发明名称 Methods and systems of four valued analogical transformation operators used in natural language processing and other applications
摘要 A system for the dynamic encoding in a semantic network of both syntactic and semantic information into a common four valued logical notation. The encoding of new information being benign to prior syntactic constructions, tests for N conditionals in time O(C) and allows for the proper quantification of variables at each recursive step. The query/inference engine constructed from such an implementation is able to optimize short term memory for maximizing long term storage in the automaton. In a parallel context this can be viewed as optimizing communication and memory allocation between processes. The self-referencing system is capable of analogically extending knowledge from one knowledge source to another linearly. Disclosed embodiments include machine translation, text summarization, natural language speech recognition natural language.
申请公布号 US9575951(B2) 申请公布日期 2017.02.21
申请号 US201314016538 申请日期 2013.09.03
申请人 Midmore Roger 发明人 Midmore Roger;Klien Sheldon
分类号 G06N7/00;G06F17/27;G06N5/04;G06N5/02 主分类号 G06N7/00
代理机构 West & Associates, A PC 代理人 West & Associates, A PC ;West Stuart;Sluman Shaun
主权项 1. A machine implemented method of executing four-valued logic in a semantic network, the method comprising: defining a logic system having a NOT logical connective denoted as “” an AND logical connective denoted as “,” and an OR logical connective denoted as “v,” wherein false values are represented with an “F” symbol, true values are represented by a “T” symbol, undefined values are represented by a “U” symbol, and defined values are represented by a “D” symbol, such that for a property P false values are in a set {P}, true values are in a set {P}, undefined values are in a set { }, defined values are in a set {P, P}, wherein said logic system is proven complete and w-consistent by a set of statements comprising: F is T, T is F, U is D, D is U; FF is F, FT is F, FU is F, FD is F; TF is F, TT is T, TU is U, TD is D; UF is F, UT is U, UU is U, UD is F; DF is F, DT is D, DU is F, DD is D; FvF is F, FvT is T, FvU is U, FvD is D; TvF is T, TvT is T, TvU is T, TvD is T; UvF is U, UvT is T, UvU is U, UvD is T; and DvF is D, DvT is T, DvU is T, DvD is D; generating a semantic network representing digitally stored information, said semantic network comprising a plurality of object nodes and a plurality of relationships between said plurality of object nodes, by: providing a first vector and a second vector in digital memory for each of said plurality of object nodes, each of said first vector and said second vector being a dynamically allocated array having a plurality of index positions each with a size of one bit;linearly encoding semantic information for an object node by storing two bits that together encode a false value, true value, undefined value, or defined value, a first bit of said two bits being stored at an index position within the first vector associated with the object node, and a second bit of said two bits being stored at the same index position within the second vector associated with the object node; linearly encoding syntactic information associated with said plurality of relationships in said digital memory; and testing conditionals or quantifying variables in said semantic network during successive recursive steps in a predicate calculus using said logic system, the false values, true values, undefined values, and defined values in the semantic information associated with said plurality of object nodes, and the syntactic information associated with said plurality of relationships, and a phrase structure rewrite rule on said plurality of object node; and using a top/down, bottom/up parser to syntactically parse grammar within said semantic network, by marking subsets of said plurality of object nodes as {S} for start, {Np} for a noun phrase, {Vp} for verb a phrase and {Pp} for a prepositional phrase (predicate participant), and marking each of said plurality of object nodes as {N} for a noun, {V} for a verb, {det} for a determinant and {prep} for a preposition, wherein linear encoding of said semantic information and said syntactic information optimizes short term memory, maximizes long term storage, and optimizes communication and memory between different knowledge sources or processes.
地址 San Francisco CA US