T 0022/85 (Document abstracting and retrieving) 05-10-1988

Summary of Facts and Submissions

I. European patent application No. 80 107 625.8 filed on 4 December 1980 (Publication No. 0 032 194) claiming a priority of 28 December 1979 (US) was refused by a decision of the Examining Division 065 dated 30 August 1984. That decision was based on Claims 1-6 filed on 31 December 1983.

II. The reason given for the refusal was that the contribution to the art resided solely in a computer program as such within the meaning of Article 52 EPC, paragraphs 2(c) and

3. Consequently, this subject-matter was not a patentable invention within the meaning of Article 52(1) EPC, in whatever form it was claimed.

III. In arriving at this conclusion the Examining Division argued on the basis that Claims 1 and 2 related to a method for automatically abstracting and storing an input document in an information storage and retrieval system and Claims 3-6 to a corresponding method for retrieving a document from the system. The claims specifically referred to a dictionary memory, input means, a main memory and a processor. These hardware elements were classical elements of an information and retrieval system (as described e.g. in: Hillman, Proceedings Spring Joint Computer Conference, 1969, pp. 447-455) and objectionable under Article 54(2) EPC as lacking novelty. According to the present description (see page 4, line 26 to page 5, line 4 and lines 22 to 28 and page 6, line 15 to page 8, line 19) the method steps were implemented by programming such a classical system. The claimed combination of steps did not imply an unusual use of the individual hardware elements involved. The claims merely defined a collocation of known hardware and new software concerned with document information to be stored but not with an unexpected or unconventional way of operating the known hardware. The differences between the prior art and the subject-matter of the present application were defined by functions to be realised by a computer program which was used to implement a particular algorithm, or mathematical method, for analysing a document. In other words the steps of the method defined operations which were based on the content of the information and were independent of the particular hardware used.

IV. The Appellant lodged an appeal against this decision on 8 November 1984. The appeal fee was paid on 9 November 1984. On 23 November 1984 the Statement of Grounds for the appeal was filed.

V. In a communication of 18 March 1987 the Rapporteur informed the Appellantthat in the provisional opinion of the Board the claimed methods did not present the technical character required for an invention and did not provide a result which could be qualified as being of a technical nature, even if technical means in the form of a suitably programmed computer were applied to carry out the said methods. The claims then on file effectively sought protection for methods in which conventional computer means are controlled by a program so as to carry out abstracting, storing or retrieving of documents in accordance with a specified set of rules for these activities. The new contribution to the art made in the present case, however, lay clearly essentially in the provision of this set of rules. Therefore the ing related to subject-matter which is excluded under Article 52(2) and (3) EPC as such.

VI. In the Statement of Grounds and in the reply of 5 November 1987 to the aforesaid communication the Appellant essentially argued as follows: The provisions of Article 52(2)(c) EPC state that computer programs are not patentable, but do not define what a computer program is. Referring to the definition given by the World Intellectual Property Organization (WIPO) a computer program is "a set of instructions capable, when incorporated in a machine-readable medium, of causing a machine having information capabilities to indicate, perform or achieve a particular function, task or result". The application does not relate to such a set of instructions but rather to a new way of operating a computer. Moreover, known hardware elements of an information and retrieval system included in the preamble of the claims are required for the implementation of the new method. Even if this method were assimilated to a program, the claims could not be refused since Article 52(2)(c)EPC excludes programs from patentability only to the extent to which the invention relates to a program as such (Article 52(3) EPC). The Appellant has always agreed that the hardware elements included in the preamble are known elements of an information storage and retrieval system. But this does not decrease the merits of the invention. Indeed, the claimed combination implies an unusual use of the hardware elements involved inasmuch as the functional relationship or "combination" of the elements is different as a result of there being a new method of operating. Besides, the test of "unexpectedness" is not different from the test of inventiveness according to Article 56 EPC and should not be used as a criterion for Article 52(2)(c) EPC. The criterion of inventive step has not been discussed in the decision to reject the application. Moreover, although the only detailed embodiment described in the specification consists in software on a programmable information and retrieval system, the information given in the claims is sufficiently clear and complete to enable a person skilled in the art to carry out the invention using logical means, in a specifically designed arrangement. The Appellant considers that the claimed invention is a technical solution to a technical problem, which is new, inventive, and capable of industrial application. In all the prior techniques for document storage and retrieval, the major intelligence burden is put on the user. Where the system aids in abstracting, it is done at the cost of voluminous cataloguing procedures and a massive data processing burden. The technical problem is therefore to provide a system which is capable of recognising and collating those language terms which may be relevant for searching a document. In other words, the problem is how to carry out automatically, without a massive data processing burden, a task which is not easy, even to a human being. The technical solution is to have a dictionary memory which contains a common fund of words, to label certain words as being of relevance such as nouns and single purpose adjectives, to compare the language terms in the input document to the dictionary in order to retain only those words which are labelled, and to consruct an abstract composed of the latter words and any words which do not appear in the dictionary such as acronyms, numerics and proper names. The claims relate to an activity which certainly brings about a change in the physiical environment insofar as, when an abstract is constructed and stored in memory, the system is modified by storing new information (the new abstract). In other words, a physical entity, which is here the information stored as electric signals, has been changed by the operation of the claimed system. Indeed, the claims do not seek protection for the content of the information (here the abstract itself as such), but for a system enabling such information to be constructed and for a method of operating such a system. That the described embodiment is implemented in software on conventional computer means is not relevant insofar as the invention is of a technical nature. The Appellant finally considers that the claimed invention is novel and shows an inventive step with respect to the prior art. The article by Hillman in the Proceedings Spring Joint Computer Conference describes a method of automatically abstracting a document which is entirely different from the claimed method. In the application under examination, the language terms are semantically processed by comparison with the language terms stored in a dictionary store. On the contrary, the method described in the article by Hillman is carried out according to a syntactic approach in which each sentence of the document is reduced to a string of syntactic categories which are then resolved into canonical substrings without comparison with terms stored in a dictionary.

VII. With his reply of 5 November 1987 the Appellant filed a set of amended Claims 1-10 and requested the grant of a European patent on the basis of these claims. New Claim 1 and appendant Claims 2-4 relate to a system for automatically abstracting an input document and storing the resulting abstract. Claim 5 relates to a method of operating the system according to Claims 1-4. Claim 6 and appendant Claims 7-9 are directed to a system for automatically retrieving from an input query a document abstract which has been constructed in accordance with the operating method of Claim 5. Claim 10 relates to a method of operating the system according to Claims 6-9. Claim 1 of this set of claims reads as follows:

1. System for automatically abstracting a document and storing the resulting abstract comprising: a dictionary memory (8) storing a dictionary of language terms commonly used in document preparation with each entry thereof for containing a language term, input means (16) for receiving the input document in machine readable form, a main memory (12), and a processor (10) connected to said dictionary memory and to said input means; said system being characterized in that said processor comprises: means for comparing the language terms of the input document to the entries in said dictionary memory, first means for selecting the language terms from said input document that do not compare to an entry in said dictionary memory, thereby being message specialization terms such as proper names, acronyms and numerics, second means for selecting the language terms from said input document that compare to an entry in said dictionary memory which has a code identifying certain ones of said language terms as selected parts of speech, first means for storing in said main memory an abstract of said input document composed of said language terms that do not compare or compare to an entry of said dictionary memory, all other terms of said document being discarded, and second means for storing in a file of said main memory a record of each selected language term including said term and several parameters determining said term with respect to said input document, said file being used to retrieve a document from terms of an input query.