Weighted system of expressing language information using a compact notation
First Claim
Patent Images
1. A method for expressing language information using compact notation, the method comprising:
- executing instructions embodied on a computer readable storage medium to direct a processor to define a compact notation for representing a large set of weighted strings via defined operators that include;
a family of first operators that are probability weighted word choice disjunction operators that allow for multiple word choices which are probabilistically weighted,a concatenation operator that forces strings within the language to be in an order stated by the operator,a probabilistically locked interleaving operator that probabilistically defines how arguments of the probabilistic operator are presented in a string;
executing instructions embodied on a computer readable storage medium to direct a processor to convert the expression to a WIDL graph which is a function of beginning and end vertices and edges, a set of vertices and edges, an edge labeling function, and a vertex ranking function.
2 Assignments
0 Petitions
Accused Products
Abstract
A special notation that extends the notion of IDL by weighted operators. The Weighted IDL or WIDL can be intersected with a language model, for example an n-gram language model or a syntax-based language model. The intersection is carried out by converting the IDL to a graph, and unfolding the graph in a way which maximizes its compactness.
92 Citations
36 Claims
-
1. A method for expressing language information using compact notation, the method comprising:
-
executing instructions embodied on a computer readable storage medium to direct a processor to define a compact notation for representing a large set of weighted strings via defined operators that include; a family of first operators that are probability weighted word choice disjunction operators that allow for multiple word choices which are probabilistically weighted, a concatenation operator that forces strings within the language to be in an order stated by the operator, a probabilistically locked interleaving operator that probabilistically defines how arguments of the probabilistic operator are presented in a string; executing instructions embodied on a computer readable storage medium to direct a processor to convert the expression to a WIDL graph which is a function of beginning and end vertices and edges, a set of vertices and edges, an edge labeling function, and a vertex ranking function. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14)
-
-
15. A method for expressing language information using compact notation, the method comprising:
-
executing instructions embodied on a computer readable storage medium to direct a processor to define a language in a notation that compactly represents a plurality of different expressions in the notation, and weights different alternatives according to a probabilistic interpretation; executing instructions embodied on a computer readable storage medium to direct a processor to carry out a text-to-text natural language application by intersecting the notation with a language model; wherein the intersecting comprises converting the notation into a graph indicative of the notation, and intersecting the graph with the language model; wherein the graph is intersected with an n- gram type language model by implementing the language model as a weighted finite state acceptor, mapping from the ideal graph to the finite state acceptor, and splitting the states to assign weights to transitions using a technique that is linear in the complexity of an input expression. - View Dependent Claims (16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31)
-
-
32. A system for expressing language information using compact notation, the system comprising:
-
a computer readable storage medium having instructions embodies thereon; and a processor for executing instructions embodied on the computer readable storage medium to direct the processor to define a compact notation for representing a large set of weighted strings via defined operators that include; a family of first operators that are probability weighted word choice disjunction operators that allow for multiple word choices which are probabilistically weighted; a concatenation operator that forces strings within the language to be in an order stated by the operator, a probabilistically locked interleaving operator that probabilistically defines how arguments of the probabilistic operator are presented in a string; wherein the processor also executes instructions embodied on the computer readable storage medium to direct the processor to intersect the set with a language model; wherein the set is intersected by, at least, converting the expression to a WIDL graph which is a function of beginning and end vertices and edges, a set of vertices and edges, an edge labeling function, and a vertex ranking function. - View Dependent Claims (33, 34, 35, 36)
-
Specification