Information Theory and Patent Documents

Akron Law Review Volume 55 Issue 2 Intellectual Property Issue Article 4 2022 Information Theory and Patent Documents W. Michael Schuster Follow this and additional works at: https://ideaexchange.uakron.edu/akronlawreview Part of the Intellectual Property Law Commons Please take a moment to share how this work helps you through this survey. Your feedback will be important as we plan further development of our repository. Recommended Citation Schuster, W. Michael (2022) "Information Theory and Patent Documents," Akron Law Review: Vol. 55: Iss. 2, Article 4. Available at: https://ideaexchange.uakron.edu/akronlawreview/vol55/iss2/4 This Article is brought to you for free and open access by Akron Law Journals at IdeaExchange@UAkron, the institutional repository of The University of Akron in Akron, Ohio, USA. It has been accepted for inclusion in Akron Law Review by an authorized administrator of IdeaExchange@UAkron. For more information, please contact , . Schuster: Information Theory and Patent Documents INFORMATION THEORY AND PATENT DOCUMENTS W. Michael Schuster * I. II. III. IV. V. VI. Introduction ..................................................................379 Quantifying Information ................................................381 A. Information of an Event ..........................................381 B. Expected Information of an Unknown Event (Entropy) ................................................................384 Message Probability ......................................................386 A. Entropy of Language using First, Second, Third, etc. Order Analysis........................................................386 B. The Reference Corpus of Information ......................387 Ambiguous Patent Language .........................................387 A. Using Information Theory to Identify Ambiguities...388 B. Ambiguity and Patent Law ......................................391 Quantifying Originality in Patent Documents .................394 Conclusion ....................................................................397 I. INTRODUCTION Recent scholarship has expanded the scope of analytical tools available to patent law researchers. Examples include the use of textual analysis to determine the similarity of two patents 1 and the use of network analysis to assess patent value. 2 This essay continues that trend by proposing a theoretical application of information theory to analyze Assistant Professor of Legal Studies, University of Georgia, Terry College of Business with a courtesy appointment at the University of Georgia School of Law. 1. See, e.g., Sam Arts, Bruno Cassiman & Juan Carlos Gomez, Text Matching to Measure Patent Similarity, 39 S TRATEGIC MGMT. J. 62, 64–65 (2018); see also W. Michael Schuster & Kristen Green Valentine, An Empirical Analysis of Patent Citation Relevance and Applicant Strategy, 59 AM. B US. L.J. (forthcoming Summer 2022) (manuscript at 231) (using the approach from Text Matching to Measure Patent Similarity to analyze backward patent citation relevance). 2. Guan-Can Yang, Gang Li, Chun-Ya Li, Yun-Hua Zhao, Jing Zhang, Tong Liu, Dar-Zen Chen & Mu-Hsuan Huang, Using the Comprehensive Patent Citation Network (CPC) to Evaluate Patent Value, 105 S CIENTOMETRICS 1319 (2015). * 379 Published by IdeaExchange@UAkron, 2022 1 Akron Law Review, Vol. 55 [2022], Iss. 2, Art. 4 380 AKRON LAW R EVIEW [55:379 textual ambiguity and identify particularly original disclosures in patent documents. Mathematician and engineer Claude Shannon published the foundations of information theory in 1948. 3 His research focused on analyzing the amount of information per second that could be transmitted and how to encode messages for efficient transmission. 4 As discussed in Section II, Shannon surmised that a message’s information content is a function of the uncertainty (also called “surprise”) of the message. 5 Highly unlikely messages convey a greater deal of information, 6 and the probability of a message can be determined by reference to earlier messages and the current context. For example, if a message thus far consists of the following characters: “I N F O R M A T I O,” then it is highly likely that the next character will be an “N.” 7 Thus, the receiver obtains very little new information from the letter “N.” Similarly, if an English language message contains a “Q,” then very little information (surprise) is received when the next character is a “U,” as a Q will be followed by a U in the vast majority of instances in English communications. 8 In contrast, we receive a great deal of information when “Q” is followed by an “F” because it is very improbable. From this recognition that not all characters (or words) convey the same amount of information, Shannon quantified the expected information content of any message through its entropy. 9 This metric (largely unrelated to the thermodynamics metric of the same name) quantifies the amount of information we expect to receive from the next part of a message, given (a) what we know about the message’s current context and (b) statistical trends in earlier bodies of collected messages. 3. C. E. Shannon, A Mathematical Theory of Communication, 27 B ELL S YS. TECH. J. 379, 623 (1948), reprinted in C LAUDE E. S HANNON & WARREN WEAVER, THE MATHEMATICAL THEORY OF C OMMUNICATION 29 (Univ. of Ill. Press Urbana ed. 1964) (10th prtg. 1964); Jeanne C. Fromer, An Information Theory of Copyright Law, 64 EMORY L.J. 71, 76–77 (2014); Alan L. Durham, Copyright and Information Theory: Toward an Alternative Model of “Authorship,” 2004 BYU L. R EV. 69, 73 (2004). 4. See Thomas M. Cover & Joy A. Thomas, ELEMENTS OF INFORMATION THEORY 1 (2d ed. 2006). 5. John R. Pierce, AN INTRODUCTION TO INFORMATION THEORY: S YMBOLS, S IGNALS, & NOISE 23–24 (2d rev. ed. 1980); Dan L. Burk, The Problem of Process in Biotechnology, 43 HOUS. L. R EV. 561, 584 (2006). 6. Ian Goodfellow, Yoshua Bengio, & Aaron Courville, DEEP LEARNING (ADAPTIVE C OMPUTATION AND MACHINE LEARNING SERIES) ILLUSTRATED EDITION 73 (2016). 7. Durham, supra note 3, at 76–77. 8. Pierce, supra note 5, at 49. 9. Ernesto Estevez-Rams, Ania Mesa-Rodriguez & Daniel Estevez-Moya, ComplexityEntropy Analysis at Different Levels of Organisation in Written Language, 14 PLOS ONE 1 (2019). https://ideaexchange.uakron.edu/akronlawreview/vol55/iss2/4 2 Schuster: Information Theory and Patent Documents 2021] INFORMATION THEORY & P ATENT DOCUMENTS 381 While seminal to modern digital technology, researchers have widely applied Shannon’s work and information theory, including within several legal studies articles. 10 In this paper, I extend that work by presenting a theoretical application of information theory to quantify several aspects of patent law, including lexical ambiguity and originality in innovation. To this end, Section II introduces Shannon’s ideas of quantifying a message’s information content and related entropy measures. Section III recogn (...truncated)