| 摘要 |
PROBLEM TO BE SOLVED: To structuralize pieces of information about medical care efficiently.SOLUTION: The present invention is configured so that analysis view point candidates are extracted from text data such as medical science documents displayed on the Internet or the like periodically. Then, pieces of data such as a disease name, an agent, an examination, and a surgery and the like included in actual clinical data are subjected to linking to the analysis view point candidates. In the analysis view point candidates, there are clinically important candidates and not clinically important candidates, including extraction error. In order to discriminate them, weighting is performed with respect to the link. At first, when evidence level of the medical science documents from which the analysis view point candidates were extracted, is high, weighting is increased. In addition, in documents of respective disease names, cooccurrence degree of terms of the analysis view point candidates and terms about an agent, an examination, and a surgery is calculated. On the basis of the cooccurrence degree, the weight of the link is increased. |
