One of the common tagging schemes is BIO, in which individual tokens are classified (B) as being at the beginning of an entity, (I) being inside the boundaries of an entity, or (O) outside the boundaries of an entity.
The annotation process followed carefully defined annotation guidelines of rules for defining what actually was considered as a chemical entity and what not, as well as how to determine the individual mention boundaries of a chemical entity in text.
We think of a boundary whenever we think of an entity demarcated from its surroundings.
Delineation: The physical means used to mark a boundary or express the magnitude of an entity.
Entity Recognition and Linking Given a piece of text and a knowledge base, the task of entity recognition aims to recognize the boundary of entity mentions in the text, and the task of entity linking is to map the entity mentions to the corresponding entities in the knowledge base.
The upper boundary of named entity recognition performance is commonly measured by comparing independent annotations carried manually by human curators.
For the cohort-patient entity type, many disagreements are due to disagreements in the boundary of the entity annotation (e.g. 'Chinese' versus 'Chinese population', or 'MSI-H CRC' versus 'CRC' or 'seven cases' versus 'cases' alone).
Given the tokenized text, the NER task can be modeled as a sequence labeling problem by assigning each token to a label to determinate the boundaries of named entities, such as B = beginning of an entity, I = inside an entity, and O = outside of an entity (see examples in Table 2).
An activity node represents the activity of an entity or an entity pool but not the entities themselves.
For instance, the IOBESf model can correctly recognize the boundary of the chemical entity "α-phenyl-N-tert-butyl nitrone" using ten tokens after fine-grained tokenization.
