"... In PAGE 18: ... Furthermore, in order to evaluate the contributions of the different features, we evaluate our system using the different combinations of the features. The results are shown in Table8 . From Table 8, several findings are concluded: 1) Based on the orthographic feature (Fo), our system achieves a basic level performance of 29.... In PAGE 18: ... The results are shown in Table 8. From Table8 , several findings are concluded: 1) Based on the orthographic feature (Fo), our system achieves a basic level performance of 29.... ..."

"... In PAGE 5: ... There are altogether 404 Medline abstracts, which are composed of 4260 sentences. The biomedical entity distribution is tabulated in Table1 . In order to fully investigate the behavior of different Meta recognition algorithms, two different training configurations were used in this work: i) 10 annotated documents for training and ii) 5 annotated documents for training.... ..."

"... In PAGE 11: ...Table 4: Baseline author named entity recognition of our evidence-based algorithm was extremely high, we decided that we could use the output of the evidence-based algorithm as the gold standard, and compared the results of the baseline algorithm to it. Results are shown in Table4 for author name identification (correctly tagging a capitalised word as part of a surname) and prefix identification (correctly tagging all prefix words that comprise part of the surname). The perfect recall score is a reflection of the fact that the gold standard data set is produced using capitalisation as a starting point, the same heuristic used by the baseline algorithm.... ..."

"... In PAGE 11: ...Table 4: Baseline author named entity recognition of our evidence-based algorithm was extremely high, we decided that we could use the output of the evidence-based algorithm as the gold standard, and compared the results of the baseline algorithm to it. Results are shown in Table4 for author name identification (correctly tagging a capitalised word as part of a surname) and prefix identification (correctly tagging all prefix words that comprise part of the surname). The perfect recall score is a reflection of the fact that the gold standard data set is produced using capitalisation as a starting point, the same heuristic used by the baseline algorithm.... ..."

"... In PAGE 13: ...rocessor running at 3.06 GHz). Processing time was 5 seconds. The precision for each data type is shown in Table4 , and recalls (i.e.... ..."

"... In PAGE 3: ... We adapted, tested and evaluated the named entity recognition and classification components during the Dumas project, and developed the corresponding components for Swedish and Finnish. Table2 shows the named entity classification scheme used in this project. The proper nouns are classified into categories: organisations, locations, individuals and unspecified name as a default category.... ..."

"... In PAGE 12: ... Named entity recognition is an established research topic25,26 and is used for various kinds of applications, such as information extraction, information retrieval, question answer- ing, and text mining. The named entity recognizer used in the PID tool extracts various types of named entities, as shown in Table3 . The types shown in bold in the table are used to identify PI.... ..."

"... In PAGE 10: ...Table 2: Alignment-based author named entity recognition the surname recogniser, we ran our citation extractor on our test corpus to output citations (including the author surname and year), and manually annotated the results; a positive result meant that the algorithm correctly identified all tokens in the author name. Performance for the named entity recognition task is shown in Table2 . For the precision and recall scores, a surname is counted as successfully recognised only if the name and all prefixes (and no additional tokens) are correctly extracted.... ..."