CN116028614B - Information processing method, device, equipment and readable storage medium - Google Patents
Information processing method, device, equipment and readable storage medium Download PDFInfo
- Publication number
- CN116028614B CN116028614B CN202310316954.2A CN202310316954A CN116028614B CN 116028614 B CN116028614 B CN 116028614B CN 202310316954 A CN202310316954 A CN 202310316954A CN 116028614 B CN116028614 B CN 116028614B
- Authority
- CN
- China
- Prior art keywords
- classification
- entity
- attribute
- dividing
- question
- 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.)
- Active
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The application discloses an information processing method, an information processing device, information processing equipment and a readable storage medium, and relates to the technical field of natural language processing so as to improve the processing efficiency of an AI dialogue system. The method comprises the following steps: acquiring input information of a user aiming at a target field; taking the input information as the input of a knowledge question-answer engine model and operating the knowledge question-answer engine model to obtain a query result; outputting the query result to the user; the knowledge question-answering engine model is obtained based on an intention tree of the target field, and the intention tree of the target field is obtained by combining a framework of the target field and a general intention tree. The embodiment of the application can improve the processing efficiency of the AI dialogue system.
Description
Technical Field
The present disclosure relates to the field of natural language processing technologies, and in particular, to an information processing method, an apparatus, a device, and a readable storage medium.
Background
With the continued breakthrough development of NLP (Natural Language Processing ) technology in recent years, AI (Artificial Intelligence ) dialog systems, in particular task-and knowledge-based dialog systems, are becoming increasingly mature in application in the fields of intelligent marketing, intelligent customer service, intelligent assistant, etc.
At present, a large number of business experts, operators, product managers, system engineers, algorithm engineers and other various role personnel are required to build a knowledge dialogue system in a collaborative mode, and even a small knowledge question-answering system needs a research and development period of weeks or even months. Therefore, the existing scheme cannot meet the requirement of autonomous and controllable and efficient construction of the AI dialogue system, and the processing efficiency of the AI dialogue system is correspondingly reduced.
Disclosure of Invention
The embodiment of the application provides an information processing method, an information processing device, information processing equipment and a readable storage medium, so as to improve the processing efficiency of an AI dialogue system.
In a first aspect, an embodiment of the present application provides an information processing method, including:
acquiring input information of a user aiming at a target field;
taking the input information as the input of a knowledge question-answer engine model and operating the knowledge question-answer engine model to obtain a query result;
outputting the query result to the user;
the knowledge question and answer engine model is obtained based on an intention tree of the target field, wherein the intention tree of the target field is obtained by combining a framework (schema) of the target field and a general intention tree.
In a second aspect, an embodiment of the present application provides an information processing apparatus, including:
the first acquisition module is used for acquiring input information of a user aiming at a target field;
the second acquisition module is used for taking the input information as the input of the knowledge question-answering engine model and running the knowledge question-answering engine model to obtain a query result;
the first output module is used for outputting the query result to the user;
the knowledge question-answering engine model is obtained based on an intention tree of the target field, and the intention tree of the target field is obtained by combining a schema and a general intention tree of the target field.
In a third aspect, embodiments of the present application further provide an electronic device, including: the information processing device includes a memory, a processor, and a program stored on the memory and executable on the processor, which when executed implements the steps in the information processing method as described above.
In a fourth aspect, embodiments of the present application also provide a readable storage medium having a program stored thereon, which when executed by a processor, implements the steps in the information processing method as described above.
In the embodiment of the application, the knowledge question-answering engine model is obtained based on an intention tree of the target field, wherein the intention tree of the target field is obtained by combining a schema and a general intention tree of the target field. Therefore, by utilizing the scheme of the embodiment of the application, the scheme of the target field and the general purpose intention tree can be combined to obtain the intention tree of the target field, and the knowledge question-answering engine model obtained based on the target field is utilized to process the input information of the user, so that the processing efficiency of the input information of the user can be improved.
Drawings
FIG. 1 is a flow chart of an information processing method provided in an embodiment of the present application;
FIG. 2 is one of the schematics provided by embodiments of the present application for building the generic intent tree;
FIG. 3 is a second schematic diagram of building the generic intent tree provided by an embodiment of the present application;
FIG. 4 is a third schematic diagram of building the generic intent tree provided by embodiments of the present application;
fig. 5 is one of schematic structural diagrams of an information processing apparatus of an embodiment of the present application;
FIG. 6 is a schematic diagram of a process of a machine collaborative building field question-answering robot according to an embodiment of the present application;
fig. 7 is a second schematic structural view of the information processing apparatus according to the embodiment of the present application.
Detailed Description
In the embodiment of the application, the term "and/or" describes the association relationship of the association objects, which means that three relationships may exist, for example, a and/or B may be represented: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship.
The term "plurality" in the embodiments of the present application means two or more, and other adjectives are similar thereto.
The following description of the technical solutions in the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
Referring to fig. 1, fig. 1 is a flowchart of an information processing method provided in an embodiment of the present application, as shown in fig. 1, including the following steps:
In the embodiment of the application, the input information includes, but is not limited to, text information, voice information, image information and the like.
And 102, taking the input information as the input of a knowledge question-answer engine model and operating the knowledge question-answer engine model to obtain a query result.
The knowledge question and answer engine comprises an intention recognition model and a knowledge extraction model. The intention recognition model is used for recognizing knowledge inquiry intention corresponding to the input information; the knowledge extraction model is used for inquiring according to the obtained knowledge inquiry intention to obtain knowledge inquiry answers or results, namely inquiry results. That is, the knowledge question-answering engine can realize that the user inputs sentences to the dialogue engine for analysis and reasoning according to the input information of the user, so as to obtain the knowledge question-asking intention of the user, and further inquire corresponding answers according to the knowledge question-asking intention of the user, and return the answers to the user.
Specifically, in this step, the input information is input to an intention recognition model, and a knowledge inquiry intention corresponding to the input information is obtained. And then, inputting the knowledge inquiry intention into the knowledge extraction model to obtain the inquiry result.
The knowledge question-answering engine model is obtained based on an intention tree of the target field, and the intention tree of the target field is obtained by combining a framework schema and a general intention tree of the target field. The target area may be any area, such as a financial area, a communication area, and the like.
The meaning of the general purpose intention tree is a general purpose intention tree constructed without distinguishing specific application fields. In the intention tree, different processing modes of knowledge question and answer types, such as different entities, attributes, relationships, different calculation modes and the like are embodied. The schema of the target domain may be understood as an entity, a relationship, an attribute of the target domain constructed by a business expert of the target domain, and further may further include a question type, a question-answer mode, etc. formed according to the entity, the attribute, the relationship. For example, for the film field, there are people, movies, company 3 entities, there are three relationship types of actors, directors, investors, people have three types of attributes of age, height and occupation, and the problem pattern of people is "who is? "how much is the question pattern combined with the attribute" height? "what occupation? "who is the highest in height? "and the like.
Optionally, the embodiment of the present application may further include: the generic intent tree is constructed. Referring to fig. 2, 3, and 4, fig. 2, 3, and 4 are schematic diagrams for constructing the general purpose intention tree provided in the embodiments of the present application. The general purpose tree can be divided into two parts including information (shown in fig. 3) and entities (shown in fig. 2 and 4), wherein the general purpose tree of the attribute part in the entities can be shown by referring to fig. 4, and the general purpose tree of the entity and the relationship part can be shown by referring to fig. 2.
In connection with fig. 2, 3, 4, in an embodiment of the present application, the generic intent tree may be constructed as follows:
the question-answer types of the known knowledge graph are classified into a question-answer type of inquiring real-time knowledge and a question-answer type of inquiring information. This partitioning result may be used as a first layer of the generic intent tree. The question-answer type of the query real-time knowledge can also be understood as a question-answer type of the query real-time knowledge, for example, how many cities are in Shandong province, etc. The question-answer type of the query information refers to a question-answer type that needs to be calculated by a certain logic operation, for example, which of the respective cities of the Shandong province has the largest area, and the like.
For the question-answer type of the query real-time knowledge, dividing according to the entity, the relation and the attribute to obtain entity classification, relation classification and attribute classification; for the question-answer type of the query information, the query information is classified into a ranking classification and a numerical calculation classification according to a processing mode. This partitioning result may be used as a second layer of the generic intent tree.
Dividing the entity classification and the attribute classification according to first reference information; and dividing the sorting classification and the numerical calculation classification according to second reference information. This partitioning result may be used as one or more layers after the third layer of the generic intent tree.
Wherein the first reference information includes: the number of entities, the number of hops between the entities, whether constraint conditions exist, the number of attributes and the number of hops between the attributes; the second reference information includes: number of hops between entities. The meaning of the first reference information and the second reference information may be further extended according to needs, which is not limited in the embodiment of the present application.
Specifically, referring to fig. 2 and fig. 4, according to the first reference information, dividing the entity class and the attribute class respectively may specifically include:
for the entity classification: dividing into single entity classification and multi-entity classification according to the number of the entities; for each single entity classification included in the single entity classification and the multi-entity classification, dividing the single entity classification and the multi-hop classification according to the hop count among the entities; for the single-hop classification, the classification is divided into: entity sub-classification, entity and attribute sub-classification, attribute sub-classification; dividing the entity sub-classification and the attribute sub-classification according to whether constraint conditions exist or not to obtain a classification result; dividing the classification result according to a mode of forming single intention, and forming actions by corresponding to the slots; and for the sub-classification of the entity and the attribute, dividing the sub-classification according to a mode of forming single intention, and forming actions by corresponding to the slots.
For the attribute classification: dividing into single attribute classification and multi-attribute classification according to the number of the attributes; for each single attribute classification included in the single attribute classification and the multi-attribute classification, dividing the single attribute classification and the multi-hop classification into single-hop classification and multi-hop classification according to the hop count among the attributes; for the single-hop classification, the classification is divided into: entity sub-classification and attribute sub-classification; dividing the entity sub-classification and the attribute sub-classification according to whether constraint conditions exist or not to obtain a classification result; on the basis of the obtained classification result, dividing according to a mode of forming single intention, and corresponding to the slot positions to form behavior actions.
Optionally, for the entity classification: for the constraint condition classification in the classification result of the entity sub-classification, performing logic operation after corresponding to the slot to form the actions; for the unconstrained condition classification in the classification result of the attribute sub-classification, performing logic operation after corresponding to the slot to form the actions; for entity and attribute sub-classifications: performing logic operation after corresponding to the slot positions to form the actions; and for the attribute classification, carrying out logic operation after corresponding to the slot to form the actions for the constraint condition classification in the classification result of the entity sub-classification. That is, in this manner, the formed actions are obtained after corresponding to the slot and performing a logical operation. Optionally, for the attribute classification, only for some cases, performing a logic operation after the corresponding slot to form the actions, for example, for the slot formed in the cases of single-intended single-entity+multi-relation- > (single-entity|attribute), single-intended single-entity+single-relation- > (single-entity|attribute), single-intended multi-entity+multi-relation- > (single-entity|attribute), performing a logic operation after the corresponding slot to form the actions. Wherein the logical operation includes, but is not limited to, an and or.
In the above-described processing of entity classes, a multi-entity class may be considered to be formed from a plurality of single-entity classes. Thus, for each single entity class it includes, reference may be made to the way in which a single entity is classified. Similarly, in the above-described processing of the attribute classification row, the multi-attribute classification may be considered to be formed of a plurality of single-attribute classifications. Therefore, the processing method for classifying each single attribute included in the image can be referred to as the processing method for classifying the single attribute.
Specifically, referring to fig. 3, fig. 3 is a schematic diagram of the information portion of the generic intent tree. Dividing the sorting classification and the numerical calculation classification according to second reference information respectively, including:
dividing the sorting classification into single-hop classification and multi-hop classification according to the hop count among the entities, and dividing the numerical calculation classification into average value classification and summation classification;
dividing the single-hop classification according to the entity, the relation and the attribute, and obtaining ordering rules under different classifications, such as sort (e 1, e2, e 3), sort (r 1, r2, r 3), sort (a 1, a2, a 3), wherein e1, e2, e3 respectively represent different entities, r1, r2, r3 respectively represent different relations, and a1, a2, a3 respectively represent different attributes;
and dividing the single-hop classification of the summation classification according to the entity, the relation and the attribute, and obtaining the summation rules under different classifications. Such as sum (e 1, e2, e 3), sum (r 1, r2, r 3), sum (a 1, a2, a 3), wherein e1, e2, e3 respectively represent different entities, r1, r2, r3 respectively represent different relationships, and a1, a2, a3 respectively represent different properties.
Through the above way, a corresponding Query sentence (Query Cypher) is formed, and then a general knowledge graph Query intention tree which is irrelevant to a specific domain, namely a general intention tree, is formed.
And combining the schema of the target field and one or more of the entity, the attribute and the relation in the general purpose intention tree on the basis of the formed general purpose tree to obtain the intention tree of the target field. For example, query pattern cases under unconstrained, computationally free, context free semantics are: entity + relationship query entity.
In the process of constructing the intention tree in the target field, by adopting the man-machine cooperation full-flow automatic construction, the field business expert can independently and controllably construct, and the participation of non-field related personnel, such as model algorithm personnel, is reduced, so that the research and development period of a knowledge question-answering robot in a certain field can be shortened, the development cost is reduced, and the like.
In this step, there are various ways of outputting the query result to the user, for example, outputting a text query result, a voice query result, a picture query result, and the like.
In the embodiment of the application, the knowledge question-answering engine model is obtained based on an intention tree of the target field, wherein the intention tree of the target field is obtained by combining a schema and a general intention tree of the target field. Therefore, by utilizing the scheme of the embodiment of the application, the scheme of the target field and the general purpose intention tree can be combined to obtain the intention tree of the target field, and the knowledge question-answering engine model obtained based on the target field is utilized to process the input information of the user, so that the processing efficiency of the input information of the user can be improved.
Based on the above embodiments, the embodiments of the present application may further extract an entity, an attribute, and a relationship set from a schema of the target domain, fill the entity, the attribute, and the relationship set into an intent tree of the target domain, obtain an instance problem set, and train the knowledge question engine model by using the training set, where the knowledge question engine includes an intent recognition model and a knowledge extraction model.
Specifically, the entity, the relationship and the attribute obtained from the schema of the target field can be filled into the entity type, the attribute type and the relationship type of the query mode of the intention tree of the target field. For example, by an entity+relationship-entity in the intent tree of the target domain, the example problem can be populated as: who is the "director" of "Duyue"? What is the "edit" of "Duyun" called?
According to the obtained example problem set, a training set is selected, further, a test set, a verification set and the like can be selected, so that training, testing and verification of the model are completed. By the method, the generation efficiency of the training set, the test set and the verification set can be improved, and therefore the intention recognition model and the knowledge extraction model can be constructed more efficiently.
Referring to fig. 5, fig. 5 is a schematic structural view of an information processing apparatus according to an embodiment of the present application. The method mainly comprises the following steps: a schema management platform 301, a question and answer capability construction platform 302 and a question and answer service output platform 303. The scheme management platform mainly realizes the collaborative construction of the field scheme of multiple people and establishes a general purpose tree irrelevant to the field; the question-answering capability construction platform 302 mainly realizes automatic mapping of a general purpose tree and a field schema, establishes a field purpose tree, automatically constructs intention sample data (training sample) on the basis, inputs intention and intention sample data with sequence labels to the model self-training platform, generates an intention recognition model and a knowledge extraction model (entity, attribute and relation) through automatic training, and configures the model into intention and slot extraction components related to a dialogue engine to form construction of a question-answering capability device; and outputting the formed question and answer capability device interface to a question and answer service output platform, and configuring a knowledge base relied on by the bottom layer and related service modules into the platform to form a knowledge question and answer robot in the new field, so that a question and answer robot device in the field is constructed, and a question and answer service of the knowledge in the professional field is provided for the outside. The question and answer service output platform 303 mainly calls a question and answer capability construction platform to perform intention recognition and knowledge extraction according to input information of a user, and provides relevant query results for the user.
Referring to fig. 6, fig. 6 is a schematic process diagram of a machine collaboration construction field question-answering robot according to an embodiment of the present application. In connection with fig. 6, the process may include:
first, a domain business expert is determined, and a schema collaboration building apparatus is provided. Business specialists can cooperate with co-building domain schemas to provide, in addition to basic entity, attribute and relationship designs, corresponding question query patterns for different types of entities and attributes, for example, what are common questions of a person's entity including what are called, how much are the heights (attributes)? What is the occupation (attribute? The corresponding query patterns include "what," "what," etc., and the question patterns may be configured with constraints according to entity and attribute types. The method specifically comprises the following steps:
the business specialists constructed by the field schema are determined, for example, a business specialist team consisting of Zhang three, li four and Wang five is determined. The established business expert cooperates with the co-construction field schema, for example, for the field of movies, the person, the movie, the company 3 entity are designed, three relationship types are provided among the three, namely, actor, director and investor, the person has three types of attributes of age, height and occupation, and the problem mode of the person has the meaning of? "how much is the question pattern combined with the attribute" height? "what occupation? "who is the highest in height? "and the like.
Secondly, preparing a knowledge base of a certain field, establishing a mapping relation between a general purpose tree (shown in fig. 2) designed by a session expert and a schema of the field to form a purpose tree of a specific field, establishing a field purpose library, and further forming a field purpose set. Meanwhile, the training data can be automatically constructed by combining the existing knowledge base.
Wherein the construction of the generic intent tree may be referred to the description of the method embodiments described above.
Specifically, when the instance problem is constructed, the entity, the relationship and the attribute obtained from the schema of the target field can be filled into the entity type, the attribute type and the relationship type of the query mode of the intention tree of the target field. For example, by an entity+relationship-entity in the intent tree of the target domain, the example problem can be populated as: who is the "director" of "Duyue"? What is the "edit" of "Duyun" called?
When training data is constructed, the entity, attribute and relation set are extracted from the knowledge graph constructed by the schema. And filling the obtained entity, relationship and attribute into the domain intention tree, and forming an instance problem set by the entity type, the attribute type and the relationship type in the obtained query mode. For example, the resulting entity+relationship-entity can fill who is the "director" for the case "unique moon? What is the "edit" of "Duyun" called? The example problem set is divided into a training set, a verification set and a test set, and training, verification, test and the like of the model are respectively carried out.
Specifically, the acquired data (training set) with the intention and sequence labels is transmitted to a model automatic training platform, and training of the intention recognition model and the knowledge extraction model is started by the training set on the model automatic training platform. After model training, an intention recognition model and a knowledge extraction model required by the knowledge question-answering engine in the field are formed, and a corresponding model service interface is formed.
And finally, inputting training data into a model self-training platform to obtain a model, and configuring the model to a knowledge question-answering engine to form a knowledge question-answering robot in the field, so as to provide knowledge question-answering service and device in the field for users.
The dialogue engine configuration module configures the intention recognition and knowledge extraction function requirements to corresponding interfaces in the knowledge question and answer engine to form NLU (Natural Language Understanding ) capability required by the question and answer system. And constructing a universal intention knowledge query interface according to the obtained universal intention tree, and starting the service.
After receiving the input information of the user, the question-answering capability interface is in butt joint with the interface of the knowledge question-answering service system, so that the user inputs sentences to a dialogue engine to conduct analysis and reasoning, and corresponding answers are queried according to the knowledge query intention of the user and returned to the user.
In the scheme of the embodiment of the application, the general intention tree is constructed and combined with the schema of the domain, so that the intention tree of the domain is formed. By the method, the research and development period of the field knowledge question-answering robot can be shortened, the enterprise cost is reduced, and meanwhile, the processing efficiency and speed of input information of a user can be improved.
Referring to fig. 7, fig. 7 is a block diagram of an information processing apparatus provided in an embodiment of the present application. As shown in fig. 7, the apparatus may include:
a first obtaining module 501, configured to obtain input information of a user for a target area; the second obtaining module 502 is configured to take the input information as input of a knowledge question-answer engine model and operate the knowledge question-answer engine model to obtain a query result; a first output module 503, configured to output the query result to the user; the knowledge question-answering engine model is obtained based on an intention tree of the target field, and the intention tree of the target field is obtained by combining a framework schema and a general intention tree of the target field.
Optionally, the apparatus may further include:
the first processing module is used for constructing the general purpose intention tree in the following way:
dividing the question-answer type of the known knowledge graph into a question-answer type of inquiring real-time knowledge and a question-answer type of inquiring information;
for the question-answer type of the query real-time knowledge, dividing according to the entity, the relation and the attribute to obtain entity classification, relation classification and attribute classification; for the question-answer type of the query information, classifying the query information into a sorting classification and a numerical calculation classification according to a processing mode;
dividing the entity classification and the attribute classification according to first reference information; dividing the sorting classification and the numerical calculation classification according to second reference information;
wherein the first reference information includes: the number of entities, the number of hops between the entities, whether constraint conditions exist or not, and the number of attributes; the second reference information includes: number of hops between entities.
Optionally, the classifying the entity and the attribute according to the first reference information includes:
for the entity classification: dividing into single entity classification and multi-entity classification according to the number of the entities; for each single entity classification included in the single entity classification and the multi-entity classification, dividing the single entity classification and the multi-hop classification according to the hop count among the entities; for the single-hop classification, the classification is divided into: entity sub-classification, entity and attribute sub-classification, attribute sub-classification; dividing the entity sub-classification and the attribute sub-classification according to whether constraint conditions exist or not to obtain a classification result; dividing the classification result according to a mode of forming single intention, and forming behavior actions by corresponding to the slots; dividing the sub-classification of the entity and the attribute according to a mode of forming single intention, and forming behavior actions by corresponding to the slot positions;
for the attribute classification: dividing into single attribute classification and multi-attribute classification according to the number of the attributes; for each single attribute classification included in the single attribute classification and the multi-attribute classification, dividing the single attribute classification and the multi-hop classification into single-hop classification and multi-hop classification according to the hop count among the attributes; for the single-hop classification, the classification is divided into: entity sub-classification and attribute sub-classification; dividing the entity sub-classification and the attribute sub-classification according to whether constraint conditions exist or not to obtain a classification result; on the basis of the obtained classification result, dividing according to a mode of forming single intention, and corresponding to the slot positions to form behavior actions.
Optionally, for the entity classification: for the constraint condition classification in the classification result of the entity sub-classification, performing logic operation after corresponding to the slot to form the actions; for the unconstrained condition classification in the classification result of the attribute sub-classification, performing logic operation after corresponding to the slot to form the actions; for the sub-classification of the entity and the attribute, performing logic operation after corresponding to the slot to form the actions;
for the attribute classification: and carrying out logic operation after corresponding to the slot positions on the constraint condition classification in the classification result of the entity sub-classification to form the actions.
Optionally, the classifying the sorting classification and the numerical calculation classification according to the second reference information includes:
dividing the sorting classification into single-hop classification and multi-hop classification according to the hop count among the entities, and dividing the numerical calculation classification into average value classification and summation classification;
dividing the single-hop classification according to the entity, the relation and the attribute, and obtaining the ordering rules under different classifications;
and dividing the single-hop classification of the summation classification according to the entity, the relation and the attribute, and obtaining the summation rules under different classifications.
Optionally, the apparatus may further include:
and the second processing module is used for combining the schema of the target field with one or more of the entities, the attributes and the relations in the general purpose intention tree to obtain the intention tree of the target field.
Optionally, the apparatus may further include:
the third acquisition module is used for extracting an entity, an attribute and a relation set from the schema of the target field;
a fourth obtaining module, configured to populate the entity, attribute, and relationship set into an intent tree in the target field, to obtain an instance problem set;
a fifth obtaining module, configured to select a training set from the set of instance questions;
and the third processing module is used for training the knowledge question-answering engine model by utilizing the training set, wherein the knowledge question-answering engine comprises an intention recognition model and a knowledge extraction model.
Optionally, the second obtaining module includes:
the first acquisition submodule is used for inputting the input information into an intention recognition model to obtain knowledge inquiry intention corresponding to the input information;
and inputting the knowledge inquiry intention into the knowledge extraction model to obtain the inquiry result.
The device provided in the embodiment of the present application may execute the above method embodiment, and its implementation principle and technical effects are similar, and this embodiment will not be described herein again.
It should be noted that, in the embodiment of the present application, the division of the units is schematic, which is merely a logic function division, and other division manners may be implemented in actual practice. In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a processor-readable storage medium. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution, in the form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (processor) to perform all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The embodiment of the application provides electronic equipment, which comprises: a memory, a processor, and a program stored on the memory and executable on the processor; the processor is configured to read a program in the memory to implement the steps in the information processing method as described above.
The embodiment of the application further provides a readable storage medium, on which a program is stored, where the program, when executed by a processor, implements each process of the above embodiment of the information processing method, and can achieve the same technical effects, so that repetition is avoided, and no further description is given here. The readable storage medium may be any available medium or data storage device that can be accessed by a processor, including, but not limited to, magnetic memories (e.g., floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical memories (e.g., CD, DVD, BD, HVD, etc.), semiconductor memories (e.g., ROM, EPROM, EEPROM, nonvolatile memories (NAND FLASH), solid State Disks (SSD)), etc.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. In light of such understanding, the technical solutions of the present application may be embodied essentially or in part in the form of a software product stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) and including instructions for causing a terminal (which may be a cell phone, computer, server, air conditioner, or network device, etc.) to perform the methods described in the embodiments of the present application.
The embodiments of the present application have been described above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those of ordinary skill in the art without departing from the spirit of the present application and the scope of the claims, which are also within the protection of the present application.
Claims (9)
1. An information processing method, characterized by comprising:
acquiring input information of a user aiming at a target field;
taking the input information as the input of a knowledge question-answer engine model and operating the knowledge question-answer engine model to obtain a query result;
outputting the query result to the user;
the knowledge question-answering engine model is obtained based on an intention tree of the target field, wherein the intention tree of the target field is obtained by combining a framework schema and a general intention tree of the target field;
wherein the method further comprises:
constructing the generic intent tree, comprising:
dividing the question-answer type of the known knowledge graph into a question-answer type of inquiring real-time knowledge and a question-answer type of inquiring information;
for the question-answer type of the query real-time knowledge, dividing according to the entity, the relation and the attribute to obtain entity classification, relation classification and attribute classification; for the question-answer type of the query information, classifying the query information into a sorting classification and a numerical calculation classification according to a processing mode;
dividing the entity classification and the attribute classification according to first reference information; dividing the sorting classification and the numerical calculation classification according to second reference information;
wherein the first reference information includes: the number of entities, the number of hops between the entities, whether constraint conditions exist, the number of attributes and the number of hops between the attributes; the second reference information includes: number of hops between entities;
wherein the dividing the entity classification and the attribute classification according to the first reference information includes:
for the entity classification: dividing into single entity classification and multi-entity classification according to the number of the entities; for each single entity classification included in the single entity classification and the multi-entity classification, dividing the single entity classification and the multi-hop classification according to the hop count among the entities; for the single-hop classification, the classification is divided into: entity sub-classification, entity and attribute sub-classification, attribute sub-classification; dividing the entity sub-classification and the attribute sub-classification according to whether constraint conditions exist or not to obtain a classification result; dividing the classification result according to a mode of forming single intention, and forming behavior actions by corresponding to the slots; dividing the sub-classification of the entity and the attribute according to a mode of forming single intention, and forming behavior actions by corresponding to the slot positions;
for the attribute classification: dividing into single attribute classification and multi-attribute classification according to the number of the attributes; for each single attribute classification included in the single attribute classification and the multi-attribute classification, dividing the single attribute classification and the multi-hop classification into single-hop classification and multi-hop classification according to the hop count among the attributes; for the single-hop classification, the classification is divided into: entity sub-classification and attribute sub-classification; dividing the entity sub-classification and the attribute sub-classification according to whether constraint conditions exist or not to obtain a classification result; on the basis of the obtained classification result, dividing according to a mode of forming single intention, and corresponding to the slot positions to form behavior actions.
2. The method according to claim 1, wherein the method further comprises:
for the entity classification: for the constraint condition classification in the classification result of the entity sub-classification, performing logic operation after corresponding to the slot to form the actions; for the unconstrained condition classification in the classification result of the attribute sub-classification, performing logic operation after corresponding to the slot to form the actions; for the sub-classification of the entity and the attribute, performing logic operation after corresponding to the slot to form the actions;
for the attribute classification: and carrying out logic operation after corresponding to the slot positions on the constraint condition classification in the classification result of the entity sub-classification to form the actions.
3. The method of claim 1, wherein the dividing the ranking class and the numerical classification based on the second reference information, respectively, comprises:
dividing the sorting classification into single-hop classification and multi-hop classification according to the hop count among the entities, and dividing the numerical calculation classification into average value classification and summation classification;
dividing the single-hop classification according to the entity, the relation and the attribute, and obtaining the ordering rules under different classifications;
and dividing the single-hop classification of the summation classification according to the entity, the relation and the attribute, and obtaining the summation rules under different classifications.
4. The method of claim 1, wherein the intent tree of the target domain is obtained from a combination of a generic intent tree and an architecture schema of the target domain, comprising:
and combining the schema of the target field and one or more of the entities, the attributes and the relations in the general purpose intention tree to obtain the intention tree of the target field.
5. The method according to claim 1, wherein the method further comprises:
extracting an entity, an attribute and a relation set from the schema of the target field;
filling the entity, the attribute and the relation set into an intention tree of the target field to obtain an instance problem set;
selecting a training set from the set of instance questions;
and training the knowledge question and answer engine model by utilizing the training set, wherein the knowledge question and answer engine comprises an intention recognition model and a knowledge extraction model.
6. The method of claim 5, wherein the taking the input information as input to a knowledge question-and-answer engine model and running the knowledge question-and-answer engine model to obtain a query result comprises:
inputting the input information into an intention recognition model to obtain a knowledge inquiry intention corresponding to the input information;
and inputting the knowledge inquiry intention into the knowledge extraction model to obtain the inquiry result.
7. An information processing apparatus, characterized by comprising:
the first acquisition module is used for acquiring input information of a user aiming at a target field;
the second acquisition module is used for taking the input information as the input of the knowledge question-answering engine model and running the knowledge question-answering engine model to obtain a query result;
the first output module is used for outputting the query result to the user;
the knowledge question-answering engine model is obtained based on an intention tree of the target field, wherein the intention tree of the target field is obtained by combining a schema and a general intention tree of the target field;
the apparatus further comprises:
the first processing module is used for constructing the general purpose intention tree in the following way:
dividing the question-answer type of the known knowledge graph into a question-answer type of inquiring real-time knowledge and a question-answer type of inquiring information;
for the question-answer type of the query real-time knowledge, dividing according to the entity, the relation and the attribute to obtain entity classification, relation classification and attribute classification; for the question-answer type of the query information, classifying the query information into a sorting classification and a numerical calculation classification according to a processing mode;
dividing the entity classification and the attribute classification according to first reference information; dividing the sorting classification and the numerical calculation classification according to second reference information;
wherein the first reference information includes: the number of entities, the number of hops between the entities, whether constraint conditions exist or not, and the number of attributes; the second reference information includes: number of hops between entities;
wherein the dividing the entity classification and the attribute classification according to the first reference information includes:
for the entity classification: dividing into single entity classification and multi-entity classification according to the number of the entities; for each single entity classification included in the single entity classification and the multi-entity classification, dividing the single entity classification and the multi-hop classification according to the hop count among the entities; for the single-hop classification, the classification is divided into: entity sub-classification, entity and attribute sub-classification, attribute sub-classification; dividing the entity sub-classification and the attribute sub-classification according to whether constraint conditions exist or not to obtain a classification result; dividing the classification result according to a mode of forming single intention, and forming behavior actions by corresponding to the slots; dividing the sub-classification of the entity and the attribute according to a mode of forming single intention, and forming behavior actions by corresponding to the slot positions;
for the attribute classification: dividing into single attribute classification and multi-attribute classification according to the number of the attributes; for each single attribute classification included in the single attribute classification and the multi-attribute classification, dividing the single attribute classification and the multi-hop classification into single-hop classification and multi-hop classification according to the hop count among the attributes; for the single-hop classification, the classification is divided into: entity sub-classification and attribute sub-classification; dividing the entity sub-classification and the attribute sub-classification according to whether constraint conditions exist or not to obtain a classification result; on the basis of the obtained classification result, dividing according to a mode of forming single intention, and corresponding to the slot positions to form behavior actions.
8. An electronic device, comprising: a memory, a processor, and a program stored on the memory and executable on the processor; the information processing method according to any one of claims 1 to 6, characterized in that the processor is configured to read a program in a memory to realize the steps in the information processing method.
9. A readable storage medium storing a program, wherein the program when executed by a processor implements the steps in the information processing method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310316954.2A CN116028614B (en) | 2023-03-29 | 2023-03-29 | Information processing method, device, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310316954.2A CN116028614B (en) | 2023-03-29 | 2023-03-29 | Information processing method, device, equipment and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116028614A CN116028614A (en) | 2023-04-28 |
| CN116028614B true CN116028614B (en) | 2023-06-16 |
Family
ID=86074382
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310316954.2A Active CN116028614B (en) | 2023-03-29 | 2023-03-29 | Information processing method, device, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116028614B (en) |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8401853B2 (en) * | 2010-09-22 | 2013-03-19 | At&T Intellectual Property I, L.P. | System and method for enhancing voice-enabled search based on automated demographic identification |
| CN107895037B (en) * | 2017-11-28 | 2022-05-03 | 北京百度网讯科技有限公司 | Question and answer data processing method, device, equipment and computer readable medium |
| CN110147436B (en) * | 2019-03-18 | 2021-02-26 | 清华大学 | Education knowledge map and text-based hybrid automatic question-answering method |
| CN111341311A (en) * | 2020-02-21 | 2020-06-26 | 深圳前海微众银行股份有限公司 | Voice conversation method and device |
-
2023
- 2023-03-29 CN CN202310316954.2A patent/CN116028614B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN116028614A (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA3040373C (en) | Deep learning techniques based multi-purpose conversational agents for processing natural language queries | |
| US11422776B2 (en) | Intelligent assistant for automating recommendations for analytics programs | |
| US10360503B2 (en) | System and method for ontology derivation | |
| US8751433B2 (en) | Efficient probabilistic reasoning over semantic data | |
| CN105975531B (en) | Robot dialog control method and system based on dialogue knowledge base | |
| CN103534697B (en) | For providing the method and system of statistics dialog manager training | |
| CN110019753B (en) | Method and device for outputting back question aiming at user question | |
| US20190220762A1 (en) | Probabilistic modeling system and method | |
| CN105631749A (en) | User portrait calculation method based on statistical data | |
| CN110580516B (en) | Interaction method and device based on intelligent robot | |
| CN106294520A (en) | The information extracted from document is used to carry out identified relationships | |
| CN116028614B (en) | Information processing method, device, equipment and readable storage medium | |
| Ménager et al. | Modeling human memory phenomena in a hybrid event memory system | |
| Ayadi et al. | A framework for knowledge models transformation: a step towards knowledge integration and warehousing | |
| US11941500B2 (en) | System for engagement of human agents for decision-making in a dynamically changing environment | |
| Mejía et al. | CHAT SPI: knowledge extraction proposal using DialogFlow for software process improvement in small and medium enterprises | |
| Ramanigopal et al. | Mind mapping and knowledge management: coding and implementation of KM system | |
| Piggott | Identification of business travelers through clustering algorithms | |
| Chen et al. | Requirement Analysis for Enterprise Information Systems-Developing an Ontological Meta-Model for Zackman Framework | |
| Buis | Applying intelligence amplification to the problem of schema matching | |
| Soffer et al. | Reusability of conceptual models: The problem of model variations | |
| Huang et al. | Reference-Limited Compositional Learning: A Realistic Assessment for Human-level Compositional Generalization | |
| Li | Analysis of requirements incompleteness using metamodel specification | |
| Ruecker et al. | Capturing design rationale in engineering contexts | |
| CN116701658A (en) | Business intention determining method and device and computer equipment |
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 |