"... In PAGE 4: ... For these reasons, we prefer to refer to it as the GENIA taxonomy henceforth. Table1 gives some examples of terms in the GENIA corpus. Figure 1 shows the fast decreasing distribution of terms in the 35 classes.... ..."

"... In PAGE 7: ... To show the effectiveness of structured retrieval more clearly, we need test collection with (struc- tured) query and lists of relevant documents, and the tag-annotated documents, for example, tags repre- senting the relation between the technical terms such as event , or taggers that can annotate such tags. Table7 and 8 show that the retrieval time in- creases corresponding to the size of the document collection. The system is efficient enough for infor- mation retrieval for a rather small document set like GENIA corpus.... ..."

"... In PAGE 7: ... To show the effectiveness of structured retrieval more clearly, we need test collection with (struc- tured) query and lists of relevant documents, and the tag-annotated documents, for example, tags repre- senting the relation between the technical terms such as event , or taggers that can annotate such tags. Table7 and 8 show that the retrieval time in- creases corresponding to the size of the document collection. The system is efficient enough for infor- mation retrieval for a rather small document set like GENIA corpus.... ..."

"... In PAGE 7: ... To show the effectiveness of structured retrieval more clearly, we need test collection with (struc- tured) query and lists of relevant documents, and the tag-annotated documents, for example, tags repre- senting the relation between the technical terms such as event , or taggers that can annotate such tags. Table7 and 8 show that the retrieval time in- creases corresponding to the size of the document collection. The system is efficient enough for infor- mation retrieval for a rather small document set like GENIA corpus.... ..."

"... In PAGE 2: ...ble to public (Ver. 1.1).1 These 670 abstracts are a subset of more than 5,000 abstracts obtained by the query human AND blood cell AND transcription factor to the MEDLINE database. Table1 shows basic statistics of the GENIA corpus. Since the GE- NIA corpus is intended to be extensive, there exist 24 distinct named entity classes in the corpus.... ..."

"... In PAGE 2: ...ble to public (Ver. 1.1).1 These 670 abstracts are a subset of more than 5,000 abstracts obtained by the query human AND blood cell AND transcription factor to the MEDLINE database. Table1 shows basic statistics of the GENIA corpus. Since the GE- NIA corpus is intended to be extensive, there exist 24 distinct named entity classes in the corpus.... ..."

"... In PAGE 5: ... Figure 3 and 4 show the F-score transition according to the size of the training set and the training time among the given models respectively. Table2 and Table 3 show the parsing performance and the training cost obtained when using 900 abstracts of the GENIA treebank. Note that Figure 4 does not include the results of the Mixture method because only the method took too much training cost as shown in Table 3.... In PAGE 5: ... It should also be noted that training time in Figure 4 includes time required for both training and development tests. In Table2 , accuracies with models other than baseline showed the significant differences from baseline according to stratified shuffling test (Cohen, 1995) with p-value BO BCBMBCBH. In the rest of this section we analyze these exper- imental results by focusing mainly on the contribu- tion of re-training lexical entry assignment models.... ..."

"... In PAGE 5: ...lind tested, i.e., after we had the results for each model, we shuffled these results randomly for each query, and the shuffled results were judged by an ex- pert in the field of biomedicine whether they were relevant or not. Table5 shows the number of the results that were judged relevant in the top ten results. The results show that our model was superior to the exact and flat models for all queries.... ..."