This book is the 19th volume belonging to a collection called ‘Terminology and Lexicography Research and Practice’, which began in 1999 and aims to provide in-depth studies in formats such as monographs, course books and state-of-the-art volumes.
The research described in this book began in 1996, directly involving about twenty-six people linked to thirteen institutions (including seven universities) and three funded projects. For twenty years, the topic of term variation has become important in the areas of terminology, computational terminology and NLP. This is the first book that presents a complete inventory of term variations and the linguistic procedures that lead to their formation. It considers multilingual applications including term variation detection in five languages: French, English, German, Spanish and Russian. Its content is addressed to both terminologists and linguists, as well as researchers in natural language processing and computer science interested in application development.
The main merit of this important work is its contribution to the specification of the recognition of variant terms in multilingual form in the development of systems in the field of NLP. Its approach is essentially the engineering for the automatic recognition of variants using principles of linguistics and terminology, and involving orthographic as well as grammatical aspects. The semantic aspect is restricted, and occurs only when so-called surface linguistic representations allow. The book is also not intended to qualify the origins of terminological variations or even to treat diachronic variations over time. The book follows instead the linguistic mechanisms that lead to the formation of a variant: derivation, composition and borrowing. It adopts the descriptive methodology as follows: collecting the terminological variations of different domains and languages, analysing and classifying the collected variations and, finally, identifying the patterns of terminological formations by variation.
The structure of the book is divided into three main parts: variant term characterisation; the automatic discovery of variant terms; and, finally, the tools and applications for terminological extraction. In addition to an extensive review of the variant term detection methods, the author adopts two investigations. First, a symbolic approach is made according to its own specifications for multitext categories,Footnote 1 as the example below shows (page 47):
\begin{eqnarray*}
&& \mathrm{A /N_{1}+V_{2} \rightarrow N_{1}\,\, N_{2}\,\, where\,\, N_{2}\,\, is\,\, derived\,\, from\,\, V_{2}}\\
&& \mathrm{(eol)\,\, En:\,\, grid}{\hbox{-}}\mathrm{connected\,\, \rightarrow \,\, grid\,\, connection}
\end{eqnarray*}
After this, an empirical method is used, where a Generic Grammar Recognition Variants is conceived, as in the following example (page 134):
\begin{eqnarray*}
&& \mathrm{Left\,\, juxtaposition\,\, (Jl)\,\, A B \rightarrow X_{1-2}\,\, A B} \\
&& \mathrm{Right\,\, deletion\,\, (Rr)\,\, A B \rightarrow A}
\end{eqnarray*}
For the two approaches above, the five languages mentioned previously were considered, and applied to three different corpora: breast cancer, diabetes and wind energy. The last part of the book presents the TermSuite tool that is designed to support multiple languages and allows the user to identify terms and their variants.
In detail, the first part of the book following the introduction consists of three chapters. In chapter 2 – Definitions, in addition to a revision of terminological linguistic properties, the author presents the difference between simple and complex terms. The focus of the study was on complex terms. After analysing the frequency of these terms, the main result was the preponderance of nominal complex terms consisting of two words. These terms have specific variations and are presented in detail in this chapter. In chapter 3 – Conceptualisation of Terminological Variants, besides presenting the main linguistic mechanisms of terminological variations, the author presents a hierarchy: conceptual variation → denominative variation → linguistic variation. In chapter 4 – Semantics of Conceptual Variants, in addition to presenting a literature review on conceptual and denominative aspects, the author introduces the relationship of predication and instance, justifying its presence in the five languages used.
The second part of the book consists of four chapters. In Chapter 5 – Primitive exploration of variants using comparable corpora, the simple n-gram method on the three corpora and five languages is used, which confirms the fragile frontier present in the literature between morphological and syntagmatic compounds. In Chapter 6 – Methods of processing for the detection of variants of bodies, in addition to the presentation of NLP techniques, one of the main difficulties highlighted by the author is the limitation of finding variations when the distance between the terms increases. In Chapters 7 and 8, Grammar of Variants and Synonym Variants, respectively, the Generic Grammar for Recognition of Variants is presented. Among the analyses made, the author empirically confirms the correlation between the conceptual and linguistic variants provided in the first part of the book. Given the difficulty of the recognition of variants by synonymy, the technique is based on the distributional analysis of the term variations.
The third part of the book is composed of two chapters. Chapter 9 – Terminology Extraction, besides describing the stages of identification of the candidate terms and their classification, the author presents as a significant empirical result that the terminology variation analysis achieves better results than the extraction of the terms alone (without their variations). In Chapter 10 – End-User Applications and Tools – the UIMA-based TermSuite tool is presented. Developed in Java and with a command line interface, this aims at extracting terminological variations.
All in all, the book is a landmark in the term variation literature, which leads also to the potential exploration of new frontiers that go beyond the book itself. The author points out that the technique of machine learning would be one of those frontiers. She presents the need for larger specialised databases than those currently available. Furthermore, we suggest that the reader could also consider extensions of the work to other languages, in addition to those five considered in the book. The multilingual aspect could acquire new perspectives with languages with different structures, such as Chinese with its radical difference using an ideographic writing system. In a complementary manner, corpora in other domains, in addition to the three studied in the book, could also be used for term variation studies. Undoubtedly, the book is robust enough to support many other studies on term variation, and is presented comprehensively and clearly in terms of theoretical, methodological and empirical aspects.
