Calculating Valence of Expressions within Documents for Searching a Document Index
AssigneeMicrosoft Corporation
InventorsLivia Polanyi, Martin Henk Van den Berg, Barney Pell
Life of this patent
Estimating the term…
Abstract
Tools and techniques related to calculating valence of expressions within documents. These tools may provide methods that include receiving input documents for processing, and extracting expressions from the documents for valence analysis, with scope relationships occurring between terms contained in the expressions. The methods may calculate calculating valences of the expressions, based on the scope relationships between terms in the expressions.
Claims20
A computer-readable storage medium comprising computer-executable instructions stored thereon that, when executed by a computer system, cause the computer system to:
- receive at least one document for processing;
- extract at least one expression from the document, the expression containing a scope relationship occurring between at least two terms contained in the expression;
- calculate a valence of the expression based on the scope relationship between terms as defined in the expression;
- and
- enter one or more facts in a semantic index representing the valence calculated for the expression, the semantic index comprising an inverted index mapping facts representing valence calculated for expressions in a plurality of documents to the document in which each expression occurs.
A computer-readable storage medium comprising computer-executable instructions stored thereon that, when executed by a computer system, cause the computer system to:
- calculate a representation of valence of an expression contained within an input document based on a scope relationship between at least two terms contained in the expression, the representation storing data representing an attitude of at least a first speaker toward a reported event;
- determine whether the representation expresses a neutral valence or a non-neutral valence;
- and
- upon determining the representation expresses a non-neutral valence, enter a fact for the representation in a semantic index comprising facts representing valence calculated for expressions in a plurality of documents, the fact comprising a role denoting the attitude of the first speaker toward the reported event.
A method comprising:
- receiving, at a computer executing a natural language engine, an input document;
- calculating, at the computer, a representation of valence of an expression included in the input document based on a scope relationship between at least two terms in the expression, the representation comprising data representing an attitude of at least a first speaker toward a reported event;
- determining, at the computer, whether the representation expresses a neutral valence or a non-neutral valence;
- and
- upon determining that the representation expresses a non-neutral valence, entering, by the computer, a fact for the representation in a semantic index comprising facts representing valence calculated for expressions in a plurality of documents, the fact comprising a role denoting the attitude of the first speaker toward the reported event.
Dependencies are inferred from the claim text.