T 313/10 – Double Mistake

This is an appeal of the applicant (Amazon.com) against the decision of the Examining Division (ED) to refuse its application relating to matching a given item description (e.g. of a book) with stored item definitions (e.g. in a website’s product catalog).

The USPTO acting as ISA found two X documents. For the supplementary search in the European phase, the search division issued a declaration of no search. The reason was that the subject-matter was a “mental act - selecting intangible items of abstract information on the basis of fuzzy but technically unspecified criteria - which is implemented by a program for computers”, both excluded under A 52(2). The overall method was seen as a method of doing business, also excluded.

In the first examination communication, the ED referred to the declaration of no search and made some other minor objections. In the reply, the applicant amended claim 1 to include the aspect of identifying candidate item. It cited T 258/03 and argued that the claims had technical character by virtue of mentioning a computer and because the identification of candidate item definitions was a technical solution to the problem of automating the matching of a large number of items.

The ED issued summons arguing in the communication that the identification of candidate item definitions was limited to the field of computer programs without a tangible technical effect going beyond the normal interaction between the computer and the program it was executing, without wilfully and directly controlling the forces of nature, without transforming or transporting a physical object and without the processed data being defined in their technical significance. The subject matter was therefore excluded from patentability. The division further stated that it was not bound by the case law. It added that the then pending referral G 3/08 showed that there was no generally accepted jurisprudence on computer program exclusions. It found that a decision could be reached on the basis of the Convention and the Guidelines alone.

In its reply, the applicant pointed out that the Guidelines in force as of April 2009 did in fact state that any claimed subject-matter defining or using technical means was an invention within the meaning of A 52(1) EPC and cited T 258/03 again. It gave arguments about which features it considered as technical and what technical problem was solved.

The applicant did not attend the oral proceedings (OPs). The application was refused on the ground of lack of patentable subject-matter (A 52(2) and (3)). The decision stated that “no further comments in the merit having arrived from the applicant …” the application had to be refused for the defects and reasons noted in the summons. The decision does not mention the Guidelines.

The applicant filed an appeal.

Claim 1 before the Board read:

A method performed by a computer system (100; 110) for identifying an item definition that matches an item description, the item definition and item description having attributes with values, the item definitions being stored in an item definition table (101), the method comprising:

providing one or more rules (211) that specify how to generate a similarity score based on similarity between the values of the attributes of an item definition and an item description, wherein at least one rule specifies a criterion for identifying candidate item definitions;

identifying one or more candidate item definitions in accordance with the rules using indexes (212) of attributes into the item definitions in the item definition table, each index for an attribute mapping values of that attribute to the item definitions;

for each of the one or more candidate item definitions, generating a similarity score for the candidate item definition and the item description in accordance with the rules (403, 902), wherein said generating comprises assigning a score to the attributes of the candidate item definition and the item description, and aggregating scores of the attributes to derive the similarity score (909); and

selecting the candidate item definition whose generated similarity score indicates it is most similar to the item description as the matching item description.

The Board found dismissed the appeal but expressed its disagreement with the ED:

[1] The first issue in this case is whether the claimed method, performed by a computer, of matching items in a table is excluded from patentability (A 52(2) and (3)). It is the established case law of the Boards of Appeal (see G 3/08, T 258/03, and T 424/03) that claimed subject-matter specifying at least one feature not falling within the ambit of A 52(2) is not excluded from patentability by the provisions of A 52(2) and (3). In this case, claims 1 to 16 are all method claims which specify that the method is “performed by a computer system”. Claims 17 to 24 are claims to a computer system and claims 25 to 41 are claims to a “computer-readable medium”. None of these features fall under the exclusions of A 52(2) and hence the claimed subject-matter of the present request is not excluded from patentability by the provisions of A 52(2) and (3).

[2] The ED argued, using their own criteria, that a method performed by a computer was excluded. This was contrary to the established jurisprudence as set out in the Guidelines for Examination at the time. After stating that they were only bound by the EPC and the Guidelines, the division ignored the applicant’s observations that their approach was in fact in breach of the Guidelines. Both of these acts were procedural violations.

[3] The next issue that arises is whether therefore the matter should be remitted to the ED for an additional search as per the appellant’s main request.

[4] It is established jurisprudence that an additional search can be dispensed with if the only technical features of the claims are considered to be “notorious”, i.e. generic and so well known that they cannot reasonably be refuted (see T 1242/04 [9.2] and T 1411/08 [4]). If this is the case, then there is no need to remit the case for an additional search. Thus the Board must examine whether there are any technical features that are not notorious.

[5] The invention is essentially to match an item description (e.g. of a book) with one in an “item definition table” (e.g. a web site’s database) using rules that define the required similarity of different attributes of the item (for a book, the attributes may be ISBN number, author, title, product type, etc.) The main aspect of the invention and the one that the appellant alleges to be technical is that, before applying the full set of matching rules, a number of “candidates” are identified (analogous to a pre-selection of candidates before a job interview). Although not claimed, this is based on a subset of the rules used in the full matching procedure. This avoids having to perform a time-consuming detailed match against all the items in the table (analogous to avoiding interviewing all applicants) [32]. The candidates are identified using indexes of the attributes.

[6] At the OPs, there was some debate about the nature of the indexes. The Board asked the representative whether they were not the same as indexes conventionally used by databases. Such an index may be a sub-table of the main table ordered by the attribute in question. Instead of searching the whole database for records with a particular value of the attribute, it is looked up in the index, which is quicker because the attributes are ordered. However, the representative explained that the index in the invention was different, being “an index into the attributes”, which was somehow different from an index into the table. Nevertheless, neither the claim, nor the description appear to provide any support for this interpretation, the claim defining the indexes as “indexes of attributes into the item definitions in the item definition table, each index for an attribute mapping values of that attribute to the item definitions”. In the Board’s view, this defines the indexes in the sense understood by the Board, namely as a mapping from attribute values to definitions in the table.

[7] According to the jurisprudence going back to T 208/84, one indication of technical character is that the method has an overall technical effect, such as controlling some physical process. This is not the case in the present invention since the effect of matching items with items in a table is either an abstract data processing one, or taking the embodiment into account (matching e.g. books and CDs), a business one.

[8] The technical character may also come from within, namely from the effect on the computer. This was the case, for example in T 424/03 where the technical effect came from “functional data structures (clipboard formats) used independently of any cognitive content … in order to enhance the internal operation of a computer system” (see point [5.2]). Such “functional data structures” were also considered to be present in the file search method that was the subject of T 1351/04. Here, the functional data was also in the form of an index containing “management information” that indicated the “start positions”, i.e. memory addresses, that pointed to the data to be searched for. This was said to control the computer by directing it to a certain memory location (see point [7.2]). This was contrasted with the looking up of addresses in T 52/85. In that case the addresses stored linguistic expressions that were equivalent to possible input expressions. The contribution there was seen to relate only to the kind of data represented, i.e. linguistic expressions. In T 1351/04, it was conjectured that the line to be drawn between the use of addresses in these two decisions is that if the method is concerned with the way that a computer performs the search, it may be technical. If, however, the kind of data is decisive it is not.

[9] It thus follows that the index may well have technical character since, using the terms of T 1351/04, it controls the computer by directing it to a certain memory location. On the other hand, the Board cannot see any technical effect in the identification of the items or candidate items in accordance with rules. This is a purely abstract process part of an algorithm to provide matching and not part of any technical process or interacting with the computer in any way.

[10] The representative essentially argued that the technical character came from being part of the technical process going on in the computer and that the identification of candidate item definitions was a technical solution to the problem of automating the matching of a large number of items using a computer. However, in the Board’s view, these are merely manifestations of the intrinsic effect of implementing an algorithm on the computer. They are analogous to “normal” physical interactions between program (software) and computer (hardware) that does not contribute in the case of a computer program (see T 1173/97 [headnote]).

[11] The representative argued that implementing the matching operation was the task of the engineer and therefore technical. The Board agrees that the implementation is indeed technical by virtue of the computer system. However, the point is that the contribution of the matching algorithm is not technical.

[12] In conclusion, the Board judges that the only features having technical character are the computer system and potentially the use of indexes. However, a computer system in general is clearly notorious. As stated above, the Board is of the view that indexes are certainly well known in the field of databases. The question is then whether they are so well known as to qualify as “notorious”. The Board judges that they are and were at the priority date of 2003. The problem of searching databases has been known for a very long time and methods of speeding this up using some kind of “index” have been implemented in virtually all commercial database products. Decision T 1411/08 requires not only that the features be notorious, but also generic, that is, features which are defined in such a way that technical details are not significant (see [headnote]). In the Board’s view, the present description of the use of the index does not specify any technical features and so is “generic” by that standard. In conclusion, the Board judges that claim 1 contains only technical features that are notorious so that no additional search is required.

[13] Accordingly, the Board cannot allow the appellant’s main request.

[14] Moreover, under these circumstances, the Board considers that there can be no inventive step since the only features that could contribute are notorious. Thus the problem to be solved would be along the lines of how to implement the matching scheme. Clearly, it would be obvious to use a computer system and to consider the use of indexes to speed up the access to the data. Claim 1 accordingly does not involve an inventive step (A 56 EPC 1973), so that the Board cannot allow the appellant’s auxiliary request either.  […]

The appeal is dismissed.

Point [2] of the reasons is interesting. Ignoring the applicant’s objection clearly amounts to a substantial procedural violation (SPV). But what exactly was the second SPV detected by the Board? Probably the fact that the ED acted in contradiction to the Guidelines. One may note, however, that sometimes Boards have stated that acting in this way did not constitute a SPV. Two examples should suffice:

• T 42/84: The Guidelines, as stated in the General Introduction to them, do not have the binding authority of a legal text. Therefore a failure by the ED to follow them is not to be regarded as a procedural violation within the meaning of R 67 unless it also constitutes a violation of a rule or principle of procedure governed by an article of the EPC or one of the Implementing Regulations. [9]

• T 647/93:  Although it is normally desirable for ED to act in accordance with the Guidelines, the Board wishes to point out that the Guidelines are guidelines, not rules of law, so that failure to follow a procedure set out there is not in itself a SPV. [4.1] 

The way in which the Board has worded point [2] of the reasons appears to suggest that what constituted the SPV is that the ED acted in contradiction to the established jurisprudence. I think this reasoning would lead us on even less solid ground, because it is not clear when case law becomes established. I have sometimes heard attorneys invoke “established case law” for which there was only one T decision (if any). And even if there was a generally accepted definition of “established jurisprudence”, I do not see anything in the EPC that would make it binding on an ED. Of course, an ED would be very unwise to go against well established case law - simply because its decision would be likely to be set aside in appeal proceedings - but I do not think that by doing so it would ipso facto commit a SPV, in particular if the Guidelines are silent on the matter. 

Any thoughts?

Should you wish to download the whole decision, just click here.

The file wrapper can be found here.

NB: This decision has also been presented on Le blog du droit européen des brevets.