[Arxiv 1703.04368] Symbol Grounding via Chaining of Morphisms

Ben Goertzel is a vocal AGI advocate whose work I don't know much (nor the et al's), but he can answer the 'show me the code' probe with his opencog work -though I think still is no AGI.

In [1] I commented about pregroup grammar and word2vec (distributional) unification being seriously worked on in Clark and Coecke via tensor nets categorically, and in a comment to [2] ('Towards deep symbolic reinforcement learning'), I liked the use of unsupervised learning (there visual autoencoders) to nucleate concepts that grounded symbols.

Two independent comments, leaving behind the need of a vertical picture subsuming both lines of thought. I can't say this paper is a solution to this harmonization problem, but It seems to tackle some aspects of it.

[1] https://www.reddit.com/r/LanguageTechnology/comments/5kzn70/starting_in_semantic_parsing_is_ccg_still_relevant/dbs0zg1/

[2] https://news.ycombinator.com/item?id=12713436