Information processing method and device

Abstract

The embodiments of this specification provide an information processing method and an apparatus, wherein the information processing method comprises: obtaining business problem information associated with a target business; encoding the business problem information to obtain a coding vector, and determining a label vector of a state label in a state label set associated with the target business; calculating the similarity between the coding vector and the label vector, and filtering a target state label in the state label set according to the similarity; and extracting business reply information in a label tree associated with the target business according to the target state label as a response to the business problem information.

Description

Technical Field

[0001] The embodiments of the present specification relate to the field of computer technology, and more particularly to information processing methods and devices. Background Art

[0002] With the development of Internet technology, intelligent question-answering technology has been applied in more and more scenarios, such as customer service question-answering services, legal question-answering services, shopping question-answering services, etc., which can provide users with more reasonable answers through semantic analysis after users ask questions. In the prior art, when users are fed back answers through intelligent question-answering technology, although some users' problems can be solved, it only supports a single round of dialogue, and in most scenarios, the question is searched by keyword matching, and the answers to the questions are also preset in the question-answering library. The question-answering link is too single, and it is difficult to feedback more favorable content to users in a single round of dialogue. In addition, there are certain requirements for the accuracy of the questions asked by users, which makes the user experience extremely low; therefore, an effective solution is urgently needed to solve the above problems. Summary of the invention

[0003] In view of this, an embodiment of this specification provides an information processing method. One or more embodiments of this specification also relate to an information processing device, a computing device, a computer-readable storage medium and a computer program to solve the technical defects existing in the prior art.

[0004] According to a first aspect of an embodiment of this specification, there is provided an information processing method, including:

[0005] Obtain business problem information associated with the target business;

[0006] Encoding the service problem information to obtain a coding vector, and determining a label vector of a state label in a state label set associated with the target service;

[0007] Calculating the similarity between the encoding vector and the label vector, and filtering a target state label from a state label set according to the similarity;

[0008] Business answer information is extracted from the tag tree associated with the target business according to the target state tag as a response to the business question information.

[0009] According to a second aspect of an embodiment of this specification, there is provided an information processing device, including:

[0010] An acquisition module, configured to acquire business problem information associated with a target business;

[0011] A determination module is configured to encode the business problem information to obtain a coding vector, and determine a label vector of a state label in a state label set associated with the target business;

[0012] A calculation module, configured to calculate the similarity between the encoding vector and the label vector, and filter the target state label from the state label set according to the similarity;

[0013] The response module is configured to extract business answer information from the tag tree associated with the target business according to the target state tag as a response to the business question information.

[0014] According to a third aspect of an embodiment of this specification, a computing device is provided, including:

[0015] Memory and processor;

[0016] The memory is used to store computer-executable instructions, and the processor is used to implement any step of the above-mentioned information processing method when executing the computer-executable instructions.

[0017] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions, and when the instructions are executed by a processor, the steps of the above-mentioned information processing method are implemented.

[0018] According to a fifth aspect of the embodiments of this specification, a computer program is provided, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the above-mentioned information processing method.

[0019] The information processing method provided in this specification, in order to quickly and accurately feedback accurate answers to users, can first encode the business question information after obtaining the business question information submitted by the user participating in the target business, so as to obtain its corresponding encoding vector, and at the same time determine the label vector of the state label associated with the target business; secondly, calculate the similarity between the encoding vector and the label vector, so as to filter out the target state label with a high correlation with the business question information in the state label set according to the similarity calculation result, and finally extract the business reply information from the label tree associated with the target business according to the target state label as a response to the business question information. The accuracy of the feedback business reply information is effectively improved, and the business reply information is determined by querying the label tree, which can further enrich the diversity of replies, so that in the subsequent interaction process, the user can be guided to make accurate choices to obtain the required answers to business questions and improve the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of an information processing method provided by an embodiment of this specification;

[0021] Figure 2 is a schematic diagram of a label tree in an information processing method provided by an embodiment of this specification;

[0022] Figure 3 is a schematic diagram of an interactive interface in an information processing method provided by an embodiment of this specification;

[0023] Figure 4 is a schematic diagram of another tag tree in an information processing method provided by an embodiment of this specification;

[0024] Figure 5 is a schematic diagram of similarity calculation in an information processing method provided by an embodiment of this specification;

[0025] Figure 6 is a schematic diagram of multiple rounds of interaction in an information processing method provided by an embodiment of this specification;

[0026] Figure 7 is a processing flow chart of an information processing method provided by an embodiment of this specification;

[0027] Figure 8 is a structural diagram of an information processing device provided by an embodiment of this specification;

[0028] Fig. 9 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0029] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0030] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0031] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0032] In this specification, an information processing method is provided. This specification also relates to an information processing apparatus, a computing device, a computer-readable storage medium and a computer program, which are described in detail one by one in the following embodiments.

[0033] Figure 1 A flowchart of an information processing method provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0034] Step S102: Acquire business problem information associated with the target business.

[0035] Specifically, the target business refers to a business that provides consulting services to users, including but not limited to legal consulting services, shopping product consulting services, medical insurance consulting services, vehicle selection consulting services, etc. This embodiment will not be described in detail here. Correspondingly, the business question information specifically refers to the business question submitted by the user when using the target business for problem consultation; the business reply information related to the target business will be selected and replied according to the business question information to solve the user's doubts.

[0036] For the convenience of description, this embodiment uses the target business of legal consulting business as an example to illustrate the information processing method. The information processing methods in other business scenarios can refer to the same or corresponding description content of this embodiment, and will not be elaborated here.

[0037] In actual applications, an accurate legal question usually contains a lot of legal terms. For ordinary people, it is difficult to clearly express the legal problems they encounter. In the existing technology, most legal consulting services only support single-round dialogue, and most of them use keyword matching to find problems. Users can only get answers if they describe their problems more accurately. This will greatly affect the accuracy and user experience.

[0038] The information processing method provided in this embodiment can, in order to quickly and accurately feedback accurate answers to users, first encode the business question information after obtaining the business question information submitted by the user participating in the target business, so as to obtain its corresponding encoding vector, and at the same time determine the label vector of the state label associated with the target business; secondly, calculate the similarity between the encoding vector and the label vector, so as to filter out the target state label with a high correlation with the business question information in the state label set according to the similarity calculation result, and finally extract the business reply information from the label tree associated with the target business according to the target state label as a response to the business question information. The accuracy of the feedback business reply information is effectively improved, and the business reply information is determined by querying the label tree, which can further enrich the diversity of replies, so that in the subsequent interaction process, the user can be guided to make accurate choices to obtain the required answers to business questions, thereby improving the user experience.

[0039] Step S104: Encode the service problem information to obtain a coding vector, and determine a label vector of a state label in a state label set associated with the target service.

[0040] Specifically, after obtaining the business problem information associated with the target business as mentioned above, in order to be able to provide users with relevant answers that meet their consulting needs, the business problem information can be encoded first to obtain an encoding vector corresponding to the business problem information; at the same time, it is also necessary to determine the label vector corresponding to each state label in the state label set associated with the target business, so as to facilitate the subsequent screening of labels with a higher degree of correlation with the business problem information from the state label set as the target state labels by calculating the vector similarity.

[0041] Among them, the encoding vector specifically refers to the vector expression obtained after encoding the business problem information; correspondingly, the state label set specifically refers to the set of all state labels that integrate the associated target business, and the state label is used to associate questions and answers, which can be understood as entities whose state labels are questions and answers; for example, in the legal consulting business, case agency, drafting contracts, tort litigation, etc. can all be understood as status labels, and through this label, the corresponding information can be selected in the label tree as the business response information of the business problem information. Different status labels have different information in the label tree, which is used to answer different business questions of users. It should be noted that the status label can be set according to actual needs in different business scenarios, and this embodiment does not make any limitation here.

[0042] Furthermore, after obtaining the business question information, considering that users have different needs for answers to the business question information, there may be cases where users chat with each other. In this process, in order to improve the user experience and popularize knowledge for users, targeted answers can be given by identifying the type of business question information. In this embodiment, the specific implementation method is as follows:

[0043] The business problem information is input into the problem detection model for processing to obtain problem detection information; when the problem detection information is of a non-business problem type, non-business answer information is extracted from a preset question and answer library according to the business problem information as a response to the business problem information; when the problem detection information is of a business problem type, execute step S104.

[0044] Specifically, the problem detection model specifically refers to a classification model that can detect whether the business problem information is a problem with a high degree of relevance to the target business. Accordingly, the problem detection information specifically refers to the information corresponding to the business problem information type, which includes non-business problem types and business problem types, wherein the business problem type is used to characterize that the business problem information is a problem related to the consulting related target business, such as in the legal consulting business, the problem the user consults is a problem related to the law; the non-business problem type is used to characterize that the business problem information is a problem related to the consulting unrelated target business, such as in the legal consulting business, the problem the user consults is a problem unrelated to the target business. Accordingly, the question-and-answer database specifically refers to a database that can respond to the user's non-business problem type questions, which stores sentences commonly used in daily interactions, and / or sentences that can popularize knowledge to users. Accordingly, the non-business reply information specifically refers to the reply information for the business problem information feedback when the business problem information is a non-business problem type.

[0045] Based on this, after obtaining the business question information, in order to be able to feed back the required answer information to the user, the business question information can be first input into a question detection model with classification capabilities for processing to obtain question detection information; when the question detection information is a business question type, it means that the business question information submitted by the user is a question related to the target business, and it needs to be answered from a professional perspective. At this time, step S104 can be executed to subsequently screen out the correct answer information and feed back to the user. When the question detection information is a non-business question type, it means that the business question information submitted by the user is a question not related to the target business, that is, the user has the possibility of chatting. Therefore, in order to improve the user experience, non-business answer information can be extracted from the preset question and answer library as a response to the business question information, avoiding discouraging the user's enthusiasm while improving the user's participation experience.

[0046] In actual applications, when extracting non-business response information from the question and answer database, in order to ensure that the content of the response in this case also has a certain relevance, the keywords of the business question information can be extracted first, and then sentences with higher relevance can be extracted from the question and answer database based on the keywords to respond, so as to meet the user's intelligent question and answer needs.

[0047] For example, the business question obtained from user input is "Hello, can I ask you a question?"; at this time, the business question can be input into the question detection model for processing, so as to determine whether the business question is a non-legal question based on the processing result. At this time, the keyword "question" of the business question can be extracted, and the answer content "Please speak" with a high match with the keyword can be extracted from the preset legal question and answer library based on the keyword, and it can be fed back to the user as a response to the business question.

[0048] In addition, when adding sentences to the preset question and answer library, in order to improve the user's knowledge in the corresponding field, sentences that can popularize knowledge can be added to the question and answer library, so that knowledge content can be provided to users in the chat mode for their learning.

[0049] In summary, by using the question detection model to determine the type of business question information and selecting the appropriate mode according to the type to provide corresponding business services to users, the user participation experience can be further improved. At the same time, accurate responses can be fed back to users' business question information to meet users' intelligent question and answer needs.

[0050] Furthermore, since the target business can provide multiple sub-business services to the user, after the user submits the business problem information, it is necessary to determine whether the user is consulting the business problem or selecting a sub-business service, so as to make an accurate decision in response to the user's business problem information. In this embodiment, the specific implementation method is as follows:

[0051] By performing intent recognition on the business problem information, determining problem intent information corresponding to the business problem information;

[0052] Determine whether the question intent information is business question and answer intent information; if not, determine the sub-target business included in the target business according to the question intent information, and use the sub-business information of the sub-target business as a response to the business question information; if so, execute step S104.

[0053] Specifically, intent recognition refers to the operation of identifying and processing the intent of business question information, which can be achieved through deep learning; correspondingly, question intent information refers to the relevant information that characterizes the intent of business question information; correspondingly, business question and answer intent information refers to the intent information that business question information belongs to the associated target business; correspondingly, sub-target business refers to the business service that the user needs to use at the current moment, which belongs to the branch business under the target business. For example, in the legal consulting business, the sub-target business can be a manual service transfer business, a lawyer search business, etc., which can be set according to the type of target business, and this embodiment does not make any restrictions here.

[0054] Based on this, in order to accurately provide feedback to users on the business services they need, after obtaining the business problem information, the business problem information can be identified for intent, so as to determine the problem intent message corresponding to the business problem information based on the identification result; and determine whether it is business question and answer intent information. If so, it means that the user needs to conduct business problem information at the current moment, and then execute step S104. If not, it means that the user needs to use other sub-target services under the target service at the current moment. At this time, the sub-target service that the user needs to use can be determined based on the problem intent information, and the sub-business information of the sub-target service can be used as a response to the business problem information. Among them, the sub-business information specifically refers to information describing the sub-target service, or the click control corresponding to the sub-target service, etc., and this embodiment does not make any limitation here. In addition, when feeding back the sub-business information to the user, it is also possible to jump directly to the sub-target service and provide the user with the corresponding sub-target service.

[0055] For example, the business question input by the user is "I want to find a lawyer". At this time, the business question is identified through the intent recognition model, the question intent information is determined, and it is determined that the question intent information is not business question intent information. At this time, the lawyer-finding business in the legal consulting business can be determined according to the question intent information, and the corresponding control selection is provided to the user. After the user clicks the control, the search for a lawyer business service interface will be jumped to enable the user to use the lawyer hiring service under this interface.

[0056] In summary, by identifying the user's current service needs, we can provide different feedback to the user based on their different needs, thereby improving the user's participation experience.

[0057] In addition, considering that the user may have gone through multiple rounds of information processes before this consultation, in order to ensure the continuity of the consultation, the historical status tag corresponding to the user can be determined first, so as to perform subsequent processing operations. In this embodiment, the specific implementation method is as follows:

[0058] Determine the identity information of the user who uploaded the business problem information; determine whether the user has a historical status tag based on the identity information; if so, extract the target business reply information from the tag tree associated with the target business based on the historical status tag as a response to the business problem information; if not, execute step S104.

[0059] Specifically, the identity information refers to the identity information corresponding to the user who uploaded the business problem information, which is used to find out whether there is a historical status tag in the historical consultation record; accordingly, the historical status tag refers to the status tag that matches the user within the historical consultation time; accordingly, the target business reply information refers to the business reply information queried according to the historical status tag.

[0060] Based on this, after determining the business problem information, considering that the user may have consulted related business problems, in order to ensure the continuity of consultation, the identity information of the user who uploaded the business problem information can be determined first, and whether there is a historical status tag associated with the user can be queried based on the identity information; if not, it means that the business problem information obtained this time is the first question submitted by the user, and step S104 can be executed. If yes, it means that the user has already consulted the business before this query, so in order to ensure the continuity of consultation, the target business reply information can be extracted from the tag tree associated with the target business based on the historical status tag as a response to the business problem information.

[0061] In summary, by querying whether the user has a historical status tag, the user's current stage can be determined, so that different information processing strategies can be provided to the user according to different stages, so that different business response information can be fed back to the user according to different information processing strategies, further improving the user's participation experience.

[0062] On this basis, considering that the time when the user consults again may be quite different from the last time, if the reply information is determined based on the previous historical status tag, the reply may be inaccurate. Therefore, it is possible to determine whether the historical status tag needs to be used as the reply basis by means of time limitation. In this embodiment, the specific implementation method is as follows:

[0063] Determine whether the user has a historical status tag within a preset time period based on the identity information; if so, execute the step of extracting target business reply information from the tag tree associated with the target business based on the historical status tag as a response to the business question information; if not, execute step S104.

[0064] For example, the legal consultation question entered by the user is "I want to consult about contract drafting". In order to provide the user with the content he needs, we can first check whether the user has a historical status tag, that is, based on the user's identity information, check whether the user has consulted other questions related to the question "I want to consult about contract drafting" within 1 month. If the test results determine that the user has a historical status tag "contract terms", then based on the historical status tag "contract terms", Figure 2 Relevant business response information is extracted from the tag tree shown and fed back to the user to realize continuous provision of legal consulting services to the user. If it is determined that the user has no historical status tags, the new status tags can be matched at this time to feed back the consulting content to the user.

[0065] In summary, by combining the time period to determine the historical status label, it is possible to decide whether to feedback continuous business answer information to the user within the specified time limit, thereby effectively improving the continuity of consultation and meeting the user's consultation needs.

[0066] Furthermore, in the case where it is determined that the user has a corresponding historical status tag based on the user's identity information, considering that the tag tree is a finite node tree structure, if the historical status tag does not have a child node in the tag tree, it is impossible to continue to feedback the corresponding business response information to the user. Therefore, it is also necessary to determine the feedback content of the business response information in combination with the child node of the historical status tag. In this embodiment, the specific implementation method is as follows:

[0067] Determine whether the historical status tag has a child node in the tag tree; if so, extract the target business reply information in the tag tree according to the status tag corresponding to the child node as a response to the business question information; if not, execute step S104.

[0068] Specifically, when it is determined that the user who uploaded the business problem information has a historical status tag associated with it, it means that when feeding back the business reply information to the user, it is sufficient to extract the information in the tag tree according to the historical status tag for feedback. In this process, considering that the historical status tag may be the terminal child node of the tag tree, it is not connected to any child node. At this time, it is impossible to determine the business reply information associated with the historical status tag based on the tag tree.

[0069] Therefore, after determining that the user has an associated historical status tag, it can be determined whether the historical status tag has a child node in the tag tree; if not, it means that it is impossible to extract the reply information from the tag tree for feedback based on the historical status tag, so a new tag can be reselected for subsequent reply based on the business question information, that is, step S104 is executed. If yes, it means that the historical status tag has a node connected to it in the tag tree, then at this time, according to the status tag corresponding to the child node, the relevant target business reply information is extracted from the tag tree and fed back to the user, so as to feedback the consultation result that the user needs to know.

[0070] In actual applications, when extracting target business response information from the tag tree based on the status tag corresponding to the subnode, the subnode corresponding to the historical status tag is first determined in the tag tree, and the position of each subnode in the tag tree is determined, and then the pre-stored information is selected from the position as the target business response information, and feedback is given to the user. The content contained in the target business response information can be a specific answer or multiple answers for the user to choose, so as to facilitate the subsequent guidance of the content that the user needs to know.

[0071] In summary, in order to be able to provide useful information to users in each round of response during the intelligent question-answering process, the child nodes of the label tree will be detected during the question-answering process to ensure that the corresponding answers can be fed back in combination with the business response information to meet the user's question-answering needs.

[0072] In addition, considering that the historical status label may be connected to multiple sub-nodes, and each sub-node corresponds to different reply content, in order to be able to provide users with reply information that meets their consultation needs, text similarity calculation can be used to complete it. In this embodiment, the specific implementation method is as follows:

[0073] In the case that there are multiple child nodes, determine the status label corresponding to each child node; calculate the text similarity between the business question information and the status label corresponding to each child node; select the status label corresponding to the target child node from the status labels corresponding to the multiple child nodes according to the text similarity, and extract the target business response information in the label tree according to the status label corresponding to the target child node.

[0074] Specifically, text similarity refers to the similarity between the business problem information and the status labels corresponding to each child node, which is used to describe the semantic similarity between texts. The higher the text similarity, the higher the similarity. Correspondingly, the target child node refers to the node corresponding to the label with the highest degree of relevance to the business problem information among multiple child nodes.

[0075] Based on this, when the historical status label is connected to multiple sub-nodes, in order to filter out the sub-nodes with a higher degree of associated business problem information from the multiple sub-nodes and feed back the business response information to them, the status label corresponding to each sub-node can be determined, and then the text similarity between the business problem information and each status label can be calculated to select the target sub-node and its corresponding status label from the multiple sub-nodes according to the text similarity, and finally the target business response information can be extracted from the label tree according to the status label corresponding to the target sub-node.

[0076] Furthermore, even if there are multiple sub-nodes, it is necessary to consider the similarity between each sub-node and the business question information so that more accurate questions can be fed back when answering. Therefore, it is necessary to determine whether there is a target sub-node. In this embodiment, the specific implementation method is as follows:

[0077] Compare each text similarity with a preset similarity threshold; determine based on the comparison result if there is a text similarity greater than the similarity threshold, then sort each text similarity, and select a target text similarity based on the sorting result; and use the child node corresponding to the target text similarity as the target child node.

[0078] That is to say, after calculating the text similarity between the business question information and the status labels of each child node, it is necessary to compare each text similarity with the preset similarity threshold. If the text similarity is greater than the similarity threshold, it means that the status labels of some child nodes are highly correlated with the business reply information, so that the business reply information can be selected more accurately; if the text similarity is not less than or equal to the similarity threshold, it means that there is no child node whose status label is highly correlated with the business reply information. If the status label of a child node is forcibly selected to determine the reply information, it may not meet the user's consultation needs. At this time, you can return to step S104 and re-select the label.

[0079] Furthermore, when it is determined that there is a text similarity greater than the similarity threshold, if there are many text similarities greater than the similarity threshold, then each text similarity can be sorted at this time, and the largest text similarity can be selected as the target text similarity according to the sorting result. Then, the child node corresponding to the target text similarity is used as the target child node, so as to facilitate the subsequent selection of the target business reply information according to the status label corresponding to the target child node for feedback to the user.

[0080] For an example, see Figure 3As shown in the schematic diagram, when the user's legal question is "how to draft a contract", the user's identity information is read to determine that the user has a historical status tag "contract". At this time, it is possible to detect whether "contract" has a child node, that is, to detect whether there is a child node of "contract" in the tag tree. According to the detection result, it is determined that there is a child node, such as Figure 2 As shown, they are "Contract Type", "Contract Effectiveness" and "Contract Composition".

[0081] Furthermore, the text similarity between the question and each node can be calculated, and based on the calculation results, the text similarity between the question and the "contract type" is determined to be S1, the text similarity between the question and the "contract effectiveness" is determined to be S2, and the text similarity between the question and the "contract composition" is determined to be S3, and the text similarity is compared with the preset similarity threshold S. According to the comparison results, if S2 and S3 are greater than the text similarity threshold S, then the label "contract type" corresponding to the text similarity S1 can be discarded. Then, the remaining "contract composition" and "contract effectiveness" are sorted according to the size of the text similarities S2 and S3, and if S3>S2, then "contract composition" is selected as the status label corresponding to the target child node, and then according to the "contract composition" in the following way: Figure 2 The information "Do you want to know what the contract is composed of?" is extracted from the tag tree shown as the reply information, and feedback is given to the user.

[0082] In summary, by determining the target business reply information based on the sub-nodes directly connected to the historical status tags, it can be ensured that when the reply information is fed back to the user, the information that meets the user's consultation needs can be extracted from the tag tree for feedback, so as to improve the user's participation experience and enable the user to get the answers they need.

[0083] Step S106, calculating the similarity between the encoding vector and the label vector, and filtering the target state label in the state label set according to the similarity.

[0084] Specifically, after determining the encoding vector of the business problem information and the label vector of the business problem information as mentioned above, further, in order to be able to select information with a higher correlation with the business problem information from the label tree for reply, the vector similarity can be calculated to determine the state label with a higher correlation with the business problem information, and select the target state label for subsequent extraction of business reply information in the label tree.

[0085] Among them, the similarity specifically refers to the similarity between vectors, which is used to characterize the similarity between business problem information and each state label; correspondingly, the target state label specifically refers to the state label corresponding to the label vector with the highest similarity with the encoding vector among multiple state labels.

[0086] Furthermore, the calculating the similarity between the encoding vector and the label vector, and filtering the target state label in the state label set according to the similarity, includes: calculating the similarity between the encoding vector and the label vector of each state label through a vector distance function; filtering the target state label in the state label set according to the maximum similarity.

[0087] Specifically, the vector distance function refers to a function that calculates the distance between vectors. The distance can reflect the similarity between business problem information and state labels. The smaller the distance value, the higher the similarity. Based on this, the distance between the encoding vector and the label vector is calculated by the vector distance function to map the similarity. Then, the target state label can be selected from the state label set according to the maximum similarity for subsequent use.

[0088] In practical applications, considering the different complexity of problems in different consulting business scenarios, such as legal consulting business, see Figure 4 In the schematic diagram shown, under the contract node, there are nodes such as contract drafting, contract type, contract entry into force, and contract terms, and under the contract type, there are nodes such as sales contract, gift contract, loan contract, lease contract, engineering contract, technical contract, and entrustment contract, which are relatively complex. Because this embodiment is convenient for explaining the information processing method clearly, only some legal issues are discussed as examples. The actual problems in actual applications can be set in response to the actual application scenarios, and this embodiment does not make any restrictions here.

[0089] For an example, see Figure 5 As shown in the schematic diagram, when the user enters the legal question as "I want to donate supplies", the legal question can be encoded through the question encoding to obtain an m-dimensional encoding vector a1, and at the same time, the n labels {donation; donation / contract; donation / contract / contract terms; donation / contract / contract effective; donation / contract / contract invalid...} are determined. After each label is processed by the label encoder, the n*m ​​matrix {b1, b2, b3,...} is obtained. Then, the phase velocity between the encoding vector a1 and the label vector is calculated by the vector distance function, so as to select the label closest to the question from the n labels according to the similarity calculation result, that is, the target state label. This is used to determine the reply information later.

[0090] In summary, by selecting the target state label from multiple labels by calculating vector similarity, it can be ensured that the similarity between it and the business question information is higher. Based on this, the reply information can be determined, and the content that meets their consulting needs can be fed back to the user, thereby guiding the user to obtain relevant business information.

[0091] Step S108: extracting business answer information from the tag tree associated with the target business according to the target state tag as a response to the business question information.

[0092] Specifically, after determining the target state tag, further, the relevant business response information can be extracted from the tag tree to provide feedback to the user, that is, as a response to the business question information. It should be noted that the business response information can be a question or an answer. The question is used to continue to guide the user to explain the consultation content, and the answer is used to respond to the user's clear questions.

[0093] Furthermore, extracting business response information from the tag tree associated with the target business according to the target state tag includes: determining the tag type of the target state tag; when the tag type is a guidance type, selecting a guidance question in the tag tree as the business response information; when the tag type is an answer type, selecting a target answer in the tag tree as the business response information.

[0094] Specifically, the guidance type refers to the type that guides the user to continue asking questions and answering them. According to the guidance type, it is necessary to extract the guidance questions in the tag tree as business response information; correspondingly, the answer type refers to the type that requires direct feedback of answers to users. According to the answer type, the target answer can be extracted from the tag tree as business response information.

[0095] Based on this, after determining the target status label, in order to be able to feedback the answer the user needs, the label type can be determined first. If the label type is a guiding type, it means that the answer the user needs cannot be clearly identified at present. In order to further guide the user to clarify the question and make an answer, the guiding question can be selected in the label tree as the business response information; if the label type is an answer type, it means that a clear answer can be directly fed back to the user at present, and the target answer can be selected in the label tree as the business response information.

[0096] For example, when the target status label is determined to be "contract", and it is determined that "contract" belongs to the guidance type, then you can select the guidance question "Do you want to know about the contract?" corresponding to "contract" in the label tree as a response to the business question information and provide feedback to the user.

[0097] When it is determined that the target status label is "Contract Type", and it is determined that "Contract Type" belongs to the answer type, then you can select "What contract types are included?" corresponding to "Contract Type" in the label tree as the answer to the business question information and provide feedback to the user.

[0098] In summary, by using the target status label to determine the type and selecting the business reply information, it can be ensured that the reply information fed back to the user satisfies a type of guidance or answer, and a reply is given based on this, thereby improving the user experience.

[0099] See also Figure 6 As shown in the schematic diagram, in the legal consultation scenario, after obtaining the business problem information, chat detection and intent recognition will be performed. If it is determined to be chat, interaction will be carried out through the question and answer library. If it is determined that the user needs a sub-service based on the intent recognition result, legal services, lawyer services, manual customer service and lawyer consultation can be provided to the user; if the intent recognition determines that the consultation question is a consulting question, an intelligent reply will be made, and the interaction will be completed through multiple rounds of interaction, and it will be determined whether there is a historical status label. If so, query tracking will be performed. If the tracking is successful, the next level status or answer will be fed back to the user. Otherwise, the business question is query encoded, and the target status label is obtained through vector query to feedback the next level status or answer based on the label.

[0100] The information processing method provided in this specification, in order to quickly and accurately feedback accurate answers to users, can first encode the business question information after obtaining the business question information submitted by the user participating in the target business, so as to obtain its corresponding encoding vector, and at the same time determine the label vector of the state label associated with the target business; secondly, calculate the similarity between the encoding vector and the label vector, so as to filter out the target state label with a high correlation with the business question information in the state label set according to the similarity calculation result, and finally extract the business reply information from the label tree associated with the target business according to the target state label as a response to the business question information. The accuracy of the feedback business reply information is effectively improved, and the business reply information is determined by querying the label tree, which can further enrich the diversity of replies, so that in the subsequent interaction process, the user can be guided to make accurate choices to obtain the required answers to business questions and improve the user experience.

[0101] The following combination Figure 7 , taking the application of the information processing method provided in this specification in a shopping scenario as an example, the information processing method is further described. Figure 7 A processing flow chart of an information processing method provided by an embodiment of the present specification is shown, which specifically includes the following steps.

[0102] Step S702: Acquire business problem information associated with the target business.

[0103] Step S704: input the business problem information into a problem detection model for processing to obtain problem detection information.

[0104] Step S706, determine whether the problem detection information is a business problem type, if not, execute step S708; if yes, execute step S710.

[0105] Step S708: extracting non-business answer information from a preset question and answer database according to the business question information as a response to the business question information.

[0106] Step S710, determining problem intent information corresponding to the business problem information by performing intent recognition on the business problem information.

[0107] Step S712, determine whether the question intention information is business question and answer intention information, if not, execute step S714; if so, execute step S716.

[0108] Step S714: determine the sub-target business included in the target business according to the question intention information, and use the sub-business information of the sub-target business as a response to the business question information.

[0109] Step S716: determine the identity information of the user who uploaded the business problem information.

[0110] Step S718, determine whether the user has a historical status tag based on the identity information, if so, execute step S720; if not, execute step S724.

[0111] Step S720, determining whether the historical status tag has no child node in the tag tree, if not, executing step S722; if yes, executing step S724.

[0112] Step S722, calculate the text similarity between the business question information and the status label corresponding to each child node, and determine whether there is a text similarity greater than a preset similarity threshold; if not, execute step S724; if so, it means that there is a text similarity greater than the similarity threshold, then select the child node corresponding to the highest text similarity as the target child node; and extract the target business reply information in the label tree according to the status label corresponding to the target child node.

[0113] That is to say, when there are multiple child nodes, determine the status label corresponding to each child node; calculate the text similarity between the business question information and the status label corresponding to each child node; select the status label corresponding to the target child node from the status labels corresponding to the multiple child nodes based on the text similarity, and extract the target business reply information in the label tree based on the status label corresponding to the target child node.

[0114] Furthermore, the target sub-node can be determined by: comparing each text similarity with a preset similarity threshold; determining based on the comparison result that if there is a text similarity greater than the similarity threshold, then sorting each text similarity and selecting the target text similarity based on the sorting result; and using the sub-node corresponding to the target text similarity as the target sub-node.

[0115] Step S724: Encode the service problem information to obtain a coding vector, and determine a label vector of a state label in a state label set associated with the target service.

[0116] Step S726, calculating the similarity between the encoding vector and the tag vector of each state tag by using a vector distance function, and filtering the target state tag in the state tag set according to the maximum similarity.

[0117] Step S728: determine the tag type of the target state tag.

[0118] Step S730: When the tag type is a guidance type, select a guidance question in the tag tree as the business response information.

[0119] Step S732: When the tag type is an answer type, select a target answer in the tag tree as the business reply information.

[0120] In summary, in order to quickly and accurately feedback accurate answers to users, after obtaining the business question information submitted by users participating in the target business, the business question information can be first encoded to obtain its corresponding encoding vector, and the label vector of the state label associated with the target business can be determined; secondly, the similarity between the encoding vector and the label vector is calculated to filter out the target state label with a high correlation with the business question information in the state label set according to the similarity calculation result, and finally, the business reply information is extracted from the label tree associated with the target business according to the target state label as a response to the business question information. The accuracy of the feedback business reply information is effectively improved, and the business reply information is determined by querying the label tree, which can further enrich the diversity of replies, so that in the subsequent interaction process, users can be guided to make accurate choices to obtain the required answers to business questions and improve the user experience.

[0121] Corresponding to the above method embodiment, this specification also provides an information processing device embodiment, Figure 8 FIG. 1 shows a schematic diagram of the structure of an information processing device provided by an embodiment of the present specification. Figure 8 As shown, the device comprises:

[0122] An acquisition module 802 is configured to acquire business problem information associated with a target business;

[0123] The determination module 804 is configured to encode the service problem information to obtain an encoding vector, and determine a label vector of a state label in a state label set associated with the target service;

[0124] A calculation module 806 is configured to calculate the similarity between the encoding vector and the label vector, and filter the target state label from the state label set according to the similarity;

[0125] The response module 808 is configured to extract business answer information from the tag tree associated with the target business according to the target state tag as a response to the business question information.

[0126] In an optional embodiment, the information processing device further includes:

[0127] The question detection module is configured to input the business question information into the question detection model for processing to obtain question detection information; if the question detection information is of a non-business question type, extract non-business answer information from a preset question and answer library according to the business question information as a response to the business question information;

[0128] In the case that the problem detection information is of a business problem type, the determination module 804 is executed.

[0129] In an optional embodiment, the information processing device further includes:

[0130] The intent recognition module is configured to determine the problem intent information corresponding to the business problem information by performing intent recognition on the business problem information; determine whether the problem intent information is business question and answer intent information; if not, determine the sub-target business included in the target business according to the problem intent information, and use the sub-business information of the sub-target business as a response to the business problem information;

[0131] If yes, the determination module 804 is executed.

[0132] In an optional embodiment, the information processing device further includes:

[0133] A historical status tag detection module is configured to determine the identity information of the user who uploaded the business problem information; determine whether the user has a historical status tag according to the identity information; if so, extract the target business reply information from the tag tree associated with the target business according to the historical status tag as a response to the business problem information;

[0134] If not, the determination module 804 is executed.

[0135] In an optional embodiment, the information processing device further includes:

[0136] A child node detection module is configured to determine whether the historical status tag has a child node in the tag tree; if so, extract the target business reply information in the tag tree according to the status tag corresponding to the child node as a response to the business question information;

[0137] If not, the determination module 804 is executed.

[0138] In an optional embodiment, the subnode detection module is further configured to:

[0139] In the case that there are multiple child nodes, determine the status label corresponding to each child node; calculate the text similarity between the business question information and the status label corresponding to each child node; select the status label corresponding to the target child node from the status labels corresponding to the multiple child nodes according to the text similarity, and extract the target business response information in the label tree according to the status label corresponding to the target child node.

[0140] In an optional embodiment, the subnode detection module is further configured to:

[0141] Compare each text similarity with a preset similarity threshold; determine based on the comparison result if there is a text similarity greater than the similarity threshold, then sort each text similarity, and select a target text similarity based on the sorting result; and use the child node corresponding to the target text similarity as the target child node.

[0142] In an optional embodiment, the calculation module 806 is further configured to:

[0143] The similarity between the encoding vector and the label vector of each state label is calculated by a vector distance function; and the target state label is filtered in the state label set according to the maximum similarity.

[0144] In an optional embodiment, the response module 808 is further configured to:

[0145] Determine the label type of the target status label; when the label type is a guidance type, select a guidance question in the label tree as the business response information; when the label type is an answer type, select a target answer in the label tree as the business response information.

[0146] In an optional embodiment, the historical status tag detection module is further configured to:

[0147] Determine whether the user has a historical status tag within a preset time period according to the identity information; if so, perform the step of extracting target business reply information from the tag tree associated with the target business according to the historical status tag as a response to the business question information;

[0148] If not, the determination module 804 is executed.

[0149] The information processing device provided in this specification can, in order to quickly and accurately feedback accurate answers to users, first encode the business question information after obtaining the business question information submitted by the user participating in the target business, so as to obtain its corresponding encoding vector, and at the same time determine the label vector of the state label associated with the target business; secondly, calculate the similarity between the encoding vector and the label vector, so as to filter out the target state label with a high correlation with the business question information in the state label set according to the similarity calculation result, and finally extract the business reply information from the label tree associated with the target business according to the target state label as a response to the business question information. The accuracy of the feedback business reply information is effectively improved, and the business reply information is determined by querying the label tree, which can further enrich the diversity of replies, so that in the subsequent interaction process, the user can be guided to make accurate choices to obtain the required answers to business questions, thereby improving the user experience.

[0150] The above is a schematic scheme of an information processing device of this embodiment. It should be noted that the technical scheme of the information processing device and the technical scheme of the above information processing method belong to the same concept, and the details not described in detail in the technical scheme of the information processing device can be referred to the description of the technical scheme of the above information processing method.

[0151] Fig. 9 The structure block diagram of a computing device 900 provided according to an embodiment of the present specification is shown. The components of the computing device 900 include but are not limited to a memory 910 and a processor 920. The processor 920 is connected to the memory 910 via a bus 930, and a database 950 is used to store data.

[0152] The computing device 900 also includes an access device 940 that enables the computing device 900 to communicate via one or more networks 960. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 940 may include one or more of any type of network interface (e.g., a network interface card (NIC)) that is wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a World Wide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, a near field communication (NFC) interface, and the like.

[0153] In one embodiment of the present specification, the above components of the computing device 900 and Fig. 9 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Fig. 9 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0154] The computing device 900 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or PC. The computing device 900 may also be a mobile or stationary server.

[0155] The processor 920 is used to execute the following computer executable instructions, which implement the steps of the above-mentioned information processing method when executed by the processor.

[0156] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above information processing method belong to the same concept, and the details not described in detail in the technical scheme of the computing device can be referred to the description of the technical scheme of the above information processing method.

[0157] An embodiment of the present specification further provides a computer-readable storage medium storing computer-executable instructions, which can implement the steps of the above-mentioned information processing method when executed by a processor.

[0158] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the above information processing method belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the above information processing method.

[0159] An embodiment of the present specification further provides a computer program, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the above-mentioned information processing method.

[0160] The above is an illustrative solution of a computer program of this embodiment. It should be noted that the technical solution of the computer program and the technical solution of the above information processing method belong to the same concept, and the details not described in detail in the technical solution of the computer program can be referred to the description of the technical solution of the above information processing method.

[0161] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0162] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0163] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0164] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0165] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can well understand and use this specification. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. An information processing method, comprising: Obtain business problem information associated with the target business; Encoding the service problem information to obtain a coding vector, and determining a label vector of a state label in a state label set associated with the target service, wherein the state label is a label corresponding to the target service in different states; Calculating the similarity between the encoding vector and the label vector, and filtering a target state label from a state label set according to the similarity; Business answer information is extracted from a tag tree associated with the target business according to the target state tag as a response to the business question information, wherein the tag tree is a tree structure.

2. The method according to claim 1, before the step of encoding the business problem information to obtain the encoding vector is performed, further comprising: Inputting the business problem information into a problem detection model for processing to obtain problem detection information; In the case where the question detection information is of a non-business question type, extracting non-business answer information from a preset question-and-answer database according to the business question information as a response to the business question information; In the case where the problem detection information is of a business problem type, a step of encoding the business problem information to obtain a coding vector is performed.

3. The method according to claim 1, before the step of encoding the business problem information to obtain the encoding vector is performed, further comprising: By performing intent recognition on the business problem information, determining problem intent information corresponding to the business problem information; Determine whether the question intention information is business question and answer intention information; If not, determining the sub-target business included in the target business according to the question intention information, and using the sub-business information of the sub-target business as a response to the business question information; If so, execute the step of encoding the business problem information to obtain a coding vector.

4. The method according to claim 1, before the step of encoding the business problem information to obtain the encoding vector is performed, further comprising: Determine the identity information of the user who uploaded the business problem information; Determining whether the user has a historical status tag according to the identity information; If so, extracting target business answer information from the tag tree associated with the target business according to the historical status tag as a response to the business question information; If not, execute the step of encoding the business problem information to obtain a coding vector.

5. The method according to claim 4, further comprising: Determine whether the historical status tag has a child node in the tag tree; If so, extracting the target business reply information from the label tree according to the state label corresponding to the child node as a response to the business question information; If not, execute the step of encoding the business problem information to obtain a coding vector.

6. The method according to claim 5, wherein extracting the target service reply information from the label tree according to the status label corresponding to the child node comprises: When there are multiple child nodes, determine a state label corresponding to each child node; Calculating the text similarity between the business problem information and the state label corresponding to each child node; The status label corresponding to the target child node is selected from the status labels corresponding to the multiple child nodes according to the text similarity, and the target business reply information is extracted from the label tree according to the status label corresponding to the target child node.

7. The method according to claim 6, wherein the determination of the target child node comprises: Compare each text similarity with a preset similarity threshold; If a text similarity greater than the similarity threshold exists according to the comparison result, each text similarity is sorted, and a target text similarity is selected according to the sorting result; The sub-node corresponding to the target text similarity is used as the target sub-node.

8. The method according to claim 1, wherein calculating the similarity between the encoding vector and the label vector and filtering the target state label from the state label set according to the similarity comprises: Calculating the similarity between the encoding vector and the label vector of each state label by using a vector distance function; The target state tag is filtered in the state tag set according to the maximum similarity.

9. The method according to claim 1, wherein extracting the service response information from the label tree associated with the target service according to the target status label comprises: Determining a label type of the target state label; In the case where the tag type is a guidance type, selecting a guidance question in the tag tree as the service answer information; In the case where the tag type is an answer type, a target answer is selected in the tag tree as the business reply information.

10. The method according to claim 4, further comprising: Determining whether the user has a historical status tag within a preset time period according to the identity information; If so, executing the step of extracting target business answer information from the tag tree associated with the target business according to the historical status tag as a response to the business question information; If not, execute the step of encoding the business problem information to obtain a coding vector.

11. An information processing device, comprising: An acquisition module, configured to acquire business problem information associated with a target business; A determination module is configured to encode the business problem information to obtain a coding vector, and determine a label vector of a state label in a state label set associated with the target business, wherein the state label is a label corresponding to the target business in different states; A calculation module, configured to calculate the similarity between the encoding vector and the label vector, and filter the target state label from the state label set according to the similarity; The response module is configured to extract business answer information from a tag tree associated with the target business according to the target state tag as a response to the business question information, wherein the tag tree is a tree structure.

12. A computing device comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method according to any one of claims 1 to 10 are implemented.

13. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions, when executed by a processor, implement the steps of the method according to any one of claims 1 to 10.

Images