Articolul precedent |
Articolul urmator |
250 3 |
Ultima descărcare din IBN: 2023-10-24 10:07 |
SM ISO690:2012 UNGUREANU, Valeriu. Application of Machine Learning in the Research of an Unknown Text. In: Conference on Applied and Industrial Mathematics: CAIM 2022, Ed. 29, 25-27 august 2022, Chişinău. Chișinău, Republica Moldova: Casa Editorial-Poligrafică „Bons Offices”, 2022, Ediţia a 29, pp. 183-184. ISBN 978-9975-81-074-6. |
EXPORT metadate: Google Scholar Crossref CERIF DataCite Dublin Core |
Conference on Applied and Industrial Mathematics Ediţia a 29, 2022 |
||||||
Conferința "Conference on Applied and Industrial Mathematics" 29, Chişinău, Moldova, 25-27 august 2022 | ||||||
|
||||||
Pag. 183-184 | ||||||
|
||||||
Descarcă PDF | ||||||
Rezumat | ||||||
For almost 34 years, WRI has been developing the Wolfram Language, and Mathematica System, as a major computation and programming environment for creative categories of people [S. Wolfram; C. Hastings and others]. The Wolfram Language permits us to apply the most advanced computations and knowledge in areas that are extremely far from programming, numeric, symbolic, and technical computations. We present an illustration how theWolfram Language, and Mathematica System, can be used to investigate poetry and prose, translations of literary works to other languages, evaluation of translation quality, discovering the author of unknown text, and highlighting plagiarism. We begin by obtaining some numerical characteristics of works in different original languages, and their translations into others, e.g., from Romanian to English, French, and Russian. A good translation must preserve most characteristics of the original works. Is this so in practice? To make more objective conclusions, some graphical, image, and sound perspectives are presented. We can also use some advanced mathematical tools such as interpolation, and curve fitting. Based on interpolation-functions (or fit-functions) that correspond to original works and their translations, we can evaluate good translation-works as ones for which interpolation functions for translated text (fit-functions) differ insignificantly from the interpolationfunctions (fit-functions) of the original works and are close one to other. We continue with the application of Machine Learning to train a function that may recognize poetry and prose texts, which may find text’s author, too. Should we train a function for every language, or is it enough to train one function for all languages? May a trained function have a “polyglot” feature? If the trained author-function “understand” more than one language, may it be applied to evaluate good work-translation? Should the trained author-function understand who is the original language author of translated work? We present answers to the above questions and highlight a series of other interesting subjects which arise in this context. The final discussion and conclusions are a good starting point to an interesting area of research: computational recognition of the original language author for a translated work. |
||||||
|