SegmentLinkingAVL

• Code for the experiments in the poster paper "Efficient navigation in learning materials: an empirical study on the linking process" published in AEID 2018
• Corpus with human judgements on topic segment similarity in the AVL domain

Running the system:

• Command:

  python ./tests/slice_tester.py <config.yaml> <tests_config.json>

• Configuration files examples can be found in AVL_trees_dataset_configs.zip, namely:

  <config.yaml> - avl.yaml

  <tests_config.json> - avl_tests.json

• After running the program the output will be in the file ./resources/tests/results.txt

Setting up the system:

1. Build a single file with all segments from all documents separated with the character sequence "==========" (see avl_segments.txt, in the AVL_trees_dataset_configs.zip, for an example)
2. Specify the path to the file in <config.yaml>, in the "input_generator" section, variable "input_directory" (see avl.yaml for an example)
3. Build a label sequence with the ground truth of the clustering of the segments. The label sequence must match the sequence of segments from step 1.
4. Specify the ground truth in <tests_config.json> file (see avl_tests.json for an example).