Semantic disambiguation using a statistical analysis
View Patent ↗A text containing a word is received by a computing device. The word is compared to inventory words in a sense inventory. The sense inventory comprises at least one inventory word and at least one concept corresponding to the at least one inventory word. Upon matching the word to an inventory word in the sense inventory, a concept for the word is identified by comparing each concept related to the inventory word to the word. The concept is assigned the word.
1. A method comprising:
receiving, by a computing device, an input natural language text including an input word;
searching a semantic register to identify a matching word corresponding to the input word, wherein the semantic register comprises a plurality of records, each record associating a word with a concept of a semantic class;
responsive to successfully identifying the matching word, identifying a first plurality of concepts associated with the matching word by the semantic register;
ranking a plurality of semantic classes associated with the identified first plurality of concepts according to a probability of the input word being associated with a respective semantic class;
selecting a pre-defined number of semantic classes having highest probabilities of the input word being associated with a respective semantic class;
iterating through a second plurality of concepts associated, by the semantic register, with the pre-defined number of semantic classes, to identify a concept corresponding to the input word; and
responsive to successfully identifying the concept, associating the identified concept with the input word.
2. The method of claim 1 , further comprising: responsive to failing to successfully identify the matching word corresponding to the input word, adding, to the semantic registry, the input word and a corresponding concept.
3. The method of claim 1 , wherein the semantic register comprises a semantic hierarchy including a plurality of semantic classes, and wherein a semantic class of the plurality of semantic classes comprises a deep model determining a semantic relationship between a parent of the semantic class and a child of the semantic class.
4. The method of claim 3 , wherein the semantic class is to inherit a deep model of a parent semantic class.
5. The method of claim 1 , wherein iterating through the plurality of concepts associated with the matching word is performed starting from a root of a semantic hierarchy associated with the semantic register.
6. The method of claim 1 , wherein the semantic registry is associated with a semantic hierarchy comprising a plurality of semantic structures, wherein a semantic structure of the plurality of semantic structures comprises a plurality of semantic classes, wherein a semantic class of the plurality of semantic classes comprises a plurality of words representing a plurality of instances of the semantic classes, and wherein an instance of the plurality of instances is associated with one or more semantic concepts.
7. The method of claim 1 , wherein identifying the concept corresponding to the input word further comprises:
identifying, in a parallel natural language text corresponding to the input natural language text, a parallel word corresponding to the input word; and
comparing a first context associated the input word to a second context associated with the parallel word in the parallel natural language text.
8. The method of claim 1 , wherein identifying the concept corresponding to the input word further comprises:
evaluating a classification function to produce a degree of association of a semantic class instance with the input word.
9. A system comprising:
a storage device; and
a processor operatively coupled to the storage device, the processor to:
receive an input natural language text including an input word;
search a semantic register to identify a matching word corresponding to the input word, wherein the semantic register comprises a plurality of records, each record associating a word with a concept of a semantic class;
responsive to successfully identifying the matching word, identify a first plurality of concepts associated with the matching word by the semantic register;
rank a plurality of semantic classes associated with the identified first plurality of concepts according to a probability of the input word being associated with a respective semantic class;
select a pre-defined number of semantic classes having highest probabilities of the input word being associated with a respective semantic class;
iterate through a second plurality of concepts associated, by the semantic register, with the pre-defined number of semantic classes, to identify a concept corresponding to the input word; and
responsive to successfully identifying the concept, associate the identified concept with the input word.
10. The system of claim 9 , wherein the processor is further to: responsive to failing to successfully identify the matching word corresponding to the input word inventory, adding a add, to the semantic registry, the input word and a corresponding concept.
11. The system of claim 9 , wherein the semantic register comprises a semantic hierarchy including a plurality of semantic classes, and wherein a semantic class of the plurality of semantic classes comprises a deep model determining a semantic relationship between a parent of the semantic class and a child of the semantic class.
12. The system of claim 11 , wherein the semantic class is to inherit a deep model of a parent semantic class.
13. The system of claim 9 , wherein iterating through the plurality of concepts associated with the matching word is performed starting from a root of a semantic hierarchy associated with the semantic register.
14. The system of claim 9 , wherein the semantic registry is associated with a semantic hierarchy comprising a plurality of semantic structures, wherein a semantic structure of the plurality of semantic structures comprises a plurality of semantic classes, wherein a semantic class of the plurality of semantic classes comprises a plurality of words representing a plurality of instances of the semantic classes, and wherein an instance of the plurality of instances is associated with one or more semantic concepts.
15. A computer-readable non-transitory storage medium comprising executable instructions to cause a processor to:
receive an input natural language text including an input word;
search a semantic register to identify a matching word corresponding to the input word, wherein the semantic register comprises a plurality of records, each record associating a word with a concept of a semantic class;
responsive to successfully identifying the matching word, identify a first plurality of concepts associated with the matching word by the semantic register;
ranking a plurality of semantic classes associated with the identified first plurality of concepts according to a probability of the input word being associated with a respective semantic class;
selecting a pre-defined number of semantic classes having highest probabilities of the input word being associated with a respective semantic class;
iterating through a second plurality of concepts associated, by the semantic register, with the identified plurality identified plurality of semantic classes, to identify a concept corresponding to the input word; and
responsive to successfully identifying the concept, associate the identified concept with the input word.
16. The computer-readable non-transitory storage medium of claim 15 , further comprising executable instructions to cause the processor to:
responsive to failing to successfully identify the matching word corresponding to the input word add, to the semantic registry, the input word and a corresponding concept.
17. The computer-readable non-transitory storage medium of claim 15 , wherein the semantic register comprises a semantic hierarchy including a plurality of semantic classes, and wherein a semantic class of the plurality of semantic classes comprises a deep model determining a semantic relationship between a parent of the semantic class and a child of the semantic class.
18. The computer-readable non-transitory storage medium of claim 17 , wherein the semantic class is to inherit a deep model of a parent semantic class.
19. The computer-readable non-transitory storage medium of claim 15 , wherein iterating through the plurality of concepts associated with the matching word is performed starting from a root of a semantic hierarchy associated with the semantic register.
20. The computer-readable non-transitory storage medium of claim 15 , wherein the semantic registry is associated with a semantic hierarchy comprising a plurality of semantic structures, wherein a semantic structure of the plurality of semantic structures comprises a plurality of semantic classes, wherein a semantic class of the plurality of semantic classes comprises a plurality of words representing a plurality of instances of the semantic classes, and wherein an instance of the plurality of instances is associated with one or more semantic concepts.