News and announcements

relex-1.3.1 released

Written for RelEx by linas on 2011-01-16

Version 1.3.1 of RelEx, the English Language dependency relation extractor, is now available.

This release consists of bug fixes, including the following:
 * Work around JWNL bug 3006600 for infinite loop --
   https://sourceforge.net/tracker/?func=detail&aid=3006600&group_id=33824&atid=409470
 * Add .n-f and .n-m linkgrammar gender tag recognition
 * Fix: anaphora resolution failure due to bad head-word identification.
 * Fix: missing frame concept error.
 * Fix: Add /usr/lib/jni to the library search path.
 * Add support for OpenNLP-1.5: launchpad bug 667949

Relex-1.3.0 is now available

Written for RelEx by linas on 2010-05-15

Version 1.3.0 of RelEx, an English-language dependency parser is now available. This is a fairly minor bug fix release; most notable changes are:
-- improved flexibility for entity tagging
-- fixes for date and money tagging
-- improved morphology handling for contractions e.g. 'd -> would, had
-- crash fix
-- minor frame output fixes.

RelEx Version 1.2.1 is now available

Written for RelEx by linas on 2009-11-03

This release provides a significant new feature: Penn-Treebank-compatible POS tagging. This feature is currently "experimental", in that it has not been evaluated for accuracy; it appears to work reasonably well. It is currently intended only for use with the Stanford parser compatibility mode (thus providing Stanford-compatible POS tagging).

Another notable change is the fallback of using the standard Java sentence splitter if the OpenNLP-based sentence splitter is not installed. There are a number of bug fixes. Some of these improve POS tagging, while others provide fixes for operation with the OpenCog embodiment code.

IMPORTANT: This release *requires* link-grammar-4.6.5 or newer!

RelEx version 1.2.0 now available

Written for RelEx by linas on 2009-10-12

Version 1.2.0 of RelEx, an English-language dependency parser is now available. Most notable in this release is the addition of a mode that provides output compatible with the Stanford dependency parser. This exercise
reinforced ideas of how RelEx differs from, and in many ways is more advanced than other dependency parsers -- RelEx provides more normalization, particularly for questions, which makes it a better starting
point for reasoning and question-answering systems.

The Stanford parser provides a defacto standard definition of dependency
relations used in parsing. Relex has a compatibility mode which
attempts to generate the same output as the Stanford parser. This is
an interesting thing to do, for two reasons:

   1) It provides a way of double-checking parse results,
   2) RelEx is nearly four times faster than the Stanford parser.

The compatibility mode is currently in "beta": it is almost fully
implemented, but still has a number of rough spots that need work.

RelEx version 1.0.0 now available

Written for RelEx by linas on 2009-06-17

There's no reason to delay version 1.0.0 any longer!

RelEx is a syntactic dependency extractor and semantic framing generator; it will parse English language sentences and return the dependency relationships between different parts of the sentence, and also provide semantic framing tags based on syntax and semantic categories.

The main RelEx website is at

    http://opencog.org/wiki/RelEx

It provides broader overview, as well as detailed documentation.
The source download and project management site is at

   https://launchpad.net/relex

New in version 1.0.0 (17 June 2009)
 * Add "ant help" and "ant javadoc" build targets
 * Upgrade to opennlp-1.4.3, maxent-2.5.2
 * Move the old, deprecated parser API to parser-old.
 * Update for link-grammar version 4.5.7
 * Fix _amod, _predadj, _advmod order in the frame rules.
 * Fix lemma form of verb 's ("That's what the man said")
 * Fix handling of word subscripts, when subscripts are multi-letter.
 * Fix OpenCog output for frame elements.
 * Improve POS tagging coverage.
 * Strip the [?], [!] and [~] suffixes from link-grammar-guessed words.
 * Add refined sentence history support for anaphora resolution.
 * Export anaphoric reference output to OpenCog.

610 of 12 results