Recommendation & Decision Making - Surveying Users for Preferences and Providing Recommendations (Dalton Sentry) - Issued Patent - PRIOR ART REQUEST

Question

This Patent has been invalidated by the US District Court in Southern District of New York.

AN OVERBROAD PATENT ON that describes a seemingly obvious process where by two parties input their preferences and a recommendation is given based on a best fit algorithm. Would greatly help alleviate the burden of defense on dozens of start-ups trying to help consumers with recommendation and matching algorithms.

QUESTION - Have you seen anything that was published before 12/23/1999 that would invalidate claim 1:

• Patent Number: US8069073
• Assignee: Dalton Sentry, LLC appears to have no business other than suing 20 operating companies such as TheLadders, Jobvite, FindTheBest, Monster, etc.
• Prior Art Date: Seeking prior Art predating Dec 23, 1999

Claim 1:

A computer-implemented method for facilitating evaluation, in connection with the procurement or delivery of products or services, in a context of at least one of (i) a financial transaction and (ii) operation of an enterprise, such context involving a first class of parties in a first role and a second class of counterparties in a second role, the method comprising:

1. in a first computer process, retrieving first preference data from a first digital storage medium, the first preference data including attribute levels derived from choices made by at least one of the parties in the first class;

2. in a second computer process, retrieving second preference data from a second digital storage medium, the second preference data including attribute levels derived from choices made by at least one of the counterparties in the second class;

3. in a third computer process, for a selected party, performing multilateral analyses of the selected party's preference data and the preference data of each of the counterparties, and computing a closeness-of-fit value based thereon;

4. and in a fourth computer process, using the computed closeness-of-fit values to derive and provide a list matching the selected party and at least one of the counterparties.

English overview (at a high level, please read original claims to understand the actual patent):

A method for evaluating parties in a transaction, comprising:

1. Retrieving preference data about individual(s) in a first group, derived in part from choices made by the individuals (e.g. "do you prefer drama or comedy?")

2. Retrieving preference data about individual(s) in a second group, derived in part from choices made by the individual (e.g. "do you prefer sushi or pasta")

3. For a selected individual, performing multi-lateral analysis (or conjoint analysis to determine closeness-of-fit between the subject individual and other individuals of either the first or the second party.

4. Using closeness-of-fit to create a list of individuals well-matched to the subject individual ("people like you are well-matched to the following list of people who also prefer comedy and sushi")

"Recommendation & Decision Making - Surveying Users for Preferences"

More background about the '073 patent and 20 lawsuits in which it appears

Answer 1

Sounds like the sort of algorithm that is used by computer dating agencies, which have been around since 1965: http://www.thecrimson.com/article/1965/11/3/operation-match-pif-you-stop-to/

Answer 2

A patent describing matching criteria and algorithms from preferences in a cable programming system is the following:

https://www.google.com/patents/US5673430

with a publication date of 9/30/97. The description in the patent goes into details regarding conflicting criteria resolution, value judgment with relative importance of criteria, determining subscriber satisfaction, and other factors and algorithmic techniques. The techniques of matching people with preferences against products with attributes using computers was not new.

Answer 3

A patent describing a sophisticated electronic marketplace using consumer and provider personal agents, decision agents, demand agents, using consumer personal preference data and a provider product information database is the following:

http://www.google.co.in/patents/WO1997026612A1?cl=en

with a publication date of 7/24/97. This patent was not cited in the continuation patents or original patent in question. This patent is particularly relevant since it is about matching preferences in both directions. Providers of products can use demand information to tailor quantity of products with specific attributes and can target advertising only to those interested in relevant products. Consumers can see only those products that are relevant to them. It is a true matching of two parties with their respective preferences in a business transactional context.

Answer 4

Regarding multilateral analysis or conjoint analysis, there is the following patent application on determining forced choice preferences using hedonic testing (evaluation of individual products on a ballot):

https://www.google.com/patents/WO1999012030A1?cl=en

with a publication date of 3/11/99 though became a U.S. patent:

https://www.google.com/patents/US6623040?cl=en

that was published 9/23/03 so the patent application must be used for the purposes of prior art. This application shows the essential rough equivalency of forced choice (conjoint analysis) techniques compared to attribute scaling. It is not directly applicable as prior art and instead may be used to weaken the novelty of forced choice (conjoint analysis) relative to more standard attribute scaling.

Answer 5

A patent application that describes discrete choice modeling analogous to conjoint analysis where it is used for individually targeting different consumers to receive different offers from a plurality of different vendors:

https://www.google.com/patents/WO1999033012A1?cl=en

with a publication date of 7/1/99. This patent application was not cited in the original or continuation patents in question. In this patent application, financial and demographic information comes from a consumer credit provider while at least one of market research, product and customer information comes from different vendors.

"Single or multiple variate regression analysis is a powerful statistical technique which enjoys low predictive error and high discrimination. It identifies the combination of characteristics that best predicts specific consumer behavior. The result of the multiple regression analysis is a regression equation, which is a tool used to store and rank customers or prospects.

The LOGIT regression, also known as the Rodbard of Probit regression is a type of discrete choice analysis that market researchers use to predict how well a product or service will be received in the market. Analogous in many ways to conjoint analysis, discrete choice analysis differs by having respondents choose one of several product packages or options presented. Discrete choice modeling can be used to answer various marketing questions. It can provide direct predictive estimates of market share for a new or existing product. It can also be used to make estimates of future market demand. Many different data collection techniques can be used to implement a discrete choice model. Examples include mail and telephone surveys."

Answer 6

A patent application for getting user preferences and requirement data for a variety of attributes with a decision engine to filter product choices is the following:

https://www.google.com/patents/WO1998035297A1?cl=en

with an 8/13/98 publication date. This was subsequently issued as the following U.S. patent:

http://www.google.com/patents/US6012051

but that has a 1/4/00 publication date so the application needs to be used for prior art purposes. The patent application was not cited in the original or continuation patent in question. Part of the summary of the invention follows:

"Thus, the invention includes a user interface which presents a sequence of input prompts to the user to gather preference and requirement data for a plurality of attributes of products in the product domain. A decision engine is coupled to the user interface that filters the product domain to present a set of products according to the gathered preference and requirement data as product choices to the user."

"The preference data comprises a variable associated with particular attributes specified by the user to have a degree of relevance to a product choice in the product domain, but not an absolute requirement. The requirement data comprises a variable associated with a particular attribute specified by the user to be required or not required for a product choice in the product domain."

"According to the present invention, the user is guided through the sequence of input prompts in order to assist the user in creating a personal profile of preferences and requirements for products in a particular product domain. The decision engine applies the requirements and the preferences to generate a list of one or more products which meet the requirements specified by the user profile, ranked according to the preferences specified by the user profile to assist the user in making a product choice."

Answer 7

Another early patent on date matching is the following:

https://www.google.com/patents/US5920845

with a 7/6/99 publication date. This patent is about the most basic of user preferences -- a binary yes/no (i.e. would like to date / would not like to date) and is done with multi-lateral pairing the prospective dates do themselves. It eliminates the one-way selection preference so is the foundation for two-way matching but where the compatibility/preference details are done outside the computer system -- that is, the computer acts as a tally identifying two-way matches.

The above patent refers to the following compatibility game patent:

https://www.google.com/patents/US5681046

with a 10/28/97 publication date that in the description goes into detail about questions, responses, categories, valuation/ranking, etc.

Answer 8

Turning to financial markets, we find the following patent application that matches a party and a counterparty based on their respective credit preference data so that they only view relevant orders which can also be matched electronically:

http://www.google.co.in/patents/WO1999019821A1?cl=en

that was published on 4/22/99 but later became the U.S. patent:

http://www.google.com/patents/US6421653

but with a publication date of 6/16/02 so the patent application is what needs to be used for prior art. The following is a quote regarding the multi-party preferences:

"The market information provided to the user is coded with credit preference data generated by referencing the complex credit preferences inputted by each user regarding all possible counterparties. Thus, potential counterparties are then able to identify which orders they are eligible to trade based upon the coded credit preference data."

There is also the following patent that matches orders but also accounts for the credit limits of both parties:

http://www.google.com/patents/US5375055

with a publication date of 12/20/94. It adds a security aspect to this credit data, but again it is added by all parties and used to determine appropriate matches (i.e. ability to trade):

"More particularly, in the preferred embodiment, the sensitive credit limit data indicating how much credit a particular client site is willing to extend to each possible counterparty is maintained only at an access node associated only with that particular client, and only a simple yes/no indication of whether the entity (for example, a trader, a trading floor, or a bank) associated with that particular access node is willing to transact business with a particular counterparty is transmitted to the other nodes of the communication network."

There is the following patent regarding multi-party risk management:

http://brevets-patents.ic.gc.ca/opic-cipo/cpd/eng/patent/2203279/summary.html

with a publication date of 6/13/96. This patent is about the formulation of multi-party risk management contracts so is again about multiple parties entering information and that being matched appropriately.

"An ordering party (13) inputs, by a data processing device (51), contract data representing an offered contract in a predetermined phenomenon, the phenomenon having a range of possible outcomes at a time of maturity, and the contract data specifying the same entitlement for each outcome due to the ordering party at maturity and a consideration due to a counterparty. The potential counterparties (14) input, by data processing means (51), registering data relating to the range of possible outcomes for the predetermined phenomenon.

"An offered contract is priced by data processing apparatus (20) by the steps of calculating a counter consideration from each counterparty's registering data and comparing the ordering party consideration with the calculated counter considerations. A match is made on the basis of the comparison."

These demonstrate that systems that did best-matching for multiple parties already existed.

There are many other patents in 1999 and earlier regarding matching financial instruments with consumer preferences, but since those are not 2-way preference selections (unlike the above that are 2-way), I will not add those.

Answer 9

An early patent application for expertly matching products, services and consumers is the following:

http://www.google.com/patents/WO1998002835A1?cl=en

with a 1/22/98 publication date. This application does not appear to have resulted in a patent, but this application is referenced by other patents, but not by the continuation or original patent in question. The following are some excerpts illustrating preference selection by both individuals and those responsible for the products or services and indicates that information may be gathered by multiple choice questions:

"The means for creating a data profile 41 creates a data profile, based on the characteristics and preferences of the individual, that defines the individual and is used to match the individual. It may comprise, for example, a list of multiple choice questions, answers to the questions, an input/output device, and a means to manipulate data. The list of multiple choice questions elicits the characteristics and preferences of the individual so as to enable the creation of the data profile."

"The means for coding codes each product or service with the types of characteristics and preferences for which the product or service is appropriate."

Answer 10

A patent about automating multi-party property equity exchange using the specifics of each owner's exchange desires and each property's characteristics:

https://www.google.com/patents/US5500793

with a publication date of 3/19/96. This patent was not cited in the original or continuation patents in question. This is an earlier patent demonstrating multi-party preference selection and matching.

Answer 11

An early patent on such computer dating matching with preferences on both sides (see Ben's answer) is the following:

https://www.google.com/patents/US5963951

with a 10/5/99 publication date and is not cited in the continuation nor the original patent. The preferences are entered through direct selection and not forced choice, but they are done by multiple parties and matches are made through multiple bilateral analysis in that multiple party-to-party comparisons are made and a closeness-of-fit (called "percentage match") is calculated. I quote from the patent below:

"In the embodiment of this example, the score is determined as follows: if a criterion matches a preference then the score is increased by the weight associated with that criterion. So, e.g., if the user's hair color criterion matches the subscriber's hair color then the score is increased by the weight for that criterion, in this case ten. The same test is applied to the user's information and the subscriber's criterion.

When all categories have been compared, the resulting score is divided by the total possible score to determine a percentage match. Note that the total possible score may depend on some of the user's criteria."

Since the patent examiner for the 8069073 continuation patent already cited from the original patent:

http://www.google.com/patents/US6915269

the following two patents:

http://www.google.com/patents/US6272467
http://www.google.com/patents/US6289340

one should look at the patent prosecution history to understand the distinctions that made this patent non-obvious to the examiner, though the cited references had publication dates after the priority date and filing date for the original patent so are not considered to be prior art.

Answer 12

An early patent matching a product with a user based on characteristics of each is the following:

https://www.google.com/patents/US5122952

with a 6/16/92 publication date. Claim 1 is the following:

1. An automated selection system for selecting and matching learning tools that possess developmental values with the individual characteristics of a user that have correspondence to such values comprising:
   
   
   • data entry means for entering user data including user static personal data and dynamic user developmental conditions;
   • learning tool list means for storing learning tool characterizing data;
   • means coupled to the data entry means and the learning tool list means for generating a preferred learning tool list by matching the user data to corresponding learning tool characterizing data.

This patent was not cited in either the continuation patent nor the original patent. This patent may not be considered prior art for rejection under anticipation unless it teaches every aspect of the claimed invention, but it could still be considered as prior art for rejection under obviousness, especially when combined with the following patent that was already cited in the original patent in question:

https://www.google.com/patents/US5124911

This patent demonstrates the basic concepts of entering attribute information about objects (learning tools) and characteristics about the individuals who would use those objects and a matching system. A questionnaire is used. The matching algorithm is very basic and binary (yes/no fit for each attribute). One can show a timeline from this patent to later ones in terms of increasing levels of sophistication and narrowing, but where at some point pulling multiple pieces together becomes obvious to those skilled in the art (and the patent in question becomes broader in some ways, which it should not do).

Answer 13

A patent comparing the pros/cons of rank order vs. forced choice that proposes somewhat of a hybrid approach is the following:

https://www.google.com/patents/US4908758

with a publication date of 3/13/90. This shows that a variety of methods were available for determining relative user preferences nearly a decade before the patent in question.

The primary attack for invalidity against the patent is one of obviousness. There were already narrower patents and applications available in the prior art describing the following:

1. storing characteristics or attributes of a product or service, storing user preferences, and matching (nearly all examples given)
2. multiple parties storing their attributes and preferences and having bilateral matching (financial markets, property equity exchange, computer dating, etc.)
3. closeness-of-fit matching algorithms (percentage matching, weighted matches, regression analysis)
4. use of multiple choice questions and discrete choice to elicit the characteristics and preferences of the individuals in the context of matching parties for delivery of advertisements, products or services
5. use of rank order, forced choice, and hybrid techniques to create weighted or ordered preferences in the context of behavioral or social sciences.

Is it truly non-obvious and novel and appropriate under patent law to know about the narrow inventions in #2, #3, #4 and #5 above and allow a broader invention that is effectively not narrowed by type of party or market ("procurement or delivery of products or services" in a context of "financial transaction" or "operation of an enterprise"), is a generic superset of well-known closeness-of-fit matching algorithms (comparing aggregate or measuring distance using generic utility functions applies to almost anything), and is only narrowed by the use of forced choice (assuming the claim language of "choice" is narrowed to "forced choice" by the description)? The patent examiner did not cite the patents I listed so this is not a question of fault given the data used in the patent application evaluation, but rather a re-evaluation in light of this prior art showing the state of knowledge by those skilled in the art at that time.