Words by the tail: Assessing lexical diversity in scholarly titles using frequency-rank distribution tail fits

Abstract

This research assesses the evolution of lexical diversity in scholarly titles using a new indicator based on zipfian frequency-rank distribution tail fits. At the operational level, while both head and tail fits of zipfian word distributions are more independent of corpus size than other lexical diversity indicators, the latter however neatly outperforms the former in that regard. This benchmark-setting performance of zipfian distribution tails proves extremely handy in distinguishing actual patterns in lexical diversity from the statistical noise generated by other indicators due to corpus size fluctuations. From an empirical perspective, analysis of Web of Science (WoS) article titles from 1975 to 2014 shows that the lexical concentration of scholarly titles in Natural Sciences & Engineering (NSE) and Social Sciences & Humanities (SSH) articles increases by a little less than 8% over the whole period. With the exception of the lexically concentrated Mathematics, Earth & Space, and Physics, NSE article titles all increased in lexical concentration, suggesting a probable convergence of concentration levels in the near future. As regards to SSH disciplines, aggregation effects observed at the disciplinary group level suggests that, behind the stable concentration levels of SSH disciplines, a cross-disciplinary homogenization of the highest word frequency ranks may be at work. Overall, these trends suggest a progressive standardization of title wording in scientific article titles, as article titles get written using an increasingly restricted and cross-disciplinary set of words.

Hot trends and impact in permafrost science

“…Given the There is, however, no objective way to accurately quantify topical diversity based on text data, such as keywords. 20,43 It is not only changes to the natural environment that explain the increase in permafrost publications. There has also been a considerable increase in publications on permafrost at the third pole (i.e., the Qinghai-Tibetan Plateau).…”

“…There is currently no consensus on which method most accurately quantifies the topical diversity of a text, which is partly related to the fact that the concept of topical diversity is elusive to define. 20 We applied three of the most commonly used methods on the keywords of papers for the 1998-2007 and 2008-2017 periods. The Measure of Textual Lexical Diversity (MTLD) method measures the mean length of a text string that maintains a certain TTR.…”

“…This metric has the disadvantage of being sensitive to the total length of the text, and therefore several methods have been proposed that apply some correction for this bias. There is currently no consensus on which method most accurately quantifies the topical diversity of a text, which is partly related to the fact that the concept of topical diversity is elusive to define . We applied three of the most commonly used methods on the keywords of papers for the 1998–2007 and 2008–2017 periods.…”

Exploração de corpora para extração e descrição de léxico de especialidade

“…Os títulos, nos vários tipos de texto típicos de cada domínio, são cada vez mais um tópico de interesse no que respeita ao estudo das línguas de especialidade (Baicchi, 2003;Roy, 2008;Moore, 2010;Soler, 2011;Méndez et al 2014), desde logo porque permitem a identificação do conteúdo dos textos sem a sua leitura (Hoek, 1981). Devido à produção e à acessibilidade massiva de conteúdos, a importância dos títulos tem, aliás, crescido, havendo provas de que muitos especialistas confiam muitas vezes apenas na informação dada no título para integrarem as obras como referências nos seus próprios textos, ou para tomarem decisões técnicas (Bérubé et al, 2018). A extração e exploração de títulos para extração de palavras-chave, léxico e terminologia, tem sido assim, desde há algum tempo, uma prática disseminada com sucesso (Hu et al, 2005;Poulimenou et al, 2014), o que motiva a sua consideração na metodologia aqui apresentada.…”

“…No entanto, tem vindo a ser recentemente explorada a aplicação de padrões léxico-sintáticos, com base em conhecimento de forma sistemática, sob a forma de filtros linguísticos que facilitam e automatizam a análise das concordâncias para a extração tanto de unidades lexicais relevantes como de relações semânticas estáveis que estruturem modelos relacionais (e.g. Wordnet, Framenet) e /ou ontologias (Amaro, 2014;Cabezas-García;Faber, 2018;Faber, 2012;Gil-Berrozpe et al, 2017;San Martín, 2018).…”