Abstract
This chapter illustrates some of the errors that translation machines may make in processing natural language. That these translation errors would never have been made by human translators raises the question as to why this is so. What is the translator’s brain doing in its handling of language that is different from the machine? Is the brain processing language in ways that have yet to be understood? Could it be that neuroscience has overlooked evidence of a cerebral language process that is different from what cognitive science and neurolinguistics have traditionally proposed for the brain? What would such a hitherto under-recognized process look like and might it be simulatable in a translation machine? These questions constitute the topic of this book.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
In November, 2016, Google released a new neural MT version (GNMT) of Google Translate. Except where otherwise noted, a number of the Google translations in this book predate this release. See Postscript 6-B of Chap. 6 for the GNMT versions of Google Translate’s SMT translations shown in this book.
- 2.
SYSTRAN is now said to be a hybrid linguistic/statistical system. Historically, its foundation is linguistic and algorithmic in nature.
- 3.
Google Translate’s new GNMT output for the third sentence in (1) is now correct: Der Klang des Herzens klingt normal.
- 4.
Logos Model translation of (1) is syntactically correct: Johns Herz ist solid . Johns Herz klingt gesund. Das Geräusch des Herzens klingt normal.
- 5.
- 6.
What remains to be answered is the question of how one is to deal with language in a generalized way without doing as Chomsky did, i.e., without reducing language to syntax. This is the question this book attempts to answer, as it has a critical bearing on the underlying linguistic competence of any MT system.
- 7.
In Sect. 8.4 The Hippocampus and Continual Learning of Chap. 8, we discuss a paper by Kumaran et al. (2016) that describes a previously unrecognized power for semantic generalization in the hippocampus, and the relevance that this unappreciated cerebral mechanism has for AI’s deep learning (and, in our view, more specifically for MT).
- 8.
Google Translate’s new neural MT (GNMT) translation of (2) into German shows some improvement: (2)(i) Die Klänge seines Herzens klingen ____. The French translation however is poorer: (2)(ii) Les sons de son son du son du cœur. See Postscript 6-B in Chap. 6 for the GNMT versions of Google SMT output shown here and there throughout this book.
- 9.
In late 2016 Microsoft also incorporated neural net technology into its Bing Translator. Bing Translator translations shown in this book are either from the SMT or NMT version, and are marked accordingly.
- 10.
Bing Translator’s new neural net system appears to have regressed in its translation of (2), both in the German and French: (2)(iii) Der Klang seines Herzens Klang klingt (2)(iv) Les sons de son cœur son.
References
Halliday MAK (2003) In: Webster J (ed) On language and linguistics: collected works of M.A.K. Halliday. Continuum, London/New York
Kumaran D, Hassabis D, McClelland JL (2016) What learning systems do intelligent agents need? Complementary learning systems theory updated. Trends Cogn Sci 20(7):512. https://doi.org/10.1016/j.tics.2016.05.004. Accessed 12 Jan. 2017
Optiz B, Friederici AD (2003) Interactions of the hippocampal system and the prefrontal cortex in learning language-like rules. NeuroImage 19:1730–1737. https://doi.org/10.1016/S1053-8119(03)00170-8. Accessed 25 Jan. 2016
Pinker S (1994) The language instinct. William Morrow and Co., New York
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2018 Springer International Publishing AG, part of Springer Nature
About this chapter
Cite this chapter
Scott, B. (2018). Introduction. In: Translation, Brains and the Computer. Machine Translation: Technologies and Applications, vol 2. Springer, Cham. https://doi.org/10.1007/978-3-319-76629-4_1
Download citation
DOI: https://doi.org/10.1007/978-3-319-76629-4_1
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-76628-7
Online ISBN: 978-3-319-76629-4
eBook Packages: Computer ScienceComputer Science (R0)