THSim v4.0.3: The Talking Heads simulation tool

THSim v4.0.3 is a simulation toolkit designed to study some aspects of language evolution computationally. It has been used by me to investigate various aspects of lexicon formation and grounding. In addition, previous versions were used in a course that I taught while I was in Maastricht and is used by students from a number of Universities that are part of the LOK project. The current version is freely downloadable to be used by researchers, students and other people who are interested.

The toolkit is written in Java using Sun's J2SE v1.4.1 and may work with later versions. It implements a simulation of the Talking Heads with a number of additional functionalities. These additional functionalities include four different types of language games, the iterated learning model and various parameters that can be used to study various aspects of language evolution.

Package Information

• Requirements: should work on any platform, Java Runtime Environment or Java 2 Platform, Standard Edition 1.4.1 (or higher)
• See a screenshot of THSim v4.0.3 thsim.jpg
• The source code is open.
• A rough guide and the paper THSim v3.2: The Talking Heads simulation tool are included, but the pdfs can be downloaded seperatedly here: roughguide THSim v3.2: The Talking Heads simulation tool.
• The API documentation of the tool is online, but it is also included in the package.
• Report any bugs, comments or problems to Paul Vogt

  DOWNLOAD THSim v4.0.3

New Release: THSim v4.0.3 (21 February 2005)

Differences with respect to earlier release:

• An essential bug-fix for HolisticAgent.java and HolisticAgent2.java. For the guessing game, the speaker evaluated the success with the wrong reference to the topic (internal meaning instead of reference). This is now fixed.

Release: THSim v4.0.2 (9 December 2004)

Differences with respect to earlier release:

• An essential bug-fix; occassionally, the contents of a string was not copied properly, allowing a hearer to interpret something into an empty string. (Note that the last two releases yield rather different results with respect to v4.0 in running a compositional agent.)

Release: THSim v4.0.1 (4 November 2004)

Differences with respect to earlier release:

• Some essential bug-repairs (especially with respect to guessing games for a compositional agent)
• Adaptation of the induction algorithm in CompositionalAgent2.java to remove an unjustified generalisation step.

Release: THSim v4.0 (August 2004)

Differences with respect to earlier releases:

• Two new types of agents:
  • Holistic agent with a hierarchical layering of conceptual spaces, as outlined in Vogt (2004). (Relates to my work on the emergence of Zipf's law.)
  • Compositional agent that can induce compositional structures, as outlined in Vogt (submitted). (Relates to my work on the emergence of compositionality.)
• Visualisation tool for inspecting the emerging languages (see Roughguide).
• Improved noise generator (see Roughguide)
• Variable memory sizes for lexicons and ontologies (see Roughguide).
• New testing methods for populations (see Roughguide and Vogt (submitted)).
• Various new forms of output (see Roughguide).

Release: THSim v3.2.2 (Spring 2004)

Improved version of THSim v3.2.1: Bug fixes:

• In v3.2.1 writing categories/meanings to the log-files could give values larger than 1 (in fact the values were between 0 and 1, but Java wrote a value of 3.5678 for values normally written in exponentials 3.5678e-6). This problem is fixed.
• The forgetting mechanism was dividing two integer values without casting them to doubles first, which caused the results be either 0 or 1 and elements with a usage-based score between 0 and 0.5 were thrown away rather than those with values smaller than 0.002. This problem is fixed.
• The memory size for the agents' lexicons is increased. If your computer now gives an out-of-memory error, try to call THSim by using: `java -Xmx256M THSim' where 256M may be smaller or larger as you wish.

New: mailing list

If you are interested in hearing about new releases or bug-fixes, please join the new mailing list by emailing Paul Vogt with in the subject: `THSim subscribe' (unsubscribe by mailing with subject `THSim unsubscribe'). For the moment this list will be moderated and handled by myself. It is intended to be a very low-trafic list. If you are interested in sending some of your experiences to the list, you can email me with in the subject: `distribute THSim: your subject'. I will then forward the mail to the list.

Some useful links:

• Paul Vogt's publications
• Language Evolution and Computation at Edinburgh
• Sony CSL Paris
• VUB AI Laboratory
• Language Evolution and Computation Bibliography and Recources

Some useful papers:

• Vogt, P. (2005) On the acquisition and evolution of compositional languages: Sparse input and the productive creativity of children. Adaptive Behavior 13(4) Special issue on Evolution and Acquisition of Language. In press.
• Vogt, P. (2005) The emergence of compositional structures in perceptually grounded language games. Artificial Intelligence 167(1-2): 206-242 (This paper describes the agents which can emerge compositional structures in language.)
• Vogt, P. (2004) Minimum cost and the emergence of the Zipf-Mandelbrot law. In Proceedings of ALife 9. The MIT Press. (This paper describes the agent with a hierarchically structured conceptual space.)
• Vogt, P. (2003). THSim v3.2: The Talking Heads simulation tool. In Proceedings of the 7th European Conference on Artificial Life, ECAL 2003 Springer-Verlag. (The paper describing THSim)
• Vogt, P. (2003) Grounded lexicon formation without explicit reference transfer: who's talking to who? In Proceedings of the 7th European Conference on Artificial Life, ECAL 2003 Springer-Verlag. (A paper reporting work done with THSim)
• Vogt, P. and Coumans, H. (2003) Investigating social interaction strategies for bootstrapping lexicon development Journal of Artificial Societies and Social Simulation 6(1). printable pdf version (a paper describing the three different language game models implemented in THSim).
• Kirby, Simon (2002) Natural Language from Artificial Life. Artificial Life. (an overview paper).
• Kirby, Simon and Hurford, James R (2001) The Emergence of Linguistic Structure: anOverview of the Iterated Learning Model, in Parisi, Domenico and Cangelosi, Angelo, Eds. ComputationalApproaches to the Evolution of Language and Communication. Springer Verlag, Berlin. (explaining the iterated learning model)
• Steels, Luc (1998) The synthetic modeling of language origins. Evolution of Communication 1(1) (an overview paper).
• Steels, L., Kaplan, F., McIntyre, A. and Van Looveren, J. Crucial factors in the origins of word-meaning. In Wray, A., editor, The Transition to Language, pages 252-271, Oxford University Press. Oxford, UK, 2002. (a paper describing the Talking Heads experiment.)
• Steels, L. and P. Vogt (1997) Grounding adaptive language games in robotic agents. In Harvey, P. and P. Husbands (eds.) (1997) Proceedings of the 4th European Conference on Artificial Life 97. (our first implementation of language games on real robots.)