Semantic extraction method and apparatus for natural language, and computer storage medium

A semantic extraction apparatus, the apparatus comprising:

• at least one memory configured to store program code;
• and
• at least one processor configured to read the program code and operate as instructed by the program code, the program code comprising:
• query code configured to cause the at least one processor to query, in a character library based on a single-character granularity, a character of a natural language that is input, to obtain first semantic vectors representing semantic descriptions of the character;
• first obtaining code configured to cause the at least one processor to obtain second semantic vectors that are obtained by using a neural network and that correspond to different permutations and combinations of the first semantic vectors;
• second obtaining code configured to cause the at least one processor to obtain candidate semantic intentions corresponding to the second semantic vectors and weight vectors corresponding to the candidate semantic intentions;
• and
• selection determining code configured to cause the at least one processor to select a candidate semantic intention as a final semantic intention of the natural language based on a preset condition, and obtain key information corresponding to the final semantic intention according to a weight vector corresponding to the final semantic intention.

A non-transitory computer storage medium storing instructions executable by at least one processor to perform:

• querying in a character library based on a single-character granularity, a character of a natural language that is input, to obtain first semantic vectors representing semantic descriptions of the character;
• obtaining candidate semantic intentions corresponding to second semantic vectors and weight vectors corresponding to the candidate semantic intentions, wherein the second semantic vectors are obtained by using a neural network and correspond to different permutations and combinations of the first semantic vectors;
• and
• selecting a candidate semantic intention as a final semantic intention of the natural language based on a preset condition, and obtaining key information corresponding to the final semantic intention according to a weight vector corresponding to the final semantic intention.