An entity is a data structured as a list of offers, fruits, type of documents, list of professions, type of contract, countries, cities ... The NER (Named Entity Recognition or Extraction of Named Entity) is the mechanism of automatic language processing to detect and extract them. The Prisme.ai platform has several NLUs, which gives you the opportunity to find the most suitable NER depending on the use case. It is possible to create your own custom entities.
How to create an entity?¶
- Click on the Entities section in Editor
- Add a name. An entity must start with a letter and end with a letter, only underscores are accepted.
- In the dictionary add an entity with its synonyms.
How to use an entity?¶
If you want to use them in natural language¶
In the training sentences of your intent, you will need to make the connection between the training sentences and your entities. To do this, simply annotate like the screenshot below. You can use entities for several reasons:
- You want to detect all possible occurrences of a particular word in an intent.
- Example with this chatbot which puts you in touch with producers where it is possible to detect different products such as vegetables, meat, cheese, strawberries, etc ...
- You want to know if a particular word (a variable) is present or not in a sentence.
- Entities influence the interpretation of intent. This helps to reduce the number of intents and avoid close intents. Refer to the confusion matrix to optimize the operation of your chatbot with entities.
If you want to use them as a stage in the route¶
- Add a parameter with the name of your entity as type.
- Put a question, check "mandatory", check "generate buttons from the elements of the entity", check if necessary "force the user to select a button"