CN111428019B - Data processing method and equipment for knowledge base questions and answers - Google Patents

Abstract

The present disclosure provides a data processing method and device for knowledge base questions and answers, where the data processing method includes: acquiring any knowledge item from a knowledge base; selecting user descriptions matched with the knowledge items from the dialogue record to form a set of user descriptions; associating the set of user utterances with the knowledge item; and training the knowledge base question-answering model by taking the associated set of user descriptions and the knowledge items as training samples so as to feed back the subsequently input user descriptions according to training results. The data processing method can improve the real-time performance of the model based on online real data optimization, and ensure the optimal model effect; the operation convenience of operators is improved, and the working efficiency is improved; the method accelerates the discovery of the defects existing in the knowledge items and promotes the continuous perfection of the knowledge base.

Description

Data processing method and device for knowledge base question answering

technical field

The present disclosure relates to the technical field of data processing, in particular to a data processing method and device for knowledge base question answering.

Background technique

The overall historical process of question answering system is from template-based question answering expert system to information retrieval-based question answering, then to community-based question answering, and then to the current knowledge base-based question answering. The question answering algorithm based on information retrieval combines information extraction and shallow semantic analysis on the basis of keyword matching. Community-based Q&A relies on the contribution of netizens, and the Q&A process relies on keyword retrieval technology. Question answering based on knowledge base is based on semantic analysis and knowledge base. Through the knowledge base question answering model, the questions entered by users are analyzed semantically, and the knowledge items that match the questions entered by users are selected in the knowledge base. Existing model optimization based on knowledge base questions and answers often needs to be performed offline, which cannot support operators to optimize the model online in real time, and the online annotation of knowledge base questions and answers is not automatic enough, and does not effectively screen, cluster and recommend a large amount of online real data , so that the operator's labeling work efficiency is low, the workload is heavy, and the repetition is high. In addition, the data of many user statements of online users is not effectively used by the model.

Contents of the invention

In order to solve or at least alleviate at least one of the above technical problems, the present disclosure provides a processing method and device for knowledge base question answering.

According to one aspect of the present disclosure, a data processing method for knowledge base question answering, the data processing method includes:

Obtain any knowledge entry from the knowledge base;

Selecting user statements matching the knowledge entry in the dialogue record to form a set of user statements;

associating the set of user sayings with the knowledge item; and

The set of associated user statements and the knowledge items are used as training samples to train the knowledge base question answering model, so as to give feedback to the subsequently input user statements according to the training results.

According to at least one embodiment of the present disclosure, the selection of user statements matching the knowledge item in the dialogue record to form a set of user statements includes:

If the knowledge item is provided to the user as an approximate answer by the knowledge base question answering model, and is selected by the user's reply or click, then the corresponding user statement in the dialogue record is set as A level;

If the knowledge item is provided to the user as an approximate answer by the knowledge base question answering model, and has not been selected by the user's reply or click, then the corresponding user statement in the dialogue record is set to B level;

If the knowledge item is neither provided to the user as the best answer nor an approximate answer by the knowledge base question answering model, but the confidence level is greater than or equal to the preset value, then the corresponding user statement in the dialogue record is set as C level ;as well as

The user statements are sorted and deduplicated in the order of priority A>B>C to form a set of user statements.

According to another aspect of the present disclosure, a data processing method for knowledge base question answering, the data processing method includes:

Clustering user statements in the dialogue records to form at least one set of user statements;

For each set of user sayings, select a set of knowledge entries matching the set of user sayings from the knowledge base;

associating the set of such user sayings with one of the knowledge items in the set of knowledge items; and

The associated set of user statements of this type and the one of the knowledge items are used as training samples to train the knowledge base question answering model, so as to give feedback to subsequent input user statements according to the training results.

According to at least one embodiment of the present disclosure, the clustering of user statements in the dialogue records to form at least one type of user statement collection includes:

In the dialogue records, the feedback content of the knowledge base question answering model includes user statements with approximate answers or no answers; grouped together.

According to at least one embodiment of the present disclosure, the clustering of user statements in the dialogue records to form at least one type of user statement collection includes:

Sorting the collection of at least one type of user sayings obtained by clustering.

According to at least one embodiment of the present disclosure, the sorting the set of at least one type of user sayings obtained by clustering includes:

The set of at least one type of user statements obtained by clustering is sorted in descending order according to the number of questions asked; wherein, the number of questions refers to the total number of unduplicated user statements in the set of each type of user statements.

According to at least one embodiment of the present disclosure, the sorting the set of at least one type of user sayings obtained by clustering includes:

The collection of at least one type of user statements with the same number of questions is arranged in ascending order according to the number of clustering questions; the number of clustering questions refers to the total number of user statements after deduplication in the set of user statements of each type.

According to at least one embodiment of the present disclosure, the sorting the set of at least one type of user sayings obtained by clustering includes:

Sorting the set of at least one type of user statements with the same number of clustering questions in order of time from near to far.

According to at least one embodiment of the present disclosure, for each set of user sayings, selecting a set of knowledge items matching the set of user sayings from the knowledge base includes:

Match the knowledge entries in the knowledge base with each user statement in the collection of each type of user statement one by one;

Selecting knowledge items whose confidence given by the knowledge base question answering model is greater than or equal to a preset value to form a set of the knowledge items; and

In the set of knowledge items, sort them in descending order according to the accumulative number of occurrences of each knowledge item and deduplicate them.

According to another aspect of the present disclosure, a processing device for knowledge base question answering, the device includes:

a memory storing instructions for execution; and

A processor, the processor executes the execution instructions stored in the memory, so that the processor executes the method described in any one of the foregoing.

Description of drawings

The accompanying drawings illustrate exemplary embodiments of the present disclosure and, together with the description, serve to explain the principles of the disclosure, are included to provide a further understanding of the disclosure, and are incorporated in and constitute this specification. part of the manual.

Fig. 1 is a schematic flowchart of an exemplary implementation of the data processing method for knowledge base question answering in the present disclosure.

Fig. 2 is a schematic flowchart of another exemplary implementation of the data processing method for knowledge base question answering in the present disclosure.

Fig. 3 is a schematic flowchart of another exemplary implementation of the data processing method for knowledge base question answering in the present disclosure.

Fig. 4 is a schematic structural diagram of an exemplary embodiment of a data processing device for knowledge base question answering according to the present disclosure.

Detailed ways

The present disclosure will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific implementation manners described here are only used to explain relevant content, rather than to limit the present disclosure. It should also be noted that, for ease of description, only parts related to the present disclosure are shown in the drawings.

It should be noted that, in the case of no conflict, the implementation modes and the features in the implementation modes in the present disclosure can be combined with each other. The present disclosure will be described in detail below with reference to the drawings and embodiments.

The knowledge base question answering system includes the knowledge base question answering model and the established knowledge base. The knowledge base includes several knowledge items, and the knowledge item is the smallest unit of the knowledge base. When the knowledge base question answering model receives a user statement (question asked by the user), the similarity calculation is performed through the semantic model, and the answer is obtained in the knowledge base and fed back to the user, which is common in the FAQ form of one question and one answer. The knowledge base question answering system can be implemented in various forms, for example, it can be set as an intelligent dialogue robot. During the online use of the knowledge base question answering system, one of the user's statements and the knowledge items fed back to the user's statement form a corresponding dialogue record.

In one application scenario, the user inputs user statements into the knowledge base question answering model, and the knowledge base question answering model can find matching knowledge items in the knowledge base for user statements to form feedback. For the same user statement, the knowledge base question answering model may feed back more than one knowledge item. And for some user statements, it is also possible that no suitable matching knowledge entry can be found in the knowledge base, so that no answer can be obtained. In this case, the knowledge base question answering model will feedback "cannot provide an answer" and similar answers. In the process of the knowledge base question answering model matching the user statement with each knowledge item in the knowledge base, for each knowledge item, the knowledge base question answering model will give the corresponding confidence, the confidence refers to the knowledge base question answering model After judging the user's statement and a corresponding knowledge item, the degree of credibility of matching between the user's statement and the corresponding knowledge item is obtained. The knowledge base question answering model will feed back the best answer, approximate answer or no answer to the user according to the confidence level of each knowledge item. The best answer means: when a user initiates a dialogue, the knowledge base question answering model can obtain a knowledge item with the highest confidence and higher than a specified value to answer his question, then the knowledge item replied at this time is regarded as the best answer. provided to the user. Approximate answer means: when a user initiates a dialogue, the knowledge base question answering model obtains several knowledge items with a confidence level within a specified range to answer their questions, and the knowledge items replied at this time are provided to the user as approximate answers. No answer means: when a user initiates a dialogue, the knowledge base question answering model cannot obtain knowledge items with a confidence level within the specified range to answer their questions, and the feedback at this time is no answer or similar statements.

During the online operation of the knowledge base question answering system, a large number of dialogue records will be generated to form a data set of dialogue records. The data of these dialogue records are all stored in the system. In the prior art, there is no dialogue record data formed online. to optimize the model. That is to say, the existing knowledge base question answering model does not effectively use the data of many user statements of online users, and cannot perform online model optimization. The optimization of the existing knowledge base question answering model often needs to be done offline, which cannot support the online real-time tuning of the model by operators.

According to one aspect of the present disclosure, refer to FIG. 1 , which is a schematic flowchart of an exemplary implementation of the data processing method for knowledge base question answering of the present disclosure. A data processing method for knowledge base question answering, which is used for processing data (such as dialogue records) generated during the knowledge base question answering process, so as to optimize the knowledge base model online. The data processing methods include:

S10. Acquire any knowledge item from the knowledge base. For example, the system automatically selects relevant knowledge items in a certain knowledge field from the knowledge base, and uses the data processing method of the present disclosure to process the knowledge items in the knowledge field one by one. Alternatively, the system may also select the acquired knowledge items with lower confidence levels according to the confidence level in the dialogue record, and process them one by one.

S20. Select user sayings that match the knowledge item in the dialogue record to form a set of user sayings. Each dialog record includes a user statement. If the knowledge base Q&A model screens out a knowledge item with a confidence level that meets the requirements in the knowledge base, the corresponding knowledge item will be included in the dialog record; if the knowledge base Q&A model does not filter out If the knowledge items whose confidence meets the requirements are found, the knowledge items are not included in the dialog record. Among all the dialogue records, select the user sayings in the dialogue records containing the knowledge items selected in step S10, and form all the selected user sayings into a set of user sayings.

S30. Associating the set of user sayings with the knowledge item. Those skilled in the art can understand that the user statements matched with the knowledge item found from all dialogue records generated during the operation of the system are all related to the knowledge item, otherwise they will not be fed back by the knowledge base question answering model Knowledge entry that is the user utterance. Through the above steps, a set of user sayings matching the knowledge item is automatically selected, and the knowledge base question answering model associates the set with the knowledge item. When linking, the knowledge base question-answering model can associate all user statements in the user statement collection with the knowledge entry, or select only some of the user statements for association. The selected user statements can make the user's session intention more clear , a user statement with a more standard questioning method, or a user statement with a high frequency of occurrence.

S40. Using the associated set of user statements and the knowledge items as training samples to train the knowledge base question answering model, so as to give feedback to the subsequently input user statements according to the training results. After a large number of dialogue records are selected and associated with a set of user statements related to knowledge items, they are input into the model as training samples. After training, the knowledge base question answering model can combine the user statements in the samples with the knowledge in the samples. The entries form an association relationship. After the next time the user enters the same or similar user statement as the sample in the knowledge base question answering model, the knowledge base question answering model can directly feed back the associated knowledge item as a reply through the association relationship in the sample, realizing It achieves the purpose of optimizing the knowledge base question answering model online. Wherein, when there is more than one same keyword in the input user statement and the user statement in the sample, it can be considered that the input user statement is similar to the user statement in the sample.

In the data processing method of the present disclosure, the system automatically screens out user statements that match the knowledge items that need to be optimized to form a set of user statements by using the data of user dialogue records generated in the production environment (during the online operation of the knowledge base question answering system). The collection of user statements and knowledge entries are associated, and the associated collection of user statements and knowledge entries are automatically imported into the knowledge base question answering model, and the online training of the knowledge base question answering model can achieve the purpose of online learning and optimization of the model. It solves the problem in the prior art that the data of many user statements of online users are not effectively used by the model. At the same time, the system automatically screens a large number of online dialogue records to form a collection of user statements, avoiding the problems of low work efficiency, heavy workload, and high repetition that exist in manual selection.

In one embodiment of the present disclosure, step S20, selecting user statements matching the knowledge item in the dialog record to form a set of user statements may include:

If the knowledge item is provided to the user as an approximate answer by the knowledge base question answering model, and is selected by the user's reply or click, then the corresponding user statement in the dialogue record is set as A level. This knowledge item is provided to the user as an approximate answer, and has been selected by the user’s reply or click, indicating that the knowledge item has a high degree of matching with the corresponding user dialogue and has a high degree of confidence. Set its priority level to A .

If the knowledge item is provided to the user as an approximate answer by the knowledge base question answering model, and has not been answered or clicked and selected by the user, then the corresponding user statement in the dialogue record is set as B level. Although the knowledge item is provided to the user as an approximate answer, it has not been replied or selected by the user, indicating that the knowledge item has a certain degree of matching with the corresponding user dialogue, and the confidence level is lower than that of A level. The priority level is set to B level.

If the knowledge item is neither provided to the user as the best answer nor an approximate answer by the knowledge base question answering model, but the confidence level is greater than or equal to the preset value when matching the user's statement, for example, the preset value can be set to 0.5, then the corresponding user statement in the dialogue record is set as C level. That is to say, the user statement to be screened should be the user statement whose confidence level is greater than the preset value when matching the knowledge item in the dialogue record, and the user statement whose confidence level is less than the preset value is no longer selected to form. In the collection of user sayings, to avoid introducing noise that does not match the knowledge entry.

The above-screened user statements are sorted and deduplicated in the order of priority A>B>C to form a set of user statements. That is to say, in the set of user statements formed, the user statements with a priority level of A are arranged at the top, and so on, and the user statements with a priority level of C are ranked last.

The above implementations are applicable to the initial stage of the knowledge base question answering system, that is, after a short period of online operation, there are already a certain number of dialogue records in the system, and the knowledge base question answering model needs to be optimized to improve system performance.

According to another aspect of the present disclosure, see FIG. 2 , which is a schematic flowchart of another exemplary implementation of the data processing method for knowledge base question answering of the present disclosure. A data processing method for knowledge base question answering, which is used for processing data (such as dialogue records) generated during the knowledge base question answering process, so as to optimize the knowledge base model online. The data processing methods include:

S110. Cluster the user statements in the dialogue records to form a set of at least one type of user statements. With the increase of the online running time of the knowledge base question answering system, there will be a large number of similar user statements for the same question in the system. The knowledge base question answering model will automatically cluster a large number of user statements existing in the dialogue records according to the degree of similarity. Multiple similar user statements related to a certain question are clustered into one category, and multiple similar user statements related to another question are clustered into one category to form a collection of different types of user statements. There may be only one category, or there may be two or three different categories of user statements.

S220. For each set of user sayings, the knowledge base question answering model selects a set of knowledge items matching the set of user sayings from the knowledge base. That is, a set of user statements and a set of knowledge items are respectively formed, and each knowledge item in the set of knowledge items has a certain degree of matching with each user statement in the set of user statements.

S330. Associate the set of such user sayings with one of the knowledge items in the set of knowledge items. Although each knowledge item in the set of knowledge items has a certain degree of matching with the user statement in the set of user statements, in order to achieve the purpose of optimizing the knowledge base question answering model and avoid introducing redundant noise, the knowledge base question answering model only One of the knowledge items is selected from the collection of knowledge items for association.

S440. Using the associated set of user statements of this type and the one of the knowledge items as training samples to train the knowledge base question answering model, so as to give feedback to the subsequently input user statements according to the training results. From a large number of dialogue records, various user statements are selected for clustering. For each set of user statements formed, the matching knowledge items are screened out and associated, and then input into the model as training samples. Knowledge base question answering After the model is trained, it can associate the collection of user statements in the sample with the knowledge items in the sample, and the next time the user enters the same or similar user as a certain type of user statement in the sample to the knowledge base question answering model. After the statement, the knowledge base question answering model can directly feed back the associated knowledge items as answers through the association relationship in the sample, realizing the purpose of optimizing the knowledge base question answering model online. Wherein, when there is more than one same keyword in the input user statement and the user statement in the sample, it can be considered that the input user statement is similar to the user statement in the sample.

In one embodiment of the present disclosure, step S110, clustering the user sayings in the dialogue records to form a set of at least one type of user sayings may include:

In the dialogue records, the feedback content of the knowledge base question answering model, including user statements with approximate answers or no answers, is clustered into one category. Or in the dialogue records, the user statements whose confidence level given by the knowledge base question answering model is less than the preset value are clustered into one category. These two clustering methods can filter out and cluster user statements with poor initial matching effects in the dialogue records, so that these user statements can be matched to better knowledge items in subsequent steps to optimize the model.

That is to say, the above two different clustering methods can be used, and one of the clustering methods can be selected according to different production environments.

In one embodiment of the present disclosure, step S110, clustering the user sayings in the dialogue records to form a set of at least one type of user sayings may include:

S111. Sorting the set of at least one type of user sayings obtained by clustering. Refer to FIG. 3 for a schematic flowchart of another exemplary implementation of the data processing method for knowledge base question answering of the present disclosure. If the clustering forms more than one set of user sayings, the sets of user sayings of each type are sorted according to certain rules, so as to facilitate subsequent further processing of these data.

In one embodiment of the present disclosure, step S111, sorting the set of at least one type of user sayings obtained by clustering may include:

The set of at least one type of user statements obtained by clustering is sorted in descending order according to the number of questions asked; wherein, the number of questions refers to the total number of unduplicated user statements in the set of each type of user statements. If more than two sets of user sayings are obtained by clustering, the set containing the largest number of user sayings is arranged in the front, and the set containing the small number of user sayings is arranged in the back.

In one embodiment of the present disclosure, step S111, sorting the set of at least one type of user sayings obtained by clustering may include:

The collection of at least one type of user statements with the same number of questions is arranged in ascending order according to the number of clustering questions; the number of clustering questions refers to the total number of user statements after deduplication in the set of user statements of each type. If there are two sets that contain the same number of user statements, then compare the number of clustering questions after deduplication and sort them, and arrange the set with fewer clustering questions (that is, the number of repeated user statements) in the front , arrange the sets with the most clustering questions (that is, the sets with the fewest repeated user statements) in the back.

In one embodiment of the present disclosure, step S111, sorting the set of at least one type of user sayings obtained by clustering may include:

Sorting the set of at least one type of user statements with the same number of clustering questions in order of time from near to far. If there are two sets that contain the same number of user statements and contain the same number of repeated user statements, the set that is closer to the current time will be placed in front of the two sets, and the set that is farther away from the current time will be ranked first. Sets come next. The time mentioned here refers to the time when the user's statement is input into the knowledge base question answering model, that is, the time when the user asks a question. In other words, it is to judge which of the user statements closest to the current time in the two sets is the closest to the current time, and rank the set corresponding to the user statement closest to the current time in front.

In one embodiment of the present disclosure, step S220, for each set of user sayings, select a set of knowledge items matching the set of user sayings from the knowledge base, which may include:

The knowledge entries in the knowledge base are matched one by one with each user statement in the collection of each type of user statement. The collection of each type of user statement contains multiple user statements, and for each user statement, find the matching knowledge entry in the knowledge base (maybe more than one knowledge entry is matched, or there may be no relevant knowledge entry). knowledge entry). That is to perform ergodic matching between each user statement in the collection of each type of user statement and the knowledge entries in the knowledge base. That is to say, in this step, each user statement in the collection of each type of user statement is re-matched in the knowledge base to filter out better matching knowledge items for association. The better matching means that the confidence level of the re-matched user statement and knowledge item in this step is higher than the confidence level of the initially matched user statement and knowledge item in the dialogue record.

A set of knowledge items is formed by selecting knowledge items whose confidence given by the knowledge base question answering model is greater than or equal to a preset value. When performing one-by-one matching in the previous step, each time a corresponding knowledge entry is matched, the knowledge base question answering model will give a confidence level for this match, and a confidence threshold can be set in advance, for example, the confidence threshold is set as 0.5, only knowledge items with a matching confidence greater than or equal to 0.5 are screened into the set of knowledge items formed to avoid noise.

In the set of knowledge items, sort them in descending order according to the accumulative number of occurrences of each knowledge item and deduplicate them. In the collection of each type of user statements, there are similar user statements, some of which have relatively clear topics and purposes, and can be matched with appropriate knowledge items in the knowledge base, while some topics and purposes are relatively vague, and there is no Able to match appropriate knowledge items. If a certain knowledge item appears more often in the set of knowledge items, it means that the matching degree of the knowledge item is better, so it is arranged in the front, otherwise, it is arranged in the back, which is convenient for further processing of the data.

In one embodiment of the present disclosure, step S220, for each set of user sayings, select a set of knowledge items matching the set of user sayings from the knowledge base, which may include:

If there is no relevant knowledge entry in the current knowledge base for a set of user statements, that is, there is no answer, then it is necessary to add relevant information to the knowledge base by using network search or manual input according to the content of the user statement. knowledge entries to supplement and improve the current knowledge base. Then, the added knowledge items are screened into the set of knowledge items formed. In the next step, the added knowledge items are associated with the matching set of user statements and then input into the knowledge base question answering model as training samples. Perform training and online optimization.

The above implementations are applicable to the middle and later stages of the knowledge base question answering system, that is, after a period of time of online operation, a large number of dialogue records are stored in the system, and the knowledge base question answering model needs to be optimized to improve system performance.

Combining the above two different implementation methods, it can be seen that the data processing method for knowledge base question answering disclosed in this disclosure can improve the real-time optimization of the model based on online real data, ensure the optimal effect of the model, and improve the operation convenience of operators. Improve work efficiency; accelerate the discovery of deficiencies in knowledge entries, and promote the continuous improvement of the knowledge base.

The present disclosure also provides a data processing device for knowledge base question answering, see FIG. 4 for a schematic structural diagram of an exemplary implementation of the present disclosure data processing device for knowledge base question answering. The device includes: a communication interface 1000 , a memory 2000 and a processor 3000 . The communication interface 1000 is used for communicating with external devices and performing interactive data transmission. Computer programs executable on the processor 3000 are stored in the memory 2000 . When the processor 3000 executes the computer program, the methods in the foregoing implementation manners are realized. The number of the memory 2000 and the processor 3000 may be one or more.

The memory 2000 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory.

If the communication interface 1000, the memory 2000, and the processor 3000 are implemented independently, the communication interface 1000, the memory 2000, and the processor 3000 may be connected to each other through a bus to complete mutual communication. The bus may be an Industry Standard Architecture (ISA, Industry Standard Architecture) bus, a Peripheral Component Interconnect (PCI, Peripheral Component) bus, or an Extended Industry Standard Architecture (EISA, Extended Industry Standard Component) bus, and the like. The bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in this figure, but it does not mean that there is only one bus or one type of bus.

Optionally, in specific implementation, if the communication interface 1000, the memory 2000, and the processor 3000 are integrated on one chip, then the communication interface 1000, the memory 2000, and the processor 3000 can communicate with each other through internal interfaces.

Any process or method descriptions in flowcharts or otherwise described herein may be understood to represent modules, segments or portions of code comprising one or more executable instructions for implementing specific logical functions or steps of the process , and the scope of preferred embodiments of the present disclosure includes additional implementations in which functions may be performed out of the order shown or discussed, including substantially concurrently or in reverse order depending on the functions involved, which shall It is understood by those skilled in the art to which the embodiments of the present disclosure belong. The processor executes the various methods and processes described above. For example, method embodiments in the present disclosure may be implemented as a software program tangibly embodied on a machine-readable medium, such as memory. In some implementations, part or all of the software program may be loaded and/or installed via memory and/or a communication interface. One or more steps in the methods described above may be performed when a software program is loaded into memory and executed by a processor. Alternatively, in other implementation manners, the processor may be configured to perform one of the above-mentioned methods in any other suitable manner (for example, by means of firmware).

The logic and/or steps shown in the flowcharts or otherwise described herein can be embodied in any readable storage medium for instruction execution systems, devices or devices (such as computer-based systems, processor-included system or other systems that may fetch and execute instructions from an instruction execution system, device, or device), or be used in conjunction with such an instruction execution system, device, or device.

As far as this specification is concerned, a "readable storage medium" may be any device that can contain, store, communicate, spread or transmit programs for instruction execution systems, devices or devices or use in conjunction with these instruction execution systems, devices or devices. More specific examples (non-exhaustive list) of readable storage media include the following: electrical connection with one or more wires (electronic device), portable computer disk case (magnetic device), random access memory (RAM), Read Only Memory (ROM), Erasable and Editable Read Only Memory (EPROM or Flash Memory), Fiber Optic Devices, and Portable Read Only Memory (CDROM). In addition, the readable storage medium may even be paper or other suitable medium on which the program can be printed, since it can be done, for example, by optically scanning the paper or other medium, followed by editing, interpretation or other suitable processing if necessary. The program is processed electronically and then stored in memory.

It should be understood that various parts of the present disclosure may be realized by hardware, software or a combination thereof. In the embodiments described above, various steps or methods may be implemented by software stored in memory and executed by a suitable instruction execution system. For example, if it is implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: Discrete logic circuits, ASICs with suitable combinational logic gates, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

Those of ordinary skill in the art can understand that all or part of the steps to realize the method of the above-mentioned embodiment can be completed by instructing related hardware through a program, and the program can be stored in a readable storage medium, and the program can be executed when executed , comprising one or a combination of the steps of the method embodiments.

In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing module, each unit may exist separately physically, or two or more units may be integrated into one module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software function modules. If the integrated modules are realized in the form of software function modules and sold or used as independent products, they can also be stored in a readable storage medium. The storage medium may be a read-only memory, a magnetic disk or an optical disk, and the like.

In the description of this specification, descriptions referring to the terms "one embodiment/mode", "some embodiments/modes", "examples", "specific examples", or "some examples" mean that the embodiments/modes are combined The specific features, structures, materials or characteristics described in or examples are included in at least one embodiment/mode or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment/mode or example. Moreover, the described specific features, structures, materials or characteristics may be combined in any one or more embodiments/modes or examples in an appropriate manner. In addition, those skilled in the art may combine and combine different embodiments/modes or examples and features of different embodiments/modes or examples described in this specification without conflicting with each other.

In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be interpreted as indicating or implying relative importance or implicitly specifying the quantity of indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include at least one of these features. In the description of the present application, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

It should be understood by those skilled in the art that the above-mentioned embodiments are only for clearly illustrating the present disclosure, rather than limiting the scope of the present disclosure. For those skilled in the art, other changes or modifications can be made on the basis of the above disclosure, and these changes or modifications are still within the scope of the present disclosure.

Claims (9)

1. A data processing method for knowledge base questions and answers, the method comprising:

acquiring any knowledge item from a knowledge base;

selecting user utterances matched with the knowledge items from the dialogue record to form a set of user utterances; the dialogue record includes: the user speaks and the knowledge items fed back to the user speaks form a corresponding dialogue record;

associating the set of user utterances with the knowledge item; and

and training the knowledge base question-answering model by taking the correlated set of user utterances and the knowledge items as training samples so as to feed back the subsequently input user utterances according to training results.

2. The data processing method of claim 1, wherein selecting the user utterance in the conversation record that matches the knowledge item forms a set of user utterances, comprising:

if the knowledge item is provided to the user as an approximate answer by a knowledge base question-answer model and is replied or clicked by the user, setting the corresponding user description in the dialogue record as a level A; the approximate answer includes: when a user initiates a dialogue, the knowledge base question-answering model acquires a plurality of knowledge items with confidence degrees within a specified range and answers questions, and the replied knowledge items are determined to be approximate answers and are provided for the user;

if the knowledge item is provided to the user as an approximate answer by the knowledge base question-answer model and is not replied or clicked by the user, setting the corresponding user description in the dialogue record as a B level;

if the knowledge item is not provided for the user as the best answer or the approximate answer by the knowledge base question-answering model, but the confidence coefficient is larger than or equal to a preset value, setting the corresponding user description in the dialogue record as a C level; and

sorting and de-duplicating the user's utterances in order of priority class A > class B > class C to form a set of the user's utterances;

the best answer includes: when a user initiates a dialogue, the knowledge base question-answering model acquires a knowledge item with highest confidence and higher than a certain designated value, answers the question, and determines the knowledge item replied at the moment as the best answer to provide for the user.

3. A data processing method for knowledge base questions and answers, the method comprising:

clustering user utterances in the conversation records to form a set of at least one type of user utterances, including: in the dialogue record, the feedback content of the knowledge base question-answer model comprises user descriptions of approximate answers or no answers to be classified; or in the dialogue record, gathering the user's description given by the knowledge base question-answer model that the confidence coefficient is smaller than the preset value into one type; the dialogue record includes: the user speaks and the knowledge items fed back to the user speaks form a corresponding dialogue record; the approximate answer includes: when a user initiates a dialogue, the knowledge base question-answering model acquires a plurality of knowledge items with confidence degrees within a specified range and answers questions, and the replied knowledge items are determined to be approximate answers and are provided for the user;

selecting, for each class of set of user utterances, a set of knowledge items from the knowledge base that match the class of set of user utterances;

associating the set of user utterances with one of the set of knowledge items; and

and training the knowledge base question-answering model by using the correlated set of the user descriptions and one of the knowledge items as a training sample so as to feed back the subsequently input user descriptions according to the training result.

4. The data processing method of claim 3, wherein clustering user utterances in the conversation records to form a set of at least one type of user utterance comprises:

and sequencing the clustered collection of at least one type of user description.

5. The data processing method of claim 4, wherein the ranking the clustered collection of at least one type of user utterance comprises:

the clustered set of at least one type of user description is arranged in descending order according to the question number; the question times refer to the total number of unrepeated user utterances in the set of each type of user utterances.

6. The data processing method of claim 5, wherein the ranking the clustered collection of at least one type of user utterance comprises:

the method comprises the steps of (1) arranging at least one group of user description sets with the same question times in ascending order according to the number of clustering questions; the clustering problem number refers to the total number of user utterances after de-duplication in the set of each type of user utterances.

7. The data processing method of claim 6, wherein the ranking the clustered collection of at least one type of user utterance comprises:

and sequencing the set of at least one type of user description with the same clustering problem number according to the sequence from the near to the far.

8. The data processing method of claim 3, wherein the selecting, for each class of user utterance, a set of knowledge items from a knowledge base that match the class of user utterance sets comprises:

matching knowledge items in the knowledge base with each user utterance in the set of each type of user utterance one by one;

selecting knowledge items with confidence degrees larger than or equal to a preset value given by a knowledge base question-answer model to form a set of the knowledge items; and

and in the knowledge item set, the knowledge items are arranged in descending order according to the accumulated times of occurrence of each knowledge item and are de-duplicated.

9. A data processing apparatus for knowledge base questions and answers, the data processing apparatus comprising:

a memory storing execution instructions; and

a processor executing the memory-stored execution instructions, causing the processor to perform the method of any one of claims 1 to 8.