An article about Esperanto that also mentions lernu HERE

Thank you for sharing that article. Reading it made me smile. I am on day three of the Duolingo course. I didn't realize that I am supposed to use cxu within a sentence instead of se. I am a member of my local Esperanto society in Sacramento and we do practice using the language but I am still a long-term beginner / eternal komencanto . . .

I copied the article into Evernote for those times when people show interest and ask me about Esperanto.

sandyduggan:Thank you for sharing that article. Reading it made me smile. I am on day three of the Duolingo course. I didn't realize that I am supposed to use cxu within a sentence instead of se. I am a member of my local Esperanto society in Sacramento and we do practice using the language but I am still a long-term beginner / eternal komencanto . . .

I am also doing Duolingo to try and beef up my Esperanto. The interactive element that pulls you up is proving useful, as well as the way it makes you read and write the language.

Nice article, however it contains some not very accurate comparisons and assumptions.

Like its vastly more successful digital cousins — C++, HTML, Python — Esperanto is an artificial language, designed to have perfectly regular grammar, with none of the messy exceptions of natural tongues

First of all, the grammar of none of the four mentioned languages is regular (type-3), but the grammars of C++ and Python are context-free (type-2), and the grammar of HTML (not XHTML) is context-sensitive (type-1).

If the author used "regular grammar" for saying "a grammar free of exceptions", then this is a tautology. No grammar can possibly contain exceptions. For example a grammar for generating English past tenses (work->worked, go->went), would need more rules than a grammar for generating Esperanto past tenses (laboras->laboris, iras->iris), but it still would have no exceptions.

However, the biggest difference is that the grammars of C++, HTML and Python are unambiguous, while Esperanto's grammar is ambiguous.

With its logical structure, Smith thought Esperanto could work as a bridge language for translation, especially between two languages like Finnish and Turkish, that are unlikely to have a large overlapping dataset that a machine translation program could use.

Questionable at least. As Esperanto is as ambiguous as for instance English, there is no real gain in using Esperanto as an intermediate language for translation (compared to e.g. English). Also actual machine translation uses probabilistic models and as there are more texts which have been translated from Finnish or Turkish to English (or vice versa), English is indeed a better candidate than Esperanto.

Having no irregular[1] plurals or tense forms is nice, but irrelevant if you want to parse this language with a machine. Unambiguity is relevant if you want an exact solution and not a state vector over some event space.
Maybe it is possible to create a grammar that accepts all unambiguous Esperanto utterances and rejects all ambiguous (and all ungrammatical to its current grammar) Esperanto utterances. The resulting language would indeed be a hot candidate for an intermediate language for machine translation, but it wouldn't be Esperanto anymore.

----
[1] In the loose sense, i.e. having more than one rule for producing plurals or having more than one rule for producing a given tense.

nornen:Also actual machine translation uses probabilistic models and as there are more texts which have been translated from Finnish or Turkish to English (or vice versa), English is indeed a better candidate than Esperanto.

This is why Google Translate is generally miserable with Esperanto - it can't recognise even a simple, obvious compound word, if that word is not found in the database.

I don't see in principle why an automated translation could not be based on some other model better suited to Esperanto's particular features (e.g., ability to break down a compound into its constituent parts).

Syntactically, I would say Esperanto is slightly less ambiguous than English due to the presence of cases, adjective-noun agreement, and clear delineation of roles by correlatives ("that" in English can play at least three diverse roles, in Esperanto you have ke, tio, tiu). But machine translation is still overrated, imo. Aside from a specially designed language like Lojban, you're not going to get a nice-fitting solution.

Tempodivalse:But machine translation is still overrated, imo.

Why don't machine translators have a grammar checker? If you translate anything via google and then put it through a grammar checker, it is pretty darn accurate.

I sent google a request to add this feature but they probably have millions of emails to read. Maybe, I could design some type of tool myself that uses the google api and then it also provides a grammar correction.

Alkanadi:Why don't machine translators have a grammar checker? If you translate anything via google and then put it through a grammar checker, it is pretty darn accurate.

I sent google a request to add this feature but they probably have millions of emails to read. Maybe, I could design some type of tool myself that uses the google api and then it also provides a grammar correction.

IMO Google Translate's main weakness is the use of English as a pivot language. While I understand the benefits (lots of translations available from English to other languages to pick data from) in English a lot of info other languages need are lost in translation. This is evident when translating from one romance language to another. The simple, straightforward French sentence "Les infirmières semblaient très pressées" (The nurses seemed to be in a hurry) loses its gender agreement on "pressées" (in a hurry) when translated into Spanish "Las enfermeras parecían muy *presionados". (To make matter worse, the correct translation for "sembler pressé" is "parecer tener prisa", not "parecer presionado" )

The lack of agreement in number in English can also be a major obstacle when translating into languages that require it. Claude Piron provided the following real-life example sentence to examplify some problems a translating machine (and even a human translator) can face: "He could not agree with the amendments to the draft resolution proposed by the delegation of India". Should "proposed" agree with "amendments" or "draft resolution"? In Esperanto it would have to be either "proponitaj" or "proponita", accordingly.