Systems and methods for identifying entities and constraints in natural language input

A system comprising:

• at least one processor;
• and
• memory, encoding computer executable instructions that, when executed by the at least one processor, cause the at least one processor to:
• receive natural language input associated with a task;
• provide the natural language input to an entity detection model, the entity detection model comprising:
• an entity encoder that performs operations comprising:
• receiving the natural language input;
• and
• processing the natural language input to generate one or more entity encodings;
• an input encoder that performs operations comprising:
• receiving the natural language input;
• and
• processing the natural language input to determining one or more contextual encodings;
• an attention engine the performs operations comprising:
• receiving a first entity encoding from the entity encoder;
• receiving a first contextual encoding from the input encoder;
• and
• generating a first contextual entity encoding based at least upon the first entity encoding and the first contextual encoding:
• and
• a scoring engine performing actions comprising:
• receiving the first entity encoding and the first contextual entity encoding;
• and
• generating a probability score for a first entity associated with the first entity encoding based at least upon the first entity encoding and the first contextual entity encoding.

A computer storage medium encoding computer executable instructions that, when executed by at least one processor, perform a method comprising:

• receiving a natural language input;
• determining, using an entity encoder, one or more entities relevant to a task associated with the natural language input, the one or more entities comprising a first entity encoding;
• determining, using an input encoder, one or more contextual encodings based upon the natural language input, the one or more contextual encodings comprising at least a first contextual encoding;
• generating one or more contextual entity encodings based at least upon the one or more entities and one or more contextual encodings, wherein the one or more contextual entity encoding comprises at least a first contextual entity encoding, the first contextual entity encoding corresponding to the first entity;
• determining one or more probability scores for the one or more entities, the probability score representing a relevance of a particular entity of the one or more entities to the task associated with the natural language input, wherein determining the probability comprises:
• determining a first probability for a first entity, wherein the first probability is determined based upon the first entity encoding and the first contextual entity encoding;
• determining whether an entity of the one or more entities is associated with a negation cue based upon the natural language input;
• when the entity is associated with a negation cue, tagging the entity as negated;
• and
• providing the one or more entities and the one or more probability scores.