CN110659356A - Information matching method and device and storage medium - Google Patents
Information matching method and device and storage medium Download PDFInfo
- Publication number
- CN110659356A CN110659356A CN201910849854.XA CN201910849854A CN110659356A CN 110659356 A CN110659356 A CN 110659356A CN 201910849854 A CN201910849854 A CN 201910849854A CN 110659356 A CN110659356 A CN 110659356A
- Authority
- CN
- China
- Prior art keywords
- information
- attribute
- classification model
- matching
- reply
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Computational Linguistics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses an information matching method, an information matching device and a storage medium. The method comprises the following steps: receiving first information to be matched; performing first processing on first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information; determining an information attribute of the first information according to the characteristic parameter, wherein the information attribute comprises a defined attribute and an undefined attribute; matching reply information for the first information according to the information attribute of the first information; when the information attribute of the first information is an undefined attribute, selecting a preset first matching mode to reply to the first information. The invention determines whether the attribute of the information is a defined attribute or an undefined attribute by calculating the characteristic parameter of the information to be matched. When the information attribute is determined to be an undefined attribute, the first information is replied by using a predetermined first matching mode, so that the condition that the replied information is matched incorrectly is avoided.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to an information matching method and device storage medium.
Background
Currently, in the intelligent question-answering task, an intention space (classification model library) is generally defined, and corresponding answer information is given by identifying the intention of the user question information.
The identification of intent may be addressed in the form of text classification. If all intent types of the user are known, the category of intent can be determined using existing classification methods, and corresponding answer information can be matched for the intent.
But in reality this is not possible because all intentions of the user cannot be predicted. When the recognition of the classification space is incomplete, that is, when an unknown sample passes through the classifier, the sample is directly classified, so that the intention of the question information outside the classification space is wrongly identified, and further, the answer information matched with the question information is wrong, and the problem of poor user experience is caused.
Disclosure of Invention
The embodiment of the invention aims to provide an information matching method, an information matching device and a storage medium, which are used for solving the problem of inaccurate information matching in the prior art.
In order to solve the technical problem, the embodiment of the application adopts the following technical scheme: an information matching method, comprising the steps of:
receiving first information to be matched;
performing first processing on the first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information;
determining an information attribute of the first information according to the characteristic parameter, wherein the information attribute comprises a defined attribute and an undefined attribute;
matching reply information for the first information according to the information attribute of the first information; when the information attribute of the first information is an undefined attribute, selecting a predetermined first matching mode to reply to the first information.
Optionally, before calculating the characteristic parameter of the obtained information, the method further includes: and preprocessing the first information to enable the first information to meet the specified conditions of the first processing.
Optionally, the method further includes: training by using sample data to obtain a plurality of classification models, and carrying out hierarchical division on each classification model to construct the first classification model library with a multilayer structure.
Optionally, the performing, on the basis of a first classification model library formed by a plurality of classification models according to a hierarchical relationship, a first process on the first information to obtain a characteristic parameter of the first information specifically includes:
respectively carrying out feature importance calculation on the first information by utilizing each classification model in the first classification model library to obtain first feature importance corresponding to each classification model;
respectively carrying out probability calculation on the first information by using each classification model of the non-bottom layer to obtain the probability that the first information is divided into each classification model of the lower layer by each classification model;
calculating to obtain second feature importance corresponding to each classification model according to the probability of each classification model and the first feature importance corresponding to each classification model;
and overlapping the second feature importance corresponding to each classification model to obtain the feature parameters of the information.
Optionally, the performing, by using each classification model in the first classification model library, feature importance calculation on the first information respectively to obtain first feature importance corresponding to each classification model specifically includes:
respectively extracting the characteristics of the first information by utilizing each classification model in a first classification model library and combining an attention mechanism to obtain characteristic information corresponding to each classification model;
and performing feature importance calculation on the feature information corresponding to each classification model to obtain first feature importance corresponding to each classification model.
Optionally, the determining, according to the characteristic parameter, an information attribute of the first information, where the information attribute includes a defined attribute and an undefined attribute, specifically includes:
comparing the characteristic parameter with a preset threshold value, and if the characteristic parameter is greater than or equal to the preset threshold value, determining that the information attribute of the first information is a defined attribute; and if the characteristic importance is smaller than a preset threshold value, determining that the information attribute of the first information is an undefined attribute.
Optionally, the determining, according to the characteristic parameter, an information attribute of the first information, where the information attribute includes a defined attribute and an undefined attribute, specifically includes:
fitting the characteristic parameters of the sample information by using a first-level support vector machine algorithm to obtain a hypersphere; wherein the information attribute of the sample information is a defined attribute;
judging the characteristic parameters of the first information based on the hypersphere, and if the characteristic parameters of the first information are within the range of the hypersphere, determining the information attribute of the first information as a defined attribute; and if the characteristic parameter of the first information is out of the range of the hypersphere, determining the information attribute of the first information as an undefined attribute.
Optionally, the selecting the predetermined first matching manner to reply to the first message includes:
and acquiring keywords of the first information, and determining the first matching mode corresponding to the first information according to the keywords so as to reply to the first information.
Optionally, the method further includes: when the information attribute is a defined attribute, selecting a predetermined second matching mode to match the first information with the reply information; wherein the second matching means comprises: and performing intention identification on the first information, determining an intention, and matching reply information corresponding to the intention for the first information according to the intention.
In order to solve the above technical problem, an embodiment of the present application provides an information matching apparatus, including:
the receiving module is used for receiving first information to be matched;
the calculation module is used for carrying out first processing on the first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information;
a determining module, configured to determine an information attribute of the first information according to the feature parameter, where the information attribute includes a defined attribute and an undefined attribute;
the matching module is used for matching reply information for the first information according to the information attribute of the first information; when the information attribute of the first information is an undefined attribute, selecting a predetermined first matching mode to reply to the first information.
Optionally, the apparatus further comprises a preprocessing module, where the preprocessing module is configured to: and preprocessing the first information to enable the first information to meet the specified conditions of the first processing.
Optionally, the apparatus further comprises a building module, wherein the building module is configured to: training by using sample data to obtain a plurality of classification models, and carrying out hierarchical division on each classification model to construct the first classification model library with a multilayer structure.
Optionally, the calculation module is specifically configured to: respectively carrying out feature importance calculation on the first information by utilizing each classification model in the first classification model library to obtain first feature importance corresponding to each classification model;
respectively carrying out probability calculation on the first information by using each classification model of the non-bottom layer to obtain the probability that the first information is divided into each classification model of the lower layer by each classification model;
calculating to obtain second feature importance corresponding to each classification model according to the probability of each classification model and the first feature importance corresponding to each classification model;
and overlapping the second feature importance corresponding to each classification model to obtain the feature parameters of the information.
Optionally, the calculation module is specifically configured to:
respectively extracting the characteristics of the first information by utilizing each classification model in a first classification model library and combining an attention mechanism to obtain characteristic information corresponding to each classification model;
and performing feature importance calculation on the feature information corresponding to each classification model to obtain first feature importance corresponding to each classification model.
Optionally, the determining module is specifically configured to:
comparing the characteristic parameter with a preset threshold value, and if the characteristic parameter is greater than or equal to the preset threshold value, determining that the information attribute of the first information is a defined attribute; and if the characteristic importance is smaller than a preset threshold value, determining that the information attribute of the first information is an undefined attribute.
Optionally, the determining module is specifically configured to:
fitting the characteristic parameters of the sample information by using a first-level support vector machine algorithm to obtain a hypersphere; wherein the information attribute of the sample information is a defined attribute;
judging the characteristic parameters of the first information based on the hypersphere, and if the characteristic parameters of the first information are within the range of the hypersphere, determining the information attribute of the first information as a defined attribute; and if the characteristic parameter of the first information is out of the range of the hypersphere, determining the information attribute of the first information as an undefined attribute.
Optionally, the matching module is specifically configured to:
and acquiring keywords of the first information, and determining the first matching mode corresponding to the first information according to the keywords so as to reply to the first information.
Optionally, the matching module is further configured to: when the information attribute is a defined attribute, selecting a predetermined second matching mode to match the first information with the reply information; wherein the second matching means comprises: and performing intention identification on the first information, determining an intention, and matching reply information corresponding to the intention for the first information according to the intention.
In order to solve the above technical problem, an embodiment of the present application provides a storage medium having a computer program stored thereon, where the computer program, when executed by a processor, implements the following steps:
receiving first information to be matched;
performing first processing on the first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information;
determining an information attribute of the first information according to the characteristic parameter, wherein the information attribute comprises a defined attribute and an undefined attribute;
matching reply information for the first information according to the information attribute of the first information; when the information attribute of the first information is an undefined attribute, selecting a predetermined first matching mode to reply to the first information.
The embodiment of the invention has the beneficial effects that: the embodiment of the invention determines whether the attribute of the information is a defined attribute or an undefined attribute by calculating the characteristic parameter of the information to be matched. When the information attribute is determined to be an undefined attribute, namely the information is determined to be information outside the intention space, and then the first information is replied by using a predetermined first matching mode, so that the situation that the information cannot be accurately identified through the intention space and wrong reply information is matched for the information is avoided, and the accuracy of information matching is improved.
Drawings
FIG. 1 is a flow chart of a method of implementing information matching in accordance with the present invention;
FIG. 2 is a flow chart of a method for matching information according to the present invention;
FIG. 3 is a schematic diagram of a hierarchical structure of a first classification model library according to an embodiment of the present invention;
FIG. 4 is a flow chart of an information matching method according to an embodiment of the present invention;
fig. 5 is a block diagram of an information matching apparatus according to an embodiment of the present invention.
Detailed Description
Various aspects and features of the present application are described herein with reference to the drawings.
It will be understood that various modifications may be made to the embodiments of the present application. Accordingly, the foregoing description should not be construed as limiting, but merely as exemplifications of embodiments. Those skilled in the art will envision other modifications within the scope and spirit of the application.
The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the application and, together with a general description of the application given above and the detailed description of the embodiments given below, serve to explain the principles of the application.
These and other characteristics of the present application will become apparent from the following description of preferred forms of embodiment, given as non-limiting examples, with reference to the attached drawings.
It should also be understood that, although the present application has been described with reference to some specific examples, a person of skill in the art shall certainly be able to achieve many other equivalent forms of application, having the characteristics as set forth in the claims and hence all coming within the field of protection defined thereby.
The above and other aspects, features and advantages of the present application will become more apparent in view of the following detailed description when taken in conjunction with the accompanying drawings.
Specific embodiments of the present application are described hereinafter with reference to the accompanying drawings; however, it is to be understood that the disclosed embodiments are merely exemplary of the application, which can be embodied in various forms. Well-known and/or repeated functions and constructions are not described in detail to avoid obscuring the application of unnecessary or unnecessary detail. Therefore, specific structural and functional details disclosed herein are not to be interpreted as limiting, but merely as a basis for the claims and as a representative basis for teaching one skilled in the art to variously employ the present application in virtually any appropriately detailed structure.
The specification may use the phrases "in one embodiment," "in another embodiment," "in yet another embodiment," or "in other embodiments," which may each refer to one or more of the same or different embodiments in accordance with the application.
An embodiment of the present invention provides an information matching method, as shown in fig. 1, including the following steps:
step S101, receiving first information to be matched;
the first information in this step may be target information that needs to be processed, and the target information may be obtained by selecting from a history database or may be information that is manually provided. The method and the system can be particularly applied to intelligent customer service scenes, such as question information of user questions, consultation information or reply information replied by the user and the like. For example, the first information is the user's question: my orders have not been received two months.
Step S102, performing first processing on the first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information;
in this step, the first classification model may be a classification model library with a tree structure, or a classification model library with a hierarchical and progressive relationship.
The characteristic parameters in this step include characteristic importance. The feature importance refers to the importance degree of feature information in the classification model.
Step S103, determining an information attribute of the first information according to the characteristic parameter, wherein the information attribute comprises a defined attribute and an undefined attribute;
the information of the defined attribute in this step represents first information that enables problem classification by the first classification model library. The information of undefined attribute indicates information that the first information cannot be subjected to problem classification by the first classification model library, that is, a problem that belongs to the outside of the first classification model library and cannot be directly subjected to intention recognition.
Step S104, according to the information attribute of the first information, matching reply information for the first information; when the information attribute of the first information is an undefined attribute, selecting a predetermined first matching mode to reply to the first information.
In this step, when it is determined that the attribute of the first information is an undefined attribute, that is, it is determined that the question belongs to a question outside the classification model library, the answer information cannot be directly matched by intention identification, but a predetermined first matching method is selected to answer the first information, for example, when it is determined that "two months of my order have not received" this first information attribute is an undefined attribute, a predetermined answer method may be matched to answer the question. Of course, a piece of reply information may be directly matched, such as: "sorry-this question i do not know how to solve so far, you can consult the sister-of-hand". So that the user knows that the first information needs to be replied in a predetermined reply manner.
The embodiment of the invention determines whether the attribute of the information is a defined attribute or an undefined attribute by calculating the characteristic parameter of the information to be matched. When the information attribute is determined to be an undefined attribute, namely the information is determined to be information outside the first classification model base (namely the defined intention space), and then the first information is replied by utilizing a predetermined first matching mode, so that the situation that the information cannot be accurately identified through the intention space and wrong reply information is matched for the information is avoided, and the accuracy of information matching is improved.
An embodiment of the present invention provides an information matching method, as shown in fig. 2, including the following steps:
step S201, training by using sample data to obtain a plurality of classification models, and performing hierarchical division on each classification model to construct the first classification model library with a multilayer structure.
In this step, the first classification model library is composed of a plurality of classification models (M00, M10, M11, M20 … M39), and the hierarchical structure of each classification model is shown in fig. 3. In this step, the model M00 is a parent model with less classification accuracy and less specific classification feature information, the models M10 and M11 are child models under the parent model M00, the classification feature information contained in M10 and M11 is more than that of M00, and the classification accuracy is higher than that of M00. The same holds true for M20 … M39.
Step S202, receiving first information to be matched;
step S203, respectively carrying out feature importance calculation on the first information by using each classification model in the first classification model library to obtain first feature importance corresponding to each classification model; respectively carrying out probability calculation on the first information by using each classification model of the non-bottom layer to obtain the probability that the first information is divided into each classification model of the lower layer by each classification model; calculating to obtain second feature importance corresponding to each classification model according to the probability of each classification model and the first feature importance corresponding to each classification model; and overlapping the second feature importance corresponding to each classification model to obtain the feature parameters of the information.
In this step, when the first feature importance corresponding to each classification model is obtained through calculation, the following method is specifically adopted: respectively extracting the characteristics of the first information by utilizing each classification model in a first classification model library and combining an attention mechanism to obtain characteristic information corresponding to each classification model; and performing feature importance calculation on the feature information corresponding to each classification model to obtain first feature importance corresponding to each classification model.
In this embodiment, the attention mechanism may be used to measure the importance of each feature from a global perspective. In a classification task, a model of the long-short-term memory neural network algorithm LSTM + attention mechanism attention can be specifically used to obtain the importance degree of each feature, thereby calculating the feature importance of the first information. For example, if the first feature importance of the first information in M20 needs to be calculated, feature extraction may be performed on the first information, for example, 4 pieces of feature information are obtained, and then the importance degrees of the 4 pieces of feature information may be calculated by using M20, so as to obtain the first feature importance of the first information.
When calculating the importance of the second feature, it is necessary to perform probability calculation on the first information by using each classification model of a non-bottom layer, so as to obtain the probability that the first information is divided into each classification model of a lower layer by each classification model. For example, the second feature importance of the first information in the classification model M10 needs to be calculated, the first feature importance of the first information needs to be calculated first, after the first feature importance is obtained, the classification model M00 needs to be used to directly calculate the probability that the first information belongs to the second classification model (i.e., the probability that the first information is divided from M00 to M20), and then the obtained probability is multiplied by the first feature importance to obtain the second feature importance.
Step S204: comparing the characteristic parameter with a preset threshold value, and if the characteristic parameter is greater than or equal to the preset threshold value, determining that the information attribute of the first information is a defined attribute; and if the characteristic importance is smaller than a preset threshold value, determining that the information attribute of the first information is an undefined attribute.
The threshold preset in this step may be set according to actual needs.
In a specific implementation process, the step of determining the information attribute of the first information may be replaced by another method, where the method includes: fitting the characteristic parameters of the sample information by using a first-level support vector machine algorithm to obtain a hypersphere; wherein the information attribute of the sample information is a defined attribute; judging the characteristic parameters of the first information based on the hypersphere, and if the characteristic parameters of the first information are within the range of the hypersphere, determining the information attribute of the first information as a defined attribute; and if the characteristic parameter of the first information is out of the range of the hypersphere, determining the information attribute of the first information as an undefined attribute.
Step S205, according to the information attribute of the first information, matching reply information for the first information; when the information attribute of the first information is an undefined attribute, selecting a predetermined first matching mode to reply to the first information.
In this step, selecting a predetermined first matching manner to reply to the first information specifically includes: and acquiring keywords of the first information, and determining the first matching mode corresponding to the first information according to the keywords so as to reply to the first information. In the implementation process, keywords can be preset, such as keywords related to before selling the product, keywords related to after selling the product, or keywords related to maintaining the product, and the like, and according to the identified keywords, the first information can be distributed to different classes of artificial customer service for reply. If the first information is identified as a question inquiring about maintenance cost and the information attribute is an undefined attribute, the first information can be distributed to a task of manual customer service corresponding to product maintenance so as to respond to the manual customer about the maintenance question to respond to the first information.
In the embodiment of the invention, the characteristic information corresponding to each classification model is extracted from the first information, and then the characteristic information is used for calculating the importance of the characteristic, so that the characteristic parameters can be accurately obtained, and a foundation is laid for determining the information attribute of the subsequent first information.
An embodiment of the present invention provides an information matching method, as shown in fig. 4, including the following steps:
step S301, training by using sample data to obtain a plurality of classification models, and performing hierarchical division on each classification model to construct the first classification model library with a multilayer structure.
Step S302, receiving first information to be matched;
step S303, preprocessing the first information to make the first information meet the specified conditions of the first processing;
step S304, performing first processing on the first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information;
s305, fitting the characteristic parameters of the sample information by using a first-level support vector machine algorithm to obtain a hypersphere; wherein the information attribute of the sample information is a defined attribute; judging the characteristic parameters of the first information based on the hypersphere, and if the characteristic parameters of the first information are within the range of the hypersphere, determining the information attribute of the first information as a defined attribute; and if the characteristic parameter of the first information is out of the range of the hypersphere, determining the information attribute of the first information as an undefined attribute.
Step S306, according to the information attribute of the first information, matching reply information for the first information; when the information attribute of the first information is an undefined attribute, selecting a preset first matching mode to reply to the first information; when the information attribute is a defined attribute, selecting a predetermined second matching mode to match the first information with the reply information; wherein the second matching means comprises: and performing intention identification on the first information, determining an intention, and matching reply information corresponding to the intention for the first information according to the intention.
In this step, when it is determined that the information attribute of the first information is an undefined attribute, selecting a predetermined first matching manner to reply to the first information specifically includes: and acquiring keywords of the first information, and determining the first matching mode corresponding to the first information according to the keywords so as to reply to the first information. In the specific implementation process, the keywords may be preset, for example: keywords related to before sale of the article, keywords related to after sale of the article, or keywords related to maintenance of the article, etc., and based on these identified keywords, the first information can be assigned to different classes of human customer service (pre-sale customer service, post-sale customer service, maintenance customer service) to be replied. If the first information is identified as a question for inquiring maintenance cost and the information attribute is an undefined attribute, the first information can be distributed to a task of manual customer service corresponding to product maintenance, so that a manual customer who answers the maintenance question can answer the first information, and the answering content is more accurate.
In this step, when the information attribute of the first information is determined to be the defined attribute, that is, the information is determined to belong to a question (a question that can be identified in the defined intention space) in the first classification model library, so that intention identification can be directly performed on the first information, an intention can be determined, and then the first information is matched with corresponding answer information directly according to the intention. Where intent recognition is performed, a classification model may be employed to identify the question type to which the first information belongs and then to match the response information thereto.
An embodiment of the present invention provides an information matching apparatus, as shown in fig. 5, including:
the receiving module 1 is used for receiving first information to be matched;
the calculation module 2 is used for performing first processing on the first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information;
a determining module 3, configured to determine an information attribute of the first information according to the feature parameter, where the information attribute includes a defined attribute and an undefined attribute;
the matching module 4 is used for matching reply information for the first information according to the information attribute of the first information; when the information attribute of the first information is an undefined attribute, selecting a predetermined first matching mode to reply to the first information.
Specifically, the information matching apparatus in this embodiment further includes a preprocessing module, where the preprocessing module is configured to: and preprocessing the first information to enable the first information to meet the specified conditions of the first processing.
Specifically, the information matching device further comprises a construction module, and the construction module is configured to: training by using sample data to obtain a plurality of classification models, and carrying out hierarchical division on each classification model to construct the first classification model library with a multilayer structure.
In an implementation process, the calculation module is specifically configured to: respectively carrying out feature importance calculation on the first information by utilizing each classification model in the first classification model library to obtain first feature importance corresponding to each classification model;
respectively carrying out probability calculation on the first information by using each classification model of the non-bottom layer to obtain the probability that the first information is divided into each classification model of the lower layer by each classification model;
calculating to obtain second feature importance corresponding to each classification model according to the probability of each classification model and the first feature importance corresponding to each classification model;
and overlapping the second feature importance corresponding to each classification model to obtain the feature parameters of the information.
Specifically, the calculation module is specifically configured to:
respectively extracting the characteristics of the first information by utilizing each classification model in a first classification model library and combining an attention mechanism to obtain characteristic information corresponding to each classification model;
and performing feature importance calculation on the feature information corresponding to each classification model to obtain first feature importance corresponding to each classification model.
In an implementation process, the determining module is specifically configured to:
comparing the characteristic parameter with a preset threshold value, and if the characteristic parameter is greater than or equal to the preset threshold value, determining that the information attribute of the first information is a defined attribute; and if the characteristic importance is smaller than a preset threshold value, determining that the information attribute of the first information is an undefined attribute.
In an implementation process, the determining module may be further specifically configured to:
fitting the characteristic parameters of the sample information by using a first-level support vector machine algorithm to obtain a hypersphere; wherein the information attribute of the sample information is a defined attribute;
judging the characteristic parameters of the first information based on the hypersphere, and if the characteristic parameters of the first information are within the range of the hypersphere, determining the information attribute of the first information as a defined attribute; and if the characteristic parameter of the first information is out of the range of the hypersphere, determining the information attribute of the first information as an undefined attribute.
Specifically, the matching module is specifically configured to:
and acquiring keywords of the first information, and determining the first matching mode corresponding to the first information according to the keywords so as to reply to the first information.
Optionally, the matching module is further configured to: when the information attribute is a defined attribute, selecting a predetermined second matching mode to match the first information with the reply information; wherein the second matching means comprises: performing intention identification on the first information, determining an intention, and matching reply information corresponding to the intention for the first information according to the intention
The embodiment of the invention determines whether the attribute of the information is a defined attribute or an undefined attribute by calculating the characteristic parameter of the information to be matched. When the information attribute is determined to be an undefined attribute, namely the information is determined to be information outside the first classification model base (namely the defined intention space), and then the first information is replied by utilizing a predetermined first matching mode, so that the situation that the information cannot be accurately identified through the intention space and wrong reply information is matched for the information is avoided, and the accuracy of information matching is improved.
The present embodiment provides a storage medium having a computer program stored thereon, the computer program, when executed by a processor, implementing the steps of:
step one, receiving first information to be matched;
secondly, performing first processing on the first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information;
determining an information attribute of the first information according to the characteristic parameter, wherein the information attribute comprises a defined attribute and an undefined attribute;
matching reply information for the first information according to the information attribute of the first information; when the information attribute of the first information is an undefined attribute, selecting a predetermined first matching mode to reply to the first information.
For the specific embodiment of the above method steps, reference may be made to the above embodiment of any information matching method, and details are not repeated here.
The embodiment of the invention determines whether the attribute of the information is a defined attribute or an undefined attribute by calculating the characteristic parameter of the information to be matched. When the information attribute is determined to be an undefined attribute, namely the information is determined to be information outside the intention space, and then the first information is replied by using a predetermined first matching mode, so that the situation that the information cannot be accurately identified through the intention space and wrong reply information is matched for the information is avoided, and the accuracy of information matching is improved.
The above embodiments are only exemplary embodiments of the present invention, and are not intended to limit the present invention, and the scope of the present invention is defined by the claims. Various modifications and equivalents may be made by those skilled in the art within the spirit and scope of the present invention, and such modifications and equivalents should also be considered as falling within the scope of the present invention.
Claims (11)
1. An information matching method, characterized by comprising the steps of:
receiving first information to be matched;
performing first processing on the first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information;
determining an information attribute of the first information according to the characteristic parameter, wherein the information attribute comprises a defined attribute and an undefined attribute;
matching reply information for the first information according to the information attribute of the first information; when the information attribute of the first information is an undefined attribute, selecting a predetermined first matching mode to reply to the first information.
2. The method of claim 1, wherein prior to calculating the characteristic parameters for obtaining the information, the method further comprises: and preprocessing the first information to enable the first information to meet the specified conditions of the first processing.
3. The method of claim 1, further comprising: training by using sample data to obtain a plurality of classification models, and carrying out hierarchical division on each classification model to construct the first classification model library with a multilayer structure.
4. The method according to claim 1, wherein the first processing is performed on the first information based on a first classification model library composed of a plurality of classification models in a hierarchical relationship to obtain the characteristic parameter of the first information, and specifically includes:
respectively carrying out feature importance calculation on the first information by utilizing each classification model in the first classification model library to obtain first feature importance corresponding to each classification model;
respectively carrying out probability calculation on the first information by using each classification model of the non-bottom layer to obtain the probability that the first information is divided into each classification model of the lower layer by each classification model;
calculating to obtain second feature importance corresponding to each classification model according to the probability of each classification model and the first feature importance corresponding to each classification model;
and overlapping the second feature importance corresponding to each classification model to obtain the feature parameters of the information.
5. The method according to claim 4, wherein the performing feature importance calculation on the first information by using each classification model in the first classification model library to obtain the first feature importance corresponding to each classification model specifically comprises:
respectively extracting the characteristics of the first information by utilizing each classification model in a first classification model library and combining an attention mechanism to obtain characteristic information corresponding to each classification model;
and performing feature importance calculation on the feature information corresponding to each classification model to obtain first feature importance corresponding to each classification model.
6. The method according to claim 1, wherein the determining an information attribute of the first information according to the characteristic parameter, the information attribute including a defined attribute and an undefined attribute, specifically includes:
comparing the characteristic parameter with a preset threshold value, and if the characteristic parameter is greater than or equal to the preset threshold value, determining that the information attribute of the first information is a defined attribute; and if the characteristic importance is smaller than a preset threshold value, determining that the information attribute of the first information is an undefined attribute.
7. The method according to claim 1, wherein the determining an information attribute of the first information according to the characteristic parameter, the information attribute including a defined attribute and an undefined attribute, specifically includes:
fitting the characteristic parameters of the sample information by using a first-level support vector machine algorithm to obtain a hypersphere; wherein the information attribute of the sample information is a defined attribute;
judging the characteristic parameters of the first information based on the hypersphere, and if the characteristic parameters of the first information are within the range of the hypersphere, determining the information attribute of the first information as a defined attribute; and if the characteristic parameter of the first information is out of the range of the hypersphere, determining the information attribute of the first information as an undefined attribute.
8. The method of claim 1, wherein said selecting a predetermined first matching means to reply to said first message comprises:
and acquiring keywords of the first information, and determining the first matching mode corresponding to the first information according to the keywords so as to reply to the first information.
9. The method of claim 1, wherein the method further comprises: when the information attribute is a defined attribute, selecting a predetermined second matching mode to match the first information with the reply information; wherein the second matching means comprises: and performing intention identification on the first information, determining an intention, and matching reply information corresponding to the intention for the first information according to the intention.
10. An information matching apparatus, comprising:
the receiving module is used for receiving first information to be matched;
the calculation module is used for carrying out first processing on the first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information;
a determining module, configured to determine an information attribute of the first information according to the feature parameter, where the information attribute includes a defined attribute and an undefined attribute;
the matching module is used for matching reply information for the first information according to the information attribute of the first information; when the information attribute of the first information is an undefined attribute, selecting a predetermined first matching mode to reply to the first information.
11. A storage medium having a computer program stored thereon, the computer program when executed by a processor implementing the steps of:
receiving first information to be matched;
performing first processing on the first information based on a first classification model library formed by a plurality of classification models according to a hierarchical relationship to obtain characteristic parameters of the first information;
determining an information attribute of the first information according to the characteristic parameter, wherein the information attribute comprises a defined attribute and an undefined attribute;
matching reply information for the first information according to the information attribute of the first information; when the information attribute of the first information is an undefined attribute, selecting a predetermined first matching mode to reply to the first information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910849854.XA CN110659356B (en) | 2019-09-09 | 2019-09-09 | Information matching method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910849854.XA CN110659356B (en) | 2019-09-09 | 2019-09-09 | Information matching method, device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110659356A true CN110659356A (en) | 2020-01-07 |
| CN110659356B CN110659356B (en) | 2022-07-26 |
Family
ID=69038046
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910849854.XA Active CN110659356B (en) | 2019-09-09 | 2019-09-09 | Information matching method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110659356B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107330021A (en) * | 2017-06-20 | 2017-11-07 | 北京神州泰岳软件股份有限公司 | Data classification method, device and equipment based on multiway tree |
| CN109615407A (en) * | 2018-10-19 | 2019-04-12 | 国家电网有限公司 | Online electric power wisdom client service method based on artificial intelligence |
-
2019
- 2019-09-09 CN CN201910849854.XA patent/CN110659356B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107330021A (en) * | 2017-06-20 | 2017-11-07 | 北京神州泰岳软件股份有限公司 | Data classification method, device and equipment based on multiway tree |
| CN109615407A (en) * | 2018-10-19 | 2019-04-12 | 国家电网有限公司 | Online electric power wisdom client service method based on artificial intelligence |
Non-Patent Citations (1)
| Title |
|---|
| 周智敏等: "《超宽带地表穿透成像雷达》", 31 March 2013, 国防工业出版社 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110659356B (en) | 2022-07-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108153780B (en) | Man-machine conversation device and method for realizing man-machine conversation | |
| US20180322416A1 (en) | Feature extraction and classification method based on support vector data description and system thereof | |
| CN106649739B (en) | Multi-round interactive information inheritance identification method and device and interactive system | |
| CN113377936B (en) | Intelligent question and answer method, device and equipment | |
| CN112163424A (en) | Data labeling method, device, equipment and medium | |
| CN113297365B (en) | User intention judging method, device, equipment and storage medium | |
| CN104750674A (en) | Man-machine conversation satisfaction degree prediction method and system | |
| CN112463949B (en) | Data recall method and system, interaction method and interaction system | |
| CN113409174A (en) | Knowledge point evaluation method and device | |
| CN111428639A (en) | Training method of gesture recognition model, gesture recognition method and device | |
| CN113342958A (en) | Question-answer matching method, text matching model training method and related equipment | |
| CN112256856A (en) | Robot dialogue method, device, electronic device and storage medium | |
| CN109753561B (en) | Automatic reply generation method and device | |
| EP4273750A1 (en) | Data processing method and apparatus, computing device, and test simplification device | |
| CN112818689B (en) | Entity identification method, model training method and device | |
| CN112288584B (en) | Insurance report processing method and device, computer readable medium and electronic equipment | |
| CN112053245B (en) | Information evaluation method and system | |
| CN105373568B (en) | Problem answers Auto-learning Method and device | |
| CN110334080B (en) | Knowledge base construction method for realizing autonomous learning | |
| CN110659356B (en) | Information matching method, device and storage medium | |
| CN117149969A (en) | Answer acquisition method and device based on intelligent question-answer model and electronic equipment | |
| CN112069833A (en) | Log analysis method, log analysis device and electronic equipment | |
| CN113946651B (en) | Maintenance knowledge recommendation method and device, electronic equipment, medium and product | |
| CN112365302B (en) | Product recommendation network training method, device, equipment and medium | |
| CN114186048A (en) | Question-answer replying method and device based on artificial intelligence, computer equipment and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |