"... In PAGE 10: ... EU POLICY TOWARDS REGIONAL COMPETITION A broad range of EU policies explicitly or implicitly aim to avoid a race-to-the-bottom. As seen in Table1 , regional policy, fiscal harmonisation and social policy can be viewed from this perspective. For each of those policy fields, the table distinguishes between three regulatory strategies.... ..."

"... In PAGE 3: ... 5500 most frequent words (at least 20 times) of the personal datasets were selected. Table1 shows that in average 87% of words in personal datasets and about 74% of the union of all personal datasets are cov- ered by the BNC. Most words that are not covered by the BNC are names and specific terms, e.... In PAGE 3: ... We selected bigrams containing words that were covered by the BNC. Table1 shows that although all words are covered, their combinations may not, which are terminologies in a specific domain. For example, (1) in the personal documents: input fusion , modality conversion and multimodal dialogue in the field of multimodal system and shallow parsing , pattern matching and speech recognition in the field of NLP - they are considered as high frequent bigrams (at least 29 times), (2) in the e-mails: employee meeting , management report and retirement plans in corporate domain and intended recipient , conference call and video con- nection in communication field, and (3) in the chat-logs: immune system , orbital path , aftertime life and underground shelters .... ..."

"... In PAGE 3: ... Which and How Useful are the Data Sources? In this step, we compared the coverage of the common English words represented by BNC Database to all personal datasets. Table1 shows that the BNC Database can cover in average 87% for each context and about 74% for the union of all personal datasets. Most words that are not covered by the BNC database from personal documents are abbreviations, names and specific terms, such as: xml , website , lexicalized and synset in the field of computer science.... ..."

"... In PAGE 6: ... The GF(36m) multiplication operation in the Kwon-BGOS Tate Pairing algorithm is car- ried out with 18 GF(3m) multiplications, while the cubing operation in GF(36m) is carried out with 6 GF(3m) cubing operations [11]. Table1 shows the complexity of GF(36m) operations for standard implementation, plus the additional overhead required for robust implementation. It should be noted that these numbers are not dependent on the field size... In PAGE 7: ...2 Hardware and Software Implementa- tions The implementation of Algorithm 1 requires one multi- plication and one cubing operation in GF(36m). Following from Table1 , multiplication is realized with 18 base field multiplications and the cubing operation is realized with 6 base field cubing operations. The overhead caused by the robust approach can also be calculated from these two ta- bles.... ..."

"... In PAGE 7: ... In addi- tion, they provide processing re- sources that message agents can use to compare themselves with one another. PARTNER agents perceive, react to, and reinforce two fields in their environment, simultaneously coordinating their activi- ties and controlling the use of resources ( Table1 ). A semantic field imitates the use of pheromones in insect societies, and guides mutually relevant messages toward each other through the network.... ..."

"... In PAGE 4: ... The cloud systems can also be distinguished in the motion field, extracted from the Meteosat-6 Rapid Scans, where the lower layers move much slower and more towards the east in contrast to the fast moving high cloud layer, moving towards north-east. The wind correction for the ATSR2 cloud-top heights which as derived from the Meteosat-6 wind field, is summarized in Table3 . The along-track wind component in this case is quite substantial and leads to an overestimation of the CTH of up to 5.... ..."

"... In PAGE 29: ... Note that significance levels here refer to our hypotheses, which are directed (as opposed to the undirected signifi- cance levels given in Table 6). --- Insert Table6 and Table 7 about here --- The first box of Table 7, regarding attitudes towards using OSS, shows significant dif- ferences between testers and developers (H2[TD]), but no support for other hypotheses, indi- cating that the differences found to be significant in Table 5 are mostly accounted for by char- acteristics other than job function. For contributing to public OSS projects, we find significant differences between the job functions of developer/project manager and architect/tester, which supports H2[TD] and H5[AD] as well as H3[TP] and H6[AP].... ..."