The legacy of Zellig Harris: Language and information into the 21st century, vol. 2: Mathematics and computability of language (review)

The present book is the second in a two-volume tribute to the memory of Zellig Harris (I reviewed the first volume in Language 81.719–36, 2005). This collection focuses on issues of formal systems, computability, and computer application, and consists of a collection of papers from a wide range of perspectives that are bound to appeal to readers from a diverse set of backgrounds, though few readers will find all of the papers accessible.

In ‘Reflections on references to mathematics in the work of Zellig Harris’, the French mathematician André Lentin offers an interesting perspective on the influence of work on finitism and constructivism on Harris’s work on rethinking the foundations of linguistics. Fernando Pereira, in ‘Formal grammar and information theory: Together again?’, gives a masterful overview of the ways in which Harrisian interests can be viewed as having evolved into the concerns of contemporary computational linguists regarding machine learning. His remark that ‘any model of language that appeals to non-observables, for instance any model that assigns syntactic analyses, will require hidden variables’ (22) will be food for thought for many of his readers.

In ‘Logics for intercalation’, Richard Oehrle develops a method for approaching Semitic-style consonant and vowel intercalation using the tools of multimodal categorial grammar. D. Terence Langendoen, in ‘Sequence structure’, defines and explores a notion of sequence structure, emphasizing the usefulness of a generalized notion of linear precedence in a system less bound to hierarchically based representations.

‘The computability of strings, transformations, and sublanguage’, by Naomi Sager and Ngô Thanh Nhàn, is quite notable for its successful effort to explicate the Harrisian notions of string analysis, transformational analysis, and sublanguage, and its lucid explanation of how these related projects have been implemented in a number of practical computational implementations, at NYU and elsewhere.

‘Hierarchical structure and sentence description’, by Aravind Joshi, continues the discussion of Harris-inspired approaches to computational analysis of natural language in which the question is addressed as to what limits can be placed on hierarchical representations in syntax, with special emphasis on tree-adjoining grammars.

Stephen Johnson, in ‘The computability of operator grammar’, presents an overview of Harris’s notion of operator grammar, which I found difficult to follow, despite Johnson’s making a useful effort to compare and contrast operator-grammar formalisms with categorial-grammar formalisms.

‘Distributional syntactic analysis and valency: Basic notions, procedures, and applications of the pronominal approach’, by Karel van den Eynde, Sabine Kirchmeier-Andersen, Piet Mertens, and Lene Schøsler, presents an overview of an interesting perspective on grammatical analysis, which is clearly inspired by Harris, as well as European valency-oriented approaches; roughly, it seeks to ground grammatical analysis explicitly in the core aspects of a language that remain invariant, we might crudely say, when all noun phrases have been ‘reduced’ to their closed-class cousins, the pronouns; case studies of French and Danish are provided.

Benoît Habert and Pierre Zweigenbaum survey some of the automatic methods being used for automatic knowledge acquisition in ‘Contextual acquisition of information categories: What has been done and what can be done automatically?’. A special effort is made to explicate the relationship between Harrisian approaches and current work in computational linguistics that falls under the rubric of unsupervised learning of grammar.

‘Text generation within sublanguages’, by Richard I. Kittredge, provides both a logical and a historical overview of large-scale computational projects that have engaged in natural-language generation of what Harris would call sublanguages, such as weather reports, medical reports, and stock-market reports; and in ‘A distributional semantics applied...