Estimating semantic change in UK constitutional discourse

Published: 30 March 2022

Alex Schwartz, Senior Lecturer in Public Law at the University of Glasgow examines informal semantic changes that may have accompanied more formal changes to the UK's constitution over the past few decades. 

Constitutional change in the United Kingdom is a fascinating but fraught topic. In the absence of a codified constitutional text, an informal conceptual scheme of interrelated principles and values plays a pronounced role in how people understand (and argue about) the boundaries of constitutional propriety. But this informal aspect of the UK’s constitution is susceptible to subtle semantic change as people (particularly members of parliament and government) use the associated concepts in new ways or with different emphases. Changes of this kind are bound to be especially difficult to track and evaluate.

Predictably, scholars disagree about the nature and extent of recent constitutional change in the UK, but many share the view that reforms during the period of New Labour government from 1997-2010 (e.g., devolution, the Human Rights Act 1998, the establishment of a new Supreme Court, etc.) had a profound and enduring effect on the constitution. Indeed, some would say that those reforms effectively created a ‘new’ constitution, one that is less centred on the principle of parliamentary sovereignty and more reliant on other concepts (e.g., human rights, the rule of law, and the separation of powers).

In my article, ‘The Changing Concepts of the Constitution’, published this year with Oxford Journal of Legal Studies, I investigate semantic change in UK constitutional discourse. Using a machine learning algorithm (colloquially known as ‘word2vec’), I generated a series of ‘word embedding’ models to estimate the extent to which some concepts have become more (or less) semantically related to the UK’s constitution since New Labour’s reforms. A word embedding model is like a map of semantic relations; each word in a corpus of text is assigned a vector that locates it relative to other words within a multidimensional mathematical space, much like latitude and longitude locate a geographic point on a two-dimensional map of the world. The algorithm optimises word embedding models to predict a target word from a surrounding window of context words or, alternatively, to predict a surrounding window of context words from a single target word. The resultant models encode semantic relations such that words with relatively similar or related meanings will have correspondingly similar vectors and thus be ‘located’ relatively ‘closer’ to one another within the embedding space. The angle that is formed between any two vectors in this space can then be used to compute a measure of how semantically related or similar the associated words are to one another within the corpus.

My analysis focused on parliamentary debates scraped from Hansard. Parliament is a salient and routine forum for discourse on constitutional affairs in the UK and, presumably, semantic change within that discourse will shape the direction of constitutional practice. I divided debates into several chronological slices according to party-political changes in government: the era of Conservative government from 1979-1997; the era of ‘New Labour’ government from 1997-2010; the era of Conservative-Liberal Democratic coalition government from 2010-2015; and the era of Conservative government from 2015 to the general election in 2019. For each of these eras, I compute measures of ‘constitutional resonance’ for various concepts of interest.  These measures are estimates of how semantically related each concept is to the more general concept of the UK’s constitution.

The results confirm some important theoretical expectations. For example, according to my analysis, the principle of parliamentary sovereignty does indeed appear to have become less semantically related to the UK’s constitution following New Labour’s reforms. Furthermore, the UK’s constitution has apparently become significantly less equated with an ‘unwritten’ form over the past 30 years. The estimates also suggest that the concept of human rights has gained greater constitutional resonance, although this change is not as dramatic or sudden as might be expected. Other expectations of semantic change are not supported by my findings. Concepts of the rule of law and the separation of powers do not appear to have gained any constitutional resonance since New Labour’s reforms. Moreover, claims that the concept of parliamentary sovereignty has been supplanted are belied by apparent semantic continuity in the discourse—yes, parliamentary sovereignty appears to have a diminished constitutional resonance but, relative to other concepts, that principle continues to be very closely related to the meaning of the constitution (at least within parliamentary debate up until the end of 2019). 

Naturally, these findings are not intended to be the last word on constitutional change in the UK. The meaning of the UK’s constitution is constantly being reproduced and reimagined through an array of discursive practices, from parliamentary debate to blog posts. Semantic change in one of these practices may or may not be matched by equivalent changes in others. Computational methods are a powerful but still underexploited resource for studying constitutional discourse in any forum; I hope that my efforts will inspire others to explore the potential of such methods to further enhance and expand public law scholarship in the UK.  

Dr Alex Schwartz is Senior Lecturer in Public Law at the University of Glasgow. His research incorporates a variety of approaches (including empirical and computational methods) to address questions in the fields of constitutional law, constitutional design, and judicial politics.

Photo by Heidi Fin on Unsplash

Estimating semantic change in UK constitutional discourse

Constitutional reflections and renewal

Managing a harder border: The bare essentials for an independent Scotland

Constitutionally Sound podcast