CN102945230B - Natural language knowledge acquisition method based on semantic matching driving - Google Patents

Abstract

The utility model discloses a natural language knowledge acquisition method based on semantic matching driving. The natural language knowledge acquisition method comprises the following steps: (1) defining a semantic model for natural language processing; (2) defining a representing method of lexical semantic; (3) defining semantic matching relations among the lexicons; (4) defining a processing method of statements; and (5) transforming analysis results into knowledge points. According to the natural language knowledge acquisition method based on the semantic matching driving, an analysis scheme better accordant with semantic logics can be selected as the final analysis result from multiple grammatical analysis schemes according to semantic matching values by adopting semantic matching information to cooperate with common syntax rules in a small quantity of natural languages. Through the adoption of the method, the natural language statement analysis can be carried out and knowledge contained in the statements of the natural languages can be acquired. Experiments indicate that the method has higher feasibility.

Description

A kind of natural language knowledge acquisition method driven based on semantic matches

Technical field

The invention belongs to computing machine natural language understanding field, particularly a kind of natural language knowledge acquisition method driven based on semantic matches.

Background technology

In Knowledge Aggregation, a large amount of knowledge is lain in natural language statement, only achieves the automatic analysis of natural language statement, could effectively obtain the knowledge contained in statement.Therefore, natural language processing technique just becomes the Key technology of Knowledge Aggregation.

Natural language processing technique mainly contains rule-based method and Statistics-Based Method two kinds of thinkings, but these two kinds of methods all do not utilize semantic information fully, are difficult to obtain high-quality result.Therefore, researcher more and more payes attention to semantic effect, there is the method analyzing natural language based on lexical semantic knowledge bases such as Wordnet, hownet, framenet, but in these knowledge bases to comprise the description granularity of semantic information too thick, do not have to represent the level to morpheme, also not accurate enough.These shortcomings cause and are difficult to utilize them to form statement disposal route and the model of set of system.

Based in the Sentence analysis of semanteme, scholars have made research to a certain degree: Yao Tianshun and had studied natural language understanding based on semantics-driven, but the description of lexical semantic is more coarse, also abundant not to the utilization of semantic information, analytical approach is inadequate system also; HPSG method drives based on lexical information, but the information spinner of vocabulary will be used to description syntax rule, very few also not accurate enough to the description of semanteme, incompatible with the feature of Chinese.Document (Tom OH, Janyce W, Exploiting SemanticRole Resources for Preposition Disambiguation [J] .ComputationalLinguistics, 2008, 35 (2): 151-184.) have studied the prepositional phrase disambiguation combined with framenet by language material treebank, document (Patwardhan S, Banerjee S, Pedersen T.Usingmeasures of semantic relatedness for word sense disambiguation [C] .Proceedings of the 4th International Conference on Intelligent TextProcessing and Computational Linguistics (CICLING-03), Mexico City, 2003 241-257.) have studied use semantic relation carry out lexical semantic disambiguation.Although these researchs obtain certain achievement, also do not form model and method that set of system land productivity semantic information carries out natural language processing.

Form the semantic model of a complete natural language processing, more adequately must represent the semanteme of vocabulary, should may meet the requirement of semantic model specific to the level special talent of morpheme, and existing lexical semantic representation does not carry out deep research to this mostly.Case grammar uses " lattice " to describe Deep Semantics relation, but case grammar is only used to carry out grammatical analysis, seldom describes the semanteme of vocabulary with lattice.Mode describes one of action executing state key factor, in natural language, the accurate semanteme of a lot of vocabulary has contained the mode of certain movement concept in itself, and existing various lexical semantic representation does not consider mode, be therefore difficult to the semanteme describing concept exactly.Body strictly can represent Concept Semantic, and body generally uses description logic to represent all concepts; But in description logic, the semantic relation between concept is carried out treating of equality, special semantic interpretation and process are not carried out to the essential characteristic of the semantic relation of action concept.

Summary of the invention

In order to can automatic analysis natural language the knowledge that contains in obtaining, propose the semantic model of a natural language processing, this model use one to suppose axiom is to judge best grammatical analysis scheme; In order to meet the semantic expressiveness demand of this model, carrying out becoming privileged process to description logic, having enabled to represent that natural language vocabulary is semantic more suitably; Based on this model, propose a kind of natural language analysis method driven based on lexical semantic coupling; The method utilizes semantic matches information, in conjunction with syntax rule common in a small amount of natural language, according to semantic matches value, can in multiple grammatical analytical plan, select a kind of compare meet semantic logic analytical plan as final analysis result, natural language Sentence analysis can be carried out by the method and obtain wherein contained knowledge; Detailed process is as follows:

(1) semantic model of natural language processing is defined; Be mainly the semantic matching relationship between definition vocabulary; Target modified in the semanteme of definition vocabulary; The semantic matches value of definition statement;

(2) method for expressing of lexical semantic is defined:

(3) semantic matching relationship between vocabulary is defined: comprise the semantic matching relationship between noun-noun; Semantic matching relationship between noun-verb; The semantic matching relationship of noun-adjective; Semantic matching relationship arranged side by side; The semantic matching relationship of adverbial word; The semantic matching relationship of measure word; The semantic matching relationship of pronoun; And syntactic match relation;

(4) definition statement disposal route; This wherein relates generally to three levels and the syntax thereof of semantic structure; Obtain the thinking of best grammatical analysis scheme; Obtain best grammatical analysis scheme; The best semantic matches of simple clause is worth a few part;

(5) analysis result is converted into knowledge point.

Feature of the present invention:

The present invention proposes the semantic model of a natural language processing; In the model, propose the concept of semantic matches, axiom judges best grammatical analysis scheme to use one to suppose; Owing to using this model must have lexical semantic representation comparatively accurately, the present invention has done to become privileged process to the semantic expressiveness mode based on description logic, and its feature is:

(1) by " lattice " and " mode " as basic semantic relation

(2) modification, restriction give special semantic interpretation have been carried out to description logic, made it the lexical semantic being convenient to represent natural language

Based on this model, the present invention proposes a kind of method of the natural language processing based on semantic matches driving, the method has following feature:

(1) the bottom-up simple clause proposed based on semantic matches sums up method

(2) object of Sentence analysis and result obtain tacit knowledge in statement

Accompanying drawing explanation

Fig. 1 is the vocabulary definitions based on semantic relation

Fig. 2 is grammar rule schematic diagram

Embodiment

Specific implementation process of the present invention is as follows:

Step 1:

(1a) semantic matching relationship between 1-definition vocabulary is defined:

Definition 1: in lexical semantic knowledge base, any two notional word W _xand W _ybetween the inherent semantic relation that has, be called semantic matching relationship; With function match (W _x, W _y) representing its level of intimate, the value of function is exactly semantic matches value; Semantic matching relationship and concrete statement have nothing to do; If W _xwith W _ybetween there is no semantic matching relationship, then match (W is set _x, W _y)=MAX, MAX is a large constant.

(1b) 2 are defined: any notional word W in statement _i(removing predicate head) all semanteme is modified in another one notional word W _gi, claim W _giw _isemantic modification target.

(1c) 3 are defined: in specific grammatical analysis option A _iwhen, suppose that V is predicate head, S is the actor of V, and O is the recipient of V, Wi be a notional word in statement and! (W _i∈ S, V, O}), W _giw _isemanteme modify target, with function match (W _i, W _gi) represent its semantic matches value, so, the semantic matches value Value of whole statement _aican represent with formula (1):

${\mathrm{Vaule}}_{\mathrm{Ai}}=K_{\mathrm{SVO}}*(\mathrm{match}(S,V)+\mathrm{match}(O,V))+K_W*\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\mathrm{match}(W_i,W_{\mathrm{Gi}})---\left(1\right)$

The number (not comprising S, V, O) of target modified in the semanteme of S and O to be V, n be notional word, K _sVOand K _wifor weights coefficient. it should be noted that value less expression semantic matches degree is larger.

(1d) the best grammatical analysis axiom of axiom 1-is supposed: suppose that a statement has m kind grammatical analysis scheme, meet the grammatical analysis option A of semantic logic most _isatisfy condition: A _i=argmin (Value _ai), namely semantic matches is worth minimum grammatical analysis scheme is best grammatical analysis scheme.

Step 2:

(2a) define 4-key concept: build set before semantic base, without the need to carry out semantical definition and at Sentence analysis and reasoning time to be carried out the limited assemble of symbol of special semantic interpretation process by system, key concept quantity is few, uses C _wB={ W _b1, W _b2... W _bkrepresent, give tacit consent to each key concept W _biinherit in root concept.

(2b) define 5-lattice: the special key concept set representing the Deep Semantics relation of correlation circumstance between action genus and other things, use C _vC={ C ₁, C ₂... C _mrepresent; Such as " reason " is the lattice representing action reason.

(2c) define 6-mode: the special key concept set representing the executing state of action genus, use C _vM={ M ₁, M ₂... M _nrepresent; Such as " completing " is the mode that an expression action has been finished.

(2d) define the semantic relation that 7-is basic: build set before semantic base, unmodifiable and at Sentence analysis time need to carry out the semantic relation of special semantic interpretation process, comprise following semantic relation:

R _c(V, C, W) lattice semantic relation: in order to represent that action genus V and concept W has the Deep Semantics relation that lattice are C, wherein C ∈ C _vC;

R _m(V, M) modal semantic relation: in order to represent that the executing state of action genus V is M, wherein C ∈ C _vM;

R _aP(W, P) attribute semantic relation: in order to represent that concept P is the semantic relation of an attribute of noun genus W;

R _aS(W, S) state semantic relation: in order to represent that concept S is the semantic relation of a state of noun genus W;

R _d(W) semantic relation is inherited: can only use once in the semantical definition formula of concept, such as W ₁inherit in W ₂;

R _p(W ₁, W ₂) finite aggregate of part of semantic relation is described: concept W is described ₂semanteme be W ₁the part of semanteme;

R _w(W ₁, W ₂) finite aggregate of whole relation is described: concept W is described ₁semanteme comprises W ₂;

R _vS(V, W) represents that the actor of action V is the semantic relation of W, R _vO(V, W) represents that the recipient of action V is the semantic relation of W.

(2e) define 8-and expand semantic relation: building the semantic relation set newly defined in semantic base process, quantity is not limit. when Sentence analysis, all expansion semantic relations have unified disposal route, do not carry out special process. use R _r(W ₁, W ₂) represent, illustrate that the concept W be defined is W ₁and W ₂between a kind of semantic relation; In non-a defined formula, use R _r(W, W ₁, W ₂) represent W ₁and W ₂between there is the expansion semantic relation of W by name.；

(2f) based on basic description logic, to its concept definition regularly 1-7 limit and convert, become Concept Semantic define method herein.

Rule 1-concept definition rule:

1) if W _b1, W _b2key concept, R ₁, R ₂semantic relation, then new symbol

R ₂(W _b2) be concept;

2) if W _b1, W _b2concept, R ₁, R ₂semantic relation, then new symbol

R ₂(W _b2) be concept;

Do not have in rule 1 concept also, concept hands over, the service regeulations of measure word, their processing mode is shown in regular 2-2; Rule 2-3 is in order to the definition mode of concept is converted into one group of semantic relation, and requires that noun meets single principle inherited.

Rule 2-concept processing rule also: if W ₁, W ₂concept, W ₁, W ₂most recent co mmon ancestor concept be W _p, when new ideas W has W=W ₁uW ₂semantic time, due to single inheritance rules, W can be defined as W=R _d(W _p) I (R _p(W _p, W ₁) U R _p(W _p, W ₂)); Such as, parents=R _d(people) I (R _p(people, father) U R _p(people, mother)).

The processing rule that rule 3-concept is handed over: if W ₁, W ₂concept, when new ideas W has W=W ₁i W ₂semantic time, due to single inheritance rules, W can be defined as W=R _d(W ₁) I R _w(W ₁, W ₂) or W=R _d(W ₂) I R _w(W ₂, W ₁).

Rule 2 Sum fanction 3 only by concept also, concept hand over converted a kind of representation, only need specify that the certain semantic of these two kinds of representations is explained, this is done to solve the flexible means of one of many succession issues and bonding inheritance principles, act on the interface be similar in java, with the speed of the retrieval and coupling of accelerating concept.

In natural language, noun all can semantically directly or indirectly inherited in key concept, in order to the more clear semanteme representing noun accurately, adopts single principle inherited when requiring that semantic nouns represents, for having the concept inheriting semanteme, adopt regular 2 Sum fanction 3 to process more; Transitive verb semantically to represent noun to another noun do the action applied; Intransitive verb then represents the one change of noun self; Adjective is at the state semantically all represented between noun or noun or attribute; Adverbial word semantically all to represent the implementation status of action or mode and correlation circumstance or lattice; Therefore, the semanteme of all kinds of vocabulary in natural language can be represented by the satisfied mode of regular 4.

Rule 4-concept classification definition rule: in natural language, concept is classified by character and is expressed as noun, verb, adjective, adverbial word; Suppose the definition representing concept W with Def (W), Num (R, W) should meet following rule for the occurrence number of semantic relation R in definition, every class vocabulary definitions:

The list of noun is inherited: satisfy condition $(R_D\left(W_P\right)\⋐\mathrm{Def}\left(W\right))\∩(\mathrm{Num}(R_D,W)=1)$ Concept W;

Verb: satisfy condition $(R_{\mathrm{VS}}\left(W_1\right)\∪(R_{\mathrm{VO}}\left(W_2\right)\⋐\mathrm{Def}\left(W\right)\∩(\mathrm{Num}(R_{\mathrm{VS}},W_1)=1)$ $\∩(\mathrm{Num}(R_{\mathrm{VO}},W_2)=1);$

Adjective: satisfy condition $(R_{\mathrm{AP}}(W,W_1)\∪R_{\mathrm{AS}}(W,W_1))\⋐\mathrm{Def}\left(W\right)\∩$ $(\mathrm{Num}(R_{\mathrm{AP}},W)=1)\∩(\mathrm{Num}(R_{\mathrm{AS}},W)=1);$

Adverbial word: satisfy condition $(R_M(W_1,W)\⋐\mathrm{Def}\left(W\right))\∩(\mathrm{Num}(R_M,W)=1).$

The processing rule of rule 5-measure word: measure word generality quantifier and existential quantifier , do not obtain special treating, by as " number of times " one of lattice semantic relation value represents the number of times of action, the value as " quantity " attribute semantic relation represents the number of noun; Because this does not affect the syntactic structure analysis of natural language, make a concrete analysis of according to the value of " number of times " lattice of action and noun " quantity " attribute when reasoning, detailed disposal route slightly.

Rule 6-example arranges rule: when defined notion W, if the concept W in definition _ioccur m time, and n that occurs for this m time referring to semantic { S ₁, S ₂... S _n, then it is available that { W, W#1...W#n-1}, n that distinguishes W is semantic, and W#i may be interpreted as example when reasoning.

Rule 7-polysemant disposes rule: have a lot of polysemant in natural language, if polysemant W has n semantic { S ₁, S ₂... S _n, then for each concrete semantical definition concept, { W@1, W@2...W@n} distinguishes and represents this n different semanteme to define n concept altogether.

(2g) suppose axiom 2: inherit semantic relation and have unidirectional delivery, lower floor's concept is inherited Upper Concept and is had semantic relation.

Theorem 1. is according to inheritance R _d, all nouns form one tree.

According to regular 1-7 and definition 4-8, the semanteme of vocabulary in natural language can be defined; Suppose to represent semantic relation with a directed line segment, according to the definition Sum fanction of theorem 1 and vocabulary, the semanteme of vocabulary W can be set by noun in one group of directed line segment represent.

Step 3:

(3a) semantic matching relationship between noun-noun is defined:

Definition 9-conjunctive word collects: the set of all vocabulary comprised in nominal definition formula, uses C _rWrepresent; The conjunctive word of such as, noun W in accompanying drawing Fig. 1 collects

C _RW＝{W，W _P，W _r1，W _r2，W _r3，W _r4，W _r5，W _v，W _vc}

In analysis hereafter with ∝ for representing succession semantic relation, W _x∝ W _yrepresent W _xinherit in W _y, and specify W ∝ W.

(1) basic semantic matching relationship

Definition 10-direct semantics matching relationship: if vocabulary W _x, W _ymeet following condition, use symbol W _xw _yrepresent:

Condition: suppose W _yconjunctive word to collect be C _wY, then

$\∃W_Z\∩(W_Z\∈C_{\mathrm{WY}})\∩(W_X\∝W_Z).$

Work as W _xw _ytime, match (W _x, W _y)=K _t* d (W _x, W _z);

K _tfor matching relationship coefficient, according to the type of mated relation R, be set to different constants, have 1≤K _t≤ 3;

Such as: { W in accompanying drawing Fig. 1 _dr1, W _dvc, W _dr2, W _d2, W _dr3, W _dr4, W _dr5in each vocabulary with W, there is direct semantics matching relationship.

Definition 11-inherits semantic matching relationship: if vocabulary W _x, W _ymeet following condition, use symbol W _x//W _yrepresent:

Condition: $\∃W_Z\∩(W_X\backslash W_Z)\∩(W_Y\∝W_Z)$

Work as W _x//W _ytime, match (W _x, W _y)=match (W _x, W _z)+d (W _y, W _z);

Such as: { W in accompanying drawing Fig. 1 _dr1, W _dvc, W _dr2, W _d2, W _dr3, W _dr4, W _dr5and W _d1, W _d2there is succession semantic matching relationship;

Definition semantic distance function d (W _x, W _y): represent two vocabulary W with inheritance _x, W _ybetween succession number of times.

(2) semantic matching relationship is comprised

The explicit semantic relation of inclusion of definition 12-: if vocabulary W _x, W _ymeet following condition, use symbol W _x⊙ W _yrepresent;

Condition: there is concept W _z, satisfy condition

$\∃W_Z(R_W(W_Y,W_Z)\⋐\mathrm{Def}\left(W_Y\right))\∩(W_X\∝W_Z)$

Work as W _x⊙ W _ytime, there is match (W _x, W _y)=K _p* (d (W _x, W _z), K _pfor comprising matching relationship coefficient.

Definition 13-implicit semantic relation of inclusion: if vocabulary W _x, W _ymeet following condition, use symbol W _xzero W _yrepresent;

Condition: there is concept W _z, satisfy condition

Work as W _xzero W _ytime, match (W _x, W _y)=K _p* (d (W _z, W _y)).

Definition 14-comprises semantic matching relationship: if vocabulary W _x, W _ymeet following condition, use symbol W _x◎ W _yrepresent:

Condition:

Work as W _x◎ W _ytime, match (W _x, W _y)=

min{match(W _X，W _Z)+match(W _Z，W _Y)，match(W _X，W _Y)}

Theorem 3: when vocabulary WX, WY meet WX ◎ WY, WY has the semantic relation of WX;

(3b) semantic matching relationship between noun-verb

The semantic matching relationship of noun-verb can be divided into two classes:

1) SVO semantic matching relationship: the recipient of actor or action may be done in noun

2) lattice semantic matching relationship: noun and verb have lattice semantic matching relationship

Suppose that verb be gerund of executing in the definition of V, V is S ₀, be O by gerund ₀; Due to define time by S ₀be set to the top noun may implementing V, O ₀be set to the top noun bearing this action, so only have same S ₀or O ₀there is the noun S of certain relation and noun O just likely to perform an action V, namely form the semantic matches of SVO; SVO semantic matches has 6 kinds of situations, and it is worth available Value _sVOrepresent, computing formula is as follows:

Value _SVO＝match(S，S ₀)+match(O，O ₀)

The conventional SVO semantic matching relationship of definition 15-: satisfy condition (S ∝ S ₀) ∩ (O ∝ O ₀).

Definition 16-heavy duty SVO semantic matching relationship: satisfy condition:

$\begin{array}{c}((S\∝S_0)\∩(\∃R_{\mathrm{VO}}(R_{\mathrm{VO}}(V,W)\⋐\mathrm{Def}\left(S\right))\∩(O\∝W))\∪((O\∝O_0)\∩\\ (\∃R_{\mathrm{VS}}(R_{\mathrm{VS}}(V,W)\⋐\mathrm{Def}\left(O\right))\∩(S\∝W))\end{array}$

For noun S and O and verb V, do not meet in the definition of V SVO coupling time, and the defined declaration of S, O they meet SVO coupling;

Example: ring=R _d(ornaments) ∩ R _vS(wearing, people) ∩ R _vO(wearing, ornaments) ∩ R _c(wearing, position, hand), owing to containing R in " ring " _vS(wearing, people), so { people, wears, ring } forms heavily loaded SVO semantic matching relationship.

Definition 17-comprises SVO semantic matching relationship: satisfy condition ((S ◎ S ₀) ∩ (O ∝ O ₀)) ∪ ((S ∝ S ₀) ∩ (O ◎ O ₀));

Example: class=R _d(set) ∩ R _w(set, student), because " student " can " eat " " meal ", " student " is the part of " class ", so { class, eats, meal } forms overall SVO semantic matching relationship.

The similar SVO semantic matching relationship of definition 18-: satisfy condition ((S ∽ S ₀) ∩ (O ∝ O ₀)) ∪ ((S ∝ S ₀) ∩ (O ∽ O ₀)).

Definition 19-likens SVO semantic matching relationship: under the following conditions, may there is metaphor in conjecture statement:

Condition 1: conventional SVO semantic matching relationship, heavily loaded SVO semantic matching relationship can be met without any noun in whole statement, comprise SVO semantic matching relationship, similar SVO semantic matching relationship;

Condition 2: there is noun S or O in statement, meets! (S ∝ S ₀) ∩ (O ∝ O ₀), guess and liken S to S ₀;

Or condition 3: there is noun S or O in statement, meet (S ∝ S ₀) ∩! (O ∝ O ₀), guess and liken O to O ₀;

For metaphor SVO semantic matching relationship, Value _sVO=K _f* (match (S, W _p)+match (O, W _p))

K _ffor weights coefficient, W _ps and S ₀most recent co mmon ancestor.

Definition 20-lattice semantic matching relationship: for noun W and verb V, meet $\∃R_C(V,C,$ $W_C)\⋐\mathrm{Def}\left(V\right)\∩(W\∝W_C);$

Match (W, V)=K _c* d (W, W _c), K _cfor weights coefficient.

(3c) semantic matching relationship of noun-adjective

For adjective W _vAwith noun W _n, meet

$\∃W((R_{\mathrm{AS}}(W,W_1)\∪R_{\mathrm{AP}}(W,W_1))\⋐\mathrm{Def}\left(W_{\mathrm{VA}}\right)\∩(W_N\∝W)),$

Match (W _vA, W _n)=K _a* d (W _n, W), K _a=1, be weights coefficient;

(3d) semantic matching relationship arranged side by side

Semantic matching relationship arranged side by side only for the judgement of parallel construction in statement, to determine the scope of conjunction.

Definition 21-semantic similitude: because nominal definition have employed single method inherited, two noun W _x, W _yalthough there is no inheritance in definition, at semantically W _xbut may be W _yone, the concept be equivalent in description logic contains, and uses symbol W _x∽ W _yrepresent; Tableau algorithm in description logic can be improved, to judge Concept Semantic similarity relation.

Definition 22-noun semantic matching relationship arranged side by side: for two noun W _x, W _yavailable match (W _x, W _y)=K _t* (d (W _x, W _e)+d (W _y, W _e)) calculate a numerical value, as heuristic information, W _ew _x, W _ynearest common ancestor's node; When meeting W _x∽ W _ytime, also may be coordination.

Definition 23-verb semantic matching relationship arranged side by side: for two verb V _x, V _yavailable match (W _x, W _y)=K _t* (d (S _x0, S _y0)+d (O _x0, O _y0)) calculate a numerical value, as heuristic information, { S _x0, S _y0, O _x0, O _y0w _x, W _yactor in definition and by dynamic person.

(3e) semantic matching relationship between other class vocabulary

The semantic matching relationship of adverbial word: about adverbs modify adjective and adverbial word, also has very complicated situation, wouldn't discuss herein, suppose adverbial word can semantic matches in verb, adjective and adverbial word, specify match (W ₁, W ₂the semantic matching relationship of)=0. measure word: if lexicon should preserve the incidence relation of measure word and noun. measure word W can modification noun W _n, then match (W, W is specified _n)=0; Otherwise match (W, W _n)=MAX; The semantic matching relationship of pronoun: refer to relation according to pronoun, replaces to corresponding noun and processes by pronoun, such as " I " by " people " process.

(3f) syntactic match relation

Pay special attention to: various semantic matching relationship are above inherent, have nothing to do with concrete statement. in concrete statement, the vocabulary of some type possible is modified mutually, but semantic relation not inherent between vocabulary itself, just may there is a kind of grammatical phenomenon of semantic modified relationship in this statement, mainly comprise following two kinds of situations:

(1) modified relationship between uncommon part of speech: between verb-verb; Between adverbial word-noun; Between adjective-verb etc.; Such as " like swimming " " to be frank " etc.; These all belong to syntactic match relation, do not have inherent semantic matching relationship, just in statement, have phraseological modified relationship between vocabulary itself; In Sentence analysis process, its semantic matches is worth available match (W _x, W _y)=MAX/K _gcalculate, K _gtype weights (generally K _g< 1.5).

(2) the flexible conversion of parts of speech, such as adjective often can be applied flexibly as adverbial word, and this situation is not considered herein.

Step 4:

(4a) three levels and the syntax thereof of semantic structure are defined

Sentence analysis to be carried out according to semantic model herein, the statement abstract representation method of applicable semantic model must be had; Any statement is all formed through iteration by statement relatively simple for structure, and phrase is seen as an ingredient in statement; In order to meet the semantic analysis needs of semantic model, the semantic structure of statement can be divided into three levels according to the complexity of semantic structure and feature: simple sentence, special simple sentence, complex sentence.

Definition 24-simple sentence: only have a verb or adjective to make the statement C of predicate _s, available grammar G ₁carry out abstractdesription;

By the thought design grammar G of case grammar ₁, mentality of designing: suppose that V is predicate, S is the actor of V; O is the recipient of V, A _bit is preposition attribute; A _ait is postpositive attributive; P _dbe the adverbial modifier or complement, be equivalent to one group of lattice in case grammar; P _cit is the lattice content of; N is noun; N _pfor noun phrase;

Grammar G ₁in the mentality of designing of key rule as follows:

1) C _s→ P _da _bsA _ap _dvP _da _boA _ap _d, the appearance order of SVO has 10 kinds, and accompanying drawing Fig. 2 is one wherein;

2) S → n|SA _aa _bs, actor done in multiple vocabulary, as the S in Fig. 2;

3)P _D→P _C|P _DP _C

S, O, A _b, A _a, P _cin the service regeulations of the vocabulary such as preposition, conjunction, auxiliary word, number, measure word can write out easily.

Definition 25-special simple sentence: there is multiple verb or adjective, but semantically not comprise the statement of subordinate clause, available grammar G ₂carry out abstractdesription;

Grammar G ₂mentality of designing: guarantee can not produce on the basis of subordinate clause, to grammar G ₁the few rule of middle interpolation can generating grammar G ₂, mainly contain following 2 kinds of situations:

1) situation of predicate made in multiple verb or adjective

2) S, O, A made in verb or adjective _b, A _a, P _csituation

Grammar G ₂key be verb phrase V _vnoun phrase N directly can not be followed in front and back _p, namely can not there is N _p+ V _vor V _v+ N _p.

Definition 26-complex sentence: in grammar G ₂the regular N of middle interpolation _p→ C _s, form grammar G ₃; Because regular N _p→ C _sillustrate that a simple sentence or special simple sentence can make any composition in a complex sentence, achieve simple sentence recurrence, therefore grammar G ₃complex sentence can be described.

(4b) thinking of best grammatical analysis scheme is obtained

(1) lexical ambiguity digestion procedure

Suppose W ₁w ₂... W _klexical semantic number be respectively n ₁, n ₂... n _k, carry out fully intermeshing for each semanteme, result is { L ₁, L ₂... L _m, then M=n ₁* n ₂* ... * n _k, suppose one of them w _mn-th meaning of a word, then L _ic _sa sequence of words without lexical ambiguity; Exhaustive each { L in parsing process ₁, L ₂... L _manalysis result, select best L _ijust lexical ambiguity can be cleared up.

(2) analytical mathematics

According to axiom 1, obtain all grammatical analysis schemes, for each grammatical analysis option A _i, calculate A according to formula 1 _icorresponding semantic matches value, and select best grammatical analysis scheme.

The simple clause of definition 27-: meet grammar G in statement ₁or G ₂substring be simple clause;

Suppose that target axiom modified in axiom 4-semanteme: supposing that target modified in the semanteme of notional word W is W _gi, then for the simple clause C meeting semantic logic in statement _s, meet (W ∈ C _s) → (W _gi∈ C _s); For the attribute A of next-door neighbour S _b, meet (W ∈ A _b) → (W _gi∈ (A _b∪ S)), similar process can be done in the attribute of next-door neighbour O; For the adverbial modifier or complement P _d, meet (W ∈ P _d) → (W _gi∈ (P _d∪ V)).

According to semantic feature of modifying target, all grammatical analysis schemes can be divided into 2 layers:

1) simple clause's level grammatical analysis scheme;

2) the grammatical analysis scheme of simple clause inside.

(4c) best grammatical analysis scheme is obtained

(1) Rule of judgment of simple clause can be summed up

For statement C _s, carry out grammar G ₁, G ₂, G ₃cYK Algorithm Analysis, the substring s (i, j) meeting table 1 conditional is the simple clause that can sum up;

Table 1 can sum up the Rule of judgment of simple clause

(2) bottom-up simple clause sums up method

Available bottom-up simple clause sums up method and asks for best subordinate clause level grammatical analysis scheme, sees algorithm 4:

The simple clause of algorithm 1-sums up method:

1) for statement C _s, according to the Rule of judgment of table 1, find out the substring set { s that can sum up corresponding to simple clause ₁, s ₂... s _m;

2) for each clause s _i, calculate (algorithm 2) simple clause s _ibest semantic matches value, by s _ibe summed up as N _p, N is set _pend semantic;

3) C is made _sequal to sum up result, by the simple clause s in recursive procedure _ithe summation of best semantic matches value, carry out the recurrence of step 1-3;

4) there is best complete simple clause s corresponding to sentence semantic matches value _iscope and end order be best grammatical analysis scheme.

The best semantic matches value calculating simple clause is the key of algorithm, and concrete grammar is shown in algorithm 2;

In the algorithm, have employed the method for exhaustion when simple clause selects, theoretic best grammatical analysis scheme can be obtained; But the calculated amount of this method is comparatively large, not easily realizes; When clause's quantity can be returned too much, can only from can the good simple clause of several semantic matches degree be selected to carry out recursive search, to ask for suboptimum grammatical analysis scheme apodosis.

(3) semanteme is summed up

In algorithm 1, by simple clause C _sbe summed up as N _pafter, N _pdo not have semanteme, the semantic matches cannot carrying out next step calculates, and the mode of solution is as follows:

1) regulation carrys out N by summing up _pcan be matched with any vocabulary W, semantic matches value is:

Match (N _p, W) and=MAX/K _c(generally there is K _c> 1)

2) if N _pmake S or O of new goal clause, then can by N _psemanteme be set to former C _sin S or O.

(4d) the best semantic matches value of simple clause is obtained

Calculate the best semantic matches value of simple clause, according to axiom 1 and axiom 4, there is multiple grammatical analysis scheme simple clause inside, all grammatical analysis schemes must be obtained, for often kind of grammatical analysis scheme, the semanteme of its notional word is modified target and is determined, just can calculate the semantic matches value under this grammatical analysis scheme according to formula 1, the grammatical analysis scheme with minimum semantic matches value is exactly required analysis result;

The grammatical analysis scheme of simple clause inside can be divided into 3 layers: 1) SVO combination level; 2) A _a, P _d, A _blevel; 3) A _a, P _d, A _binner grammatical analysis scheme; Wherein best grammatical analysis scheme is selected by algorithm 2;

The best semantic matches value of the simple clause of algorithm 2-:

1) if simple clause is special simple sentence, all methods being summed up as simple sentence are found

2) for often kind of resolution principle, special simple sentence is summed up as simple sentence

3) for this simple sentence, all possible SVO combined method is found out

4) for often kind of SVO combined method, by C _sbe segmented into { L ₁, L ₂..L _n; If S or O is phrase, then carry out algorithm 3

5) each segmentation L _iinside A can be comprised at most _a, P _d, A _bthree partial contents, find out L _iin all A _a, P _d, A _bdivision methods

6) for often kind of A _a, P _d, A _bdivision methods, the means combined by syntax and semantics the matching analysis, are determined that target modified in the semanteme of each notional word, make A _a, P _d, A _bsemantic matches value minimum

7) ask for the semantic matches value of full sentence, select the minimum corresponding analytic process of semantic matches value as the grammatical analysis scheme of the best

Suppose for simple clause C _scarry out grammar G ₁the operation result of CYK algorithm, expression can generate the grammar symbol collection of substring s (i, j).

(1) the grammatical analysis scheme of SVO combination level

In simple sentence, suppose noun W ₁and W ₂meet SVO with verb V to mate, then { W ₁, V, W ₂a SVO combination, but S or O may be a phrase, when there is { W in sentence ₁, V, W ₃and { W ₂, V, W ₃sVO coupling, and V, W ₃not at W ₁and W ₂centre, and W ₁and W ₂middle substring s (m, n) meets: time, at W ₁+ s (m, n)+W ₃the phrase formed is S, in like manner can obtain longer S or O.

Algorithm 3-S or O segmentation:

1) obtain phrase S or O, find out in S all nouns meeting SVO coupling, be assumed to be { n ₁, n ₂..n _m}

2) according to { n ₁, n ₂..n _mphrase S is divided into m-1 section, according to regular S → n|SA _aa _bs is known each not for empty segmentation may comprise A _aa _b

(2) A _a, P _d, A _bthe division methods of level

Suppose segmentation L _isubstring be s (m, n), then meet p, q is grammatical A _a, P _d, A _bdivision methods, segmentation result is: A _a=s (m, p), P _d=s (p, q) A _b=s (q, n);

(3) A _a, P _d, A _binner best grammatical analysis scheme

Theorem 2:A _a, P _d, A _binner best grammatical analysis scheme works as A _a, P _d, A _bthe grammatical analysis scheme that interior each notional word is corresponding under having best semantic modification target conditions;

Definition 28-simple noun phrase: do not comprise verb and adjectival noun phrase is exactly simple noun phrase;

Theorem 3: the attribute A of simple sentence _aor A _bbest grammatical analysis many equivalents in the best grammatical analysis scheme of a simple noun phrase.

Prove: only comprise a verb due to simple sentence or predicate V made in adjective, therefore A _a, A _bin do not comprise verb and adjective, according to axiom 4, A _a, A _bgrammatical analysis many equivalents in simple noun phrase N _pgrammatical analysis scheme, N _p∈ { (A _b+ S), (A _a+ S), (A _b+ O), (A _a+ O) };

Simple noun phrase N _pdifferent grammatical analysis schemes only by conjunction, preposition, auxiliary word, measure word impact; The key of grammatical analysis is selected conjunction, preposition, auxiliary word, the scope of measure word and their end order;

The determination of A, scope: in conjunction, preposition, auxiliary word, measure word, suppose w _bfor preposition type, w _mfor preposition type, then its scope can be summed up as two kinds of forms 1) ..N _bn..N _b1... w _m..N _a1..N _am..; 2) w _b..N _bn..N _b1... w _m..N _a1..N _am...; Wherein { N _b1, N _b2... N _bnthe noun of first half in scope, { N _a1, N _a2... N _amit is the noun of latter half in scope;

Custom modified in backward semanteme according to Chinese, can at { N _a1, N _a2... N _amin find out grammatical and N _b1there is the noun N of best semantic matches value _ajas boundary after scope; The scope prezone of form 1 can be determined by similar method;

(2) determination of end order: conjunction, preposition, auxiliary word, measure word and scope thereof should be summed up by certain grammar rule, the available method of exhaustion obtains their the best end order; Generally, after simple sentence statement to carry out repeatedly segmentation, A _a, A _bin comprise conjunction, preposition, auxiliary word, measure word number n be generally less than 4, there is calculating feasibility.

Definition 29 (noun sequences): do not exist conjunction, preposition, auxiliary word, measure word simple noun phrase be noun sequence.

After conjunction, preposition, auxiliary word, measure word are all summed up, simple noun phrase has just been known as a noun sequence, and conjunction, preposition, auxiliary word, scope of quantifier inside may also exist one or two noun sequence in addition; In noun sequence, only noun affects semantic modified relationship, and custom modified in the backward semanteme according to Chinese, supposes that noun sequence is L _n=W ₁w ₂... W _m; Then determine L by semanteme _nin arbitrarily noun to modify target concrete grammar (approximation method) as follows:

Target modified in the best semanteme of algorithm 4-noun sequence:

Set C is set _wfor sky, for L _nin each noun W _iif, match (W _i, W _m) < MAX, by W _ibe added to C _w, do to operate as follows:

1) C is supposed _welement be W by order successively ₁-W ₂-...-W _n, be then done as follows: by L _nbe divided into n+1 section, the semanteme modification target arranging them is W _m, and recurrence is carried out to each section;

2) C is worked as _win when only having a noun, carry out step 3) and step 4);

3) forward direction modified relationship is set: for any segmentation, if there is W _xw _x+1... W _y-1w _y, satisfy condition: 1. any W _x+1... W _ybetween noun and W _yafter the semantic matches value of noun be MAX; 2. match (W _y, W _x) < MAX; Then W is set _ysemanteme modify target be W _x; Then W is set _x+1... W _y-1between the modification target W of noun _y;

4) if L _nin also have noun W _ydo not modify target, then arranging its modification target is W _y+1.

P _danalytical approach be similar to A _a, A _b, key carries out boundary line delimitation according to preposition, the content in preposition scope is also converted into a simple noun phrase; Due to comparatively loaded down with trivial details, do not discuss in detail here.

(4) disposal route of special simple sentence

Sum up all Non-Definite Verb/adjectives, statement is converted into simple sentence, select best end scheme, disposal route is as follows:

1) statement is carried out to the CYK algorithm of grammar G 2, find the verb or Adjective Phrases that likely do predicate, may kinds of schemes be had.

2) for each scheme, sum up remaining verb (or adjective), choose semantic matches and be worth minimum analytical plan.

Also need when summing up Non-Definite Verb or adjective to arrange to sum up semanteme, such as " beautiful necklace ", summing up semanteme is " necklace ", does not discuss in detail here.

Step 5:

According to the grammatical analysis result with best semantic matches value, simple sentence can be converted into a knowledge point, each simple clause of complex sentence is converted into knowledge point, and whole complex sentence is converted into one group of knowledge point.

Example: " moulding is that the green bronze ornaments of insect are welcome by Brazilian girl very much to statement." knowledge five-tuple, in table 2:

The more educated example of table 2 statement

After statement being converted into the knowledge point of depositing with structural data form, just can carry out various Intelligent Information Processing to these knowledge datas.

Above-described embodiment is only for the invention example is clearly described, and the restriction not to the invention embodiment.For those of ordinary skill in the field, can also make other changes in different forms on the basis of the above description.Here exhaustive without the need to also giving all embodiments.And thus the apparent change of amplifying out or variation be still among the protection domain of the invention claim.

Claims (1)

1., based on the natural language knowledge acquisition method that semantic matches drives, mainly comprise following process:

(1) semantic model of natural language processing is defined; Be mainly the semantic matching relationship between definition vocabulary; Target modified in the semanteme of definition vocabulary; The semantic matches value of definition statement;

(2) method for expressing of lexical semantic is defined:

(3) semantic matching relationship between vocabulary is defined; Be mainly the semantic matching relationship between noun-noun; Semantic matching relationship between noun-verb; The semantic matching relationship of noun-adjective; Semantic matching relationship arranged side by side; The semantic matching relationship of adverbial word; The semantic matching relationship of measure word; The semantic matching relationship of pronoun; And syntactic match relation;

(4) definition statement disposal route; This wherein relates generally to three levels and the syntax thereof of semantic structure, comprises the thinking obtaining best grammatical analysis scheme; Obtain best grammatical analysis scheme; The best semantic matches value of simple clause;

(5) analysis result is converted into knowledge point;

Wherein step (1) performs according to the following procedure:

(1a) 1 is defined---the semantic matching relationship between definition vocabulary:

In lexical semantic knowledge base, any two notional word W _xand W _ybetween there is inherent semantic relation, be called semantic matching relationship; With function match (W _x, W _y) representing its level of intimate, the value of function is exactly semantic matches value; Semantic matching relationship and concrete statement have nothing to do; If W _xwith W _ybetween there is no semantic matching relationship, then match (W is set _x, W _y)=MAX, MAX is a constant;

(1b) 2 are defined---target modified in the semanteme between definition vocabulary:

Any notional word W outside removing predicate head in statement _iequal semanteme is modified in another one notional word W _gi, claim W _giw _isemantic modification target;

(1c) 3 are defined---the matching value of definition statement:

In specific grammatical analysis option A _iwhen, suppose that V is predicate head, S is the actor of V, and O is the recipient of V, Wi be a notional word in statement and! (W _i∈ S, V, O}), W _giw _isemanteme modify target, with function match (W _i, W _gi) represent its semantic matches value, so, the semantic matches value Value of whole statement _aican represent with formula (1):

${\mathrm{Vaule}}_{\mathrm{Ai}}=K_{\mathrm{SVO}}*(\mathrm{match}(S,V)+\mathrm{match}(O,V))+K_W*\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{match}(W_i,W_{\mathrm{Gi}})---\left(1\right)$

The semanteme modification target of S and O is V, n is do not comprise S, the number of the notional word of V and O, K _sVOand K _wifor weights coefficient, it should be noted that value less expression semantic matches degree is larger;

(1d) axiom 1 is supposed---best grammatical analysis axiom: suppose that a statement has m kind grammatical analysis scheme, meet the grammatical analysis option A of semantic logic most _isatisfy condition: A _i=argmin (Value _ai), namely semantic matches is worth minimum grammatical analysis scheme is best grammatical analysis scheme;

Wherein step (2) performs according to the following procedure:

(2a) define 4---key concept: build set before semantic base, without the need to carry out semantical definition and at Sentence analysis and reasoning time to be carried out the limited assemble of symbol of special semantic interpretation process by system, key concept quantity is few, uses C _wB={ W _b1, W _b2w _bkrepresent, give tacit consent to each key concept W _biinherit in root concept;

(2b) 5 are defined---lattice: the special key concept set representing the Deep Semantics relation of correlation circumstance between action genus and other things, use C _vC={ C ₁, C ₂c _mrepresent;

(2c) 6 are defined---mode: the special key concept set representing the executing state of action genus, use C _vM={ M ₁, M ₂m _nrepresent;

(2d) define 7---basic semantic relation: build set before semantic base, unmodifiable and at Sentence analysis time need to carry out the semantic relation of special semantic interpretation process, comprise following semantic relation:

R _c(V, C, W) lattice semantic relation: in order to represent that action genus V and concept W has the Deep Semantics relation that lattice are C, wherein C ∈ C _vC;

R _m(V, M) modal semantic relation: in order to represent that the executing state of action genus V is M, wherein C ∈ C _vM;

R _aP(W, P) attribute semantic relation: in order to represent that concept P is the semantic relation of an attribute of noun genus W;

R _aS(W, S) state semantic relation: in order to represent that concept S is the semantic relation of a state of noun genus W;

R _d(W) semantic relation is inherited: can only use once in the semantical definition formula of concept;

R _p(W ₁, W ₂) finite aggregate of part of semantic relation is described: concept W is described ₂semanteme be W ₁the part of semanteme;

R _w(W ₁, W ₂) finite aggregate of whole relation is described: concept W is described ₁semanteme comprises W ₂;

R _vS(V, W) represents that the actor of action V is the semantic relation of W, R _vO(V, W) represents that the recipient of action V is the semantic relation of W;

(2e) 8 are defined---expansion semantic relation: building the semantic relation set newly defined in semantic base process, quantity is not limit; When Sentence analysis, all expansion semantic relations have unified disposal route, do not carry out special process; Use R _r(W ₁, W ₂) represent, illustrate that the concept W be defined is W ₁and W ₂between a kind of semantic relation; In non-a defined formula, use R _r(W, W ₁, W ₂) represent W ₁and W ₂between there is the expansion semantic relation of W by name;

(2f) based on basic description logic, to its concept definition regularly 1-7 limit and convert, become Concept Semantic define method herein;

Rule 1---concept definition rule:

1) if W _b1, W _b2key concept, R ₁, R ₂semantic relation, then new symbol $W=\&Not;W_{B1}|\&Not;R_1\left(W_{B1}\right)|\&Not;W_{B1}|\&Not;R_1\left(W_{B1}\right)|R_1\left(W_{B1}\right)\&cup;R_2\left(W_{B2}\right)|R_1\left(W_{B1}\right)\&cap;$ $R_2\left(W_{B2}\right)$ It is concept;

2) if W _b1, W _b2concept, R ₁, R ₂semantic relation, then new symbol $W=\&Not;W_{B1}|\&Not;R_1\left(W_{B1}\right)|\&Not;W_{B1}|\&Not;R_1\left(W_{B1}\right)|R_1\left(W_{B1}\right)\&cup;R_2\left(W_{B2}\right)|R_1\left(W_{B1}\right)\&cap;$ $R_2\left(W_{B2}\right)$ It is concept;

Do not have in rule 1 concept also, concept hands over, the service regeulations of measure word, their processing mode is shown in regular 2-3; Rule 2-3 is in order to the definition mode of concept is converted into one group of semantic relation, and requires that noun meets single principle inherited;

Rule 2---concept processing rule also: if W ₁, W ₂concept, W ₁, W ₂most recent co mmon ancestor concept be W _p, when new ideas W has W=W ₁∪ W ₂semantic time, due to single inheritance rules, W can be defined as W=R _d(W _p) ∩ (R _p(W _p, W ₁) ∪ R _p(W _p, W ₂));

The processing rule that rule 3---concept is handed over: if W ₁, W ₂concept, when new ideas W has W=W ₁∩ W ₂semantic time, due to single inheritance rules, W can be defined as W=R _d(W ₁) ∩ R _w(W ₁, W ₂) or W=R _d(W ₂) ∩ R _w(W ₂, W ₁)

Rule 4---concept classification definition rule: in natural language, concept is classified by character and is expressed as noun, verb, adjective and adverbial word; Suppose the definition representing concept W with Def (W), Num (R, W) should meet following rule for the occurrence number of semantic relation R in definition, every class vocabulary definitions:

The list of noun is inherited: satisfy condition $(R_D\left(W_P\right)\&Subset;\mathrm{Def}\left(W\right))\&cap;(\mathrm{Num}(R_D,W)=1)$ Concept W;

Verb: satisfy condition $(R_{\mathrm{VS}}\left(W_1\right)\&cup;R_{\mathrm{VO}}\left(W_2\right))\&Subset;\mathrm{Def}\left(W\right)\&cap;(\mathrm{Num}(R_{\mathrm{VS}},W_1)=1)$ $\&cap;(\mathrm{Num}(R_{\mathrm{VO}},W_2)=1);$

Adjective: satisfy condition $(\mathrm{Num}(R_{\mathrm{AP}},W)=1)\&cap;(\mathrm{Num}(R_{\mathrm{AS}},W)=1);$

Adverbial word: satisfy condition $(R_M(W_1,W)\&Subset;\mathrm{Def}\left(W\right))\&cap;(\mathrm{Num}(R_M,W)=1);$

Rule 5---the processing rule of measure word: measure word does not obtain special treating, by as " number of times " one of lattice semantic relation value represents the number of times of action, the value as " quantity " attribute semantic relation represents the number of noun;

Rule 6---example arranges rule: when defined notion W, if the concept W in definition _ioccur m time, and n that occurs for this m time referring to semantic { S ₁, S ₂s _n, then available { W, W#1 ... W#n-1}, n that distinguishes W is semantic, and W#i may be interpreted as example when reasoning;

Rule 7---polysemant disposes rule: have a lot of polysemant in natural language, if polysemant W has n semantic { S ₁, S ₂s _n, then for each concrete semantical definition concept, define n concept { W@1, W@2 altogether ... W@n} distinguishes and represents this n different semanteme;

(2g) axiom 2 is supposed: inherit semantic relation and have unidirectional delivery, lower floor's concept inherits semantic relation that Upper Concept has;

Theorem 1. is according to inheritance R _d, all nouns form one tree;

According to regular 1-7 and definition 4-8, the semanteme of vocabulary in natural language can be defined; Suppose to represent semantic relation with a directed line segment, according to the definition Sum fanction of theorem 1 and vocabulary, the semanteme of vocabulary W can be set by noun in one group of directed line segment represent;

Wherein step (3) performs according to the following procedure:

(3a) semantic matching relationship between noun-noun is defined:

Definition 9---conjunctive word collects: the set of all vocabulary comprised in nominal definition formula, uses C _rWrepresent;

In analysis hereafter with ∝ for representing succession semantic relation, W _x∝ W _yrepresent W _xinherit in W _y, and specify W ∝ W;

(1) basic semantic matching relationship

Definition 10---direct semantics matching relationship: if vocabulary W _x, W _ymeet following condition, use symbol W _xw _yrepresent:

Condition: suppose W _yconjunctive word to collect be C _wY, then

$\&Exists;W_Z\&cap;(W_Z\&Element;C_{\mathrm{WY}})\&cap;(W_X\&Proportional;W_Z).$

Work as W _xw _ytime, match (W _x, W _y)=K _t* d (W _x, W _z);

K _tfor matching relationship coefficient, according to the type of mated relation R, be set to different constants, 1≤K _t≤ 3;

Definition 11---inherit semantic matching relationship: if vocabulary W _x, W _ymeet following condition, use symbol W _x//W _yrepresent:

Condition: $\&Exists;W_Z\&cap;(W_X\&Proportional;W_Z)\&cap;(W_Y\&Proportional;W_Z)$

Work as W _x//W _ytime, match (W _x, W _y)=match (W _x, W _z)+d (W _y, W _z);

Definition semantic distance function d (W _x, W _y): represent two vocabulary W with inheritance _x, W _ybetween succession number of times;

(2) semantic matching relationship is comprised

Definition 12---explicit semantic relation of inclusion: if vocabulary W _x, W _ymeet following condition, use symbol W _x⊙ W _yrepresent;

Condition: there is concept W _z, satisfy condition

$\&Exists;W_Z(R_W(W_Y,W_Z)\&Subset;\mathrm{Def}\left(W_Y\right))\&cap;(W_X\&Proportional;W_Z)$

Work as W _x⊙ W _ytime, there is match (W _x, W _y)=K _p* (d (W _x, W _z), K _pfor comprising matching relationship coefficient;

Definition 13---implicit semantic relation of inclusion: if vocabulary W _x, W _ymeet following condition, use symbol W _xzero W _yrepresent;

Condition: there is concept W _z, satisfy condition

$\&Exists;W_Z(R_P(W_Z,W_X)\&Subset;\mathrm{Def}\left(W_X\right))\&cap;(W_Z\&Proportional;W_Y)$

Work as W _xo W _ytime, match (W _x, W _y)=K _p* (d (W _z, W _y));

Definition 14---comprise semantic matching relationship: if vocabulary W _x, W _ymeet following condition, use symbol W _x◎ W _yrepresent:

Condition: work as W _x◎ W _ytime, match (W _x, W _y)=

min{match(W _X，W _Z)+match(W _Z，W _Y)，match(W _X，W _Y)}

Theorem 3: when vocabulary WX, WY meet WX ◎ WY, WY has the semantic relation of WX;

(3b) semantic matching relationship between noun-verb

The semantic matching relationship of noun-verb can be divided into two classes:

1) SVO semantic matching relationship: the recipient of actor or action may be done in noun

2) lattice semantic matching relationship: noun and verb have lattice semantic matching relationship

Suppose that verb be gerund of executing in the definition of V, V is S ₀, be O by gerund ₀; Due to define time by S ₀be set to the top noun may implementing V, O ₀be set to the top noun bearing this action, so only have same S ₀or O ₀there is the noun S of certain relation and noun O just likely to perform an action V, namely form the semantic matches of SVO; SVO semantic matches has 6 kinds of situations, and it is worth available Value _sVOrepresent, computing formula is as follows:

Value _SVO＝match(S，S ₀)+match(O，O ₀)

Definition 15---conventional SVO semantic matching relationship: satisfy condition (S ∝ S ₀) ∩ (O ∝ O ₀);

Definition 16---heavily loaded SVO semantic matching relationship: satisfy condition:

$(\left({S\&Proportional;S}_0\right)\&cap;(\&Exists;R_{\mathrm{VO}}(R_{\mathrm{VO}}(V,M)\&Subset;\mathrm{Def}\left(S\right))\&cap;(O\&Proportional;W))\&cup;((O\&Proportional;O_0)\&cap;$

$(\&Exists;R_{\mathrm{VS}}(R_{\mathrm{VS}}(V,M)\&Subset;\mathrm{Def}\left(O\right))\&cap;(S\&Proportional;W))$

For noun S and O and verb V, do not meet in the definition of V SVO coupling time, the defined declaration of S, O they meet SVO coupling;

Definition 17---comprise SVO semantic matching relationship:

Definition 18---similar SVO semantic matching relationship:

Definition 19---, may there is metaphor in conjecture statement in metaphor SVO semantic matching relationship: under the following conditions:

Condition 1: conventional SVO semantic matching relationship, heavily loaded SVO semantic matching relationship can be met without any noun in whole statement, comprise SVO semantic matching relationship, similar SVO semantic matching relationship;

Condition 2: there is noun S or O in statement, meets! (S ∝ S ₀) ∩ (O ∝ O ₀), guess and liken S to S ₀;

Or condition 3: there is noun S or O in statement, meet (S ∝ S ₀) ∩! (O ∝ O ₀), guess and liken O to O ₀;

For metaphor SVO semantic matching relationship, Value _sVO=K _f* (match (S, W _p)+match (O, W _p))

K _ffor weights coefficient, W _ps and S ₀most recent co mmon ancestor;

Definition 20---lattice semantic matching relationship: for noun W and verb V, meets match (W, V)=K _c* d (W, W _c), K _cfor weights coefficient;

(3c) semantic matching relationship of noun-adjective

For adjective W _vAwith noun W _n, meet

$\&Exists;W((R_{\mathrm{AS}}(W,W_1)\&cup;R_{\mathrm{AP}}(W,W_1))\&Subset;\mathrm{Def}\left(W_{\mathrm{VA}}\right)\&cap;(W_N\&Proportional;W)),$

Match (W _vA, W _n)=K _a* d (W _n, W), K _a=1, be weights coefficient;

(3d) semantic matching relationship arranged side by side

Semantic matching relationship arranged side by side only for the judgement of parallel construction in statement, to determine the scope of conjunction;

Definition 21---semantic similitude: because nominal definition have employed single method inherited, two noun W _x, W _yalthough there is no inheritance in definition, at semantically W _xbut may be W _yone, the concept be equivalent in description logic contains, and uses symbol W _x∽ W _yrepresent; Tableau algorithm in description logic can be improved, to judge Concept Semantic similarity relation;

Definition 22---noun semantic matching relationship arranged side by side: for two noun W _x, W _yavailable match (W _x, W _y)=K _t* (d (W _x, W _e)+d (W _y, W _e)) calculate a numerical value, as heuristic information, W _ew _x, W _ynearest common ancestor's node; When meeting W _x∽ W _ytime, the W in statement _xwith W _yit may be coordination;

Definition 23---verb semantic matching relationship arranged side by side: for two verb V _x, V _yavailable match (W _x, W _y)=K _t* (d (S _x0, S _y0)+d (O _x0, O _y0)) calculate-individual numerical value, as heuristic information, { S _x0, S _y0, O _x0, O _y0w _x, W _yactor in definition and by dynamic person;

(3e) semantic matching relationship between other class vocabulary

The semantic matching relationship of adverbial word: suppose adverbial word can semantic matches in verb, adjective and adverbial word, regulation match (W ₁, W ₂)=0; The semantic matching relationship of measure word: lexicon should preserve the incidence relation of measure word and noun; If measure word W can modification noun W _n, then match (W, W is specified _n)=0; Otherwise match (W, W _n)=MAX; The semantic matching relationship of pronoun: refer to relation according to pronoun, replaces to corresponding noun and processes by pronoun;

(3f) syntactic match relation

In concrete statement, the vocabulary of some type is modified mutually, but semantic relation not inherent between vocabulary itself, just may have a kind of grammatical phenomenon of semantic modified relationship in this statement, i.e. phraseological modified relationship, comprises following situation:

(1) modified relationship between uncommon part of speech: between verb-verb; Between adverbial word-noun; Between adjective-verb; These all belong to syntactic match relation, do not have inherent semantic matching relationship, just in statement, have phraseological modified relationship between vocabulary itself; In Sentence analysis process, its semantic matches is worth available match (W _x, W _y)=MAX/K _gcalculate, K _gtype weights, K _g< 1.5;

Wherein step (4) performs according to the following procedure:

(4a) three levels and the syntax thereof of semantic structure are defined

Sentence analysis to be carried out according to semantic model herein, the statement abstract representation method of applicable semantic model must be had; Any statement is all formed through iteration by statement relatively simple for structure, and phrase is seen as an ingredient in statement; In order to meet the semantic analysis needs of semantic model, the semantic structure of statement can be divided into three levels according to the complexity of semantic structure and feature: simple sentence, special simple sentence, complex sentence;

Definition 24---simple sentence: only have a verb or adjective to make the statement C of predicate _s, available grammar G ₁carry out abstractdesription;

By the thought design grammar G of case grammar ₁, mentality of designing: suppose that V is predicate, S is the actor of V; O is the recipient of V, A _bit is preposition attribute; A _ait is postpositive attributive; P _dbe the adverbial modifier or complement, be equivalent to one group of lattice in case grammar; P _cit is the lattice content of; N is noun; N _pfor noun phrase;

Grammar G ₁in fuzzy rules be no less than two, mentality of designing of its crucial rule is as follows:

1)C _S→P _DA _BSA _AP _DVP _DA _BOA _AP _D

2)S→n|SA _AA _BS

3)P _D→P _C|P _DP _C

S, O, A _b, A _a, P _cin preposition, conjunction, auxiliary word, number, measure word service regeulations can write out easily;

SVO in simple sentence has 6 kinds of different order altogether: SVO, SOV, VSO, OSV, VOS, OVS; SV, VS; Dispense S or O sometimes in simple sentence, have 4 kinds of different order: SV, VS, OV, VO; When predicate made in intransitive verb or adjective, this adjective is expressed as V, has 2 kinds of different order: SV, VS; There are 10 kinds of orders altogether; The concrete representation of the G1 of the syntax is as follows:

Definition 25---special simple sentence: there is multiple verb or adjective, but semantically not comprise the statement of subordinate clause, available grammar G ₂carry out abstractdesription;

Grammar G ₂mentality of designing: guarantee can not produce on the basis of subordinate clause, to grammar G ₁the few rule of middle interpolation can generating grammar G ₂, mainly contain following 2 kinds of situations:

1) situation of predicate made in multiple verb or adjective

2) S, O, A made in verb or adjective _b, A _a, P _csituation

Grammar G ₂key be verb phrase V _vnoun phrase N directly can not be followed in front and back _p, namely can not there is N _p+ V _vor V _v+ N _p;

The concrete representation of the G2 of the syntax is as follows:

Definition 26---complex sentence: in grammar G ₂the regular N of middle interpolation _p→ C _s, form grammar G ₃; Because regular N _p→ C _sillustrate that a simple sentence or special simple sentence can make any composition in a complex sentence, achieve simple sentence recurrence, therefore grammar G ₃complex sentence can be described;

(4b) thinking of best grammatical analysis scheme is obtained

(1) lexical ambiguity digestion procedure

Suppose W ₁w ₂... W _klexical semantic number be respectively n ₁, n ₂... n _k, carry out fully intermeshing for each semanteme, result is { L ₁, L ₂... L _m, then M=n _1*n _2*... _*n _k, suppose one of them w _mn-th meaning of a word, then L _ic _sa sequence of words without lexical ambiguity; Exhaustive each { L in parsing process ₁, L ₂... L _manalysis result, select best L _ijust lexical ambiguity can be cleared up;

(2) analytical mathematics

According to axiom 1, obtain all grammatical analysis schemes, for each grammatical analysis option A _i, calculate A according to formula 1 _icorresponding semantic matches value, and select best grammatical analysis scheme;

Definition 27---simple clause: meet grammar G in statement ₁or G ₂substring be simple clause;

Suppose axiom 4---semantic modification target axiom: supposing that target modified in the semanteme of notional word W is W _gi, then for the simple clause C meeting semantic logic in statement _s, meet (W ∈ C _s) → (W _gi∈ C _s); For attribute A _b, meet (W ∈ A _b) → (W _gi∈ (A _b∪ S)); For the adverbial modifier or complement (P _d), meet (W ∈ P _d) → (W _gi∈ (P _d∪ V));

According to semantic feature of modifying target, all grammatical analysis schemes can be divided into 2 layers:

1) simple clause's level grammatical analysis scheme;

2) the grammatical analysis scheme of simple clause inside;

(4c) best grammatical analysis scheme is obtained

(1) Rule of judgment of simple clause can be summed up

For statement C _s, carry out grammar G ₁, G ₂, G ₃cYK Algorithm Analysis, the substring s (i, j) meeting table 1 conditional is the simple clause that can sum up;

Table 1 can sum up the Rule of judgment of simple clause

(2) bottom-up simple clause sums up method

Available bottom-up simple clause sums up method and asks for best subordinate clause level grammatical analysis scheme, sees algorithm 4:

Algorithm 1---simple clause sums up method:

1) for statement C _s, according to the Rule of judgment of table 1, find out the substring set { s that can sum up corresponding to simple clause ₁, s ₂... s _m;

2) for each clause s _i, calculate simple clause s with algorithm 2 _ibest semantic matches value, by s _ibe summed up as N _p, N is set _pend semantic;

3) C is made _sequal to sum up result, by the simple clause s in recursive procedure _ithe summation of best semantic matches value, carry out the recurrence of step 1-3;

4) there is best complete simple clause s corresponding to sentence semantic matches value _iscope and end order be best grammatical analysis scheme;

The best semantic matches value calculating simple clause is algorithm 2;

In the algorithm, have employed the method for exhaustion when simple clause selects, theoretic best grammatical analysis scheme can be obtained; But the calculated amount of this method is comparatively large, not easily realizes; When clause's quantity can be returned too much, the good simple clause of k semantic matches degree can be only selected to carry out recursive search, wherein k < m, to ask for suboptimum grammatical analysis scheme;

(3) semanteme is summed up

In algorithm 1, by simple clause C _sbe summed up as N _pafter, N _pdo not have semanteme, the semantic matches cannot carrying out next step calculates, and the mode of solution is as follows:

1) regulation carrys out N by summing up _pcan be matched with any vocabulary W, semantic matches value is:

Match (N _p, W) and=MAX/K _c, wherein K _c> 1

2) if N _pmake S or O of new goal clause, then can by N _psemanteme be set to former C _sin S or O;

(4d) the best semantic matches value of simple clause is obtained

Calculate the best semantic matches value of simple clause, according to axiom 1 and axiom 4, there is multiple grammatical analysis scheme simple clause inside, all grammatical analysis schemes must be obtained, for often kind of grammatical analysis scheme, the semanteme of its notional word is modified target and is determined, just can calculate the semantic matches value under this grammatical analysis scheme according to formula 1, the grammatical analysis scheme with minimum semantic matches value is exactly required analysis result;

The grammatical analysis scheme of simple clause inside can be divided into 3 layers: 1) SVO combination level; 2) A _a, P _d, A _blevel; 3) A _a, P _d, A _binner grammatical analysis scheme; Wherein best grammatical analysis scheme is selected by algorithm 2;

Algorithm 2---the best semantic matches value of simple clause:

1) if simple clause is special simple sentence, all methods being summed up as simple sentence are found

2) for often kind of resolution principle, special simple sentence is summed up as simple sentence

3) for this simple sentence, all possible SVO combined method is found out

4) for often kind of SVO combined method, by C _sbe segmented into { L ₁, L ₂..L _n; If S or O is phrase, then carry out algorithm 3

5) each segmentation L _iinside A can be comprised at most _a, P _d, A _bthree partial contents, find out L _iin all A _a, P _d, A _bdivision methods

6) for often kind of A _a, P _d, A _bdivision methods, the means combined by syntax and semantics the matching analysis, are determined that target modified in the semanteme of each notional word, make A _a, P _d, A _bsemantic matches value minimum

7) ask for the semantic matches value of full sentence, select the minimum corresponding analytic process of semantic matches value as the grammatical analysis scheme of the best

Suppose for simple clause C _scarry out grammar G ₁the operation result of CYK algorithm, expression can generate the grammar symbol collection of substring s (i, j);

(1) the grammatical analysis scheme of SVO combination level

In simple sentence, suppose noun W ₁and W ₂meet SVO with verb V to mate, then { W ₁, V, W ₂a SVO combination, but S or O may be a phrase, when there is { W in sentence ₁, V, W ₃and { W ₂, V, W ₃sVO coupling, and V, W ₃not at W ₁and W ₂centre, and W ₁and W ₂middle substring s (m, n) meets: time, at W ₁+ s (m, n)+W ₃the phrase formed is S, in like manner can obtain longer S or O;

Algorithm 3---S or O segmentation:

1) obtain phrase S or O, find out in S all nouns meeting SVO coupling, be assumed to be { n ₁, n ₂..n _m}

2) according to { n ₁, n ₂..n _mphrase S is divided into m-1 section, according to regular S → n|SA _aa _bs is known each not for empty segmentation may comprise A _aa _b

(2) A _a, P _d, A _bthe division methods of level

Suppose segmentation L _isubstring be s (m, n), then meet p, q is grammatical A _a, P _d, A _bdivision methods, segmentation result is: A _a=s (m, p), P _d=s (p, q) A _b=s (q, n);

(3) A _a, P _d, A _binner best grammatical analysis scheme

Theorem 2:A _a, P _d, A _binner best grammatical analysis scheme works as A _a, P _d, A _bthe grammatical analysis scheme that interior each notional word is corresponding under having best semantic modification target conditions;

Definition 28---simple noun phrase: do not comprise verb and adjectival noun phrase is exactly simple noun phrase;

Theorem 3: the attribute A of simple sentence _aor A _bbest grammatical analysis many equivalents in the best grammatical analysis scheme of a simple noun phrase;

Prove: only comprise a verb due to simple sentence or predicate V made in adjective, therefore A _a, A _bin do not comprise verb and adjective, according to axiom 4, A _a, A _bgrammatical analysis many equivalents in simple noun phrase N _pgrammatical analysis scheme, N _p∈ { (A _b+ S), (A _a+ S), (A _b+ O), (A _a+ O) };

Simple noun phrase N _pdifferent grammatical analysis schemes only by conjunction, preposition, auxiliary word, measure word impact; The key of grammatical analysis is selected conjunction, preposition, auxiliary word, the scope of measure word and their end order;

The determination of A, scope: in conjunction, preposition, auxiliary word, measure word, suppose w _bfor preposition type, w _mfor preposition type, then its scope can be summed up as two kinds of forms 1) ..N _bn..N _b1... w _m..N _a1..N _am..; 2) w _b..N _bn..N _b1... w _m..N _a1..N _am...; Wherein { N _b1, N _b2... N _bnthe noun of first half in scope, { N _a1, N _a2... N _amit is the noun of latter half in scope;

Custom modified in backward semanteme according to Chinese, can at { N _a1, N _a2... N _amin find out grammatical and N _b1there is the noun N of best semantic matches value _ajas boundary after scope; The scope prezone of form 1 can be determined by similar method;

(2) determination of end order: conjunction, preposition, auxiliary word, measure word and scope thereof should be summed up by certain grammar rule, the available method of exhaustion obtains their the best end order; Generally, after simple sentence statement to carry out repeatedly segmentation, A _a, A _bin the number n of { conjunction/preposition/auxiliary word/measure word } that comprises be generally less than 4, there is calculating feasibility;

Definition 29-noun sequence: do not exist conjunction, preposition, auxiliary word, measure word simple noun phrase be noun sequence;

After conjunction, preposition, auxiliary word, measure word are all summed up, simple noun phrase has just been known as a noun sequence, and conjunction, preposition, auxiliary word, scope of quantifier inside may also exist one or two noun sequence in addition; In noun sequence, only noun affects semantic modified relationship, and custom modified in the backward semanteme according to Chinese, supposes that noun sequence is L _n=W ₁w ₂... W _m; Then determine L by semanteme _nin arbitrarily noun to modify target concrete grammar as follows:

Target modified in the best semanteme of algorithm 4---noun sequence:

Set C is set _wfor sky, for L _nin each noun W _iif, match (W _i, W _m) < MAX, by W _ibe added to C _w, do to operate as follows:

1) C is supposed _welement be W by order successively ₁-W ₂-...-W _n, be then done as follows: by L _nbe divided into n+1 section, the semanteme modification target arranging them is W _m, and recurrence is carried out to each section;

2) C is worked as _win when only having a noun, carry out step 3) and step 4);

3) forward direction modified relationship is set: for any segmentation, if there is W _xw _x+1... W _y-1w _y, satisfy condition: 1. any W _x+1... W _ybetween noun and W _yafter the semantic matches value of noun be MAX; 2. match (W _y, W _x) < MAX; Then W is set _ysemanteme modify target be W _x; Then W is set _x+1... W _y-1between the modification target W of noun _y;

4) if L _nin also have noun W _ydo not modify target, then arranging its modification target is W _y+1;

P _danalytical approach be similar to A _a, A _b, key carries out boundary line delimitation according to preposition, the content in preposition scope is also converted into a simple noun phrase; Due to comparatively loaded down with trivial details, do not discuss in detail here;

(4) disposal route of special simple sentence

Sum up all Non-Definite Verb/adjectives, statement is converted into simple sentence, select best end scheme; Disposal route is as follows:

1) statement is carried out to the CYK algorithm of grammar G 2, find the verb or Adjective Phrases that likely do predicate, may kinds of schemes be had;

2) for each scheme, sum up remaining verb or adjective, choose semantic matches and be worth minimum analytical plan;

Also need when summing up Non-Definite Verb or adjective to arrange to sum up semanteme,

Wherein step (5) performs according to the following procedure:

According to the grammatical analysis result with best semantic matches value, simple sentence is converted into a knowledge point, each simple clause of complex sentence is converted into knowledge point, whole complex sentence is converted into one group of knowledge point;

After statement being converted into the knowledge point of depositing with structural data form, various Intelligent Information Processing is carried out to these knowledge datas.