Back to results
US 11,455,494 B2Granted
Automated building of expanded datasets for training of autonomous agents
AssigneeOracle International Corporation
InventorBoris Galitsky
Life of this patent
Estimating the term…
Abstract
Improved systems and methods for generating training data for classification models are disclosed. In an example, a training application accesses two fragments of text. The application represents each fragment of text as a parse thicket. The parse thickets jointly represent syntactic and discourse information. From the parse thickets, the application generalizes the text by identifying common entities or common rhetorical relations between parse thickets. The generalized text is added to a training data set, thereby increasing the coverage of the training set.
Claims20
A method for extending a training data set for a classification model, the method comprising:
- constructing, for a first body of text comprising fragments, a first syntactic tree and a first discourse tree;
- constructing, for a second body of text comprising fragments, a second syntactic tree and a second discourse tree, each discourse tree comprising terminal nodes and nonterminal nodes, wherein each terminal node is associated with one of the fragments of the respective body of text and each nonterminal node represents a rhetorical relationship between two of the fragments;
- merging the first syntactic tree and the first discourse tree into a first parse thicket;
- merging the second syntactic tree and the second discourse tree into a second parse thicket, wherein each parse thicket identifies one or more of the following:
- (i) two nouns in a different fragment represented by the respective syntactic trees, wherein each noun represents a common entity, or (ii) an additional rhetorical relation between fragments represented by the respective discourse trees;
- identifying, between the first parse thicket and the second parse thicket, one or more common components, wherein the one or more common components include one or more of (i) a common part of speech, (ii) a common communicative action, (iii) a common rhetorical relation, or (iv) a common verb signature;
- creating a generalized fragment comprising text that is associated with the one or more common components;
- and
- adding the generalized fragment to a training data set.
6 more elements
A system comprising:
- a non-transitory computer-readable medium storing computer-executable program instructions;
- and
- a processing device communicatively coupled to the non-transitory computer-readable medium for executing the computer-executable program instructions, wherein executing the computer-executable program instructions configures the processing device to perform operations comprising:
- constructing, for a first body of text comprising fragments, a first syntactic tree and a first discourse tree;
- constructing, for a second body of text comprising fragments, a second syntactic tree and a second discourse tree, each discourse tree comprising terminal nodes and nonterminal nodes, wherein each terminal node is associated with one of the fragments of the respective body of text and each nonterminal node represents a rhetorical relationship between two of the fragments;
- merging the first syntactic tree and the first discourse tree into a first parse thicket;
- merging the second syntactic tree and the second discourse tree into a second parse thicket, wherein each parse thicket identifies one or more of the following:
- (i) two nouns in a different fragment represented by the respective syntactic trees, wherein each noun represents a common entity, or (ii) an additional rhetorical relation between fragments represented by the respective discourse trees;
- identifying, between the first parse thicket and the second parse thicket, one or more common components, wherein the one or more common components include one or more of (i) a common part of speech, (ii) a common communicative action, (iii) a common rhetorical relation, or (iv) a common verb signature;
- creating a generalized fragment comprising text that is associated with the one or more common components;
- and
- adding the generalized fragment to a training data set.
9 more elements
A non-transitory computer-readable storage medium storing computer-executable program instructions, wherein when executed by a processing device, the program instructions cause the processing device to perform operations comprising:
- constructing, for a first body of text comprising fragments, a first syntactic tree and a first discourse tree;
- constructing, for a second body of text comprising fragments, a second syntactic tree and a second discourse tree, each discourse tree comprising terminal nodes and nonterminal nodes, wherein each terminal node is associated with one of the fragments of the respective body of text and each nonterminal node represents a rhetorical relationship between two of the fragments;
- merging the first syntactic tree and the first discourse tree into a first parse thicket;
- merging the second syntactic tree and the second discourse tree into a second parse thicket, wherein each parse thicket identifies one or more of the following:
- (i) two nouns in a different fragment represented by the respective syntactic trees, wherein each noun represents a common entity, or (ii) an additional rhetorical relation between fragments represented by the respective discourse trees;
- identifying, between the first parse thicket and the second parse thicket, one or more common components, wherein the one or more common components include one or more of (i) a common part of speech, (ii) a common communicative action, (iii) a common rhetorical relation, or (iv) a common verb signature;
- creating a generalized fragment comprising text that is associated with the one or more common components;
- and
- adding the generalized fragment to a training data set.
6 more elements
Dependencies are inferred from the claim text.