Ordering a lexicon network for automatic disambiguation

摘要

Described systems and methods allow a computer system to employ a lexicon network for word sense disambiguation (WSD). In an exemplary embodiment, each node of the lexicon network represents a gloss of a lexicon entry, while an edge represents a lexical definition relationship between two glosses. The lexicon network is ordered prior to WSD, wherein ordering the lexicon network comprises arranging the nodes of the lexicon network in an ordered sequence, and removing a set of edges to eliminate loops. In some embodiments, the position of each node within the ordered sequence is determined according to a direction and a weight of an edge connected to the respective node. The weight may represent a semantic importance of the respective edge relative to other edges of the network.

主权项

1. A computer system comprising at least one processor configured to transform a lexicon network stored in memory, the lexicon network comprising a plurality of nodes and a plurality of edges, wherein:
each node of the lexicon network comprises a gloss of a lexicon entry of a natural language; each edge of the lexicon network comprises a direction and a weight; wherein creating the lexicon network comprises receiving an input indicative of a direction of an edge connecting a first node of the lexicon network to a second node of the lexicon network; wherein transforming the lexicon network comprises ordering the lexicon network by distributing the plurality of nodes among a plurality of levels, the plurality of levels arranged in an ordered sequence, so that all edges of the ordered lexicon network point in the same direction relative to the ordered sequence of levels; wherein ordering the lexicon network comprises assigning the first node of the lexicon network to a first level of the plurality of levels the first level selected according to the direction of the edge connecting the first and second nodes, and further according to a first penalty score determined for a first subset of edges of the lexicon network, the first subset of edges selected for removal from the lexicon network according to a tentative assignment of the first node to the first level, the first penalty score determined according to the weight of at least one edge of the first subset of edges, wherein the weight of the at least one of the first subset of edges is indicative of a semantic importance of the at least one of the first subset of edges compared to other edges of the lexicon network; and wherein the at least one processor is further configured to transmit at least part of the ordered lexicon network to a second computer system, wherein the second computer system is configured to:
receive electronic natural language text,perform, via a word sense disambiguation engine, word sense disambiguation on the received electronic natural language text according to the at least part of the ordered lexicon network, andoutput a disambiguation indicator based on the word sense disambiguation to a computer-based application.