https://d2ng68cf36kxsk.cloudfront.net/store/public/post/1266/post_images/Slide408.JPG">

The solution Google was proposing: “shared semantic representations (interlingua) between languages … , a universal interlingua.” Of the thirty-two citations in this paper,¹³⁵ most were recent – except, standing out oddly in the references, there was “Richens (1958).”^§AS2.2.1c Interlingua was back, now in hybrid relation to statistical methods.¹³⁶

The mechanism for generating the interlingua was to be algorithmic – consistent with Margaret Masterman’s insistence that the interlingua be generated empirically, and Halliday’s that language would have to do the work.^§AS2.2.1c But by 2017, the process was called “neural nets,”^§AS1.4.6c where the input and output might be a word or phrase, but the connecting “neuron” is a number representing an intermediate meaning.¹³⁷ In our grammar, numbers can reference instances and concepts, so an intermediate number might as well be a word.^§MS1.1.1d

Inspired by Halliday, we had been working for some time on the logic and processes for what we termed an “interlanguage.”¹³⁸ We had applied for a patent which, despite all the difficulties related to software as intellectual property, made it through the U.S. Patent Office.¹³⁹ Our focus was not machine translation, but the interoperability of partially overlapping ontologies. However, the underlying principle of creating an interlanguage, which we called Common Ground Markup Language, was the same. Whether it is machine translation or ontology alignment, the most practical way to achieve many-to-many mappings is to use a middle term which is a semantic abstraction. As soon as you do that, it suggests there could be a parallel term in any/all other possible languages or ontologies.

We wrote to Google Brain, noting the similarities with our work. One of the authors of the articles wrote back saying we should take this up with the patent department at Google, so we sent an email and never received a reply. Such is the futility of patents in practice for all but rich corporations, aside from their problems in principle.

---

¹³¹ https://medium.freecodecamp.org/the-mind-blowing-ai-announcement-from-google-that-you-probably-missed-2ffd31334805#.dytec365b

¹³² https://research.googleblog.com/2006/04/statistical-machine-translation-live.html

¹³³ Yonghui Wu, Mike Schuster, Zhifeng Chen, Quoc V. Le, and Mohammad Norouzi, 2016, “Google’s Neural Machine Translation System: Bridging the Gap between Human and Machine Translation,” arXiv: 1609.08144v2.

¹³⁴ https://research.googleblog.com/2016/09/a-neural-network-for-machine.html

¹³⁵ Melvin Johnson, Mike Schuster, Quoc V. Le, Maxim Krikun, Yonghui Wu, Zhifeng Chen, Nikhil Thorat, Fernanda Viégas, Martin Wattenberg, Greg Corrado, Macduff Hughes, and Jeffrey Dean, 2017, “Google’s Multilingual Neural Machine Translation System: Enabling Zero-Shot Translation,” arXiv.org: 1611.04558 [cs.CL], pp. 1–2, 15.

¹³⁶ Marta R. Costa-jussà and José A.R. Fonollosa, 2015, “Latest Trends in Hybrid Machine Translation and Its Applications,” Computer Speech and Language 32:3–10.

¹³⁷ Kyunghyun Cho, Bart van Merrienboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio, 2014, “Learning Phrase Representations Using Rnn Encoder-Decoder for Statistical Machine Translation,” arXiv: 1406.1078v3 [cs.CL].

¹³⁸ Bill Cope and Mary Kalantzis, 2011, “Creating an Interlanguage of the Social Web,” pp. 371–428 in Towards a Semantic Web: Connecting Knowledge in Academic Research, edited by B. Cope, M. Kalantzis, and L. Magee, Cambridge: Elsevier.

¹³⁹ Bill Cope, 2011, “Method for the Creation, Location and Formatting of Digital Content,” edited by U.S. Patent Office.

---

Kalantzis, Mary, and Bill Cope. 2020. Adding Sense: Context and Interest in a Grammar of Multimodal Meaning. Cambridge UK: Cambridge University Press. pp. 226–228. https://doi.org/10.1017/9781108862059

Google Rediscovers “Interlingua” §AS2.2.1d