Vi le Sritoe pe la Cefli

that is, on the desk of the CEO of the Loglan Institute

Welcome to the doubtless somewhat untidy web presence of Randall Holmes, qua "chief executive officer" of the Loglan Institute (the title seems excessively grand, but that is what it is, officially). I prefer la Cefli, in Loglan :-)

The Loglan Institute (TLI) is the official organization which coordinates the development of the artifical human language Loglan. The main Loglan web site is here: that is the place to look for lots of general information and resources on the language. There is a mirror of this site (not always perfect) here at Boise State.

Here is a good place to start finding out what Loglan is, Loglan I (fourth edition) by James Cooke Brown.

Here is the current public agenda of the Institute (which I wrote in 2013): Public Agenda of TLI

Here is a new agenda document which is intended be essentially a complete report on the state of the language from my perspective as of Fall 2015. It has reached the status of a first draft. It includes a list of issues that I have worked on, an updated text of the reference grammar, a list of current proposals (including new ones), the complete PEG grammar (I am still updating annotations to it), and the trial.85 grammar. It is a substantial document:

Fall 2015 Report on the State of the Loglan Language.

There is a mailing list for people interested in Loglan. It is not an open list (due to the usual problems with spammers), but if you look here you will see how to apply to get on it.

On the rest of this page you will find my current Loglan projects and some general information about the Loglan language, its history, and its relationships with other artificial languages.

My current Loglan projects


I am currently developing a parser for Loglan, which has involved investigating and sometimes proposing modifications to all sorts of features of the language. The problem is that LIP (the old parser) is based on a now publicly available formal grammar which is provably unambiguous (as per our aims) on the level of grammar, but which is much more opaque (and demonstrably has errors) in the area of parsing individual words. It is legacy software, and I have no way to maintain it conveniently, and further it is now outdated. So I have written and am maintaining my own parser.

The parser uses PEG (Parsing Expression Grammars), a scheme for developing parsers (developed by Bryan Ford) described here. This differs from a use of a BNF grammar disambiguated using yacc which underlies LIP, and does raise questions about ambiguity (or more correctly, unintended parse) problems.

The parser has versions written in two different languages, Standard ML and Python. The Standard ML version is far faster, but getting and installing Python (and using it) is easier. I am still trying to figure out the bottlenecks which cause the Python version to take minutes instead of moments on some parses.

The ML version: Here find the PEG parser generator (written by me, updated 11/6/2015 with ability to log activity to a file) which I use to generate the provisional PEG parser for Loglan (written by me) implemented here (this is now the phonetic parser). Both of these are files to be run using the Moscow ML interpreter, which is free and runs on many platforms. The files are each documented at their front ends. The ML version is MUCH faster than the Python version; it is very much worth the extra effort to install Moscow ML. If someone can fix to work faster I would be grateful (I have done some fine-tuning but there are still bottlenecks that I do not understand in performance). I reiterate that the programming issue here is with the Python version of the PEG parser engine; it is not an issue with Loglan grammar as such.

The Python version: Here is a Python module for parsing Loglan, (now the latest phonetic version) which calls a PEG parsing Python module You can parse Loglan utterances using my grammar in this environment. All functions are working pretty well now, but of course there may be bugs. Please tell me about any problems that you find! The Python version has the besetting problem that it is slow: I believe this is because I used my ML programming style in writing it, and ML has optimizations beyond the imagination of Python for writing functional code. I have made some efforts to fine tune it but it is still very very slow on some examples.

An attempt at documentation: Here is a file containing an introduction to both of my Loglan parsers, intended to help you start using one (I used to suggest using the Python version, but the speed advantages of the ML version are becoming overwhelming: it's easy to get and set up Python and more involved to get ML; Python 3 will load the grammar much faster than Python 2, but the ML runs much faster even than Python 3 on large examples [moments versus minutes in the worst cases]).

Here is the Loglan PEG source in Ford's exact syntax in a text format generated by the ML parser engine. This might be useful if someone wants to try this with different PEG software.

I am no longer providing alternative parsers at this time. At the moment I have turned off pause/GU equivalence, though it would be fairly easy to turn it on, or bits of it. I am providing the fancy definition of gasent (requiring that the final GA terms component contain one argument or all of them) though once again this is easily converted to the older behavior by using code in a comment. The latest version allows one to pause almost freely inside NI words without grammatical risk; it does have the effect that some peculiar sentences parsed under earlier versions now parse quite differently. Further, one can now pause inside PA words. The reference grammar has been updated to reflect the new conditions.

Old test parsers with short scope PO

Here find the ML parser and here find the Python parser for the old test version of the parser with PO having short scope.

Here find the NB3 corpus in ML and Python versions for the old PO approach.

These documents are now quite out of date, as further revisions have happened and naturally will not be made here.

Sample texts

I set up the teaching corpus in Notebook 3 as input to the parser.

Here is a corpus of parsed Loglan sentences (run this Python file in the presence of the other two), this now contains the entire teaching corpus in NB3 modified to work with the current grammar.

Here is the corpus as a use file to run under ML; it runs in a minute or two! Here is a text log file containing the NB3 utterances and their parses.

Here is my Technical Report on the now completed process of parsing the NB3 corpus.

More sample material (these are also Python files to be run in the presence of the Python parser files): Lots of parsed examples from Loglan 1 ; Some examples of complex predicates and phonetic parses

Here is a translation of a small opening segment of the Old English poem Beowulf.

I am parsing the text of the Visit to Loglandia, the novel by Alex Leith which is the largest single Loglan text. It may be useful to note that my input files include comments on the revisions I had to make in the text to make it parse.

More installments to appear.

Dictionary Work (with Peter Hill)

Here are the current trial drafts of the English-to-Loglan and Loglan-to-English dictionaries with my latest trial words added. Here is my latest version of the database (for use with Peter Hill's dictionary program). I need to do dictionary work (or someone does :-)

work on Notebook 3

Here is a version of Notebook 3, one of our founding documents, which has been scanned and OCR'ed by Gleki Arxokuna and post-OCR proofread and edited by Peter Hill. This is the unmodified original. This is a version to which I am in the process of adding comments (I will be updating this as I work through it). These are Microsoft Word documents. I am also planning to use this to add the English translations to the source file for the NB3 corpus.

Reference Grammar project

Here is a zeroth draft of a complete (in intention) reference grammar for the language, written in English (with unavoidable TLI grammatical jargon but without PEG notation). It now covers essentially all of the grammar, but still lacks full discussion of semantics, and needs more examples, motivation and pragmatics of actually using the grammar to say what we want to say. The pointer to the reference grammar document has been replaced with a pointer to the Fall 2015 report, which contains the reference grammar and which is where it will continue to be updated.

La Keugru: the Loglan Academy

The Loglan Academy is a small group which makes official decisions about updates to the formal definition of the language.

Here is a pointer to Appendix H of Loglan I, which details the past work of the Academy.

Here is the official list of proposals before the Loglan Academy (la Keugru) This agenda document for the Loglan Academy now also contains my notes on the PEG parser, which now cover the entire phonology, lexicology, and grammar in detail. This document is intention superseded by my Fall 2015 Report, but I leave it here because there may be useful material in its older annotations on the PEG grammar. It will however no longer be updated: I will henceforth upgrade the copy of the PEG and annotations in the Fall 2015 Report.

It is important to note that my PEG grammar includes features which are still under consideration by the Academy. However, it should parse most Loglan sentences a learner will produce correctly, and indeed LIP will still do this in most cases.

Institutional and historical information

The Loglan project was founded by James Cooke Brown (JCB) in about 1955: it was originally intended as a tool to test the Sapir-Whorf hypothesis. To test the Sapir-Whorf hypothesis in the laboratory, he needed a language which was manageably small, but nonetheless a human language, and which was extreme in some respect, so that it might be expected to have Whorfian effects. He chose to make the language extremely logical; he also chose to make it syntactically unambiguous. The aim was that the structure of a Loglan sentence can always be determined precisely, and indeed that Loglan can be parsed by a computer.

At Brown's death in 2000, Robert A McIvor was CEO of the Institute; in 2008 when McIvor retired from this role he appointed me. I am not a member of the original group that built the language, though I had some personal contact with JCB and served as the logical consultant (la Lodtua) for a while in the 1990's; I am always interested to learn more about the earlier history of the project from those who were there. The Academy which made official decisions about the development of the language had made its last ruling in 1996 and was essentially defunct until I reconstituted it in 2013.

There is a sister language, Lojban, whose main web site is here, Ido to our Esperanto. The schism between the two languages happened about 1987 and culminated, alas, in a lawsuit, hinging on the fact that JCB claimed copyright on the word "Loglan" and on the Loglan vocabulary. He lost the point about the word, and the other language as a result has (or had originally) quite similar grammar and semantics and an unintelligible vocabulary (though even the vocabulary is formally similar in important ways). The two languages have drifted apart since, but are still quite closely related. Lojban has a much larger community and more activity. I think the original language still has something to offer, and the existence of two languages may provide some opportunities as well. Whatever the situation in the past, TLI is on good terms with the LLG which steers the development of Lojban, and though we do claim that essential materials of Loglan are the intellectual property of TLI and/or of their individual authors, we make them freely available to anyone with a private interest in learning the language or experimenting with modifications of the language (including Lojbanists!). Before undertaking any other use of the materials (and certainly before making any claim that your favorite modification is a feature of TLI Loglan!), please talk to us.

Here is another related language. And here is yet another!