Unsupervised focus-driven graph-based content extraction
AssigneeThe Regents of the University of Michigan
InventorsVeronica Perez-Rosas, Ashkan Kazemi, Rada Mihalcea
Life of this patent
Estimating the term…
Abstract
Systems and methods for processing natural language text using a graph obtain a natural language text and a query text, and parse that the natural language text into the plurality of text units, associating each with a graph node, and removing information leak text units from the plurality of text units. Connecting relationship between at least two of the remaining set of the plurality of text units are determined and associated with a graph edge between graph nodes. Based on the probabilistic relations between each graph node and the query text, graph node restart probabilities are determined for one or more of the graph nodes. The graph nodes that can be ranked.
Claims25
A method of processing natural language text using a graph, the method comprising:
- obtaining, at one or more processors, natural language input text comprising a plurality of text units and obtaining a query text indicating a focus for the graph;
- parsing, at the one or more processors, the natural language input text into the plurality of text units;
- removing, at the one or more processors, information leak text units from the plurality of text units to generate a remaining set of the plurality of text units;
- associating, at the one or more processors, an embedding of each of the remaining set of the plurality of text units with a graph node of a plurality of graph nodes;
- determining, at the one or more processors, at least one connecting relationship between at least two of the remaining set of the plurality of text units and associating the at least one connecting relationship with at least one graph edge connecting at least two of the plurality of graph nodes;
- associating, at the one or more processors, a probabilistic relation between each graph node and the query text, the probabilistic relation indicating a similarity between a respective graph node and the focus for the graph;
- constructing, at the one or more processors, a graph using the plurality of graph nodes that correspond to one of the remaining set of the plurality of text units and the at least one graph edge;
- based on the probabilistic relations between each graph node and the query text, (i) determining random restart probabilities for one or more of the plurality of graph nodes associated with each of the remaining set of the plurality of text units and (ii) generating a query text-based biased graph of the remaining set of the plurality of text units, the query text-based biased graph containing the one or more of the plurality of graph nodes;
- and
- ranking the one or more of the plurality of graph nodes of the query text-based biased graph by applying a graph-based ranking algorithm biased towards the random restart probabilities of the one or more of the plurality of graph nodes using the probabilistic relation associated with each respective graph node and the query text.
Dependencies are inferred from the claim text.