T 1523/11 - Skilled person and inventive step

Key points

• In this examination appeal, the Board adds further evidence of the publication date of the document used as closest prior art of its own motion (" In the EPO's database, the Board has located document D13, which is another copy of the same document [D1]." 
• In connection with inventive step, the Board finds that the technical field of the skilled person can be defined by the problem addressed.
• " According to the appellant, the skilled person starting from document D1 would not look for a solution to this problem in the technical field of database searching. The Board considers, however, that if the technical problem relates to index structures for searches, then the skilled person on the basis of whose knowledge and ability the obviousness of the solution is to be assessed is a person with knowledge of index structures, i.e. a person skilled in the field of database management systems. " 

EPO T 1523/11 - link

3. Document D1

3.1 Document D1 bears the heading "SP003 V1.3 part B - System issues" but mentions neither an author nor a date. The Examining Division cited it as "1st Draft of Metadata Specification SP003v1.3", authored by "EVAIN J P" and published on 11 June 2002. It apparently assumed that it corresponded to the document cited in the application as disclosing a single key index structure (see point 2.2 above).

3.2 In the EPO's database, the Board has located document D13, which is another copy of the same document. Its bibliographic details suggest that document D1/D13 is indeed the document cited in the application and was published in June 2002 in connection with the 17th TV-Anytime Forum meeting (see document D17). These details are consistent with the information on "SP003v13" given on page 2 of document D14, which reports on the meeting.

3.3 The Board further notes that document D18 confirms that TV-Anytime Forum specifications at all stages of their development have been publicly available from the Internet address "ftp://tva:tva@ftp.bbc.co.uk/Specifications/" (page 2, section "Relevant ancillary documentation"). The Board notes that document D17 mentions the same address.

3.4 In view of these findings, the Board considers the content of document D1 to be part of the prior art under Article 54(2) EPC. The appellant has not contested this and in its letter of 16 December 2016 in fact stressed that it accepts the document's publication date to be 11 June 2002.

4. Main request - inventive step

4.1 It is undisputed that document D1 discloses the "single key index structure" for searching TV-Anytime metadata which is discussed in the application and summarised in point 2.2. 2 above. Document D1 indeed discloses in section 2.3 an index structure comprising a "key index list" (section 2.3.2), a "key index" (section 2.3.3) and a "sub key index" (section 2.3.4). The metadata is provided as an XML document (see section 2.3.1) and hence is "arranged in a semistructured schema".

4.2 The subject-matter of claim 1 of the main request differs from this prior-art single key index structure in that

- the keys are not single-attribute keys ("single keys") but multi-attribute keys ("multi-keys"); and
- values of a multi-key are compared by prioritising the fields (or attributes) making up the multi-key and comparing the fields in sequence starting from the field having the highest order of priority, wherein the values of a field are compared on an arithmetic basis if the field has numerical values or on a lexicographic basis if the field has alphabetical values.

4.3 The objective problem addressed by these features may be regarded as providing an efficient index structure for compound condition searches, i.e. searches on two or more fields.

4.4 According to the appellant, the skilled person starting from document D1 would not look for a solution to this problem in the technical field of database searching. The Board considers, however, that if the technical problem relates to index structures for searches, then the skilled person on the basis of whose knowledge and ability the obviousness of the solution is to be assessed is a person with knowledge of index structures, i.e. a person skilled in the field of database management systems (cf. decision T 26/98 of 30 April 2002, reasons 6.2 and 6.3). For the problem posed it is irrelevant that the metadata was retrieved in the form of a segmented stream.

4.5 The skilled person in the field of database management systems is well aware of the possibility of creating an index on a combination of multiple fields and he knows that such an index will speed up compound condition searches on those fields. Indeed, document D19, which is a textbook on database management systems and is illustrative of the common general knowledge of the skilled person, discusses in section 8.4.4 indexes for "composite" or "concatenated" search keys which contain several fields. The value of such a search key is a tuple of values of the combined fields.

The skilled person would therefore, on the basis of his common general knowledge, consider adapting the prior-art single-key index structure to work with such composite keys, i.e. multi-keys.

4.6 In doing so, the skilled person would realise that an order has to be defined on the values of multi-keys. Indeed, section 2.3.3 of document D1 explains that the key index structure lists the values of the highest key of each of the ranges of key values sorted by increasing key value, which means that it must be possible to say for any two key values which value is "higher". In the case of multi-keys these key values are tuples of values.

A well-known order on tuples of values is given by what is known in the computer-science literature as the "lexicographic order". This order is a generalisation of the alphabetic or lexicographic order on strings of characters (i.e. sort first on the first character, then on the second character, and so on). According to the lexicographic order on tuples of values, first the tuples are ordered by the value of their first ("highest priority") component; in case of a tie they are ordered by the value of the second component, and so on. In this comparison process, values of tuple components are compared in the normal way: if they are numerical values, they are typically compared on an arithmetic basis; if they are alphabetical values, they are typically compared lexicographically (i.e. alphabetically).

The examples given in section 8.4.4 of document D19 in fact silently use the lexicographic order on tuples: the values of the "<age, sal>" composite key are ordered (11, 80), (12, 10), (12, 20), (13, 75), i.e. first by age, then by salary ("sal"), and the values of the "<sal, age>" composite key are ordered (10, 12), (20, 12), (75, 13), (80, 11), i.e. first by salary, then by age. While the appellant is correct that document D19 does not "hand" the lexicographic order to the reader, the point is that document D19 presupposes that its skilled reader is well aware of the lexicographic order on tuples of values.

The order on multi-key values specified in claim 1 is in fact this lexicographic order on tuples.

4.7 Hence, the skilled person would adapt the prior-art single-key index structure to work with composite keys. In doing so, he would apply the lexicographic order when comparing values of multi-keys. He would thereby arrive at the subject-matter of claim 1 without the exercise of inventive skill.

4.8 In the statement of grounds of appeal, the appellant submitted that it is "precisely the prioritized arrangement of the metadata in the fields [that] allows an improvement in the search efficiency and speed" and that the prioritised order "allows a sequential search to yield results sooner (earlier in the sequence than half-way)".

However, in the application and in the claims the term "prioritization" refers merely to a ranking or ordering of the attributes that make up the multi-key. Such a ranking is necessary to define the lexicographic order on tuples of attribute values. The "prioritization" feature is thus part of the well-known "multi-key index" solution for speeding up compound condition searches.

4.9 Hence, the subject-matter of claim 1 lacks inventive step (Article 56 EPC).