Multiple name strings that do not exactly match may refer to the same hospital for the following reasons.
- The strings refer to former names and aliases of the hospital.
- The strings follow different conventions, e.g., half/full-width characters, "
0..9"/"零..九", simplified/traditional Chinese characters. - The strings are structured differently and with different syntax, e.g.,
- the string may be a single name,
- the string may encode multiple aliases using a certain syntax,
- the string may encode annotations using a certain syntax.
- The strings may have typos.
Given two lists of strings: 1) names (external names) and 2) labels (internal names), the task is to label each string in names with a string in labels such that the two refer to the same hospital, or none if no match is found.
Given a list of strings names (external names), the task is to cluster strings that refer to the same hospital and output the most informative string in each cluster.
The command line tool Label.exe operates in either classification or clustering mode. In either mode, output file would be UTF-8 encoded.
Label.exe <names> <labels> <output>
| Parameter | Description |
|---|---|
names |
Path to the file containing the strings to be labeled, each in a separate line. |
labels |
Path to the file containing the label strings, each in a separate line. |
output |
Path to the file to write result to. The output is a two-column table with each row in a separate line and each column being tab separated. First column exactly matches names in order. Second column contains corresponding label from labels or an empty string if no match is found. |
Label.exe <names> <output>
| Parameter | Description |
|---|---|
names |
Path to the file containing the strings to be clustered, each in a separate line. |
output |
Path to the file to write result to. The output will contain the most informative string for each cluster, each in a separate line. |
If the tool is invoked with incorrect input, a GUI dialog will show up to help collect input. When LabelsFilePath is specified, the tool runs in classification mode and otherwise in clustering mode.
Classification mode supports concurrency and the tool runs with max(logical processor count - 1, 1) thread(s) by default. Concurrency can only be tuned with GUI. Clustering mode does not support concurrency and always runs single threaded.