Stance classification of multi-perspective consumer health information
View Patent ↗Method, system, and apparatus for automatic stance classification. Propositions can be collected that are relevant to a query. A classifier can classify the stance of each proposition based on whether the proposition supports the query, opposes the query, or is neutral with respect to the query in order to thereafter provide substantive data for decision making based on and extracted from the query. The stance can be classified based on, for example, an SVM-SC (SVM Based Stance Classification) approach and/or an NN-SC (Neural Network Stance Classification Approach).
1. A method for automatic stance classification, said method comprising:
collecting a plurality of propositions relevant to a query wherein each proposition among said plurality of propositions comprises labeled data;
subjecting said labeled data to a seed word set operation and a non-seed word set operation;
generating a seed lexicon after processing of said seed word set operation; and
classifying a stance of each proposition among said plurality of propositions using said seed lexicon and based on whether said each proposition including said labeled data supports said query, opposes said query or is neutral with respect to said query in order to thereafter provide substantive data for decision making based on and extracted from said query; and
wherein said stance comprises a medical stance based on word level stance vectors comprising vector representations of a stance context of text and said query comprises a medical query and wherein said word level stance vectors are generated via unsupervised emotion detection that includes a use of a fraction of annotated labeled data and stance vectors created for a seed set of NAVA (Noun, Adjective, Verb, and Adverb) words that form a seed lexicon.
2. The method of claim 1 wherein said stance is classified with a SVM (Support Vector Machine).
3. The method of claim 1 wherein said stance is classified based on an SVM-SC (SVM Based Stance Classification) approach.
4. The method of claim 1 wherein said stance is classified with a classifier comprising a neural network.
5. The method of claim 1 wherein said stance is classified based on an NN-SC (Neural Network Stance Classification) approach comprising a three layer neural network operable to predict a stance category.
6. The method of claim 5 wherein said stance is classified based on an SVM-SC (SVM Based Stance Classification) approach and on said NN-SC approach comprising said three layer neural network.
7. The method of claim 6 further comprising extracting a textual entailment between said query and said at least one proposition and inputting said textual entailment to a classifier for said classifying said stance.
8. The method of claim 1 further comprising extracting a textual entailment between said query and said at least one proposition and inputting said textual entailment to a classifier for said classifying said stance.
9. The method of claim 8 wherein said stance vectors for seed words are constructed by measuring an alignment of each NAVA seed word among said seed set of NAVA words towards three stance categories of: support, oppose, and neutral.
10. The method of claim 1 wherein said stance is attributable to a sentence that is in turn attributable to NAVA words in said sentence.
11. An apparatus for automatic stance classification, said apparatus comprising:
a classifier for classifying a stance of each proposition among a plurality of propositions relevant to a query and using a seed lexicon wherein each proposition among said plurality of propositions comprises labeled data, based on whether said each proposition supports said query, opposes said query, or is neutral with respect to said query in order to thereafter provide substantive data for decision making based on and extracted from said query;
wherein said labeled data is subject to a seed word set operation and a non-seed word set operation and said seed lexicon is generated after processing of said seed word set operation; and
wherein said stance comprises a medical stance based on word level stance vectors comprising vector representations of a stance context of text and said query comprises a medical query and wherein said word level stance vectors are generated via unsupervised emotion detection that includes a use of a fraction of annotated labeled data and stance vectors created for a seed set of NAVA (Noun, Adjective, Verb, and Adverb) words that form a seed lexicon.
12. The apparatus of claim 11 wherein said stance is classified by said classifier with a SVM (Support Vector Machine).
13. The apparatus of claim 11 wherein said stance is classified by said classifier based on an SVM-SC (SVM Based Stance Classification) approach.
14. The apparatus of claim 11 wherein said stance is classified by said classifier with a classifier comprising a neural network.
15. The apparatus of claim 11 wherein said stance is classified by said classifier based on an NN-SC (Neural Network Stance Classification Approach) comprising a three layer neural network.
16. The apparatus of claim 11 wherein a textual entailment is extracted between said query and said at least one proposition and inputting said textual entailment to a classifier for said classifying said stance.
17. The apparatus of claim 11 wherein said stance is attributable to a sentence that is in turn attributable to NAVA words in said sentence.
18. A system for automatic stance classification, said system comprising:
at least one processor; and
a non-transitory computer-usable medium embodying computer program code, said computer-usable medium capable of communicating with said at least one processor, said computer program code comprising instructions executable by said at least one processor and configured for:
collecting a plurality of propositions relevant to a query wherein each proposition among said plurality of propositions comprises labeled data;
subjecting said labeled data to a seed word set operation and a non-seed word set operation;
generating a seed lexicon after processing of said seed word set operation; and
classifying a stance of each proposition among said plurality of propositions using said seed lexicon and based on whether said each proposition including said labeled data supports said query, opposes said query or is neutral with respect to said query in order to thereafter provide substantive data for decision making based on and extracted from said query; and
creating annotated labeled data and stance vectors for a seed set of NAVA (Noun, Adjective, Verb, and Adverb) words that form a seed lexicon, wherein said stance comprises a medical stance based on word level stance vectors comprising vector representations of a stance context of text and said query comprises a medical query and wherein said word level stance vectors are generated via unsupervised emotion detection that includes a use of a fraction of the annotated labeled data and the stance vectors created for the seed set of the NAVA (Noun, Adjective, Verb, and Adverb) words that form the seed lexicon.
19. The system of claim 18 wherein said stance is classified with a SVM (Support Vector Machine).
20. The system of claim 18 wherein said stance is classified with a classifier comprising a neural network.