Research

Hi there. This is the research page of Hendrik Maryns. From 2005 until 2008 I worked at the Eberhard Karls Universität Tübingen. Here is some information on what I did there. As you can see, it is rarely updated now.

On this page:

MonaSearch
Paul Hertz
Publications
Slides
Java APIs

MonaSearch

In my time at the SFB441 at the Universität Tübingen, I wrote a query tool for linguistic tree banks. The goal was to write a query tool which can handle queries of some formalism stronger than first order logic.

Since I am no longer employed at the university, development on MonaSearch has stalled. If anyone is interested in continuing development, feel free to contact me for getting started.

The approach is to use results of the 70s to convert an MSO formula into a tree automaton, and to use those tree automata to query the trees.

The Javadoc.
The project is now hosted in Launchpad, where you can get the source, file bugs and help translate.

Paul Hertz

I did some work for the department Logik und Sprachtheorie at the Wilhelm Schickard-Institut für Informatik, also at Universität Tübingen, on the work of Paul Hertz, a physician, logician and philosopher of the early 20th century, and the relevance of his logical work to current-day mathematical logic. Unfortunately, my ideas haven’t ripened into something publishable, and the world now asks other things from me. Ask if interested, though.

Publications

An article about the problems using the naive approach to implementing the transition function of the automata, using simple hash maps, has appeared in the proceedings of TLT’06 in Prague: On the Implementation of Tree Automata: Limitations of the Naive Approach (TeX source, BibTeX).
Remark 18 June 2007: Since then, I discovered a serious flaw in the code, such that the results in the article are not representative any more. Minimizing automata at the right times proved to be essential. Right now, memory is no longer the main problem.
A second approach was to implement the transition functions as MDDs. This proved mildly successful. An article about the algorithms used is in progress: Non-Deterministic Multi-valued Decision Diagrams with Default Value and their Application in Tree Automata (TeX source, BibTeX).
We finally discovered a way to interact with an existing automaton package, Mona. This approach worked leading to the release of MonaSearch. My colleague and I wrote an article about it: MonaSearch – A Tool for Querying Linguistic Treebanks (TeX source, BibTeX), in the proceedings of TLT’09 in Groningen.

My general bibliography file is used in compiling most of the above.

I am ever and always grateful for any, be it tiny, correction to any of the above files.

Slides

The presentation accompanying the first TLT article (never held, unfortunately ☹) (TeX source).
An introduction to Dynamic Syntax, held at the Seminar for linguistics at Tübingen University (in German) (TeX source).
A talk about the implementation issues of MonaSearch (in German) (TeX source). The same talk in English (TeX source).
A general talk about MonaSearch (TeX source).

The university logo and other common images are used in the slide shows.

Java APIs

In the process of writing the above tool, some modules came into existence which I believe might be interesting. These might be outdated, please contact me for newer versions if you would like to use them.

A formula API
The formula module of the query tool program. Designed to handle MSO formulas, but also useful for first order logic formulas.
Javadoc
Combinatorics utility classes
Some classes that perform combinatoric operations, such as computing all permutations of a list, or the Cartesian product of a multidimensional list.
Javadoc