Saturday, May 30, 2015

Relevant Logic Revisited

In 1992 I tried to read Anderson and Belnalp's book `Entailment, Vol. 1: The Logic of Relevance and Necessity' (1976) to help Maria Claudia Mere with her PhD thesis. 

She completed her work `Logicas Relevantes: formalismo e semantica'
in 1993, and I was a bit disappointed with relevant logics. I was hoping for prettier mathematics, cleaner proof-theory and clearer philosophical views. Now that my standards have changed a little, maybe I should re-read this work.

Wednesday, May 27, 2015

Maude for Textual Entailment

This is Maude Fealy, a star of silent movies, and no, I don't mean this Maude, but instead Maude the rewriting system developed by Messeguer and others, first at SRI then at Illinois.

They say in their Maude webpage "Maude is a high-performance reflective language and system supporting both equational and rewriting logic specification and programming for a wide range of applications."

Vivek Nigam and I have thought that it should be easy to use rewriting logic to prove formally entailment and contradiction of formulas arising from text, using the PARC's  Bridge system, as a black box. We wrote a preliminary note, and I talked about it at LSFA 2014.

The note `Towards a Rewriting Framework for Textual Entailment' is available, as are the Maude files used.

Transfer vs. Glue Semantics?

The short version of this post is Glue=principled, based on Linear Logic, Transfer=efficient, based on (linear) Rewriting.

But much, much more can and should be written.

The reason for the post is collect a few references. And to post a new(ish)  paper of Dick Crouch, which shows that Transfer semantics is totally independent of the grammatical theory that it originates from,  Lexical Functional Grammar (LFG). The paper is Transfer Semantics for the Clear Parser, it was presented at Natural Language and Computer Science (NLCS 2014), the workshop I organized with Larry Moss and Christian Retore as part of the Vienna Summer of Logic.

There are also two other papers on Transfer Semantics, these use LFG and the the XLE parser. The  papers can be found in this blog post. One shows how to construct transfer semantics from f-structures, the other how to rewrite transfer semantics into knowledge representation. Now I need to add the references for Glue Semantics.

Friday, May 15, 2015

Remembrance of things long past

 I first talked in Nuance on October 26th  2012 about my work on NL-based KR representations at PARC  and how I wanted to improve it. 
I first gave a talk similar to this in Feb 2010 at SRI (slides in slideshare, paper).
Then I tried to think a bit more along the lines of expressivity vs. complexity, following the work of Larry Moss
First at Stanford Workshop on Natural Logic, Proof Theory, and Computational Semantics, March 2011 and  then at RAIN in Austin, Texas, June 2012, when I called it Little Engines of Inference. Further I have written a paper for CommonSense2014, Contexts for Quantification.

Logic in the Pub, again?

Can repetition of non-information create new information?
My intuition was to say NO WAY! But look below...