Systems and methods for learning to extract relations from text via user feedback
According to some embodiments, a system and method are provided to extract relationships from unstructured text documents. The method comprises receiving a training set of sentences that comprise labeled objects and subjects for creating an initial relationship model. A set of unlabeled sentences may be received. Objects and subjects from the set of unlabeled sentences are determined based on the initial model and the determined objects and subjects from the set of unlabeled sentences are displayed to a user for feedback and approval. An indication of whether the determined objects and subjects from the set of unlabeled sentences are correct is received and the initial relationship model is updated based on the received indication.
1. A method to extract relationships from text, the method comprising:
receiving a training set of sentences comprising labeled objects and subjects for creating an initial relationship model;
receiving a set of unlabeled sentences;
determining, via a processor, objects and subjects from the set of unlabeled sentences based on the initial model;
displaying the determined objects and subjects from the set of unlabeled sentences to a user for feedback and approval;
receiving an indication of whether the determined objects and subjects from the set of unlabeled sentences are correct; and
updating the initial relationship model based on the received indication,
wherein creating the initial relationship model comprises executing a concept tagger against the training set of sentences to determine a domain and range of the sentences, wherein the domain and range is determined based upon a most frequently occurring subject-object semantic type pair.
2. The method of claim 1 , wherein the training set of sentences comprises less than thirty sentences.
3. The method of claim 1 , wherein creating the initial relationship model further comprises extracting word tokens from the training set of sentences wherein word tokens comprise words that are located between the subjects and the objects.
4. The method of claim 3 , wherein creating the initial relationship model further comprises determining extended word tokens comprising synonyms of the extracted word tokens.
5. The method of claim 4 , wherein the extended word tokens are weighted based on a similarity to the extracted word tokens.
6. A non-transitory computer-readable medium comprising instructions that when executed by a processor perform a method to extract relationships from text, the method comprising:
receiving a training set of sentences comprising labeled objects and subjects for creating an initial relationship model;
receiving a set of unlabeled sentences;
determining, via a processor, objects and subjects from the set of unlabeled sentences based on the initial model;
displaying the determined objects and subjects from the set of unlabeled sentences to a user for feedback and approval;
receiving an indication of whether the determined objects and subjects from the set of unlabeled sentences are correct; and
updating the initial relationship model based on the received indication,
wherein creating the initial relationship model comprises executing a concept tagger against the training set of sentences to determine a domain and range of the sentences, wherein the domain and range is determined based upon a most frequently occurring subject-object semantic type pair.
7. The medium of claim 6 , wherein the training set of sentences comprises less than thirty sentences.
8. The medium of claim 6 , wherein creating the initial relationship model further comprises extracting word tokens from the training set of sentences wherein word tokens comprise words that are located between the subjects and the objects.
9. The medium of claim 6 , wherein creating the initial relationship model further comprises determining extended word tokens comprising synonyms of the extracted word tokens.
10. The medium of claim 9 , wherein the extended word tokens are weighted based on a similarity to the extracted word tokens.
11. A system to determine an asset event, the system comprising:
a processor; and
a non-transitory computer-readable medium comprising instructions that when executed by the processor perform a method to extract relationships from text, the method comprising:
receiving a training set of sentences comprising labeled objects and subjects for creating an initial relationship model;
receiving a set of unlabeled sentences;
determining objects and subjects from the set of unlabeled sentences based on the initial model;
displaying the determined objects and subjects from the set of unlabeled sentences to a user for feedback and approval;
receiving an indication of whether the determined objects and subjects from the set of unlabeled sentences are correct; and
updating the initial relationship model based on the received indication,
wherein creating the initial relationship model comprises executing a concept tagger against the training set of sentences to determine a domain and range of the sentences, wherein the domain and range is determined based upon a most frequently occurring subject-object semantic type pair.
12. The system of claim 11 , wherein the training set of sentences comprises less than thirty sentences.
13. The system of claim 11 , wherein creating the initial relationship model further comprises extracting word tokens from the training set of sentences wherein word tokens comprise words that are located between the subjects and the objects.
14. The system of claim 13 , wherein creating the initial relationship model further comprises determining extended word tokens comprising synonyms of the extracted word tokens.
15. The system of claim 14 , wherein the extended word tokens are weighted based on a similarity to the extracted word tokens.