In this decision, the European Patent Office did not grant a patent on a system for structured queries in social networks, particularly for mapping natural language text input to elements of a social-graph database. Here are the practical takeaways of the decision T 1089/17 (Ambiguous queries on online social networks/FACEBOOK) of 7.2.2020 of Technical Board of Appeal 3.5.07:
This European patent application, filed by Facebook, Inc., generally relates to social graphs and performing searches for objects within a social-networking environment.
The invention proposes a social-networking system that may, in response to a text query received from a user, generate structured queries that include references to particular social-graph elements. By providing suggested structured queries in response to a user’s text query, the social-networking system may provide a way for users of an online social network to search for elements represented in a social graph based on their social-graph attributes and their relation to various social-graph elements.
For example, when a user enters the ambiguous search term “facebook” as part of the search query “people who like facebook” (a shown in the figure below), the online social network may propose several matching elements of the social graph to the user for selection. If the user then selects one of the proposed matching elements (for example, “Facebook”), the system proposes a set of queries related to the selected graph element (for example, “People who work for Facebook”, “People who like Facebook and Stanford”), from which the user can select one to be performed.
Here is how the invention is defined in claim 1:
Claim 1 (main request)A method of operating an online social network (160), comprising:
accessing (510), by a computing device (160), a database (164) of the online social network (160), the database (164) implemented as a social-graph database (164) comprising a plurality of nodes (202, 204) and a plurality of edges (206) connecting the nodes (202, 204), each of the edges (206) between two of the nodes (202, 204) representing a single degree of separation between them, the nodes (202, 204) comprising:
a first node corresponding to a first user associated with the online social network (160); and
a plurality of second nodes that each correspond to a concept or a second user associated with the online social network (160);
by the computing device:
– receiving (520) from a client system (130) of the first user an unstructured text query comprising an ambiguous n-gram;
– identifying (530) in the social-graph database (164) a plurality of second nodes or a plurality of edges corresponding to the ambiguous n-gram;
– generating (540) a first set of structured queries, each structured query from the first set of structured queries corresponding to an identified second node or identified edge, the structured query comprising a reference to the identified second node or identified edge;
– receiving (550) from the client system (130) of the first user a selected first structured query from the first set of structured queries, the first structured query corresponding to a selected second node or selected edge from the identified second nodes or identified edges, respectively; and
– generating (560) a second set of structured queries, each structured query of the second set of structured queries comprising a reference to the selected second node or selected edge.
Is it technical?
Already in the European search opinion, the application was considered to be directed “at a cognitive problem”. In essence, the claim elements related to the social graph were found to be “purely social-administrative”, and the elements related to the query processing were considered to have “a purely cognitive character”. Following the EPO’s appproach for mixed-type inventions, the examiner held that the closest prior art was a general-purpose computer system, and that implementing the non-technical method on such notoriously known hardware was obvious. The examining division essentially maintained this position throughout the prosecution procedure, so that the application was refused for lack of inventive step in first instance.
On appeal, the applicant argued that the operation and function of the claimed search engine may be considered to lie in a technical field. The board agreed as far as none of the cited documents was considered to disclose a search process which maps natural language text input to elements of a social-graph database in the context of a social network. The board therefore took document D5 (a patent application by Facebook itself) as the starting point for assessing inventive step.
The board found that the invention as claimed differed from D5 by “generating a second set of structured queries, each structured query of the second set of structured queries comprising a reference to the selected second node or selected edge”. Here is how the board interpreted these features:
The distinguishing features have the effect of generating a set of exemplary queries regarding a selected element of the social graph. In this respect, the Board observes that the application discloses ranking the structured queries based on advertising sponsorship (description, paragraph ), which seems to suggest that the generated queries may serve a non-technical purpose and result from business considerations.
The claimed generation of queries according to features 8 and 8.1 does not contribute to a “further” technical effect. In particular, the Board does not agree that the human-machine interaction is improved, as no interaction between the user and the system after the disambiguation of the entered data is claimed. The distinguishing features do not involve any further interaction or the display of the generated second set of structured queries to the user.
Moreover, the Board considers that defining a natural language query is per se not a technical task, but lies in a non-technical field. As the claim is entirely silent regarding the user interface for entering search queries, the Board is not convinced that the claimed method solves a technical problem in the area of user interfaces.
The applicant also argued that the fact that claimed subject-matter had some relation to “semantic” aspects did not inevitably imply that it was of non-technical character. Non-technical character could only be affirmed if absolutely no further technical considerations, for example related to the database/search engine and their structure and/or function, were involved. But also this argument did not persuade the board:
The Board is not convinced that the distinguishing features involve any “further technical considerations” (see opinion G 3/08, OJ EPO 2011, 10, Reasons 13.5.1), as the generation of the second set of queries is based on non-technical considerations regarding the desired query semantics in the context of the social graph. The Board observes that the social graph, which is known from document D5, constitutes social data not serving a technical purpose, and the distinguishing features do not define whether or how the social graph as a data structure is used to generate the second set of structured queries.
Lastly, the applicant argued that the generated search term related to the underlying graph structure of the database, so that the invention takes the technical structure of the underlying database and search engine into account, which makes a technical contribution. But the board also did not follow this argument:
However, the social graph is already known from document D5, and the distinguishing features 8 and 8.1 define the generation of the second set of structured queries only in terms of non-technical semantic aspects. Consequently, the Board considers that features 8 and 8.1 are not about exercising technical control over the functioning of the search engine in the sense of decision T 2230/10. Hence, the Board is not persuaded by the appellant’s arguments.
As a result, the board held that the distinguishing features do not contribute to the solution of a technical problem and cannot be considered for the assessment of inventive step. Claim 1 was found to lack an inventive step.
You can read the whole decision here: T 1089/17 (Ambiguous queries on online social networks/FACEBOOK)