This duo of decisions dealt with an application at the UKIPO claiming a system that would trigger alerts when a trend reversal in data was detected. The way the trend reversal was detected involved a simple operation with two thresholds and was said to be simpler and more robust than prior art methods. The first decision found that some of the fields of use were non-technical, and hence the applicant was invited to limit to technical fields. When the applicant chose to continue the argument rather than limit claims to the technical fields, the second decision refused the application.
The first decision makes for good reading to illustrate how the UKIPO applies the "signposts" from case law to decide whether a claim is excluded subject matter or not. This involves identifying the contribution the claim makes without reference to the state of the and then asks a series of questions, for example, whether there is an effect on a process outside a computer, whether the result is a better computer, whether a technical problem is solved and so forth (the “sign posts). This approach is more similar to the US approach than the current EPO approach, which is focused on inventive step. Considering whether the contribution was more than just a computer program or mathematical method, the hearing officer found that is not the case for any date in general, so it became necessary to consider whether the nature of the data made a difference.
The claim in question recited that the data (“variable”) “represents a value in the field of meteorology, climatology, seismology, economy, population dynamics or cosmology”. The hearing officer observed that the data could be grouped into scientific data (meteorology, climatology, seismology and cosmology) and non-scientific data (finance, economy, population dynamics). While triggering an alert based on the latter would not take the contribution out of the exclusion as a computer-implemented business or administrative method, limiting the claim to scientific data would do so.
The hearing office observed that “there is instinctively a hint of something technical about an invention which relates to triggering an alert as a result of analysing data obtained from observations or measurements about the physical world.” This was because it will then relate to triggering an alert in response to determining a reversal of a trend in observations or measurements that pertain to the physical world. The applicant was given a month to amend the claims to limit to the scientific data. The claim would then be sent back to the examiner to search and examine the application for novelty and inventive step.
This is where the second decision came in. Instead of amending the claims as suggested by the hearing officer, the applicant amended the claims to recite “ones corresponding to a quantity of a technical nature”. The hearing officer observed that “the use of technical data does not make it inevitable that an invention falls outside an excluded field. Something more is needed, for example, an application or effect in a technical field.” Limiting to triggering an alert on specific data about the real world would have done the trick. Limiting vaguely to a quantity of a technical nature did not. Since business and administrative methods were not excluded by the amendment, the application was refused.
If you want to find out more about the point made in the second decision, this was dealt with in the High Court decision regarding Cappelini’s application, where the data in question was geographical way points.
While the algorithm in question here was straightforward (indeed, this was part of the benefit claimed by the applicant), the same line of reasoning should apply when machine learning is used to monitor observations or measurements of the physical world to trigger an alert or similar event. As the hearing officer said: “there is instinctively a hint of something technical about an invention which relates to triggering an alert as a result of analysing data obtained from observations or measurements about the physical world.” Thus, the decisions here may be helpful when prosecuting machine learning techniques that involve monitoring observations or measurements about the physical world.