T 2431/18 (Target user determination/HUAWEI) of 16.12.2021

Summary of Facts and Submissions

I. The appeal lies from the decision of the examining division to refuse European patent application No. 14879270.8, which was filed as international application PCT/CN2014/095612 and published under Article 153(4) EPC as EP 3 001 332 A1, for lack of inventive step in the subject-matter of claim 1 of the main request and first and second auxiliary requests over a notoriously known network server computer.

The following documents were cited in the decision under appeal (in the following the board cites the printed version which was annexed to the board's communication):

D1: "Recommender system", Wikipedia, the free encyclopedia, 29 July 2014, pages 1 to 13, retrieved from the Internet: https://en.wikipedia.org/w/index.php?title=Recommender_system&oldid=618921071;

D2: "Server (computing)", Wikipedia, the free encyclopedia, 24 July 2014, pages 1 to 8, retrieved from the Internet: https://en.wikipedia.org/w/index.php?title=Server_(computing)&oldid=618342099.

II. In the statement of grounds of appeal, the appellant requested that the decision under appeal be set aside and that a patent be granted on the basis of the main request or of one of the two auxiliary requests considered in the decision under appeal and re-submitted with the statement of grounds of appeal.

III. The appellant was summoned to oral proceedings before the board. In a subsequent communication sent in advance of the oral proceedings, the board expressed its preliminary opinion that the subject-matter of claim 1 of each of the requests lacked inventive step over a well known recommendation system as illustrated by D1. The board was of the view that the underlying method for determining a target user for a target service, i.e. to whom a target service can be "successfully" promoted, was not technical. Documents D1 and D2 were annexed to the communication.

IV. With a letter of reply the appellant presented new arguments in favour of patentability of the requests.

V. After a further letter by the appellant announcing that it would not attend the oral proceedings, the board cancelled the oral proceedings.

VI. The final requests were therefore that the decision under appeal be set aside and that a patent be granted on the basis of the main request or of one of the two auxiliary requests considered in the decision under appeal and re-submitted with the statement of grounds of appeal.

VII. Claim 1 of the main request reads as follows:

"A method for determining a target user, characterized by comprising:

for any target service, acquiring (101), by a network server, user behavior data generated according to multiple behavior objects that belong to a same service type as the target service, wherein each piece of the user behavior data comprises a user identifier and a behavior object identifier;

determining (102), by the network server, according to the user identifiers and behavior object identifiers that are comprised in the acquired user behavior data, a correspondence between user identifiers of different users and behavior object identifiers of different behavior objects of the same service type, wherein the correspondence is used to represent an operating and operated relationship between a user corresponding to a user identifier and a behavior object corresponding to a behavior object identifier;

according to a behavior object comprised in the target service, assigning (103), by the network server, an initial value to each user identifier in the correspondence, and assigning an initial value to each behavior object identifier in the correspondence;

using (104), by the network server, the correspondence to construct a data model that is used for score transferring, wherein elements of the constructed data model comprise the user identifier and the behavior object identifier that are in the correspondence;

calculating (105), by the network server, based on the data model and the initial values and by using a value update rule, a value of an element comprised in the data model to obtain a value of a probability that a user corresponding to each user identifier becomes a target user corresponding to the target service, and selecting, according to the value of the probability, a target user of the target service."

VIII. Claim 1 of the first auxiliary request differs from claim 1 of the main request in that the following text has been added at the end:

"wherein the data model is a transfer matrix, and elements comprised in the transfer matrix comprise the user identifier and the behavior object identifier that are in the correspondence; and

the calculating, by the network server, based on the data model and the initial values and by using a value update rule, an iterative operation on a value of an element comprised in the data model to obtain, by means of calculation, a value of a probability that a user corresponding to each user identifier becomes a target user corresponding to the target service comprises:

performing, by the network server, according to the initial values and the value update rule, an iterative operation on a value of a matrix element comprised in the transfer matrix to obtain, by means of calculation, a convergence value of each user identifier, and using the convergence value as the value of the probability that the user corresponding to each user identifier becomes the target user corresponding to the target service."

IX. Claim 1 of the second auxiliary request differs from claim 1 of the first auxiliary request in that the following text has been added at the end of the claim:

"wherein the performing by the network server, according to the initial values and the value update rule, an iterative operation on a value of a matrix element comprised in the transfer matrix to obtain, by means of calculation, a convergence value of each user identifier, and using the convergence value as the value of the probability that the user corresponding to each user identifier becomes the target user corresponding to the target service comprises:

obtaining, by the network server, by means of calculation, a convergence value in the transfer matrix element comprised in the transfer matrix; and

determining, by the network server, a matrix element corresponding to each user identifier, and using a convergence value corresponding to a determined matrix element as a convergence value of a user identifier corresponding to the matrix element, wherein the convergence value is obtained by means of calculation; wherein

the convergence value in the transfer matrix element comprised in the transfer matrix is obtained by means of calculation in the following manner:

FORMULA/TABLE/GRAPHIC; wherein

R(n)m indicates convergence values of n elements in the transfer matrix that are obtained by means of the M**(th)iterative operation, R(n)m-1 indicates convergence values of n elements in the transfer matrix that are obtained by means of the (M-1)**(th)iterative operation, alpha is a diminution factor, T is a transfer matrix, R(n)0 comprises an initial value of each user identifier and an initial value of each behavior object identifier, n is a natural number and indicates that the transfer matrix comprises n elements, a value of n is a sum of a quantity of user identifiers and a quantity of behavior object identifiers, wherein the user identifiers and the behavior object identifiers are comprised in the acquired user behavior data, m is a natural number and indicates a quantity of times of performing an iterative operation, and a value of m is determined by whether R(n)m obtained by means of calculation is convergent."

X. The appellant's arguments, where relevant to this decision, are addressed in detail below.

Reasons for the Decision

Application

1. The application concerns the determination of a target user for a target service for the purposes of recommending a product to a user, for instance for determining a user who needs a smartphone designed exclusively for students and recommending the smartphone to the user (see paragraphs [0003] and [0007] of the amended translation of the description filed upon entry into the European phase). According to the description, the invention improves accuracy of and efficiency in determining a target user (paragraph [0041]).

1.1 The target user for a target service is determined by acquiring, at a network server, user historical behaviour data, e.g. a user purchase of product B from vendor A (paragraphs [0040] to [0052], Figure 1). Behaviour data is generated by multiple behaviour objects that belong to the same service type as the target service and is used for establishing correspondences between individual users and behaviour objects of the service type. Each established correspondence is modelled as an "element of the data model" or a "piece of user behavior data" including a user identifier and a behaviour object identifier (paragraphs [0040], [0041], [0045] and [0063] to [0068]).

1.2 An iterative algorithm is then used to obtain the probability that a user becomes the target user of the target service. The target user of the target service is selected according to the value of the probability (paragraphs [0041] and [0092] to [0112], Figure 1, steps 103 to 105). The iterative algorithm may use a "transfer matrix" to model the correspondences (paragraphs [0102] to [0110]), and a "value update rule" to calculate convergence values. Users corresponding to convergence values greater than a given threshold value may be selected as target users (paragraphs [0112] to [0117]).

Main request

2. Inventive step - claim 1

2.1 At the priority date of the present application, different recommendation systems were well known, as described in document D1. In particular, recommendation systems were well known which acquired user behaviour data in order to determine target products for a user (see e.g. page 1, second paragraph, to page 2, fourth full text paragraph, "Recommender systems have become extremely common [...] that are similar to the original seed)."). Such systems use a network server for acquiring and processing behaviour data.

2.2 The subject-matter of claim 1 differs from that well-known prior art in the particular method for determining a target user for a target service.

In its reply to the board's communication, the appellant agreed with this result of the novelty assessment but contested the board's preliminary opinion that the distinguishing features were non-technical and served a purely non-technical business purpose of finding a user to whom a target service could be successfully promoted.

2.3 Business methods, mathematical methods and computer programs constitute as such non-technical subject-matter which is not patentable under Article 52(2) and (3) EPC. Such subject-matter is not to be taken into account for inventive step, unless it makes a technical contribution by interacting with the technical features of the invention to solve a technical problem, bringing about a technical effect (see e.g. T 697/17, Reasons 5.2.2).

2.4 In its statement of grounds of appeal the appellant argued that data processing operations which provided the advantage of enabling data processing steps that could otherwise not be carried out were considered to be technical as for instance decided in T 598/14. In that decision, so the appellant, the board had considered that the translation, with the aim of enabling the linguistic analysis to be done automatically by a computer, could be seen as involving, at least implicitly, technical considerations. According to the appellant, the technical problem of the present invention was to analyse user needs.

The board does not agree with the appellant's interpretation of decision T 598/14. The appellant is citing a passage of that decision out of its context since in the following sentences and paragraphs (see Reasons 2.3, last two paragraphs and Reasons 2.4) the competent board concluded that, however, the considerations enabling the linguistic analysis by a computer were not "further technical considerations". The board drew a parallel with G 3/08 (OJ EPO 2011, 10) which considered that "although it may be said that all computer programming involves technical considerations since it is concerned with defining a method which can be carried out by a machine, that in itself is not enough to demonstrate that the program which results from the programming has technical character; the programmer must have had technical considerations beyond 'merely' finding a computer algorithm to carry out some procedure". The appellant has also mentioned this under point 20 of its statement of grounds of appeal.

The board further notes that analysing user needs is as such not a technical problem.

2.5 In point 15 of the statement of grounds of appeal the appellant also argued that in the method of claim 1 the target service could be a promotion of a 4G-based Internet surfing service belonging to a data service type which also included, for example, a tariff package and an Internet surfing package. Information about tariff packages and Internet surfing packages that were used by different users had to be acquired to find a user who was likely to use the 4G-based Internet surfing service. Analysing behaviour object identifiers representing user behaviour was technical because technical data was processed and correspondences were obtained from the acquired user behaviour data.

The board notes, however, that tariff packages and Internet surfing packages in the context of the present invention are non-technical business data. Claim 1 does not mention any technical specifications of such packages, or of the "target service" or the "service type", and therefore the claimed data processing does not take into account any such technical specifications. The non-technical business data is acquired automatically by technical means of the network server but that is not a distinguishing feature since it is already used in the well-known prior art as described in document D1 (see e.g. page 1, penultimate line, to page 2, fifth paragraph, referring to "build a model from a user's past behavior (items previously purchased or selected and/or numerical ratings given to those items) as well as similar decisions made by other users [...]", "observing what bands and individual tracks that the user has listened to", "requires a large amount of information on a user in order to make accurate recommendations [...]"; it is implicit from those passages that the behaviour information is obtained automatically).

2.6 In its reply to the board's preliminary opinion, the appellant argued that the invention was used in a commercial context, but had technical effects on data transmission, data processing, and the mobile network. The invention's purpose of finding a user to whom a target service could be successfully promoted was to be regarded as a further technical use within the meaning of decision G 1/19 of the Enlarged Board of Appeal. Referring to point 137 of the reasons of that decision, the appellant argued that the subject-matter of claim 1 also had "underlying models that form boundaries, which may be technical or non-technical" and that the outcome of the claimed method had an impact on the physical reality because the value of the probability was used to select the target user of the target service and to effectively improve accuracy of and efficiency in determining a target user. In this way, a network connection of electronic devices was changed. The method resulted in efficient computation using limited computing resources. As was discussed in point 126 of the reasons of G 1/19, a computer-implemented invention for use in a commercial context could be technical and provide a technical effect if for instance data processing and data transmission were affected by the results of the calculation. Therefore, improving accuracy of and efficiency in determining a target user was a technical effect solving a technical problem. The invention solved the technical problem of determining a target user based on data and was inventive over the prior art.

The board does not find these arguments convincing. The purpose of selecting a target user for a target service is to sell a product to the target user, which is a business purpose. As pointed out in the decision under appeal, no use of the selected target user is apparent from the claim, let alone a technical one. The distinguishing features do not contribute to any technical effect which could serve as a basis for an inventive step. The correspondence between identifiers is part of the abstract data model, not a technical solution to a technical problem. The alleged increased accuracy in determining the target user merely contributes to a different match between services and users based on the user's behaviour, which is not a further technical effect. Finding a user to whom a target service can be successfully promoted is thus not a further technical use within the meaning of G 1/19, Reasons 124 and 137.

Contrary to the appellant's argument referring to G 1/19, Reasons 126, the board is not convinced that the method is (more) efficient or that the distinguishing features are based on technical considerations relating to the implementation of the data processing in a computer or computer network (G 1/19, Reasons 125 and 126). The distinguishing features are based on non-technical business considerations regarding how to find out which user may be most interested on a target service based on the user's behaviour and how to program, in terms of the abstract algorithm, the non-technical method in a computer.

2.7 Since the distinguishing features are not technical and do not contribute to a technical effect, the method of claim 1 is not inventive (Article 56 EPC).

Auxiliary requests

3. Claim 1 of the first auxiliary request includes all the features of claim 1 of the main request and additional features essentially specifying that the data model is a transfer matrix and that the calculation of a value of an element in the data model comprises performing an iterative operation on a value of a matrix element of the transfer matrix to obtain, by means of calculation, a convergence value for each user identifier, and using the convergence value as the value of the probability that the user becomes the target user corresponding to the target service.

4. Claim 1 of the second auxiliary request adds to claim 1 of the first auxiliary request features essentially specifying how a convergence value R(n)m in the transfer matrix element in the Mth iteration is calculated as a function of the convergence value R(n)m-1, initial value R(n)0, the transfer matrix T, a diminution factor alpha, and the sum n of the number of user identifiers and behaviour object identifiers by the following formula:

FORMULA/TABLE/GRAPHIC**(.)

5. Inventive step - claim 1

5.1 The additional features concern the algorithm used to calculate the convergence value and are solely based on non-technical considerations regarding how to determine how likely it is that a user is interested in a target service and how to implement in a computer program the calculations of the convergence value.

5.2 In its statement of grounds of appeal the appellant argued that performing, by the network server, according to the initial values and the value update rule, an iterative operation on a value of a matrix element comprised in the transfer matrix and obtaining, by means of calculation, a convergence value for each user identifier and using it as a probability that a user becomes a target user for a target service were technical.

The board does not recognise that those features are technical. As explained above, they concern either the business aspects of finding a target user for the purposes of promoting a service or the non-technical programming aspects of finding the data structures and mathematical calculations to determine the probability of a user being interested in a service.

5.3 The appellant's alleged effect of increasing accuracy in determining a target user is not technical with regard to the additional features of the auxiliary requests either since neither the result of the selection of a target user nor the accuracy in that selection contribute to a further technical effect (see also decision T 755/18, Reasons 3.2).

5.4 In the board's opinion, none of the additional features defined in claim 1 of the auxiliary requests is technical, contributes to a further technical effect or is based on technical considerations. Therefore, claim 1 of the first and second auxiliary requests is not inventive either (Article 56 EPC).

Final remark

6. Since none of the requests is allowable, the appeal is to be dismissed.

Order

For these reasons it is decided that:

The appeal is dismissed.