Training guidelines update

grobid-ner/doc/training-guidelines.md

@@ -1,19 +1,19 @@
-# Guidelines for annotation of NAmed Entities Recognition
+# Guidelines for annotation of Named Entities Recognition
 
-The creation of annotated corpus for Named Entities is the process of find the correct class of named entities for words based on the context. 
-Grobid-NER can automatically generate training data from text files ( [Link to Page] ), recognising the best named entities with the model currently used. 
+The creation of annotated corpus for Named Entities is the process of find the correct class of named entities for words based on the context.
+Grobid-NER can automatically generate training data from text files ( [Link to Page] ), recognising the best named entities with the model currently used.
 The goal of the annotator is to correct the generated entities by: (1) changing them, (2) extending them to the proximity tokens or (3) removing them.
 
 ### Format 
-        
-The format the training data is managed is the [CONLL 2003 format](http://www.cnts.ua.ac.be/conll2003/ner/), which is a 2 column tab separated file. 
+
+The format the training data is managed is the [CONLL 2003 format](http://www.cnts.ua.ac.be/conll2003/ner/), which is a 2 column tab separated file.
 The first column is the token, the second column is the class:
 ```
 token           B-CLASS
 token           CLASS
 ```
 
-The `B-` prefix is used to indicate the beginning of the class. This is important when the same class is repeated for two adiacent entities, **normally this is a very rare event**. 
+The `B-` prefix is used to indicate the beginning of the class. This is important when the same class is repeated for two adiacent entities, **normally this is a very rare event**.
 
 During training it's mandatory not to modify the token for any reason. Only the column of the class can be changed. 
 
@@ -26,20 +26,24 @@ Composed concept should be considered instead of simple concept. Usually extende
 
   1. the token _british_: 
     `_british_ is tagged with class NATIONAL`
-  but 
+    but 
     `_british_ referendum it's an EVENT`
     `_british_ government it's an INSTITUTION`
 
   2. composed token like European Union should be considered as a whole (please note that INSTITUTION could vary based on the context): 
 
+    ```
     European     B-INSTITUTION
-    Union           INSTITUTION 
+    Union        INSTITUTION 
+    ```
 
     instead the following case, which is not correct because it consider the two tokens separately: 
 
+    ```
     European     B-NATIONAL
-    Union           B-CONCEPT 
-
+    Union        B-CONCEPT 
+    ```
+
     3. more realworld case is when the entity precede an object that make the object specific: 
 
     European          B-EVENT