to vote Democrat; if it is below 0.5, the individual is forecast to
`
`vote Republican.
`
`44.
`
`It is also possible to report the marginal effect of different independent variables.
`
`This marginal effect reflects the change in the probability of an outcome given a change of one
`
`unit in the selected independent variable. To continue the example above, it would be possible to
`
`determine the change in the probability of an individual voting Democrat or Republican resulting
`
`from an increase of one year in their age.
`
`45.
`
`Logit and probit regression generally produce similar results, and are used
`
`interchangeably. The difference between them lies in the assumption of the distribution of error
`
`terms. Logit regression assumes that the errors follow a logistic distribution, while probit
`
`regression assumes that the errors follow a normal distribution. These two distributions are
`
`similar, and form a bell-shaped curve (i.e. a small number of very high and very low values, and
`
`a large number of values towards the center; an example is height, with small numbers of
`
`
`CHICAGO/#2726795.1
`
`13
`
`
`
`Ocean Tomo Ex. 1003-013
`
`
`
`extremely tall and extremely short people, and most people being in the mid-range). The logistic
`
`distribution does have fatter tails (i.e. a larger number of extreme values), but only marginally.
`
`The similarity of the results means that the selection of logit or probit is generally an individual
`
`preference, rather than one test being notably superior to the other in most cases.
`
`46.
`
`Both logit and probit regression have a long history of use in statistical analysis
`
`(and standard multiple regression has an even longer history). Probit regression was introduced
`
`by Chester Bliss in 1934 (Bliss, 1934) to analyze data from bioassays. Logit regression was also
`
`first used to analyze bioassay data, by Joseph Berkson in 1944 (Berkson, 1944). While probit
`
`was initially more popular, logit regression gained acceptance and popularity due to its simpler
`
`calculation (especially at a time when many statistical tests were carried out with pencil and
`
`paper).
`
`47.
`
`Since this initial work, both logit and probit models have moved well beyond
`
`biomedicine, and are tools used widely across many disciplines. A useful history of probit and
`
`logit models can be found in Cramer (2002). Table 1 in this paper (reproduced below) is
`
`revealing as to the extent of usage of logit and probit models over time. Specifically, this table
`
`shows the number of articles in statistical journals that use the terms ‘probit’ or ‘logit’. As
`
`Cramer notes, these numbers are a fraction of the total number of articles that use logit and probit
`
`models, since they only count articles in twelve major statistics journals. These models are also
`
`widely used in economics, biomedicine, social sciences etc, and papers in journals from those
`
`disciplines are not included in the counts in the table. Indeed, a simple search in Google Scholar
`
`for scientific papers published since 2010 returned 31,500 articles that use the term ‘probit’ and
`
`48,500 articles that use the term ‘logit’. These are both relatively specific terms, so most of the
`
`articles are likely to refer to probit and logit models.
`
`
`CHICAGO/#2726795.1
`
`14
`
`
`
`Ocean Tomo Ex. 1003-014
`
`
`
`Logit
`Probit
` Time Period
`0
`6
`1935-39
`1
`3
`1940-44
`6
`22
`1945-49
`15
`50
`1950-54
`23
`53
`1955-59
`27
`41
`1960-64
`41
`43
`1965-69
`61
`48
`1970-74
`72
`45
`1975-79
`147
`93
`1980-84
`215
`98
`1985-89
`311
`127
`1990-94
`The article counts above reflect the breadth of usage of logit and probit models,
`
`48.
`
`which are the standard method for predicting the probability of a binary outcome based on a
`
`given set of independent variables. They can be found in statistics textbooks and in any standard
`
`statistical software package, such as R, SAS or SPSS.
`
`VI.
`
`LEGAL CONTEXT OF THE ‘849 PATENT
`
`49.
`
`The discussion above is designed to provide some background on the tools and
`
`techniques with particular relevance to the ‘849 patent. This background provides the broader
`
`technical context for the opinions expressed in this Declaration.
`
`50.
`
`This section of the Declaration focuses on the legal context related to the ‘849
`
`patent. The opinions expressed in this Declaration are based on my analysis of the ‘849 patent in
`
`the context of patent law related to software and business methods patents and the technology
`
`associated with bibliometric and statistical analysis. I am not an attorney, but I work extensively
`
`with patent data. I thus track closely any significant events related to patents, since these events
`
`may affect our current and future client engagements. Below is a brief overview of my
`
`
`CHICAGO/#2726795.1
`
`15
`
`
`
`Ocean Tomo Ex. 1003-015
`
`
`
`understanding of current law related to software and business methods patents. This
`
`understanding forms the basis for my opinions outlined in this Declaration.
`
`51.
`
`The basic rule for determining whether a new type of invention is patentable is
`
`that ‘anything under the sun that is made by man’ may qualify, following language provided by
`
`Congress. This does not include laws of nature and natural phenomena, which are discovered,
`
`rather than ‘made’. Based on this rationale, courts have consistently stated that laws of nature,
`
`natural phenomena and abstract ideas (including abstract mathematics) cannot be considered
`
`patentable subject matter.
`
`52.
`
`In recent decades, there have been a series of court decisions related to whether
`
`software and business methods should be patentable subject matter. In its 1998 State Street Bank
`
`vs. Signature Financial Group decision, the Federal Circuit stated that software should be
`
`patentable as long as it yields a ‘useful, concrete, and tangible result’. Following this decision, a
`
`large number of patents related to software, and particularly business methods software, were
`
`filed and granted. These include the parent application of the patent family of which the ‘849
`
`patent is a member.
`
`53.
`
`In its 2008 In re Bilski decision, the Federal Circuit rejected the earlier ‘useful,
`
`concrete and tangible result’ guideline for software patents set forth in State Street Bank,
`
`replacing it with a ‘machine or transformation’ test. This requires a process to be tied to a
`
`particular machine or apparatus, or physically transform an article into a different state or thing,
`
`in order to be patentable subject matter. In 2010, the Supreme Court partially overturned the
`
`Federal Circuit decision in Bilski vs. Kappos. It stated that the machine-or-transformation test is a
`
`‘useful and important clue’ as to whether particular subject matter should be patentable, rather
`
`than the sole test.
`
`
`CHICAGO/#2726795.1
`
`16
`
`
`
`Ocean Tomo Ex. 1003-016
`
`
`
`54.
`
`Following Bilski vs. Kappos, processes that fail the machine-or-transformation
`
`test are likely to be unpatentable for preempting an abstract idea, while processes that pass the
`
`test are likely to be patentable. The test effectively establishes a preliminary conclusion, which
`
`can be overcome with arguments to the contrary. In other words, the test establishes who has the
`
`burden of proof as to whether a process preempts an abstract idea or not.
`
`55.
`
`In 2014, in its Alice Corp. vs. CLS Bank International decision, the Supreme
`
`Court established a two-part test for determining whether a claim describes patentable subject
`
`matter. In the first part, courts must determine whether a claim is directed to a patent-ineligible
`
`concept, such as a natural law or abstract idea. The Court declined to define an ‘abstract idea’,
`
`but it did describe the concept of the patents at issue in the case (namely intermediated
`
`settlement, or escrow), as ‘a fundamental economic practice long prevalent in our system of
`
`commerce’ and a ‘building block of the modern economy’, thus making it an ‘abstract idea’ that
`
`is not patentable.
`
`56.
`
`If it is determined that a claim does indeed describe a patent-ineligible concept,
`
`the second part of the Alice test considers whether the claim contains an ‘inventive step’. This
`
`‘inventive step’ must be sufficient to ensure that, in practice, the patent amounts to more than a
`
`claim on the abstract idea itself. The Court also emphasized that ‘the mere recitation of a generic
`
`computer cannot transform a patent-ineligible abstract idea into a patent-eligible invention’.
`
`Therefore, processes that use a generic computer to implement an abstract idea are not
`
`pat