Open Source in CL

• University research should be open in principle:
  • ... the principle of openness in research - the principle of freedom of access by all interested persons to the underlying data, to the processes, and to the final results of research - is one of overriding importance.
    Openness in Research (Stanford University, Research Policy Handbook 2.6)
• Making resources Open Sources ensures that the researchers do not run into any institutional difficulties in distributing work or in continuing work at another institution (academic or otherwise). This is of most concern to those without permanent academic jobs.
• CL needs serious resources, which cannot be built and maintained by a single university group (or even industry group, unless the group is of reasonable size and persists over many years). Making resources Open Source is the only practical way of achieving flexible multi-group collaboration, especially if groups are on different continents.
• CL needs standards and historically the successful ones have been created bottom-up. Once a system is available and becomes reasonably widely used, other systems may well be built which use the same interfaces etc. At this point there is a de facto standard.
• Seeing one's work used by other groups is very rewarding. Students and other researchers are generally enthusiastic about contributing to work that is going to be used outside a small group. While this can be achieved without making the sofware Open Source, the question of individual rights can be very complex if the work is not open.
• Academics are evaluated by the impact that their work has. Open Source work generally has more impact.
• Licensing attempts are extremely time-consuming.

• Open Source systems cannot be licensed to companies for monetary gain. However, I think that the number of times an opportunity to make a worthwhile gain has been missed is massively outweighed by the amount of time that has been wasted in chasing licensing deals. Very few academic resources are ever licensed for worthwhile sums of money. Nearly all require the active participation of the developers to be of use. Given this, I think that it makes more sense to ask companies for consultancy and/or research support for the developers than to expect them to pay just to license technology.
• University personnel in research administration can become very unhappy about Open Source. It is very important to know your university's official policy and to be prepared to stand your ground. Remember that, although administrators like to think of academics as unworldly, the chances are that you know far more about commercial prospects in your area of research than they do, simply because they are not specialists. If you want to make a resource that you are developing Open Source, specify this on any connected grant applications.
• Some researchers are worried that others may take advantage of their work and compete for grants or get earlier publication of results. This is a possibility, but in most cases, it will be too difficult for other researchers to exploit work in this way without the active participation of the developers. Note that Open Source does not mean that the author loses copyright and obviously use of someone else's research work requires acknowledgement in any event.
• For Open Source to be of full benefit, researchers have to allow others to make modifications to their code and accept that they will sometimes screw up.

There are, however, some serious issues to think about in terms of ongoing support. A system is only really useful if there is a commitment to keep maintaining it, including supporting users who are making use of aspects of the system that you are no longer interested in yourself. Some users will have completely unrealistic expectations, others may have reasonable expectations that you simply cannot meet because of time pressure. Development of code and `lingware' does not have the same status as writing papers. A piece of code used by 1000 people does not seem to carry the same number of brownie points as a published paper cited by 10 people even though it may require many times more effort to produce. One possible (though admittedly partial) solution is to write decent documentation for the software and an accessible explanation of the underlying theory and to publish that as a book. (In this context, please see the CSLI CL series: we are very interested in publishing books with associated software.) Another (partial) solution is to link up with other people doing related work and agree to jointly maintain the software.