Entity Attribute List (ATR)
Salience configures entities either through files found the subdirectories in data/salience/entities
or data/user/salience/entities
or by setting entities through the "User Entity List" option. Fpr either case, you can associate entity attributes with their entities by using an entity attribute list. The following describes how to create and install entity attribute lists.
Entity Attribute Lists
An entry in an entity attribute list one of these two formats:
entity name<tab>attribute label<tab>attribute value<tab>entity label
entity name<tab>attribute label<tab>attribute value
Note that the entity label field is optional. If left unspecified, it will default to the directory containing the ATR file ("Company", "Place", etc.) in the case of named entities. For user entities, an unspecified entity label field remains undefined. Entity labels allow us to distinguish between entities that share the same name but are of different entity types, such as "Charles Schwab" the person versus "Charles Schwab" the company. In the case of user entities, an entity attribute with no entity label field specified will match any entity's normalized name regardless of its type or label. Here is an example of entity attributes defined for two types of drugs, "Tecfidera" and "Hizentra".
Tecfidera Generic Name Dimethyl Fumarate Drug
Tecfidera Manufacturer Biogen Drug
Tecfidera Method of Action Unknown Drug
Tecfidera Route of Administration Oral Drug
Hizentra Generic Name Immune Glubulin Subcutaneous Drug
Hizentra Manufacturer Bayer Drug
Hizentra Route of Administration Intramuscular Drug
If an entity can have many variant names ("Tecfidera", "dimethyl fumarate", "DMF"), it's important that you create a normalized form of the the name ("Tecfidera") that can be used in the first field of an attribute file entry. Otherwise you would be forced to replicate attribute entries for all variants of the name. Normalizing entity names is easily done through CDL files for named entities
Tecfidera Drug Tecfidera
dimethyl fumarate Drug Tecfidera
DMF Drug Tecfidera
or through user entity lists for user-defined entities
Tecfidera normalized=Tecfidera, label=Drug
dimethyl fumarate normalized=Tecfidera, label=Drug
DMF normalized=Tecfidera, label=Drug
Entity attribute lists are maintained in files ending with the extension ".atr", e.g. "drugs.atr". For named entities, their ATR files are placed in the same directory as their CDL files. For user-defined entities, their attributes are set up by setting the "User Entity List" option to their ATR file.
Updated about 1 year ago