R1 says CLOnE isn't really a controlled language. We disagree: CLOnE is a sublanguage of English with restricted syntax and a partly restricted lexicon. It is designed to be easy to learn, so it has a partly open lexicon and a looser syntax than many other controlled languages. R1 says this isn't "information extraction", that we are dealing only with a very specific aspect of knowledge management (ontology engineering), and that our title is misleading. We accept that it isn't IE in the usual sense (from free text) and we have changed the title to "CLOnE: Controlled Language for Ontology Editing". R1 says the evaluation needs more explanation. R1 and R3 say the meaning of Table 2 (confidence intervals) is unclear. R1 says Table 3 should have highlight the interesting figures (as discussed in the text), and R2 and R3 ask for an explanation of correlation coefficients. So we have added footnotes (with references) to explain confidence intervals and correlation coefficients, as well as another column to Table 3. We are constrained by the page limit to present only pithy summaries of the implementation and our evaluation measures, but we have added a reference to the project deliverable, which has more details. R1 complains that SUS scores are not comparable to the baseline, but has misunderstood (confusing the pre-test scores with the reference baseline for SUS scores in general). R2 figured this out but says it is confusing. We have rewritten the text accompanying Table 1 to make this clearer. R1 notes that our statistics suggest using the CL interface is not faster than using Protégé; we make no claims about speed! R1 says we should mention Orakel and provides a URL. We're aware of it but the original paper provides little data about the type of Controlled Language used, if any. Orakel it automatically generates its general purpose Lexicon using LTAG and the LopPar parser but no evaluation was provided. We could mention it just for completeness but there is no room in the page limit! R1 says the reference list is sloppy. I've cleaned it up as suggested. R2 says writing is awkward at times. I've edited sections 1, 2 & 3 to make them less stylistically inconsistent (that was an artefact of splicing various people's writing together) and shorter (to make room for other details the reviewers requested). R2 says we should show more about how the system works, and include a screenshot --- no room for a screenshot, but I've added a little more about how the system works. R2 says the task times are not directly reported --- no room in 14 pages, but we are adding a reference to the deliverable. R2 says we should explain why users liked our system in the evaluation. R2 says the last section should spend less space on projects and more on the general ways CLOnE can be applied. I've tried to tighten this section up to make applications clearer in the context. R3 says the experimental task is artificial and the tasks are too much like CL data entry. But we had to write the tasks simply so that users could carry them out in both tools in a reasonable amount of time without losing interest. We also provided a mini-manual that covered just enough Protégé to do the tasks without confusing additional information. R3 says the scope of the CL is not clear in the paper, but he then states exactly what it's scope is (taxonomic and property information) and understands that it is narrower than ACE's scope. 13 August. Discussed with Kalina. Edited and shortened "Related work". Changed all CLIE to CLOnE (less confusing). Fiddled with tables and reduced section headings to get more space, then added the list of syntactic rules.
