CN115129878A

CN115129878A - Conversation service execution method, device, storage medium and electronic equipment

Info

Publication number: CN115129878A
Application number: CN202211052703.XA
Authority: CN
Inventors: 赵闻飙; 应缜哲; 王昊天; 林金镇; 王维强
Original assignee: Alipay Hangzhou Information Technology Co Ltd
Current assignee: Alipay Hangzhou Information Technology Co Ltd
Priority date: 2022-08-31
Filing date: 2022-08-31
Publication date: 2022-09-30
Anticipated expiration: 2042-08-31
Also published as: CN115129878B

Abstract

In the conversation service execution method, the conversation service execution device, the storage medium and the electronic device provided by the present specification, the reply content input by the user in the previous round can be acquired, and the type of the dialect transmitted to the user in the previous round can be determined; determining the intention of the user in a pre-constructed intention library according to the reply content; determining a combination of the determined intent and the utterance class as a current logical combination; determining a rule matched with the current logic combination in a pre-constructed rule base; determining the type of the dialect to be sent to the user in the current round according to the rule; determining the dialect to be sent to the user in a pre-constructed dialect library according to the type of the dialect, and sending the determined dialect to the user. When the method is adopted to execute the conversation service, the conversation transmitted to the user in the next round can be determined through the form of the rule, and when the conversation logic needs to be modified, the modification can be completed only by adding or modifying the specific rule, so that the threshold of updating and maintaining is greatly reduced.

Description

Conversation service execution method, device, storage medium and electronic equipment

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to a method and an apparatus for executing a dialog service, a storage medium, and an electronic device.

Background

Nowadays, users are increasingly paying attention to protecting private data. With the development of artificial intelligence, intelligent customer service is widely used in performing a dialogue service with a user. Currently, most intelligent customer services are implemented based on routing trees. When setting intelligent customer service, inputting possible dialogs sent to the user in one conversation, contents possibly input by the user and logic relations among the dialogs into the intelligent customer service in a tree form; in other words, all possible situations in the whole dialog are pre-stored in the intelligent service in the form of a tree-like flow chart, i.e. a routing tree. When the intelligent customer service device has a conversation with the user, the intelligent customer service device can continuously search the corresponding branch nodes in the routing tree according to the content input by the user and reply to the user by adopting the dialogs specified in the routing tree. And when no other node exists after the node corresponding to the transmitted one dialog, ending the one dialog.

At present, along with the rapid iteration of various kinds of information, the dialect and logic in the intelligent customer service need to be updated frequently. However, it is not flexible to update and maintain the existing intelligent customer service based on the routing tree. Because each node in the routing tree may have a certain logical relationship, the whole routing tree must be rearranged during modification, and this process requires that an operator can use a professional tool, and has strong logical arrangement capability and high threshold. It can be seen that the route tree mode causes that the update and maintenance process of the intelligent customer service is difficult, and general operators are difficult to finish alone without the help of professionals.

Therefore, intelligent customer service realized in other ways is needed at the present stage, and the problem of high threshold for updating and maintaining the intelligent customer service is solved.

Disclosure of Invention

The present specification provides a method, an apparatus, a storage medium, and an electronic device for executing a dialog service, which partially solve the above problems in the prior art.

The technical scheme adopted by the specification is as follows:

the present specification provides a method for executing a dialog service, including:

acquiring reply content input by a user in the previous round, and determining the type of the voice operation sent to the user in the previous round;

determining the intention of the user according to the reply content;

determining a combination of the intention and a language category of a language transmitted to the user in a previous round as a current logical combination;

determining rules matched with the current logic combination in each rule contained in a pre-constructed rule base;

determining candidate dialect types of the candidate dialects sent to the user according to the rule;

and determining the dialect to be sent to the user in a pre-constructed dialect library according to the candidate dialect types, and sending the dialect to the user.

Optionally, storing each dialect and each dialect category in the dialect library;

determining the dialect sent to the user in a pre-constructed dialect library according to the candidate dialect types, specifically comprising the following steps:

determining each candidate phone operation corresponding to the candidate phone operation type in a phone operation library constructed in advance;

and selecting one of the determined candidate dialogs as the one to be sent to the user.

Optionally, the rule includes: a correspondence between a logical combination and a conversational class, wherein a logical combination is a combination of a conversational class and at least one intention.

Optionally, a rule base is constructed in advance, and the method specifically includes:

displaying an input interface to an operator, wherein the input interface at least comprises a first input column, a second input column and a third input column;

responding to data input by the operator in the first input field, the second input field and the third input field respectively, and determining a first target language technique, a target reply and a second target language technique;

determining an input combination according to the sequence of the first target dialect, the target reply and the second target dialect;

combining the inputs into an input model, and generating a target rule through the rule generation model;

and constructing the rule base by adopting the target rule.

Optionally, generating the target rule through the rule generation model specifically includes:

determining a first target class of the first target utterance and a second target class of the second target utterance by class partitioning layers in the rule generation model;

determining at least one target intention according to the target reply through an intention analysis layer in the rule generation model;

and generating a target rule according to the target first target type, the target intention and the second target type through a generation layer in the rule generation model.

Optionally, the target intent is a plurality of intents;

determining at least one target intention according to the target reply through an intention analysis layer in the rule generation model, wherein the method specifically comprises the following steps:

determining, by an intent analysis layer in the rule generation model, each target intent and a logical relationship between the target intents according to the target replies in the input combination, wherein the logical relationship comprises at least one of AND and or OR.

Optionally, generating a target rule according to the target first target category, the target intention, and the second target category, specifically including:

determining a combination of the first target category and the target intent as a target logical combination;

and generating a target rule according to the target logic combination and the second target type, wherein the target rule is that the target logic combination and the second target type have a corresponding relation.

Optionally, determining the intention of the user according to the reply content specifically includes:

determining at least one intention of the user and a logical relationship among the intentions according to the reply content through the intention analysis layer.

Optionally, the constructing the rule base by using the target rule specifically includes:

when the specified rule containing the target logic combination exists in the rule base, replacing the specified rule with the target rule;

when no rule containing the target logic combination exists in the rule base, the target rule is added to the rule base.

Optionally, the pre-training rule generating model specifically includes:

acquiring a first sample conversation, a sample reply, a second sample conversation and a labeling rule;

determining a sample input combination according to the sequence of the first sample grammar, the sample reply and the second sample grammar;

inputting the sample input combination into a rule generating model to be trained, and generating a rule to be optimized through the rule generating model;

and training the rule generation model by taking the minimum difference between the rule to be optimized and the labeling rule as an optimization target.

This specification provides a conversation service execution apparatus, including:

the acquisition module is used for acquiring reply contents input by the user in the previous round and determining the type of the dialect sent to the user in the previous round;

the intention determining module is used for determining the intention of the user in a pre-constructed intention library according to the reply content;

a combination determination module for determining the combination of the intention and the type of the dialect transmitted to the user in the previous round as the current logic combination;

the rule matching module is used for determining a rule matched with the current logic combination in all rules contained in a rule base constructed in advance;

the category determination module is used for determining candidate speech technology categories of the candidate speech technologies sent to the user according to the rules;

and the sending module is used for determining the dialect sent to the user in a pre-constructed dialect library according to the candidate dialect types and sending the dialect to the user.

The present specification provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the above-described dialogue service execution method.

The present specification provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the above-mentioned dialog service execution method when executing the program.

The technical scheme adopted by the specification can achieve the following beneficial effects:

in the dialog service execution method provided by the present specification, the reply content input by the user in the previous round can be obtained, and the type of the dialect sent to the user in the previous round is determined; determining the intention of the user in a pre-constructed intention library according to the reply content; determining a combination of the determined intent and the utterance class as a current logical combination; determining a rule matched with the current logic combination in a rule base established in advance; determining the type of the dialect to be sent to the user in the current round according to the rule; determining the dialect to be sent to the user in a pre-constructed dialect library according to the type of the dialect, and sending the determined dialect to the user. When the method is adopted to execute the conversation service, the conversation operation sent to the user in the next round can be determined through the form of the rule, and when the conversation logic needs to be modified, the modification can be completed only by adding or modifying the specific rule, so that the threshold of updating and maintaining is greatly reduced.

Drawings

The accompanying drawings, which are included to provide a further understanding of the specification and are incorporated in and constitute a part of this specification, illustrate embodiments of the specification and together with the description serve to explain the specification and not to limit the specification in a non-limiting sense. In the drawings:

fig. 1 is a schematic flow chart of a method for executing a session service provided in this specification;

fig. 2 is a schematic diagram of a dialogue service execution device provided in the present specification;

fig. 3 is a schematic diagram of an electronic device corresponding to fig. 1 provided in the present specification.

Detailed Description

In the intelligent customer service realized based on the routing tree, the conversation with the user is completely carried out according to the flow in the preset routing tree. Each node in the routing tree represents a dialog, each edge represents a reply of a user, and each edge is used for connecting two nodes. The edges branched off under one node represent a plurality of replies which can be given by the user when the dialect of the node is sent to the user; and the other node connected to each edge represents the next dialect sent to the user for the user reply represented by the edge.

It can be seen that in the intelligent customer service implemented based on the routing tree, since each node and the root node have a path, that is, when a problem occurs in which one of the nodes is not expected, the cause of the problem may occur on any edge or node in the routing tree. At this time, if the intelligent customer service is to be updated, the reason has to be located and the logical relationship of the whole routing tree has to be edited again. The process is high in threshold, time-consuming, labor-consuming and very inflexible.

In order to solve the above problems, the present specification provides a new method for executing a dialog service.

In order to make the objects, technical solutions and advantages of the present disclosure more clear, the technical solutions of the present disclosure will be clearly and completely described below with reference to the specific embodiments of the present disclosure and the accompanying drawings. It is to be understood that the embodiments described are only a few embodiments of the present disclosure, and not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present specification without any creative effort belong to the protection scope of the present specification.

The technical solutions provided by the embodiments of the present description are described in detail below with reference to the accompanying drawings.

Fig. 1 is a schematic flowchart of a method for executing a dialog service provided in this specification, including the following steps:

s100: and acquiring the reply content input by the user in the last round, and determining the type of the voice operation sent to the user in the last round.

In the present specification, an execution subject for implementing the method for executing the dialog service may refer to a designated device such as a server installed on the service platform, and for convenience of description, the present specification only takes the server as the execution subject, and describes a method for testing a code provided in the present specification.

In this specification, for convenience of description, artificial intelligence that makes a conversation with a user when executing a conversation service is referred to as an intelligent customer service, and corresponds to a "virtual person" who makes a conversation with the user. It will be appreciated by those skilled in the art that the intelligent customer service itself does not exist as an entity, let alone as an executing subject.

In this specification, the term "conversation" refers to a general term of contents that an intelligent customer service sends to a user.

Currently, there are two main types of intelligent customer service when performing conversational services. One is an active question type, that is, in each round of conversation with a user, an intelligent customer service is the party who asks questions first, and the user is the party who answers the questions; this type of intelligent customer service is mainly used in scenarios such as auditing during leasing and lending, or conducting research on users, and trading, and collects necessary information under the condition of approving the users. The other is a passive answer type, namely in each round of conversation with the user, the intelligent customer service answers according to questions put forward by the user; this type of intelligent customer service is mainly applied in scenarios such as information inquiry, pre-sale and post-sale of goods, etc. The conversation service execution method provided by the specification is mainly applied to intelligent customer service of an active question asking type.

In the method for executing the dialog service provided by the present specification, each turn means that a question and a response are made between the intelligent customer service and the user, that is, after the intelligent customer service sends a dialog to the user, the user replies to the dialog again, and thus a turn is made. In each round, the intelligent customer service firstly initiates a question, and the user replies to the question raised by the intelligent customer service in the current round. The input content of the user includes, but is not limited to, text content, audio content, and image content.

Of course, in general, when a conversation starts, the intelligent customer service can call the user and send a word for asking for a good word; likewise, at the end of the conversation, the intelligent customer service may send a dialog to the user indicating bye. It is conceivable that the conversation service execution method provided by this specification does not take into account the above-described process of cold talk, i.e., the above-described process of cold talk will not affect the progress of normal conversation, regardless of whether the user replies to a session transmitted for small packets.

In the process of the conversation between the intelligent customer service and the user, the intelligent customer service asks the questions once in each round, and the user answers once, and the round is ended. In each round, the intelligent customer service firstly obtains the reply content input by the user in the last round, and meanwhile, determines the type of the words sent to the user in the last round for determining the words sent to the user in the current round in the subsequent steps. The type of dialect is used to characterize what the dialect expresses, and in general, the dialects expressing the same meaning belong to the same type of dialect.

S102: and determining the intention of the user according to the reply content.

At each turn, the user's intent may be determined from the user's reply content. The user's intention can characterize the attitude, demand, purpose, etc. of the user, for example, the user's intention can be affirmative, negative, uncertain, or want to do something, or can indicate that it is not heard, hope that the intelligence customer service repeats talk operation once, etc.

In the dialogue service execution method provided in the present specification, there are a pre-constructed dialect library and an intention library. The dialect library is used for storing all dialects which can be sent by the intelligent customer service in the process of carrying out conversation with the user and the type of each dialect; the intent library is used to store all intents that a user may present during a conversation. In other words, the intelligent customer service selects a dialog from the dialog library to be sent to the user in each round, and the intention obtained by the user according to the reply content in each round is also in the intention library.

It is worth mentioning that, in general, in order to ensure diversity in conversation with a user, there are a plurality of dialects expressing the same meaning, that is, the same kind of dialects, in the dialect library. On the other hand, many times, the user's reply may express more than one intention, so that there may be a plurality of intentions according to the user's reply content, and the logical relationship between the intentions may be and or not.

Typically, the data from a historically conducted dialogue service is used to construct a dialogistic library and an ideogram library. Specifically, a historical dialog may be obtained from the dialog service that is executed historically; determining available dialogs and available intentions according to the historical dialogs; and constructing a dialect library by using the available dialect, and constructing an intention library by using the available intention. When determining available dialogs and available intentions according to historical dialogs, the determination may be performed in a manual operation manner, or may be performed by using a trained model, which is not limited in this specification. It is worth mentioning that different historical dialogue data can be selected according to the intelligent customer service applied to different scenes, and different dialogue libraries and intention libraries are constructed in advance. In other words, according to the application scene of the intelligent customer service, the historical conversations under the same application scene are selected to build the dialect library and the intention library.

Additionally, when the user's intention is determined according to the input content, a pre-trained intention model can be used for the determination. Specifically, the reply content may be input into a pre-trained intention model, through which the user's intention is determined. When the intention model is trained, sample reply contents can be obtained, and the annotation intention is determined according to the sample reply contents; inputting the sample reply content into an intention model to be trained, and obtaining an intention to be optimized through the intention model; and training an intention model by taking the minimum difference between the intention to be optimized and the labeling intention as an optimization target. It should be noted that the intention model is not a model bound to intelligent customer service. In other words, the intent model can be trained and used separately from the intelligent customer service.

S104: determining a combination of the intent and a type of utterance sent to the user in a previous round as a current logical combination.

In the conversation business execution method provided in the present specification, generally, the flow of the conversation between the intelligent customer service and the user is that, in each turn, the intelligent customer service first asks questions, and then the user answers to the questions of the intelligent customer service. In a dialogue, certain logic relation is necessarily existed between dialogs of each turn, and the dialogs are not independent questions and answers without relevance; in most cases, the closer the conversation is, the stronger the relevance is. In other words, the smart customer service transmits the words to the user in the next round, and is largely determined by the words transmitted to the user in this round by the smart customer service and the replies to the user in this round. Therefore, when determining the next round of the intelligent customer service to the user, the other rounds with smaller influence can be ignored, and the type of the next round of the voice to the user can be determined according to the type of the voice to the user in the current round and the intention of the user in the current round.

In addition to the dialoging library and the intention library, the dialog service execution method provided by the specification also has a pre-constructed rule library. In the method, rules may characterize the conversion relationships between different conversational classes, i.e. how to derive one conversational class from another conversational class. Specifically, each rule may include: a correspondence between a logical combination and a conversational class, wherein a logical combination is a combination of a conversational class and at least one intention.

And determining the candidate dialect types of the candidate dialects transmitted to the user in the current round according to the dialect types of the dialects transmitted to the user in the last round and the intention of the user. In short, "another speech category is obtained by adding several intentions to one speech category", which can be expressed as "speech category 1+ several intentions → speech category 2". Wherein, several intents can be a single intention A, or an intention A and an intention B, or an intention A or an intention B; it is conceivable that several intents may also be a combination of more intents, which are not repeated here. In practice, however, this translation is not clear, and in practice, talkclass 1 and talkclass 2 may be the same. Thus, the "one utterance class + intents" can be made to exist as a whole, i.e., one utterance class plus intents is determined as a logical combination. Thus, a new translation relationship "from a logical combination, a conversational class" may be obtained that is both unambiguous and more consistent with the operating logic of the computing device.

After the session of the previous round is finished and before the session of the current round is started, in order to determine the type of the dialect transmitted to the user from the current round, the type of the dialect transmitted to the user from the previous round and the intention of the user from the previous round can be determined as the current logical combination.

S106: and determining the rule matched with the current logic combination in each rule contained in the pre-constructed rule base.

In step S104, the conversion relation "a speech category is determined according to a logical combination", which is a rule used in the speech service execution method provided in this specification, is obtained.

There are a large number of rules in the rule base, each rule including a correspondence between a logical combination and a conversational class, i.e., a conversational class is obtained from a logical combination. Therefore, the rule containing the current logic combination, that is, the rule matched with the current logic combination can be found in the rule base, and the dialect type corresponding to the current logic combination can be obtained.

S108: and determining the candidate dialect types of the candidate dialects sent to the user according to the rules.

As mentioned in step S102, the utterances of the same utterance class are expressed in the same meaning, so that to determine the utterance candidates to be transmitted to the user, the utterance candidates to be transmitted to the user may be first determined.

In the dialog service execution method provided in the present specification, one logical combination can only have a correspondence with one utterance class, and one utterance class can have a correspondence with a plurality of logical combinations. In other words, according to one logical combination, only one kind of dialogies can be obtained; and the same dialect class may be obtained according to different logical combinations.

Thus, when the current logical combination is determined, the type of the candidate dialect to be transmitted to the user in the next round is determined. The rule matching the current logic combination can be found in the rule base, and the only one rule has a corresponding relation with the current logic combination.

S110: and determining the dialect to be sent to the user in a pre-constructed dialect library according to the candidate dialect type, and sending the dialect to the user.

In the method, all dialogs sent to the user by the intelligent customer service are existing dialogs in the dialogs library. When the conversation is determined to be performed, the determined conversation can be sent to the user, and the conversation is continued.

Since there are a plurality of dialects in one category of the dialects, one of them can be selected to be transmitted to the user when determining the dialects to be transmitted to the user in the pre-constructed dialects library. Specifically, each candidate utterance corresponding to the candidate utterance category may be determined in a pre-constructed utterance library; and selecting one of the determined candidate dialogs as the one to be sent to the user. Actually, the meaning of each utterance in the same utterance category is the same, so when selecting an utterance in each candidate utterance, one of the candidate utterances can be arbitrarily selected and sent to the user, and the one of the candidate utterances can be selected randomly or in a specific manner, which is not limited in the specification.

When the dialogue service execution method provided by the specification is adopted to execute the dialogue service, the dialect sent to the user in the current round and the intention of the user in the current round can be used as the current logic combination, the type of the dialect sent to the user in the next round is determined according to the current logic combination by the rules in the rule base, and the dialect sent to the user is selected from the dialect base according to the type of the dialect. Compared with the traditional dialogue method realized based on the routing tree, the dialogue method has the advantages that the relation among all dialogues in the dialogue at each turn in the method does not need to be considered, and the dialogue logic is flexible; meanwhile, when the conversation content of the intelligent customer service needs to be modified, only the rules in the rule base need to be modified, the logics of all conversations do not need to be rearranged, and the updating is flexible.

Further, the present specification additionally gives here how to pre-construct a rule base. Specifically, an input interface can be displayed for an operator, wherein the input interface at least comprises a first input field, a second input field and a third input field; responding to data input by the operator in the first input field, the second input field and the third input field respectively, and determining a first target conversation, a target reply and a second target conversation; determining an input combination according to the sequence of the first target dialect, the target reply and the second target dialect; combining the inputs into an input model, and generating a target rule through the rule generation model; and constructing the rule base by adopting the target rule.

Additionally, it is conceivable that the device for executing the dialog service execution method is updated when the dialog contents of the intelligent customer service need to be modified, and the method can also be implemented. It should be noted that the dialogs and intents employed in building or updating the rule base should likewise be from the dialogs base as well as the intents base.

It should be noted that the first target dialect and the second target dialect should be dialects existing in a dialect library, and the target intention should be intentions existing in an intention library; similarly, the logical combination contained in each rule in the rule base is the combination of a grammar in a grammar base and an intention in an intention base.

When an operator wishes to build or update a rule base, it is actually desirable to generate new rules. During actual operation, an input interface comprising the first input field, the second input field and the third input field is displayed to an operator. In this specification, the first input field receives the first target utterance by default, the second input field receives the target reply by default, and the third input field receives the second target utterance by default. Meanwhile, after the operator completes all the inputs, the system automatically determines the inputs of the operator as input combinations according to the sequence of the first target dialect, the target reply and the second target dialect, and inputs the input combinations with the sequence into the rule generation model so as to generate the rules through the model.

In this specification, a rule generation model trained in advance is employed to generate each rule. Specifically, a first target category of the first target utterance and a second target category of the second target utterance may be determined through a category classification layer in the rule generation model; determining at least one target intention according to the target reply through an intention analysis layer in the rule generation model; and generating a target rule according to the target first target type, the target intention and the second target type through a generation layer in the rule generation model.

Since more than one target intention may be determined according to the target reply, and the logical relationship between the target intentions is different, the finally generated rules are also different. For example, when there are a target intention X and a target intention Y, a rule of "first target kind + (target intention X and target intention Y) → second target kind", and possibly "first target kind + (target intention X or target intention Y) → second target kind" may be obtained. Therefore, in the intention analysis layer in the rule generation model, the logical relationship between the target intentions can also be determined. Specifically, the intention analysis layer in the rule generation model can determine each target intention and a logical relationship between the target intentions according to the target reply in the input combination, wherein the logical relationship comprises at least one of AND and or OR.

It is worth mentioning that, in actual use, the model for analyzing the intention of the user to reply to the content may be the same as the intention analysis layer in the rule generation model. Namely, at least one intention of the user and a logic relation between the intentions are determined according to the reply content through the intention analysis layer.

The rules include the correspondence between logical combinations and the types of dialects, but in practice, the logical combinations are artificially defined for the convenience of understanding and calculation of the machine, and it is the dialects and intentions that are input into the model for the operator who creates the library or updates, and the logical combinations are determined by the model. Since the contents input by the operator are input to the model in the form of a combination in a fixed order, the model can automatically determine a logical combination according to the order in the input combination and generate a rule. Specifically, a combination of the first target category and the target intent may be determined as a target logical combination; and generating a target rule according to the target logic combination and the second target type, wherein the target rule is that the target logic combination and the second target type have a corresponding relation.

When the generated rule is added into the rule base, the existing rule can be modified, or a new rule can be added. Specifically, when a specified rule containing the target logic combination exists in the rule base, replacing the specified rule with the target rule; adding the target rule to the rule base when a rule containing the target logical combination does not exist in the rule base.

For example, when there is already a rule "type 3+ intention C → type 0" (type 0 may be any type) in the existing rules of the rule base, the new incoming rule may be used to replace the original rule because the logical combination "type 3+ intention C" included in the rule is the same as the logical combination included in the new incoming rule "type 3+ intention C → type 4". In the replacement, the whole rule may be directly and completely replaced, or only two different parts may be replaced, that is, the "dialect type 4" is used to replace the "dialect type 0", and the present specification is not limited herein. Alternatively, when there are no rules in the rule base that contain the logical combination "tactical category 3+ intent C", then the new incoming rule may be added directly to the rule base.

Similarly, the rule generating model may be pre-trained. Specifically, a first sample conversation, a sample reply, a second sample conversation and a labeling rule are obtained; determining a sample input combination according to the sequence of the first sample conversation, the sample reply and the second sample conversation; inputting the sample input combination into a rule generating model to be trained, and generating a rule to be optimized through the rule generating model; and training the rule generation model by taking the minimum difference between the rule to be optimized and the labeling rule as an optimization target. The obtained labeling rule can be generated directly in a manual operation mode.

It is worth mentioning that the rule generation model can be trained and used independently without the intelligent customer service. That is, the dialog service execution method provided by the present specification can directly adopt the trained model, and the intelligent customer service itself does not need to be trained additionally when in use.

In fact, the intelligent customer service in this specification is provided for the merchant to use for a dialogue with the user of the merchant. When the intelligent customer service leaves the factory, the intelligent customer service system is provided with a default dialect library, an idea library and a rule library. In fact, the dialog library, like the rule library, may be updated. That is, each merchant may set its own personalized grammar library and rules library. The updating of the rule base is already given in the above description, and when the word base is updated, only the word to be added to the word base and the existing word category corresponding to the word need to be input in the corresponding input interface. Furthermore, the merchant can share the personalized language library and the personalized rule library of the merchant to other users of the intelligent customer service as long as the merchant wishes. That is, the merchant using the intelligent customer service can directly apply the dialect library and the rule library shared by other people to the intelligent customer service. The difficulty of the intelligent customer service in use and adjustment is greatly reduced, and the flexibility of the conversation service processing method provided by the specification is further improved.

Based on the same idea, the present specification further provides a corresponding dialog service execution device, as shown in fig. 2.

Fig. 2 is a schematic diagram of a session service execution apparatus provided in this specification, including:

the obtaining module 200 obtains the reply content input by the user in the previous round, and determines the type of the dialect transmitted to the user in the previous round;

an intention determining module 202, determining the intention of the user according to the reply content;

a combination determination module 204, which determines the combination of the intention and the type of the dialect transmitted to the user in the previous round as the current logical combination;

a rule matching module 206, which determines a rule matching the current logic combination from the rules contained in the pre-constructed rule base;

a category determination module 208, for determining the candidate utterance category of the candidate utterance transmitted to the user according to the rule;

the sending module 210 determines a dialect to be sent to the user in a pre-constructed dialect library according to the candidate dialect category, and sends the determined dialect to the user.

the sending module 210 is specifically configured to determine, in a pre-constructed utterance library, each candidate utterance corresponding to the candidate utterance category; and selecting one of the determined candidate dialogs as the one to be sent to the user.

Optionally, the apparatus further includes a building module 212, specifically configured to present an input interface to the operator, where the input interface includes at least a first input field, a second input field, and a third input field; responding to data input by the operator in the first input field, the second input field and the third input field respectively, and determining a first target language technique, a target reply and a second target language technique; determining an input combination according to the sequence of the first target dialect, the target reply and the second target dialect; inputting the input combination into a pre-trained rule generating model, and generating a target rule through the rule generating model; and constructing the rule base by adopting the target rule.

Optionally, the building module 212 is specifically configured to determine a first target category of the first target utterance and a second target category of the second target utterance through a category classification layer in the rule generation model; determining at least one target intention according to the target reply through an intention analysis layer in the rule generation model; and generating a target rule according to the target first target type, the target intention and the second target type through a generation layer in the rule generation model.

Optionally, the target intent is a plurality of intents;

the building module 212 is specifically configured to determine, through an intent analysis layer in the rule generation model, each target intent and a logical relationship between the target intents according to the target replies in the input combinations, where the logical relationship includes at least one of and or.

Optionally, the building module 212 is specifically configured to determine a combination of the first target category and the target intent as a target logical combination; and generating a target rule according to the target logic combination and the second target type, wherein the target rule is that the target logic combination and the second target type have a corresponding relation.

Optionally, the belonging intention determining module 202 is specifically configured to determine, by the intention analysis layer, at least one intention of the user and a logical relationship between the intentions according to the reply content.

Optionally, the building module 212 is specifically configured to, when a specified rule including the target logical combination exists in the rule base, replace the specified rule with the target rule; adding the target rule to the rule base when a rule containing the target logical combination does not exist in the rule base.

Optionally, the apparatus further includes a training module 214, specifically configured to obtain a first sample utterance, a sample reply, a second sample utterance, and a labeling rule; determining a sample input combination according to the sequence of the first sample grammar, the sample reply and the second sample grammar; inputting the sample input combination into a rule generating model to be trained, and generating a rule to be optimized through the rule generating model; and training the rule generation model by taking the minimum difference between the rule to be optimized and the labeling rule as an optimization target.

The present specification also provides a computer readable storage medium storing a computer program, which is operable to execute a method of conversational service execution as provided in fig. 1 above.

This description also provides a schematic block diagram of an electronic device corresponding to that of fig. 1, shown in fig. 3. As shown in fig. 3, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, but may also include hardware required for other services. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs the computer program to implement the method for performing the dialog service described in fig. 1. Of course, besides the software implementation, the present specification does not exclude other implementations, such as logic devices or a combination of software and hardware, and the like, that is, the execution subject of the following processing flow is not limited to each logic unit, and may be hardware or logic devices.

In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an Integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Hardware Description Language), traffic, pl (core universal Programming Language), HDCal (jhdware Description Language), lang, Lola, HDL, laspam, hardward Description Language (vhr Description Language), vhal (Hardware Description Language), and vhigh-Language, which are currently used in most common. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic of the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may thus be considered a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.

The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functions of the various elements may be implemented in the same one or more software and/or hardware implementations of the present description.

As will be appreciated by one skilled in the art, embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The description has been presented with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the description. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

It should also be noted that 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 an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

This description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

All the embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only an example of the present specification, and is not intended to limit the present specification. Various modifications and alterations to this description will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification should be included in the scope of the claims of the present specification.

Claims

1. A method for performing a dialogue service, comprising:

acquiring reply content input by a user in the previous round, and determining the type of the dialect transmitted to the user in the previous round;

determining the intention of the user according to the reply content;

2. The method of claim 1, wherein each utterance and an utterance class of each utterance are stored in the utterance library;

3. The method of claim 1, the rule comprising: a correspondence between a logical combination and a conversational class, wherein a logical combination is a combination of a conversational class and at least one intention.

4. The method of claim 3, wherein the pre-constructing of the rule base specifically comprises:

displaying an input interface to an operator, wherein the input interface at least comprises a first input field, a second input field and a third input field;

inputting the input combination into a pre-trained rule generation model, and generating a target rule through the rule generation model;

and constructing the rule base by adopting the target rule.

5. The method according to claim 4, wherein generating the target rule through the rule generation model specifically comprises:

6. The method of claim 5, the target intent being a plurality of intents;

determining at least one target intention according to the target reply through an intention analysis layer in the rule generation model, and specifically comprising:

7. The method according to claim 5, wherein generating the target rule according to the target first target category, the target intent, and the second target category specifically comprises:

8. The method according to claim 6, wherein determining the user's intention according to the reply content specifically comprises:

determining at least one intention of the user and a logic relation among the intentions according to the reply content through the intention analysis layer.

9. The method of claim 4, wherein constructing the rule base using the target rule specifically comprises:

adding the target rule to the rule base when a rule containing the target logical combination does not exist in the rule base.

10. The method of claim 4, wherein pre-training the rule generating model specifically comprises:

11. A conversation service execution apparatus comprising:

the acquisition module is used for acquiring reply contents input by a user in the previous round and determining the type of the dialect transmitted to the user in the previous round;

the category determining module is used for determining the candidate speech technology category of the candidate speech technology sent to the user according to the rule;

12. A computer-readable storage medium, storing a computer program which, when executed by a processor, implements the method of any of claims 1 to 10.

13. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of any of claims 1 to 10 when executing the program.