Supporting Information S1 - Using Typed Dependencies to Study and Recognise Conceptualisation Zones in Biomedical Literature

The experiments described in our manuscript have generated a substantial amount of data, which we cannot directly include in a document. Parts of this data are made available in the supporting_information archive, while the entire raw classification data is available on request. Below we briefly describe the content of this archive: Table set S1. Basic rule set statistics on the ART corpus, including the sizes of the rule sets for each ranking mechanism we have used, for each rhetorical type at different thresholds: 1%, 5%, 10% and 15%. Table set S2. Basic rule set statistics on the Wilbur corpus, including the sizes of the rule sets for each ranking mechanism we have used, for each rhetorical type at different thresholds: 1%, 5%, 10% and 15%. Table set S3. Corpus-driven inter-type comparison on the ART corpus, looking at the overlaps between the rule sets of the different rhetorical types. Table set S4. Corpus-driven inter-type comparison on the Wilbur corpus. Table set S5. Corpus-driven intra-type comparison on the ART corpus, looking at the overlaps between the different serialization types in the context of a rhetorical type. Table set S6. Corpus-driven intra-type comparison on the Wilbur corpus. Table set S7. Cross-corpus type comparison, listing the overlaps in rule sets between the rhetorical types of the ART and Wilbur corpora. (ZIP)