The Patents Act of 1902 introduced a new examination element in Great Britain to determine whether all or any part of an invention had been previously patented. This change rested almost entirely on evidence of such ‘anticipation’ presented by the Patent Office to a Parliamentary Inquiry in 1901, claiming 42% of patents granted in the preceding three years had been anticipated, and were thus invalid. Evidence provided by the Institute of Patent Agents, inventors and consumers of patented goods could not deter the Committee from finding the issue of anticipated patents a ‘serious evil, inasmuch as it attends to the restraint of trade and to the embarrassment of honest traders and inventors.’
Detailed examination of Parliamentary and Patent Office papers reveals, however, the evidence on which this momentous decision was based is far from robust, representing an extremely limited sample of specifications subject to human error. This paper exploits the entire population of patent specifications for the years 1883-1913, subjecting them to modern textual analysis techniques using machine learning to test the evidence of the Patent Office and Committee findings. We provide falsifiable evidence to assess the extent of anticipation in patent specifications before and after the drafting of the Patents Act into law in 1905.
Our results determine whether examination to prevent anticipation was indeed necessary, and if it was successful. If successful in lowering anticipation, then we can demonstrate the 1902 Patents Act did indeed force novelty in patenting, raising further questions concerning whether this meant patented inventions after the Act were more valuable. If no significant evidence of anticipation is found, then the considerable financial and institutional resources invested in the changes were effectively wasted, calling into question the Inquiry, actions of the Committee and ultimately the Act itself.