Parallel treebanks

  1. The experiments with the small parallel treebank on Sophie's World have been rewarding and we think that this undertaking could evolve from a feasibility study into a testbed for exploiting the possibilities of parallel treebanks in more depth.
  2. Parallel treebanks require special interfaces for parallel tree displays. Lars Nygaard intends to look into this.
  3. Exploiting the parallelism requires the alignment of parallel trees below the sentence level (and above the word level). Yvonne Samuelsson is planning her D-uppsats project in this direction (for a DE-SV parallel treebank).
  4. Parallel treebanks could profit from work done on parallel grammar development. Koenraad de Smedt will contact the people in the ParGram project and will investigate whether there is an intersection of interests.
  5. Transfering information from a treebank in one language (e.g. EN) to a parallel language (e.g. Amharic) is an interesting endeavour. In particular when there are little resources and tools for the second language. Atelach Alemu (a GSLT student of Stockholm University's CS department) wants to look into this topic for a project in the Treebank Course. Martin contacted her and she is considering working on Sophie's World (rather than the Bible as she had intended), if there is an Amharic version of Sophie's World.
  6. In order to have more material to experiment with (and to find more interesting phenomena) we ask every site to enlarge their annotation of Sophie's World to cover the sentences of the first chapter. All sites should make sure that the first 50 sentences are sentence-aligned to the Norwegian original. And we (Lars?) will try to automatically align the additional sentences.