CN110688458A - Combat model retrieval method - Google Patents

Abstract

The invention discloses a combat model retrieval method, which comprises the following steps: obtaining a model query framework of a user, the model query framework comprising: a model component descriptor; inquiring according to the model inquiry frame, calculating the matching degree of the operation model in the inquiry result and the model inquiry frame, judging whether the matching degree is greater than a threshold value, if so, calculating the satisfaction degree of the operation model in the inquiry result, and sequencing operation model components according to the satisfaction degree to form the final result of multi-facet inquiry of the operation model; if the synonym is smaller than the preset synonym, constructing a synonym dictionary query model according to the synonyms of the terms in the model component descriptor; and querying according to the synonym dictionary query model. The invention reduces the complexity of the system, ensures the flexibility of application and enables the facet retrieval algorithm to be realized more easily.

Description

Combat model retrieval method

Technical Field

The invention relates to the technical field of model retrieval, in particular to a combat model retrieval method.

Background

With the development of science, informatization is more and more common. Information systems are also an important direction of development at the military level. The whole life operation model management system supporting the development, representation, storage, modification, model creation, linkage, integration, selection, operation, maintenance and reuse of operation models is an important component of the operation model system construction.

In the prior art, the precision and the efficiency of a retrieval result cannot meet the requirements.

Disclosure of Invention

The invention provides a combat model retrieval method to overcome the technical problems.

The invention relates to a combat model retrieval method, which comprises the following steps:

obtaining a model query framework of a user, the model query framework comprising: a model component descriptor, the model component descriptor comprising: at least one facet or/and at least one respective level term;

inquiring according to the model inquiry frame, calculating the matching degree of the operation model in the inquiry result and the model inquiry frame, judging whether the matching degree is greater than a threshold value, if so, calculating the satisfaction degree of the operation model in the inquiry result, and sequencing operation model components according to the satisfaction degree to form the final result of multi-facet inquiry of the operation model; if the synonym is smaller than the preset synonym, constructing a synonym dictionary query model according to the synonyms of the terms in the model component descriptor;

and querying according to the synonym dictionary query model.

Further, after the query is performed according to the query framework of the synonym dictionary query model, the method further includes:

calculating the matching degree of the combat model in the query result and the model query frame, judging whether the matching degree is greater than a threshold value, if so, calculating the satisfaction degree of the combat model in the query result, and sequencing combat model components according to the satisfaction degree to form the final result of multi-facet query of the combat model; if the number is less than the preset number, constructing a hierarchical network dictionary query model by adopting a hierarchical network;

and inquiring according to the hierarchical network dictionary inquiring model.

Further, after the querying according to the third model query framework, the method further includes:

and displaying the query result in a list form.

Further, the constructing a synonym dictionary query model according to synonyms of terms in the model component descriptor includes:

cutting words of the terms in the model component descriptor to obtain a set of words corresponding to the terms in the model component descriptor;

removing conjunctions, auxiliary words and adverbs in the set;

merging word sets corresponding to two terms in the model component descriptor; and deriving therefrom the corresponding vector T, T ═ T₁，t₂，…，t_n) Where n ═ a |,1 ≦ i ≦ n, 1 ≦ j ≦ n, and for any i, j, there is t_iE.g. A, and when i ≠ j, t_i≠t_jA is a word set corresponding to the merged two terms;

two terms in the model building block descriptor are expressed in vector form:

V_i＝(b₁，b₂，…，b_n)，V_j＝(c₁，c₂，…，c_n)

wherein i is more than or equal to 1 and less than or equal to n, j is more than or equal to 1 and less than or equal to n, and V_iIn the form of a vector corresponding to the first term, said V_jIn the form of a vector corresponding to the second term, said b₁Vector elements corresponding to the first term, said c₁Vector elements corresponding to the second term;

computing two term vectors (V)_i，V_j) The cosine value of (2) to obtain the relevance R in terms of the term grammar₁＝cos(V_i，V_j) The term B is a set of corresponding terms obtained by segmenting the second term in the model component descriptor and removing conjunctions, auxiliary words and adverbs in the set, and the term C is a set of corresponding terms obtained by segmenting the second term in the model component descriptor and removing conjunctions, auxiliary words and adverbs in the set;

constructing a Bayesian network model according to the description of the same model component in two sets of words corresponding to the model component descriptors, and obtaining the correlation degree of terms in semantics according to conditional probability;

and constructing a synonym dictionary query model according to the grammar relevance and the semantic relevance.

Further, the obtaining the correlation degree of terms in the model component descriptor in the aspect of semantics by the conditional probability comprises:

using conditional probability formulae

R₂＝P(T_k/S_j)＝P(T_kS_j)/P(S_j)

R＝ω₁×R₁+ω₂×R₂

Calculating the semantic relatedness of two terms in the descriptor of the model component, wherein T_k，S_jIs a descriptive term for the same member of two sets of words corresponding by model member descriptors, P (T)_k/S_j) Is a conditional probability, R₂Being the degree of correlation of semantics, omega₁+ω₂＝1；ω₁≥0；ω₂≥0；ω₁And ω₂Is R₁And R₂The weight of (c).

Further, the constructing a hierarchical network dictionary query model by using a hierarchical network includes:

and adopting a classification tree to establish the upper-layer and lower-layer relation among terms in the model component descriptor, wherein the upper layer comprises a plurality of lower layers, and the lower layers refine the upper layers corresponding to the lower layers from different angles.

Further, the calculating the queried model matching degree and satisfaction degree includes:

determining to adopt an optimal strategy or satisfaction according to the complexity of the operation models, if the optimal strategy is adopted, calculating the matching degrees of the operation models in all the selected retrieval ranges, and sequencing according to the matching degrees;

if a satisfactory strategy is adopted, the retrieval can be stopped when the matching degree of the retrieved combat model meets a preset threshold value;

the model complexity includes: the number of submodels included in the model, and the code amount of the model.

According to the invention, the initial query is carried out through the model component query model, if the initial query can not meet the requirement, the query is carried out through the synonym dictionary query model, the evaluation strategy of the matching degree is adjusted, the complexity of the system is reduced, the flexibility of the application is ensured, and the facet retrieval algorithm is easier to realize.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.

FIG. 1 is a flow chart of a method for battle model retrieval according to the present invention;

FIG. 2 is a diagram of a model component descriptor structure according to the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Fig. 1 is a flowchart of a battle model search method according to the present invention, and as shown in fig. 1, the method of the present embodiment includes:

step 101, obtaining a model component query framework of a user, wherein the model query framework comprises: a model component descriptor, the model component descriptor comprising: at least one facet or/and at least one respective level term.

For example, the facets include: military warfare, command hierarchy, operational fields, etc., the terms corresponding to the above facets include a plurality of levels, as shown in fig. 2, wherein:

first-order terms corresponding to the military species of the facet military include: navy, air force, etc.;

the primary term navy corresponds to a secondary term that includes: surface vessels, submarines, aviation soldiers, etc.;

the secondary terms corresponding to the primary term air force include: bombers, fighters, etc.;

the first-level terms of the corresponding facet conductor level include: campaign level, tactical level, etc.;

the first level term campaign level corresponds to a second level term comprising: amphibious formation, aircraft carrier formation, and the like;

first-order term tactical-order corresponds to second-order terms including: destroyer, warship, etc.;

the first-level terms for the corresponding facet battle fields include: marine combat, air combat, and the like.

The primary term marine combat corresponds to a secondary term comprising: threat judgment, situation analysis and the like;

the primary term air combat corresponds to the secondary term comprising: air combat, air patrol, etc.

The model query framework can form a team for all terms corresponding to the facet military and a second-level term aircraft carrier under the facet command level. Or any level of term combination under any facet.

A facet is a fixed set used to describe some aspect or perspective of a model component, each facet may classify the component from a different side, each facet having a set of terms. The model building descriptor is an ordered N-tuple, each of which is a term selected from facets according to the user query requirement. Therefore, a tree structure is adopted to represent the facet classification mode, wherein main facets and sub facets are respectively mapped to corresponding parent nodes and child nodes in the tree, and corresponding terms are mapped to leaf nodes of the tree.

The construction of the model component query framework (the relationship between model component descriptors) generally follows the following rules:

rule 1 if a facet is selected by the user at the time of retrieval, the component descriptor is the set of all terms in the term space of all sub-facets under that facet.

Rule 2 if two or more facets, which term spaces do not intersect, are selected by the user at the time of retrieval, the component descriptor is the set of all terms in the term space of these facets.

Rule 3 if a user selects a term for a non-leaf node when retrieving, the component descriptor is only a collection of this term and does not include its underlying terms.

Rule 4 if a user selects multiple terms in the same facet term space when retrieving, then the component descriptor is a collection of these terms.

Rule 5 if a user selects multiple terms in different facet term spaces when retrieving, the component descriptor is a collection of these terms.

102, inquiring according to the model member inquiry model, calculating the matching degree of the combat model in the inquiry result and the model inquiry frame, judging whether the matching degree is greater than a threshold value, if so, calculating the satisfaction degree of the combat model in the inquiry result, and sequencing the combat model members according to the satisfaction degree; if the synonym is smaller than the preset synonym, constructing a synonym dictionary query model according to the synonyms of the terms in the model component descriptor;

specifically, the term of the selected component descriptor in the model query framework is a secondary term situation analysis corresponding to a primary term sea combat, and synonyms of the term can be situation judgment, situation analysis, situation judgment and the like.

And 103, inquiring according to the synonym dictionary inquiry model.

Further, after the query is performed according to the query framework of the synonym dictionary query model, the method further includes:

calculating the matching degree of the combat model in the query result and the model query frame, judging whether the matching degree is greater than a threshold value, if so, calculating the satisfaction degree of the combat model in the query result, and sequencing combat model components according to the satisfaction degree to form the final result of multi-facet query of the combat model; if the number is less than the preset number, constructing a hierarchical network dictionary query model by adopting a hierarchical network;

and inquiring according to the hierarchical network dictionary inquiring model.

Specifically, facets corresponding to terms of selected component descriptors in the model query framework have N levels, and the matching degree of the query result of the terms of ① N-1 level is smaller than a threshold value, then the next level of terms are automatically queried by adopting a hierarchical network, ② the matching degree of the N level of terms or the query result of the facets is smaller than the threshold value, then N +1 level of terms are generated according to the N level of terms, and the N +1 level of terms are used for querying.

Further, after the querying according to the third model query framework, the method further includes:

and displaying the query result in a list form.

Further, the method for constructing the synonym query according to the synonym of the terms in the model component descriptor,

And performing word segmentation on the terms in the model component descriptor to obtain a set of words corresponding to the terms in the model component descriptor.

Removing conjunctions, auxiliary words and adverbs in the set;

merging word sets corresponding to two terms in the model component descriptor; and deriving therefrom the corresponding vector T, T ═ T₁，t₂，…，t_n) Where n ═ a |,1 ≦ i ≦ n, 1 ≦ j ≦ n, and for any i, j, there is t_iE.g. A, and when i ≠ j, t_i≠t_jA is a word set corresponding to the merged two terms;

two terms in the model building block descriptor are expressed in vector form:

V_i＝(b₁，b₂，…，b_n)，V_j＝(c₁，c₂，…，c_n)

wherein i is more than or equal to 1 and less than or equal to n, j is more than or equal to 1 and less than or equal to n, and V_iIn the form of a vector corresponding to the first term, said V_jIn the form of a vector corresponding to the second term, said b₁Vector elements corresponding to the first term, said c₁Vector elements corresponding to the second term;

computing two term vectors (V)_i，V_j) The cosine value of (2) to obtain the relevance R in terms of the term grammar₁＝cos(V_i，V_j) The term B is a set of corresponding terms obtained by segmenting the second term in the model component descriptor and removing conjunctions, auxiliary words and adverbs in the set, and the term C is a set of corresponding terms obtained by segmenting the second term in the model component descriptor and removing conjunctions, auxiliary words and adverbs in the set;

constructing a Bayesian network model according to the description of the same model component in two sets of words corresponding to the model component descriptors, and obtaining the correlation degree of terms in semantics according to conditional probability;

and constructing a synonym dictionary query model according to the grammar relevance and the semantic relevance.

Further, the obtaining the correlation degree of terms in the model component descriptor in the aspect of semantics by the conditional probability comprises:

using conditional probability formulae

R₂＝P(T_k/S_j)＝P(T_kS_j)/P(S_j)

R＝ω₁×R₁+ω₂×R₂

Calculating the semantic relatedness of two terms in the descriptor of the model component, wherein T_k，S_jIs a descriptive term for the same member of two sets of words corresponding by model member descriptors, P (T)_k/S_j) Is a conditional probability, R₂Being the degree of correlation of semantics, omega₁+ω₂＝1；ω₁≥0；ω₂≥0；ω₁And ω₂Is R₁And R₂The weight of (c).

Further, the constructing a hierarchical network dictionary query model by using a hierarchical network includes:

and establishing the upper-layer and lower-layer relation between the model component descriptors by adopting a classification tree, wherein the upper layer comprises a plurality of lower layers, and the lower layers refine the upper layers corresponding to the lower layers from different angles.

Specifically, in the facet classification model, each facet is associated with a structured set of legal terms (term spaces) from which the terms used in the classification and retrieval of the component come. The structure of the term space reflects the semantic relation among the terms, so the term space can be regarded as a semantic net, and the components are a set of facet terms from the external view. There are two relationships of terms in term space, namely hierarchical relationships and synonym relationships.

The system is divided into two or more layers, wherein the first layer is the highest layer and is used for describing the most general concept, the concept of the upper layer is refined layer by layer, and the number of the layers is determined by the subdividability of the concept. Then, in the brother nodes of the same layer, the lateral relation is added, and the distance value is marked to represent the similarity. Smaller distance values indicate a higher degree of similarity.

Further, the calculating the queried model matching degree and satisfaction degree includes:

determining to adopt an optimal strategy or satisfaction according to the complexity of the operation models, if the optimal strategy is adopted, calculating the matching degrees of the operation models in all the selected retrieval ranges, and sequencing according to the matching degrees;

if a satisfactory strategy is adopted, the retrieval can be stopped when the matching degree of the retrieved combat model meets a preset threshold value;

the model complexity includes: the number of submodels included in the model, and the code amount of the model.

Specifically, the factors of the merit function: model function matching degree, model interface matching degree, model complexity and model credit; the optimal strategy is to calculate the matching degrees of all the operation models in the selected retrieval range, select the operation models of which the matching degrees of the operation models accord with a certain threshold value, and sort the operation models according to the satisfaction degrees of the operation models. This method may be used to ensure that an optimal solution is obtained when the number of operational models in the operational model library is small. However, when there are many models in the battle model library, the speed of searching the optimal strategy model is often slow. And the credit of the combat model is the evaluation level of the user on the combat model. Calculated in a weighted manner

The complexity of the battle model is generally comprehensively expressed by the factors of the number of submodels contained in the model, the code amount of the model and the like; the model evaluation function can be obtained by comprehensively considering various factors and adopting a weighting mode. Namely, setting: the weights of all factors of the satisfaction degree evaluation function of the combat model are W ═ W₁,W₂…W_n}; the data value of each factor in the model satisfaction is X ═ X respectively₁,X₂…X_n}. The satisfaction m (r) of the battle model is: m ═ W₁×X₁,W₂×X₂…W_n×X_n}. The optimal strategy is to calculate the matching degrees of all the operation models in the selected retrieval range, select the operation models of which the matching degrees of the operation models accord with a certain threshold value, sort the operation models according to the satisfaction degrees of the operation models, and then select the optimal operation models as the final selection result. This method may be used to ensure that an optimal solution is obtained when the number of operational models in the operational model library is small. The method for satisfying strategy is to calculate the match degree of the operation model in the selected search range in sequence, if the match degree of the operation model meets a certain threshold valueThe operational model estimates the satisfaction degree of the operational model, and if the satisfaction degree meets the requirement, the operational model is the final selected result. Because the battle model base models are numerous, the optimal strategy often causes the combined explosion of model solution, so that the retrieval speed of the optimal strategy model is often slow, and therefore, the optimal strategy is often replaced by a satisfactory strategy in practice.

The invention adopts a multi-facet retrieval mode, ensures the richness and the integrity of facet retrieval, effectively avoids errors caused by uncertain multiplexing requirements of users on the system in the query and improves the query effect. The invention constructs the satisfaction evaluation function of the weighted operation model, provides a sorting mechanism for retrieval, improves the quality of output results, improves the efficiency of searching the operation model by a user and improves the credibility of the retrieval.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.

Claims (7)

1. A battle model retrieval method is characterized by comprising the following steps:

obtaining a model query framework of a user, the model query framework comprising: a model component descriptor, the model component descriptor comprising: at least one facet or/and at least one respective level term;

inquiring according to the model component inquiry model, calculating the matching degree of the combat model in the inquiry result and the model inquiry frame, judging whether the matching degree is greater than a threshold value, if so, calculating the satisfaction degree of the combat model in the inquiry result, and sequencing the combat model components according to the satisfaction degree to form the final result of the multi-facet inquiry of the combat model; if the synonym is smaller than the preset synonym, constructing a synonym dictionary query model according to the synonyms of the terms in the model component descriptor;

and querying according to the synonym dictionary query model.

2. The method according to claim 1, wherein after said querying according to the synonym dictionary query model query framework, further comprising:

calculating the matching degree of the combat model in the query result and the model query frame, judging whether the matching degree is greater than a threshold value, if so, calculating the satisfaction degree of the combat model in the query result, and sequencing combat model components according to the satisfaction degree to form the final result of multi-facet query of the combat model; if the number is less than the preset number, constructing a hierarchical network dictionary query model by adopting a hierarchical network;

and inquiring according to the hierarchical network dictionary inquiring model.

3. The method of claim 2, wherein after querying according to the third model query framework, further comprising:

and displaying the query result in a list form.

4. The method of claim 1, wherein constructing a synonym dictionary query model from synonyms of terms in the model component descriptor comprises:

cutting words of the terms in the model component descriptor to obtain a set of words corresponding to the terms in the model component descriptor;

removing conjunctions, auxiliary words and adverbs in the set;

merging word sets corresponding to two terms in the model component descriptor; and deriving therefrom the corresponding vector T, T ═ T₁，t₂，…，t_n) Where n ═ a |,1 ≦ i ≦ n, 1 ≦ j ≦ n, and for any i, j, there is t_iE.g. A, and when i ≠ j, t_i≠t_jA is a single term corresponding to the combination of two termsA set of words;

two terms in the model building block descriptor are expressed in vector form:

V_i＝(b₁,b₂,…,b_n),V_j＝(c₁,c₂,…,c_n)

wherein i is more than or equal to 1 and less than or equal to n, j is more than or equal to 1 and less than or equal to n, and V_iIn the form of a vector corresponding to the first term, said V_jIn the form of a vector corresponding to the second term, said b₁Vector elements corresponding to the first term, said c₁Vector elements corresponding to the second term;

computing two term vectors (V)_i,V_j) The cosine value of (2) to obtain the relevance R in terms of the term grammar₁＝cos(V_i,V_j) The term B is a set of corresponding terms obtained by segmenting the second term in the model component descriptor and removing conjunctions, auxiliary words and adverbs in the set, and the term C is a set of corresponding terms obtained by segmenting the second term in the model component descriptor and removing conjunctions, auxiliary words and adverbs in the set;

constructing a Bayesian network model according to the description of the same model component in two sets of words corresponding to the model component descriptors, and obtaining the correlation degree of terms in semantics according to conditional probability;

and constructing a synonym dictionary query model according to the grammar relevance and the semantic relevance.

5. The method of claim 4, wherein obtaining the semantic relevance of terms in the model building block descriptor with conditional probability comprises:

using conditional probability formulae

R₂＝P(T_k/S_j)＝P(T_kS_j)/P(S_j)

R＝ω₁×R₁+ω₂×R₂

Calculating semantic relatedness of terms in the model component descriptor, wherein T_k,S_jIs a descriptive term for the same member of two sets of words corresponding by model member descriptors, P (T)_k/S_j) Is a conditional probability, R₂Being the degree of correlation of semantics, omega₁+ω₂＝1；ω₁≥0；ω₂≥0；ω₁And ω₂Is R₁And R₂The weight of (c).

6. The method of claim 2, wherein constructing the hierarchical network dictionary query model using the hierarchical network comprises:

and adopting a classification tree to establish the upper-layer and lower-layer relation among terms in the model component descriptor, wherein the upper layer comprises a plurality of lower layers, and the lower layers refine the upper layers corresponding to the lower layers from different angles.

7. The method of claim 1, wherein calculating the queried model match and satisfaction comprises:

determining to adopt an optimal strategy or satisfaction according to the complexity of the operation models, if the optimal strategy is adopted, calculating the matching degrees of the operation models in all the selected retrieval ranges, and sequencing according to the matching degrees;

if a satisfactory strategy is adopted, the retrieval can be stopped when the matching degree of the retrieved combat model meets a preset threshold value;

the model complexity includes: the number of submodels included in the model, and the code amount of the model.

Images