CN115640386A - Method and apparatus for conducting dialogs based on recommended dialogs - Google Patents

Abstract

The present application proposes a method and device for dialogue based on recommended utterances. The method includes: for the current dialogue topic among one or more dialogue topics involved in the real-time dialogue, based on the real-time dialogue data and the prior information of the dialogue participants, selecting the first one from at least one recommended speech template corresponding to the current dialogue topic Recommending a speech template, wherein the dialogue topic is the intention or purpose of the real-time dialogue; generating dialogue content based on the first recommended speech template; obtaining responses from dialogue participants to the generated dialogue content, and updating real-time dialogue data based on the reply to generate the following A dialog content, wherein the dialog topic and at least one recommended speech template corresponding to the dialog topic are determined based on historical dialog data of historical dialogs.

Description

Method and device for dialogue based on recommended utterances

technical field

The present application relates to intelligent information interaction, and more particularly, to a method, device, and computer-readable medium for intelligent dialogue based on recommended speech techniques.

Background technique

With the development of speech recognition and semantic matching technology, there are more and more applications of using dialogue robots to communicate with human users based on speech recommendation algorithms. Common intelligent dialogue solutions often use fixed recommended speech techniques, which cannot obtain the dialogue needs of human users in real time and respond in a timely manner, making the effect of automated dialogue unsatisfactory.

Therefore, there is a need to improve existing intelligent dialog solutions.

Contents of the invention

The present application proposes a method and device for selecting recommended words based on prior information in historical dialogue data and real-time dialogue data.

According to one aspect of the present application, a method for dialogue based on recommended speech is proposed, including:

For the current dialogue topic among the one or more dialogue topics involved in the real-time dialogue, based on the real-time dialogue data and the prior information of the dialogue participants, a first recommended speech template is selected from at least one recommended speech template corresponding to the current conversation topic , where the conversation topic is used to indicate the intent or purpose of the live conversation;

Generate dialogue content based on the first recommended speech template;

Obtaining responses from dialogue participants to the generated dialogue content, and updating real-time dialogue data based on the responses to generate the next dialogue content,

Wherein, the dialog topic and at least one recommended utterance template corresponding to the dialog topic are determined based on historical dialog data of historical dialogs.

According to another aspect of the present application, a device for dialogue based on recommended speech techniques is proposed, including: an interaction unit configured to acquire real-time dialogue data and output dialogue content to dialogue participants; and a recommendation unit configured to For the current dialogue topic among the one or more dialogue topics involved in the real-time dialogue, based on the real-time dialogue data and the prior information of the dialogue participants, a first recommended speech template is selected from at least one recommended speech template corresponding to the current conversation topic , wherein the dialogue topic is used to indicate the intention or purpose of the real-time dialogue; generate the dialogue content based on the first recommended speech template; and obtain the reply of the dialogue participants to the generated dialogue content, and update the real-time dialogue data based on the reply to generate the next conversation content,

Wherein, the dialog topic and at least one recommended utterance template corresponding to the dialog topic are determined based on historical dialog data of historical dialogs.

According to still another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program includes executable instructions. When the executable instructions are executed by a processor, the method as described above is implemented.

According to yet another aspect of the present application, an electronic device is proposed, including a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the executable instructions to implement the above-mentioned method.

The dialogue scheme based on recommended speech techniques proposed in this application can combine the prior information associated with the dialogue participants and dialogue scenes and the real-time dialogue data of the current dialogue to timely and accurately obtain the dialogue needs and feedback information of human dialogue participants , choose a more appropriate recommended speech template to ask or reply to the dialogue participants, quickly obtain the information needed for the dialogue task from the dialogue, and continue to maintain a good relationship with the human dialogue participants until the dialogue task is completed. communication effect. On the basis of instantly and accurately obtaining and responding to the interlocutor's needs, the solution of this application can also periodically return real-time dialogue data to update historical dialogue data according to the dialogue effect of the recommended speech template, and iteratively update and improve the recommended speech template database. , so as to obtain a better intelligent dialogue effect.

Description of drawings

The above and other features and advantages of the present application will become more apparent by describing in detail exemplary embodiments thereof with reference to the accompanying drawings.

Fig. 1 is a schematic flowchart of a process for generating a database of recommended speech templates according to an embodiment of the present application.

Fig. 2 is a schematic flow chart of a method for dialogue based on recommended utterances according to an embodiment of the present application.

Fig. 3 is a schematic block diagram of a device for dialogue based on recommended utterances according to an embodiment of the present application.

Fig. 4 is a schematic block diagram of an electronic device according to an embodiment of the present application.

Detailed ways

Exemplary embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many forms and should not be construed as limited to the embodiments set forth herein; rather, these The concepts of are fully conveyed to those skilled in the art. In the drawings, the size of some elements may be exaggerated or deformed for clarity. The same reference numerals in the drawings denote the same or similar structures, and thus their detailed descriptions will be omitted.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided in order to give a thorough understanding of the embodiments of the application. However, those skilled in the art will appreciate that the technical solutions of the present application may be practiced without one or more of the specific details, or that other methods, elements, etc. may be employed. In other instances, well-known structures, methods, or operations are not shown or described in detail to avoid obscuring aspects of the application.

The intelligent dialogue solution based on the speech skill recommendation algorithm proposed in this application can be applied to various intelligent dialogue scenarios. For example, in a smart recruitment scenario, the recruiter first sends recruitment information (such as job advertisements, etc.) about the job to the public or a specific group of candidates. , You can use an intelligent dialogue system such as a dialogue robot to conduct initial communication with candidates, so as to obtain the candidate information required for recruitment as well as the candidate's needs and views on the recruitment position as much as possible. Therefore, automated dialogue in intelligent recruitment can be used as a means of supplementing candidate information and performing preliminary matching and screening between candidates and recruiting positions. For another example, in intelligent customer service scenarios such as intelligent handling of repair reports and complaints, in order to reduce the work pressure of customer service specialists and improve the accuracy and professionalism of services, an intelligent dialogue system can be used to communicate with users first and guide users to describe accurately Repair information or complaint content, in order to find professionals who can solve the problem. It is even possible to solve the customer service needs of users through the intelligent dialogue system and reduce the workload of customer service specialists. In addition, the intelligent dialogue system can also be used to identify the user's emotions, properly address the user's needs and improve the user's communication experience.

The intelligent dialogue scene may include chatting type, task type and other dialogue type dialogue scenes according to its purpose. The dialog scheme based on recommended speech involved in this application needs to obtain as much information as possible from human dialog participants associated with the dialog topic in the dialog scenario, which belongs to the intelligent dialog scenario of the task type. In the real-time dialogue process, at least two dialogue participants are included, one is a human dialogue participant (such as a candidate in a smart recruitment scenario, or a user or customer in a smart customer service scenario), and the other is generated based on a recommended speech template An intelligent dialogue system (such as a dialogue robot) that uses dialogue content to ask questions of human dialogue participants or answer questions from human dialogue participants. The real-time dialog process can also include more parties. In this paper, the dialogue participant specifically refers to the human dialogue participant who communicates with the intelligent dialogue system, that is, the dialogue partner of the intelligent dialogue system.

The prior information of the intelligent recruitment scenario mentioned in this application refers to the information related to the dialogue participants and the information related to the intelligent dialogue scenario. The information related to the dialogue participants can represent the attributes of the dialogue participants themselves. For example, in an intelligent recruitment scenario, the candidate’s resume information can be the information part of the prior information related to the dialogue participant (candidate) himself, and the recruitment information (such as a job advertisement) is the part of the prior information related to the dialogue scene information section of the . The candidate may have sent the resume to the recruiter before the automated dialogue process of the intelligent recruitment, assuming that the resume data has been sent to the intelligent dialogue system. In the intelligent customer service scenario, the user has electronically submitted preliminary information about the fault description and complaint content as prior information through a questionnaire or template before talking to the intelligent dialogue system.

At least one of the prior information and the historical dialog data of the historical dialog is used to generate a default recommended utterance template database used by the intelligent dialog system. Prior information and historical dialogue data can be of various data types. For example, the prior information such as a candidate's resume may be a paper resume, or it may be digital resume data acquired through a web interface or a mobile program (APP) provided by the recruiter. Non-digital information such as paper resumes can be converted into digital data through digital processing. It is also possible to convert different types of digitized data into specific types of digitized data required by the intelligent dialogue system. For example, historical conversation data that includes conversations between human conversation participants (such as phone calls or face-to-face interview data between HR and candidates) or conversations between intelligent conversation systems and human conversation participants is often recorded voice or video Data can also be text or image data that processes voice or video data, then graphics, images, or voice/video data can be converted into text data for system analysis and use.

In intelligent dialogue scenarios of task type, the default dialogue process can be set according to the scene requirements. In the dialogue process, the intelligent dialogue system needs to ask questions to the human dialogue participants to obtain enough information expected from the dialogue task. At the same time, the intelligent dialogue system should also give corresponding responses according to the questions raised by the human dialogue participants, so as to meet the peer-to-peer needs of the human dialogue participants to obtain information. Taking the intelligent recruitment scenario as an example, the role of the intelligent dialogue system (dialogue robot) is equivalent to that of human HR. During the dialogue, human HR needs to complete the tasks of asking candidates and answering candidate inquiries during the interview process.

During the conversation, the chatbot may need to talk to the candidate on at least one topic. Here, one or more topics in the real-time dialogue process are dialogue topics in the dialogue scene, which may also be called dialogue intent, and are used to indicate the intention or purpose of the dialogue. For example, in an intelligent recruitment scenario, a dialogue robot replaces HR to ask candidates (dialogue participants) questions about each dialogue topic, or a robot (HR) asks questions at each dialogue node first, and waits for the candidate to reply before viewing the dialogue topic. Depending on the degree of information expected to be obtained and the status of the candidate, continue to ask questions again under the topic of the dialogue, or wait for the candidate to ask the dialogue robot a question and give a targeted reply. Here, the concept of a dialog node can also be called a topic node or an intent node, which is a dialog opportunity or link used to obtain information related to a certain dialog topic, which includes information from the dialog robot and/or candidate nodes corresponding to the dialog topic. One or more questions and corresponding responses from a person. Therefore, the number of question-type or answer-type recommended utterance templates that can be used by the dialogue robot in each dialogue node is one or more. Multiple recommended speech templates are used to solve the problem that a single question is not enough to obtain all the information related to the conversation topic, or when the conversation content generated by using the recommended speech template is rejected by the candidate when asking or replying to the conversation topic. Replace the recommended speech template to generate the dialogue content again.

If all the information expected by the current dialogue topic has been obtained and the candidate has no further questions to ask the dialogue robot, the real-time dialogue session for this dialogue topic ends, and the dialogue process enters the next dialogue topic, that is to say The current real-time dialogue enters the dialogue node of the next dialogue topic. Otherwise, the real-time dialog remains at the current dialog node.

As mentioned above, the default dialogue flow of the intelligent dialogue scene may be divided into one or more dialogue links for one or more dialogue topics by dialogue nodes. The intelligent dialogue system (dialogue robot) selects and recommends one or more dialogue topics involved in the dialogue based on the real-time dialogue data of the current dialogue topic and the prior information related to the dialogue participants in the corresponding dialogue node. The speech template generates dialogue content to complete questions or answers to dialogue participants. Repeat the dialog link for each dialog topic in each dialog node until all the desired information of the dialog topic is obtained, or the dialog is terminated by the intelligent dialog system or the dialog participants when the preset termination conditions are met .

According to the embodiment of the present application, the dialogue process with human dialogue participants is usually carried out in the form of voice or video, but the technology for generating dialogue content based on recommended words usually adopts an algorithm or model based on text data. Therefore, when the dialogue robot asks the dialogue participants or replies to the latter's questions, the generated dialogue content (that is, historical dialogue data or real-time dialogue data from the intelligent dialogue system/dialogue robot) can be converted into voice data or information. Correspondingly, when obtaining replies or questions from human dialogue participants (that is, historical dialogue data or real-time dialogue data from dialogue participants), the audio/video data can be converted into text data. There is a two-way conversion of data formats between text data and audio/video data.

The solution of this application will be introduced below by taking the recruitment dialogue process between a dialogue robot and a human candidate in an intelligent recruitment scene as an example with reference to accompanying drawings 1 to 3 . Those skilled in the art will understand that this embodiment is only exemplary, rather than limiting the protection scope of the scheme of the present application.

The method for dialogue based on recommended speech skills of the present application first generates a default recommended speech template database based on historical dialogue data and/or prior information of historical dialogues, and then selects recommendations based on real-time dialogue data and prior information during the real-time dialogue process Speech templates to generate dialogue content in real time. At the same time, the real-time dialogue data of the real-time dialogue process can be returned periodically or when the preset conditions are met to train offline and update the default recommended speech template database to achieve closed-loop iterative update of recommended speech.

Figure 1 shows the process of generating and updating the default recommended speech template database. Those skilled in the art will understand that when using a dialogue robot for an intelligent recruitment dialogue for the first time (also referred to as a cold start of an intelligent dialogue system), audio or video The data is processed into the historical dialogue data of the historical dialogue in the form of text, and the first version of the default recommended speech template database is generated. After using the recommended speech skills generated by the default recommended speech template database for intelligent dialogue, the real-time dialogue data of the continuously recorded real-time dialogue can be added to the historical dialogue data collection as new historical dialogue data 101, and updated to obtain a new version of the recommendation Database of speech templates.

The process of using the default recommended speech generation algorithm for the default recommended speech template database mainly includes the step S110 of extracting the conversation topic, the conversation label and the historical conversation data from the historical conversation data, the step S120 of determining the slot and the corresponding relationship, and removing The step S130 is to generate a default recommended speech template database through content items in the historical dialogue data.

In step S110, use algorithms such as unsupervised clustering and keyword mining to extract and summarize the questions and corresponding answers raised by HR and candidates in historical dialogues, and obtain relevant dialogues in the recruitment dialogue scene Topics, conversation tags related to conversation topics, and historical conversation materials corresponding to conversation topics.

In the default dialogue flow, it usually starts with a question from one party to the dialogue, followed by a reply from the other party. Therefore, extracting relevant data and information from (historical or real-time) dialogue data needs to detect the type of the current dialogue content in the historical dialogue data or real-time dialogue data, that is, the dialogue content is from the dialogue participant who asks the question (in the historical dialogue, it can be a question In the real-time dialogue of the intelligent dialogue process, it can be a dialogue robot or candidate) or from the answering dialogue participant (in the historical dialogue, it can be the HR or candidate who asked the question, in the real-time dialogue of the intelligent dialogue process, it can be conversational bot or candidate). In this paper, the dialogue types of real-time dialogues are distinguished by question types and answer types. During the dialogue, the candidate can give a reply including specific information to questions from HR or the dialogue robot, and can also give affirmative or negative responses to questions such as asking for tendentious opinions to indicate whether the candidate agrees to come from HR or the dialogue Bot-related inquiries. Usually, when HR or the dialogue robot receives a positive or non-negative reply from the candidate, it means that the information that needs to be confirmed in the question just raised has been verified, but if it receives a negative reply, it means that at least one of the information that needs to be confirmed is Incorrect or Candidate Negative. At the same time, a negative reply may also indicate that the interlocutor (candidate) is not interested in or has negative emotions in the conversation topic of the conversation process or the current real-time conversation content. In the process of human dialogue, HR can also give introductory dialogue content to candidates such as recruitment information (such as job content, salary benefits, working hours, etc.) in response to the negative reply just now, so as to attract the candidate's attention or reverse its negative emotions, thereby maintaining the conversation until the end of this conversation. The content of the introductory dialogue usually adopts the template of the introductory speech type, which is still the revised dialogue content of the previous question or reply in essence, so the content of the introductory dialogue can be classified as a real-time dialogue with the previous question or reply The same dialogue type, such as dialogue content that is still considered a question type or an answer type.

When extracting information from historical dialogue data (and real-time dialogue data described below), it is necessary to separate the complete dialogue data given by the questioning party or the replying party belonging to the question type or reply type from the recorded continuous voice or video data. The complete dialogue data covered by a complete and continuous question or reply output by the same dialogue participant is called the dialogue material in the dialogue data, which is the smallest dialogue unit of the dialogue process. A dialogue material may include one or more sentences, or even include one or more paragraphs with multiple sentences. Dialogue materials can be accurately segmented from continuous speech or video data (and converted from continuous text-type dialogue data) through relevant speech recognition algorithms. For example, a complete dialogue material can be determined by identifying whether the dialogue data is output by the same dialogue partner, or identifying modal particles (such as interrogative words) and/or punctuation marks (such as periods or question marks) in the dialogue data. Usually, in the interaction between two parties or even multiple parties in a dialogue, multiple questions continuously output by the same dialogue party can be divided into the same dialogue data from the dialogue data.

After the dialogue material segmentation is completed, the dialogue material can be clustered using an unsupervised clustering algorithm for two different sets of dialogue material of question type and answer type, and the dialogue material can be divided into different topic categories. , and determine the name of each topic category, that is, the dialogue topics involved in the dialogue materials in this class, so as to determine one or more dialogue topics involved in the historical dialogue data and the historical dialogue materials under the dialogue topics. This step is used to determine which categories of questions need to be communicated with candidates during the dialogue process, and what is the theme of each category of questions. Since the two different types of dialogue material collections of question type and answer type are clustered respectively, the historical dialogue materials belonging to the same dialogue theme category can be divided into historical dialogue materials of question type (question type historical dialogue data) and Answer type historical dialogue data (answer type historical dialogue data), this classification method facilitates the generation and use of subsequent recommended speech templates.

For intelligent recruitment dialogue scenarios, the clustered dialogue topics can include job content, salary and benefits, working hours, promotion mechanism, etc. Among them, salary and benefits can be further subdivided into two dialogue topics, insurance and salary. In clustering, the historical dialogue materials involving topics such as four insurances and one housing fund and five insurances and one housing fund can be clustered into insurance dialogue topic categories, and historical dialogue materials involving annual salary, monthly salary, hourly salary, bonus, etc. Conversation topic classes clustered into payroll. For example, historical dialogue materials of the question type asking about job content or candidates’ satisfaction with the job content, and historical dialogue materials of the answer type responding to candidates’ job content questions all belong to the dialogue topic category of job content.

For the clustered conversation topics and the historical conversation data in each conversation topic class, at least one conversation label related to the conversation topic is extracted using a keyword harvesting algorithm. Dialogue labels are used to characterize the attributes of each dimension of the dialogue topic, which is equivalent to the keywords of the dialogue topic or intent. For example, for the dialogue topic of insurance, five insurances and one housing fund and four insurances and one housing fund can be extracted as corresponding dialogue tags. For example, in a conversation topic related to working time, related conversation tags may include time-related keywords and other related high-frequency keywords.

Step S110 converts the historical dialogue data obtained through clustering and keywords into a historical dialogue corpus consisting of a collection of historical dialogue materials with corresponding dialogue topics and related dialogue labels. After obtaining updated historical dialogue data, the historical dialogue corpus of this application can perform clustering and keyword retrieval algorithms again to update dialogue topics, related dialogue labels, and corresponding historical dialogue materials, so as to be the default recommended speech template database. The update provides data support.

Next, in step S120, for the results of clustering and keyword extraction, use NER (Named Entity Recognition) and RE (Relation Extractio, relationship extraction) algorithms to extract and identify the dialogue tags in the historical dialogue materials. entity. The entity indicates the content and location of the slot for removal from the historical conversation corpus for both the question type and the answer type. Further, the slot can be determined and extracted through NER, and the corresponding relationship between the slot and the dialogue label can be determined through the RE algorithm.

The NER algorithm can identify entities in the historical dialogue material (question or answer type) in the dialogue material referred to by the dialogue label. An entity is a content item corresponding to or referred to by a dialog tag, and may be text data in the form of a character string. These entities are removed from the historical dialogue data, and the slots corresponding to the dialogue labels in the historical dialogue data are obtained.

The RE algorithm is used to find the context of the slot in the historical dialogue data. Contextual relations include referential relations and modification relations. For example, for the dialog tag A, after the NER algorithm determines the corresponding slots A1 and A2 of the dialog tag A in the historical dialog data, the RE algorithm finds the context relationship between the slots A1 and A2. The reference relationship includes A1 referring to A2 or A2 referring to A1, and the modification relationship includes A1 modifying A2 or A2 modifying A1. The slot corresponding to the dialog label A can be found through the context determined by the RE algorithm.

Therefore, the slots corresponding to the dialog tags in the dialog data can be extracted through NER, and the corresponding relationship between the slots and the dialog tags can be determined through the RE algorithm. According to the embodiment of the present application, the determination of the slot and the corresponding relationship between the slot and the dialogue label are all carried out for the historical dialogue materials belonging to the same dialogue type under the same dialogue topic category, so that the recommended speech template generated according to the slot Properties with the same conversation topic and conversation type. This screening mechanism only marks the dialogue tags in the dialogue material of the same dialogue topic and the same dialogue type, and only determines the slot corresponding to the dialogue label for a fixed context relationship. Its role is to generate recommended speech templates more conveniently and accurately. Reduce computational burden. For example, for candidates who do not accept work schedules such as shifts, the recommended script templates can be found only in the historical dialogue materials that do not accept working hours, thereby improving the accuracy and efficiency of the script recommendation.

The determined slots can be reviewed by business experts to assist in checking whether the slot settings in the historical dialogue data conform to the language habits.

In step S130, the corresponding entity (content item) in the historical dialogue material is removed based on the determined slot to form a slot to be filled, thereby converting the historical dialogue material into a recommended speech template. In this way, a default recommended speech template database with slots is generated from the historical dialogue corpus. The recommended speech template database includes not only the recommended speech templates, but also the dialogue topics and dialogue types corresponding to the recommended speech templates, as well as the dialogue labels related to the dialogue topics and the correspondence between dialogue labels and slots.

An example of the generated default recommended speech template is as follows:

If the candidate gives a negative reply to the question about working hours, that is, does not accept the working hours system such as shift work, the default recommended speech template generated can be: "Seeing your ____ related work experience, I believe you also I can understand that the time in the service industry is almost the same, but we can ensure that ____, the working hours are more flexible when shifting shifts, just have time to do other things. Although our working hours are relatively early, it is also in ____ Go to work around 10:00, and leave work at ____ in the afternoon. In fact, the evening shift also starts to work at ____ in the afternoon, and usually goes up to around ____. And I see that you are also living in: ____, our working place It is also allocated nearby to reduce commuting time and difficulty, can you give it a comprehensive consideration?" The underline is the slot corresponding to the conversation label related to the conversation topic during working hours.

If the real-time dialog data from the real-time dialog is used as the update data of the historical dialog data 101 in the subsequent process, steps S110 to S130 can be performed again to obtain an updated default recommended speech template database. According to the embodiment of the present application, the updated default recommended speech template database includes not only the updated recommended speech template, but also updated conversation topics after re-clustering, conversation tags related to the updated conversation topics, and the conversation Updated historical dialogue materials under the topic category, including updated slot settings and the correspondence between slots and dialogue labels.

Fig. 2 shows a schematic flow of a method for dialogue based on recommended speech according to an embodiment of the present application.

The method flow mainly includes the step S210 of selecting a recommended speech template based on the real-time dialogue data and the prior information of the dialogue participants for the current dialogue topic among one or more dialogue topics involved in the current real-time dialogue so as to generate the dialogue content in real time and acquiring the dialogue content. A step S220 of updating real-time dialogue data based on the responses of the participants (such as candidates) to the generated dialogue content so as to generate the next dialogue content in real time for the intelligent dialogue system (such as a dialogue robot).

For example, step S210 may ask candidates at corresponding dialogue nodes for dialogue topics such as education background, age, and graduation time, and obtain or confirm information such as the candidate's education background, age, and graduation time. Specifically, in step S210, firstly, in substep S211, the dialogue type of the real-time dialogue is determined according to the current dialogue topic, such as question type and answer type. The process of determining the dialogue type from the real-time dialogue data is similar to the process of determining the dialogue type from the historical dialogue data introduced above with reference to FIG. 1 , and will not be described in detail here.

Next, in sub-step S212 , the conversation tags (also referred to as real-time conversation tags) related to the real-time conversation of the current conversation topic are extracted from the real-time conversation data and prior information. Dialogue labels mainly come from dialogue labels related to real-time dialogue content (dialog content labels) 201 and dialogue labels related to prior information (priori dialogue labels), wherein the dialogue labels related to prior information further include information from dialogue participants Related dialog tags (dialogue participant tags) 202 and dialog tags (dialogue scene tags) 203 related to intelligent dialog scenarios. These dialog tags are combined to form a real-time dialog tag related to the current real-time dialog.

The dialog content tag 201 is a dialog tag extracted from real-time dialog content. The dialog content tag 201 may be a dialog tag selected from the current dialog topic-related dialog tags, wherein the topic-related dialog tags have been pre-selected through the keyword harvesting algorithm in the process of generating the default recommended speech template database shown in FIG. 1 . Sure. For example, when the dialog topic of the current dialog node is salary, the dialog tags related to the salary topic include current salary, expected salary, graduation time, and so on. Therefore, relevant information items such as current salary, expected salary and graduation time can be extracted from the real-time conversation content as the conversation content tag 201 .

As mentioned above, in the intelligent recruitment dialog scenario, the dialog participant tags 202 related to the dialog participants come from the candidate's resume information, so the dialog participant tags 202 may also be called candidate tags. Conversation participant labels 202 may be obtained based on resume information obtained prior to the smart recruiting dialogue using a resume parsing algorithm such as clustering. The resume parsing algorithm can extract candidate-related information items in the resume information as candidate labels. Resume information can be resume data in structured or unstructured form. Compared with structured forms, resume parsing algorithms can process unstructured resume data to generate structured resume data, from which to extract information such as the candidate's name, age, education, expected work location (such as city, region), current job Location (such as city, region), expected job category, current job category, work experience and other information items are used as candidate labels.

The dialogue scene label 203 related to the dialogue scene is related to the intelligent dialogue scene, for example, the dialogue scene label 203 of the intelligent recruitment dialogue is recruitment information. In the intelligent recruitment scenario, the dialog scene tag 203 related to the dialog scene is extracted from the previously released recruitment information through a recruitment parsing algorithm such as clustering, so it can also be called a recruitment information tag. If the recruiter has a wide range of business and a large market distribution (such as a national enterprise), there are differences in salary and benefits, work content and working hours in each market area of the country. The recruitment parsing algorithm can extract the information most relevant to the dialogue topic of this intelligent recruitment dialogue from the recruitment information in different regions as the dialogue scene label 203 . The recruitment information database can be established in advance, and the recruitment information in each recruitment advertisement comes from a part of the recruitment information database, wherein the recruitment information label can be selected from a set of recruitment information labels in the recruitment information database. The recruitment information tag may include, for example, the job content, working hours, job location, job category, job salary and benefits of the job to be recruited.

In actual use, quite a few job advertisements have a unified format, such as "job content: xxx, working hours: xxx". At this time, the job information label can be obtained by extracting the content item corresponding to the corresponding information item. If the job advertisements are not issued digitally, you can use the OCR recognition algorithm to perform text recognition on non-digital job advertisements in a uniform style (such as job posters in image format) to obtain text data about the job information, and then use NER and RE algorithms to extract relevant information. The recruitment information tag, that is, the dialog scene tag 203. In fact, when the pre-acquired candidate's resume information is in a non-digital format, the resume information in the form of text data can also be obtained in a similar manner to further extract candidate labels.

Similar to the dialog content tag 201, in resume information parsing and recruitment information parsing, the dialog participation is generally only extracted from the dialog tags that have been predetermined through the keyword harvesting algorithm in the process of generating the default recommended speech template database shown in Figure 1 The speaker tag 202 and the dialog scene tag 203, the purpose of which is to use the dialog tag, candidate tag, and recruitment information tag including the real-time dialog content to match the recommended speech template to obtain as much information as possible related to the dialog topic concerned by the current dialog node.

After obtaining the dialog content tag 201, the dialog participant tag 202 and the dialog scene tag 203, the data alignment operation can be performed, and the dialog tags (that is, the dialog content tag, the dialog participant tag) relevant to the dialog content, the dialog participant and the dialog scene and dialogue scene label) are merged into a dialogue label (real-time dialogue label) related to the real-time dialogue, and the latest and most comprehensive personalized information of the current real-time dialogue is obtained. The above combined operations do not require clustering, because the clustering operation is used to pre-generate the default recommended speech template database and/or extract candidate labels and recruitment information labels from resume information and recruitment information, in order to improve recommendation during real-time dialogue. Efficiency and speed, only extract and update the dialog content tag 201 related to the real-time dialog data of the current dialog in each dialog node.

Next, in sub-step S213 , select the recommended speech template that best matches the dialogue requirement or purpose of the current conversation topic from the default recommended speech template database to generate the dialogue content that the dialogue robot will output to human candidates. The currently optimal recommended speech template can be determined according to the degree of matching between the real-time conversation tag related to the real-time conversation and the slots of multiple recommended speech templates in the default recommended speech template database. The relationship between all the dialog tags in the dialog content tag 201, the dialog participant tag 202 and the dialog scene tag 203 in the personalized information obtained after the data alignment operation and the slot positions of all the templates in the default recommended speech template database can be calculated. The matching degree of the corresponding relationship. For example, for all dialogue tags in the personalized information of the current dialogue, the number of slots that have a corresponding relationship with these dialogue labels is based on the number of slots included in the default recommended speech template—that is, the number of slots that can be used in these dialogues The number of corresponding slots filled by tags—the ratio of the total number of slots of the recommended speech template is used as the matching rate or matching degree to sort all the templates. The recommended speech template with the highest matching rate can be selected as the optimal template for the real-time dialog of the current dialog. For example, if recommended speech template 1 can fill 80% of the slots and recommended speech template 2 can fill 100% of the slots, then select recommended speech template 2 as the recommended speech template for the current real-time dialogue , called the first recommended speech template or the optimal recommended speech template. When determining the matching rate, it may be based only on the corresponding relationship between the dialogue label and the slot, without using the context relationship between the slots.

In sub-step S214, from the dialogue content label 201, dialogue participant label 202 and dialogue scene label 203 of the candidate's personalized information, select a Those dialogue tags corresponding to the relationship are filled into the corresponding slots, and a complete dialogue material is generated as the dialogue content output by the dialogue robot to the candidate in the real-time dialogue.

In step S220, after the dialog content for the candidate is generated and output in sub-step S221, the candidate's reply to the dialog content may be acquired. The reply may be to provide further specific information for the information inquiry in the question, or may be an affirmative or negative reply or confirmation for the information inquiry in the question.

If it is judged in the sub-step S222 that the dialogue participant's reply is a negative reply or includes a negative reply (branch "yes"), it proves that the dialogue content generated in the step S210 has not obtained the expected speech recommendation effect and produces a negative reply of the candidate . The method proceeds to sub-step S224, in the dialog node for the current dialog topic, select another recommended speech template corresponding to the dialog type from the recommended speech template database based on the dialog content tag 201, the dialog participant tag 202 and the dialog scene tag 203 Generate the next conversation content.

As mentioned above, a negative reply indicates that at least one of the information to be confirmed is incorrect, or indicates that the candidate is not interested in the content of the current dialogue process or the real-time dialogue content of the current dialogue or has negative emotions. Such a negative reply can be regarded as real-time dialogue data including negative or negative information, wherein the negative information is negative or reverse information or data to the information involved in the dialogue content just generated. The negative information shows that at least in the personalized information, the dialog content tag 201 related to the real-time dialog needs to be updated and changed. In some cases, the information in the question may also involve the confirmation of the candidate's resume information and/or recruitment information, and a negative reply may also be used for the dialogue participant label 202 and the dialogue scene label 203 of the current real-time dialogue. make an impact.

After updating the dialog content tag 201, the dialog participant tag 202, and the dialog scene tag 203 in the personalized information, in substep S224, select the recommended speech that best matches the dialog requirement of the current dialog from the default recommended speech template database. Technical templates are used to generate new dialogue content for the dialogue bot to output to human candidates. At this time, the most suitable recommended speech template can be selected as a new recommended speech template, that is, the second recommended speech template, by selecting a recommended speech template with a different matching rate in the previous matching rate ranking. The second recommended speech template may have a lower matching rate than the first recommended speech template, for example, it has the second highest matching rate in the matching rate ranking. A new recommended speech template can also be selected based on other criteria. According to the embodiment of the present application, since the negative reply updates the personalized information to obtain a new set of conversation tags related to the real-time conversation, it is also possible to use the new conversation tag set (for example, a new real-time conversation tag) to re-determine all conversations The degree of matching between the label and the slots of all templates in the default recommended speech template database.

After the second recommended speech template is determined, in substep S225, the dialog tags are filled into the slots in the second recommended speech template to generate new dialog content and in substep S221, candidates are again acquired for the new dialog content reply.

For example, in the dialogue topic related to working hours, when the candidate is dissatisfied with the working hours-related questions output by the dialogue robot and gives a negative reply, the newly generated dialogue content of the question type can further be:

"Are you dissatisfied with the morning shift system, the evening shift system or the entire shift system?"

Candidate replies are retained during data alignment operations as real-time dialog-related label information. Correspondingly, according to the candidate's resume information, the content related to several key conversation tags can be analyzed, such as: whether the current city of residence is consistent with the city of application, whether the city of expected work is consistent with the city of current application, whether there is work experience related to catering, etc. . At the same time, the conversation tags related to the specific commuting time information can be found from the post-related recruitment information submitted by the candidates.

Then, if the candidate applies for the same city as his current residence, has relevant work experience, and is dissatisfied with the morning shift, the corresponding dialogue content generated based on the second recommended speech template can be: "Seeing that you have relevant work experience, I believe you can also understand this kind of working hours system. And although our working hours are relatively early, we also go to work around 6 or 7 o'clock, and get off work at 3 or 4 pm. You can go back to rest earlier, and I see that you are also living In: Shanghai , which is the same as the applicant city, and our work location is also assigned nearby, which reduces the time and difficulty of the job, you can think about it comprehensively.”

According to an embodiment of the present application, a strategy for repeatedly receiving negative replies from candidates in a dialog node of the same dialog topic may be set. For example, after receiving negative replies twice in a row, the robot ends the intelligent dialogue process and ends the interview.

If it is judged in sub-step S222 that the reply of the dialogue participant is not a negative reply (branch "no"), it proves that the dialogue content generated in step S210 has obtained the expected recommendation effect, and relevant specific information can be obtained from the candidate's reply or obtained. Acknowledgment by the candidate of the content of the question asked. Next, in sub-step S223, the reply of the current dialogue content is used as new dialogue data to update the real-time dialogue data, and the method returns to sub-step S211, and the dialogue robot asks the candidate about the current dialogue topic or the next dialogue topic again.

It can be seen that in the real-time dialogue in each dialogue node, the recommended speech skills determination algorithm must be run for each question link and reply link to generate real-time dialogue content dedicated to candidates.

The method can also use the dialogue robot to conduct an intelligent dialogue each time, according to the affirmative or negative reply (accepted or not accepted) made by the candidate, in substep S226, the real-time dialogue data involved in the affirmative reply that the candidate expresses acceptance The records are sent back to the generation and update process of the default recommended speech template database shown in Figure 1, and the algorithm model is iteratively upgraded to form a closed-loop update. The new historical dialogue data 101 formed by the real-time dialogue data of the affirmative reply is actually equivalent to the verified correct recommended speech samples, which can improve the performance of the algorithm model.

While using conversational bots to complete intelligent conversations, human services can also be retained. By counting the acceptance degree of various types of candidates for the dialogue content of the recommended speech, for those types of candidates whose acceptance of the dialogue content generated based on the recommended speech template does not meet the preset conditions, it can be continuously supplemented. For the artificial dialogue data of candidates of this type, re-run the default recommended speech template database generation algorithm to update the proprietary template database of candidates of this type. The preset condition may be, for example, that for a certain type of candidate, the acceptance degree of the dialogue content related to the corresponding dialogue type, dialogue topic, etc. is lower than the average acceptance degree of the overall recommended speech skills. Wherein, the types of candidates may include types based on the candidate's current city of residence (for example, candidates living locally and candidates living in other places), whether they are fresh graduates, and other characteristics. According to the statistical data of speech recommendation results of different candidate types and related parameters (such as high-frequency keywords appearing in speech recommendation), specific conversation tags corresponding to the candidate type can be found, and specific conversation tags corresponding to the specific speech tags can also be found. The specific slot in the template corresponding to the dialogue label, so as to update the dialogue material database generated based on the historical dialogue data and the corresponding dialogue label and slot for this type of candidate.

Existing speech recommendation schemes usually use past dialogue data or even artificially constructed dialogue templates to determine recommended speech templates, and the recommended speech templates are not updated after generation, so it is impossible to use real-time dialogue content in each dialogue node Select the most matching recommended speech template, and the database of recommended speech templates cannot be automatically updated based on the closed loop.

In contrast, the scheme for dialogue based on recommended speech skills proposed in this application can combine the prior information associated with the dialogue participants and dialogue scenes, as well as the real-time dialogue data of the current dialogue, to obtain the real-time and accurate information of human dialogue participants in a timely and accurate manner. Dialogue requirements and feedback information, select a more appropriate recommended speech template to ask or reply to the dialogue participants, quickly obtain the information required for the dialogue task from the dialogue, and continue to maintain communication with humans until the dialogue task is completed Good communication between the dialogue participants. On the basis of instantly and accurately obtaining and responding to the interlocutor's needs, the solution of this application can also periodically return real-time dialogue data to update historical dialogue data according to the dialogue effect of the recommended speech template, and iteratively update and improve the recommended speech template database. , so as to obtain a better intelligent dialogue effect.

Fig. 3 shows a device 300 for dialogue based on recommended utterances according to an embodiment of the present application. The device 300 may include an interaction unit 310 and a recommendation unit 320 .

The interaction unit 310 is used to obtain the real-time dialogue data of the current dialogue between the intelligent dialogue system (such as a dialogue robot) and the human dialogue participant 301 such as a candidate, and output the dialogue content generated according to the recommended speech algorithm to the dialogue participant 301 . The interaction unit 310 can also convert the dialogue content in the text data format provided by the dialogue robot to the dialogue participant 301 into audio/video data audible or visible to the dialogue participant 301, and convert the audio (such as voice) from the dialogue participant 301 ) or video (eg, real-time video chat) formatted data into text data that can be processed by the recommendation unit 320 of the device 300. The interaction unit 310 may also include a display unit or an input unit that facilitates dialog exchanges with the human dialog participants 301, wherein the display unit is used to provide the dialog participants 301 with information or prompts related to the real-time dialog, and the input unit may be used for dialog participation. Or 301 provides other information input channels besides voice or video. The interaction unit 310 may also provide an interaction interface in the form of a web page or App.

The recommendation unit 320 is configured to select the current dialogue topic from the default recommended speech template database based on the real-time dialogue data and the prior information of the dialogue participants 301 for the current dialogue topic among one or more dialogue topics included in the current dialogue. The most matching (first) recommended speech template; Generate the dialogue content based on the selected recommended speech template; Obtain the reply of the dialogue participant 301 for the generated dialogue content and update the real-time dialogue data based on the reply to generate the following A dialogue content. For example, when the reply of the conversation participant 301 is a non-negative reply, it is confirmed that the generated conversation content is accurate and appropriate. While extracting the information in the real-time dialogue data contained in the reply, it is judged whether the information required by the current dialogue topic has been completely obtained, and based on this judgment, it is decided whether to continue to communicate with the dialogue participants in the dialogue node of the current dialogue topic 301 asks or answers the question raised by it, or enters the dialogue node corresponding to the next dialogue topic. If a negative reply is obtained from the conversation participant 301, it can be determined that the conversation content generated just now is inaccurate or the conversation participant 301 is not interested in the real-time conversation content of the current conversation. Correspondingly, based on the negative information or data contained in the negative reply, the (second) recommended speech template is selected to generate more accurate and appropriate dialogue content. The recommendation unit 320 can also complete the function of generating a default recommended speech template database based on historical dialogue data as described above in conjunction with FIG. 1 , and complete the functions of other steps and details of generating real-time dialogue content described in conjunction with FIG. 2 .

It should be noted that although several modules or units of the device and system for dialogue based on recommended utterances are mentioned in the above detailed description, this division is not mandatory. Actually, according to the embodiment of the present application, the features and functions of two or more modules or units described above may be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided to be embodied by a plurality of modules or units. Components shown as modules or units may or may not be physical units, may be located in one place, or may be distributed over multiple network elements. Part or all of the modules can be selected according to actual needs to realize the purpose of the solution of the present application. It can be understood and implemented by those skilled in the art without creative effort.

In an exemplary embodiment of the present application, there is also provided a computer-readable storage medium, on which a computer program is stored, the program includes executable instructions, and when the executable instructions are executed by, for example, a processor, any one of the above-mentioned The steps of the method for conducting dialogue based on the recommended words described in the embodiments. In some possible implementation manners, various aspects of the present application can also be implemented in the form of a program product, which includes program code, and when the program product is run on a terminal device, the program code is used to make the The terminal device executes the steps according to various exemplary embodiments of the present application described in the method for conducting a dialogue based on recommended speech in this specification.

The program product for implementing the above method according to the embodiment of the present application may adopt a portable compact disk read-only memory (CD-ROM) and include program codes, and may run on a terminal device such as a personal computer. However, the program product of the present application is not limited thereto. In this document, a readable storage medium may be any tangible medium containing or storing a program, and the program may be used by or in combination with an instruction execution system, device, or device.

The program product may reside on any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: electrical connection with one or more conductors, portable disk, hard disk, random access memory (RAM), read only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

The computer readable storage medium may include a data signal carrying readable program code in baseband or as part of a carrier wave traveling as a data signal. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A readable storage medium may also be any readable medium other than a readable storage medium that can send, propagate or transport a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the readable storage medium may be transmitted by any suitable medium, including but not limited to wireless, cable, optical cable, RF, etc., or any suitable combination of the above.

Program codes for performing the operations of the present application may be written in any combination of one or more programming languages, including object-oriented programming languages—such as Java, C++, etc., as well as conventional procedural programming languages. Programming language - such as "C" or a similar programming language. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server to execute. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., using an Internet service provider). business to connect via the Internet).

In an exemplary embodiment of the present application, there is also provided an electronic device, which may include a processor, and a memory for storing executable instructions of the processor. Wherein, the processor is configured to execute the steps of the method for conducting a dialogue based on recommended words in any one of the above embodiments by executing the executable instructions.

Those skilled in the art can understand that various aspects of the present application can be implemented as a system, method or program product. Therefore, various aspects of the present application can be specifically implemented in the following forms, that is: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which can be collectively referred to herein as "circuit", "module" or "system".

An electronic device 400 according to this embodiment of the present application is described below with reference to FIG. 4 . The electronic device 400 shown in FIG. 4 is only an example, and should not limit the functions and scope of use of the embodiments of the present application.

As shown in FIG. 4, electronic device 400 takes the form of a general-purpose computing device. Components of the electronic device 400 may include, but are not limited to: at least one processing unit 410, at least one storage unit 420, a bus 430 connecting different system components (including the storage unit 420 and the processing unit 410), a display unit 440, and the like.

Wherein, the storage unit stores program codes, and the program codes can be executed by the processing unit 410, so that the processing unit 410 executes the method according to the present application described in the method for dialogue based on the recommended words in this specification. Steps of Various Exemplary Embodiments. For example, the processing unit 410 may execute the steps shown in FIG. 1 and FIG. 2 .

The storage unit 420 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 4201 and/or a cache storage unit 4202 , and may further include a read-only storage unit (ROM) 4203 .

The storage unit 420 may also include a program/utility 4204 having a set (at least one) of program modules 4205, such program modules 4205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include the realization of the network environment.

The bus 430 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local area using any of a variety of bus structures. bus.

The electronic device 400 can also communicate with one or more external devices 500 (such as keyboards, pointing devices, Bluetooth devices, etc.), and can also communicate with one or more devices that enable the user to interact with the electronic device 400, and/or communicate with Any device (eg, router, modem, etc.) that enables the electronic device 400 to communicate with one or more other computing devices. Such communication may occur through input/output (I/O) interface 450 . Moreover, the electronic device 400 can also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN) and/or a public network such as the Internet) through the network adapter 460 . The network adapter 460 can communicate with other modules of the electronic device 400 through the bus 430 . It should be appreciated that although not shown, other hardware and/or software modules may be used in conjunction with electronic device 400, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives And data backup storage system, etc.

Through the description of the above implementations, those skilled in the art can easily understand that the example implementations described here can be implemented by software, or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of the present application can be embodied in the form of software products, which can be stored in a non-volatile storage medium (which can be CD-ROM, U disk, mobile hard disk, etc.) or network In the above, several instructions are included to make a computing device (which may be a personal computer, a server, or a network device, etc.) execute the method for conducting a dialogue based on a recommended speech according to an embodiment of the present application.

Other embodiments of the present application will readily occur to those skilled in the art from consideration of the specification and practice of the disclosure herein. This application is intended to cover any modification, use or adaptation of the application, these modifications, uses or adaptations follow the general principles of the application and include common knowledge or conventional technical means in the technical field not disclosed in the application . The specification and examples are to be considered exemplary only, with a true scope and spirit of the application indicated by the appended claims.

Claims (17)

1. A method for dialogue based on recommended speech, comprising: For the current dialogue topic among the one or more dialogue topics involved in the real-time dialogue, based on the real-time dialogue data and the prior information of the dialogue participants, a first recommended utterance is selected from at least one recommended utterance template corresponding to the current dialogue topic a technical template, wherein the conversation topic is used to indicate the intent or purpose of the real-time conversation; generating dialogue content based on the first recommended speech template; obtaining a reply from the conversation participant to the generated conversation content, and updating the real-time conversation data based on the reply to generate the next conversation content, Wherein, the dialog topic and at least one recommended speech template corresponding to the dialog topic are determined based on historical dialog data of historical dialogs. 2. The method according to claim 1, wherein the first recommended speech template is selected from at least one recommended speech template corresponding to the current conversation topic based on the real-time dialog data and prior information of the dialog participants include: determining a dialog type of the real-time dialog based on the real-time dialog data, wherein the dialog type includes a question type and an answer type; extracting a real-time dialog tag related to the real-time dialog from the real-time dialog data and the prior information; The first recommended speech template corresponding to the dialog type is selected based on the real-time dialog tag. 3. The method according to claim 2, wherein the conversation topic is related to at least one conversation label, and extracting the real-time conversation label relevant to the real-time conversation from the real-time conversation data and the prior information comprises : Based on the conversation tags related to the current conversation topic, extracting the conversation tags related to the conversation content from the real-time conversation data to generate the conversation content tags; Based on the dialog tags related to the current dialog topic, extracting dialog tags related to the dialog participants from the prior information to generate dialog participant tags and/or dialog tags related to the dialog scene to generate dialog scene label; Combining the dialog content tag, the dialog participant tag and the dialog scene tag into the real-time dialog tag. 4. The method according to claim 3, wherein the recommended speech template includes at least one slot to be filled, the slot has a corresponding relationship with the dialog tag, and is selected based on the real-time dialog tag The first recommended speech template corresponding to the dialogue type includes: The first recommended speech template is determined based on a degree of matching between the real-time dialog tag and the slot having the corresponding relationship with the real-time dialog tag. 5. The method according to claim 4, wherein the recommended speech template with the highest matching degree is selected as the first recommended speech template. 6. The method according to claim 4, wherein generating dialogue content based on the first recommended speech template comprises: Filling the real-time dialogue tags corresponding to the slots of the first recommended speech template into the slots to generate the dialogue content. 7. The method according to claim 1, wherein updating the real-time dialogue data based on the reply to generate the next dialogue content comprises: In response to the fact that the reply includes a negative reply, selecting from the at least one recommended speech template corresponding to the current conversation topic based on the real-time conversation label related to the real-time conversation corresponds to the conversation type of the real-time conversation The second recommended speech template for ; and generating the next dialogue content based on the second recommended speech template, Wherein the second recommended speech template has a lower matching degree than the first recommended speech template. 8. The method of claim 7, wherein the conversation topic is associated with at least one conversation label, and the real-time conversation label is derived from the real-time conversation data and Extracted from the prior information, the recommended speech template includes at least one slot to be filled, the slot has a corresponding relationship with the dialogue label, Selecting the second recommended speech template corresponding to the dialogue type of the real-time dialogue includes: Based on the matching degree of the real-time dialog tag and the slot having the corresponding relationship with the real-time dialog tag, determine the second recommended speech template, wherein the second recommended speech template has a higher ratio than the The matching degree of the slot of the first recommended speech template is lower; and Generating the next dialogue content based on the second recommended speech template includes: Filling the real-time dialog tags corresponding to the slots of the second recommended speech template into the slots to generate the next dialog content. 9. The method according to any one of claims 1 to 8, characterized in that the method further comprises: for the current conversation topic, the information extracted in the reply meets the requirements of the current conversation topic In the case of , the dialogue content is generated for the next dialogue topic. 10. The method according to any one of claims 1 to 8, wherein the historical dialog data includes question type historical dialog data and answer type historical dialog data, and the dialog topic corresponds to the dialog topic At least one of the recommended speech templates is determined in the following manner: Extracting the dialogue topic, the dialogue label related to the dialogue theme and the historical dialogue data related to the dialogue theme from the historical dialogue data; determining a content item related to the dialog tag in the historical dialog data as the slot, and determining a correspondence between the slot and the dialog tag; The content items are removed from the historical dialogue data to generate recommended speech templates of question type and answer type related to the dialogue topic. 11. The method according to claim 10, characterized in that the conversation topic and the conversation label are determined by clustering and/or keyword retrieval. 12. The method according to any one of claims 1 to 8, characterized in that, using the real-time dialogue data and/or manual dialogue data for which the reply is a non-negative reply to update the historical dialogue data to update the A dialog topic and at least one recommended speech template corresponding to the dialog topic. 13. The method according to claim 1, wherein the dialogue scene includes a recruitment scene, the dialogue participants include candidates, and the prior information includes resume information and/or recruitment information of the candidates . 14. The method according to claim 1, wherein the dialogue scene includes a customer service scene, the dialogue participants include users, and the prior information includes maintenance information and/or complaint information submitted by the user . 15. A device for dialogue based on recommended speech, comprising: an interaction unit configured to acquire real-time dialogue data and output dialogue content to dialogue participants; The recommendation unit is configured to, for a current conversation topic among one or more conversation topics involved in the real-time conversation, select from at least one corresponding to the current conversation topic based on the real-time conversation data and the prior information of the conversation participants. Selecting a first recommended speech template from among the recommended speech templates, wherein the conversation topic is the intention or purpose of the real-time conversation; generating the conversation content based on the first recommended speech template; and acquiring the conversation participants replying to the generated dialog content, and updating the real-time dialog data based on the reply to generate the next dialog content, Wherein, the dialog topic and at least one recommended speech template corresponding to the dialog topic are determined based on historical dialog data of historical dialogs. 16. A computer-readable storage medium, on which a computer program is stored, the computer program comprising executable instructions, when the executable instructions are executed by a processor, the implementation according to any one of claims 1 to 14 method. 17. An electronic device, characterized in that it comprises: processor; and a memory for storing executable instructions of the processor; Wherein, the processor is configured to execute the executable instructions to implement the method according to any one of claims 1-14.

Images