Interactive Information Processing Method, Apparatus, Device, and Storage Medium

Abstract

The present application provides an interactive information processing method, apparatus, device, and storage medium. The method includes: obtaining the interaction information between a customer service and a user, and forming a text file according to the interaction information; generating a plurality of key-value pairs according to the text file; the key-value pairs include questions and answers; determining a plurality of question-and-answer categories according to the plurality of key-value pairs, wherein at least one key-value pair among the plurality of key-value pairs corresponds to each question-and-answer category, and each key-value pair corresponds to one question-and-answer category; performing a deduplication process on each key-value pair under each question-and-answer category to obtain the remaining key-value pairs under each question-and-answer category; generating question-answering information according to the remaining key-value pairs under each question-and-answer category. This solution can generate question-answering information, so that the customer service can quickly answer the questions asked by the user with reference to the generated question-answering information, improving the efficiency of the customer service.

Description

Technical Field

[0001] This application relates to the field of Internet technologies, and in particular, to an interactive information processing method, apparatus, device, and storage medium. Background Art

[0002] With the continuous development of Internet technologies, major Internet e-commerce platforms have been widely developed and applied.

[0003] In the prior art, each Internet e-commerce platform provides an online customer service system for users. When a user wants to consult a question, a customer service representative, such as a human customer service or an intelligent customer service, can be assigned to the user through the online customer service system, and the assigned customer service representative can answer the user's question.

[0004] However, in the prior art, when a user asks a question to a human customer service or an intelligent customer service, each answer requires the human customer service or the intelligent customer service to organize language and then answer the user's question with the organized language. In this way, since each answer requires the customer service representative to organize language by themselves, the efficiency of the customer service is low. Summary of the Invention

[0005] This application provides an interactive information processing method, apparatus, device, and storage medium, which can generate question answer information. Thus, the customer service representative can quickly answer the user's question with reference to the generated question answer information, greatly improving the efficiency of the customer service.

[0006] In a first aspect, this application provides an interactive information processing method, including:

[0007] Obtain the interaction information between the customer service representative and the user, and form a text file according to the interaction information;

[0008] Generate a plurality of key-value pairs according to the text file; the key-value pairs include questions and answers;

[0009] Determine a plurality of question-answer categories according to the plurality of key-value pairs, where at least one key-value pair in the plurality of key-value pairs corresponds to each question-answer category, and each key-value pair corresponds to a question-answer category;

[0010] Perform a duplicate removal process on each key-value pair under each question-answer category to obtain the remaining key-value pairs under each question-answer category;

[0011] Generate question answer information according to the remaining key-value pairs under each question-answer category.

[0012] Further, determining a plurality of question-answer categories according to the plurality of key-value pairs includes:

[0013] Select a key-value pair from the multiple key-value pairs, determine the selected key-value pair as the initial centroid, and determine a question-and-answer category according to the selected key-value pair;

[0014] Repeat the following steps until there are no unselected key-value pairs: Select an unselected key-value pair from the multiple key-value pairs, and calculate the first similarity between the currently selected key-value pair and each determined centroid; When it is determined that each first similarity is less than a preset first threshold, determine the currently selected key-value pair as a new centroid, and determine a new question-and-answer category according to the currently selected key-value pair; When it is determined that among the first similarities, there is at least one first similarity greater than or equal to the first threshold, then determine that the currently selected key-value pair belongs to the question-and-answer category corresponding to the centroid corresponding to the largest first similarity.

[0015] Further, when there are at least two key-value pairs corresponding to each question-and-answer category, perform deduplication processing on each key-value pair under each question-and-answer category to obtain the remaining key-value pairs under each question-and-answer category, including:

[0016] Calculate the second similarity between every two key-value pairs under each question-and-answer category;

[0017] If the second similarity is less than a preset second threshold, then retain the corresponding two key-value pairs; if the second similarity is greater than or equal to the second threshold, then remove one key-value pair from the corresponding two key-value pairs to obtain the remaining key-value pairs under each question-and-answer category;

[0018] Wherein, when there are at least two remaining key-value pairs under the question-and-answer category, the second similarity between any two remaining key-value pairs under the question-and-answer category is greater than or equal to the second threshold.

[0019] Further, after obtaining the remaining key-value pairs under each question-and-answer category, it further includes:

[0020] According to a preset clustering algorithm, determine a core key-value pair from each of the remaining key-value pairs under each question-and-answer category, determine each core key-value pair as a new centroid, and store each centroid.

[0021] Further, after storing each centroid, it further includes:

[0022] Obtain other interaction information, and form other text files according to the other interaction information;

[0023] Generate multiple other key-value pairs according to the other text files; the other key-value pairs include questions and answers;

[0024] Repeat all the following steps until there are no other key-value pairs that have not been selected: Select an unselected other key-value pair from the multiple other key-value pairs, and calculate the third similarity between the currently selected other key-value pair and each stored centroid; When it is determined that each third similarity is less than a preset third threshold, determine the currently selected other key-value pair as a new centroid, and determine a new Q&A category according to the currently selected other key-value pair; When it is determined that among the third similarities, there is at least one third similarity greater than or equal to the third threshold, determine that the currently selected other key-value pair belongs to the Q&A category corresponding to the centroid corresponding to the maximum third similarity;

[0025] For each other key-value pair under each Q&A category, perform a deduplication process to obtain the remaining other key-value pairs under each Q&A category;

[0026] Generate other question answer information according to the remaining other key-value pairs under each Q&A category.

[0027] Further, after generating the question answer information, it further includes:

[0028] Judge whether the question answer information exists in a preset knowledge base. If not, store the question answer information in the knowledge base.

[0029] Further, after generating the question answer information according to the remaining key-value pairs under each Q&A category, it includes:

[0030] Obtain question information;

[0031] Determine the answer corresponding to the question information from the question answer information, and provide the determined answer to the corresponding user.

[0032] In a second aspect, the present application provides an interactive information processing device, including:

[0033] A first acquisition unit, configured to acquire the interaction information between the customer service and the user, and form a text file according to the interaction information;

[0034] A first generation unit, configured to generate a plurality of key-value pairs according to the text file; the key-value pairs include questions and answers;

[0035] A first determination unit, configured to determine a plurality of Q&A categories according to the plurality of key-value pairs, where at least one key-value pair in the plurality of key-value pairs corresponds to each Q&A category, and each key-value pair corresponds to a Q&A category;

[0036] A first processing unit, configured to perform a deduplication process on each key-value pair under each Q&A category to obtain the remaining key-value pairs under each Q&A category;

[0037] A second generation unit, configured to generate question answer information according to the remaining key-value pairs under each of the question and answer categories.

[0038] Further, the first determination unit is specifically configured to:

[0039] Select a key-value pair from the multiple key-value pairs, determine the selected key-value pair as the initial centroid, and determine a question and answer category according to the selected key-value pair;

[0040] Repeat the following steps until there are no unselected key-value pairs: Select an unselected key-value pair from the multiple key-value pairs, calculate the first similarity between the currently selected key-value pair and each determined centroid; when it is determined that each first similarity is less than a preset first threshold, determine the currently selected key-value pair as a new centroid, and determine a new question and answer category according to the currently selected key-value pair; when it is determined that at least one of the first similarities is greater than or equal to the first threshold, determine that the currently selected key-value pair belongs to the question and answer category corresponding to the centroid corresponding to the maximum first similarity.

[0041] Further, when there are at least two key-value pairs corresponding to each question and answer category, the first processing unit is specifically configured to:

[0042] Calculate the second similarity between every two key-value pairs under each question and answer category; if the second similarity is less than a preset second threshold, retain the corresponding two key-value pairs; if the second similarity is greater than or equal to the second threshold, remove one key-value pair from the corresponding two key-value pairs to obtain the remaining key-value pairs under each question and answer category; wherein, when there are at least two remaining key-value pairs under the question and answer category, the second similarity between any two remaining key-value pairs under the question and answer category is greater than or equal to the second threshold.

[0043] Further, the apparatus further includes:

[0044] A second determination unit, configured to, after the first processing unit obtains the remaining key-value pairs under each question and answer category, determine a core key-value pair from the remaining key-value pairs under each question and answer category according to a preset clustering algorithm, determine each core key-value pair as a new centroid, and store each centroid.

[0045] Further, the apparatus further includes:

[0046] A second acquisition unit, configured to, after the second determination unit stores each centroid, acquire other interaction information, and form other text files according to the other interaction information;

[0047] A third generation unit, configured to generate a plurality of other key-value pairs according to the other text files; the other key-value pairs include questions and answers;

[0048] A second processing unit, configured to repeatedly execute the following steps until there are no unselected other key-value pairs: select an unselected other key-value pair from the plurality of other key-value pairs, and calculate a third similarity between the currently selected other key-value pair and each stored centroid; when it is determined that each third similarity is less than a preset third threshold, determine the currently selected other key-value pair as a new centroid, and determine a new question-and-answer category according to the currently selected other key-value pair; when it is determined that at least one of the third similarities is greater than or equal to the third threshold, determine that the currently selected other key-value pair belongs to the question-and-answer category corresponding to the centroid corresponding to the maximum third similarity;

[0049] A third processing unit, configured to perform a deduplication process on each of the other key-value pairs under each question-and-answer category to obtain the remaining other key-value pairs under each question-and-answer category;

[0050] A fourth generation unit, configured to generate other question answering information according to the remaining other key-value pairs under each question-and-answer category.

[0051] Further, the second generation unit is further configured to, after generating the question answering information, determine whether the question answering information exists in a preset knowledge base, and if not, store the question answering information in the knowledge base.

[0052] Further, the apparatus further includes:

[0053] A fourth processing unit, configured to, after the second generation unit generates question answering information according to the remaining key-value pairs under each question-and-answer category, obtain question information; determine an answer corresponding to the question information from the question answering information, and provide the determined answer to the corresponding user.

[0054] In a third aspect, the present application provides an interactive information processing device, including: a memory and a processor;

[0055] The memory is configured to store a computer program;

[0056] Wherein, the processor executes the computer program in the memory to implement the method according to any one of the embodiments of the first aspect.

[0057] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to implement the method according to any one of the embodiments of the first aspect.

[0058] The present application provides a method, apparatus, device and storage medium for processing interactive information. First, the interactive information between customer service and users is obtained, and a text file is formed based on the interactive information. Second, based on the text file, multiple key-value pairs including questions and answers are generated, each key-value pair including questions and answers. Then, based on the multiple key-value pairs generated, multiple question and answer categories are determined, wherein each question and answer category corresponds to at least one key-value pair from the multiple key-value pairs. Then, each key-value pair under each question and answer category is deduplicated to obtain the remaining key-value pairs under each question and answer category. Finally, question and answer information is generated based on the remaining key-value pairs under each question and answer category. This case generates question and answer information by processing the interactive information generated between customer service and users, so that when customer service receives questions asked by users, they can use the generated question and answer information as a reference to quickly answer user questions, thereby greatly improving customer service efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0060] Figure 1 A schematic diagram of the architecture of the interactive system provided for this application;

[0061] Figure 2 A flowchart of the interactive information processing method provided in Example 1 of the present application;

[0062] Figure 3 A flowchart of the interactive information processing method provided in Example 2 of the present application;

[0063] Figure 4 This is a schematic diagram of the interface of the user-side device provided in Example 2 of the present application;

[0064] Figure 5 A schematic diagram of the structure of the interactive information processing device provided in Example 3 of the present application;

[0065] Figure 6 A schematic diagram of the structure of an interactive information processing device provided in Example 4 of the present application;

[0066] Figure 7 This is a structural diagram of the interactive information processing device provided in Example 5 of the present application.

[0067] The above drawings illustrate specific embodiments of the present disclosure, which will be described in more detail below. These drawings and textual descriptions are not intended to limit the scope of the present disclosure in any way, but rather to illustrate the concepts of the present disclosure to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0068] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. Apparently, the described embodiments are some, but not all, of the embodiments of this application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the scope of protection of this application.

[0069] Figure 1 It is a schematic architecture diagram of an interaction system provided by this application. The interaction system includes a customer service side device 101 and a user side device 102. Among them, the user can send the questions they want to ask to the customer service side device through the user side device 102. After the customer service views the questions consulted by the user through the customer service side device 101, the customer service can send the answer information of the questions to the user side device 102 through the customer service side device 101, and the user views the answer information through the user side device 102.

[0070] Figure 2 It is a schematic flowchart of an interaction information processing method provided by Embodiment 1 of this application. As Figure 1 shown, the method includes:

[0071] Step 201: Obtain the interaction information between the customer service and the user.

[0072] In practical applications, the execution subject of this embodiment can be an interaction information processing device. The interaction information processing device can be a program software or a medium storing relevant computer programs, such as a USB flash drive, etc.; or, the interaction information processing device can also be an entity device integrated or installed with relevant computer programs, such as a chip, a smart terminal, a computer, a server, etc.

[0073] In this embodiment, the interaction information can be the voice conversation record information between the customer service and the user, the conversation record information in the live broadcast platform, etc. To make the subsequent generated question answer information more referenceable, the interaction information between the customer service and the user within a preset time period can be obtained, where the preset time period can be set according to actual needs, such as the recent half year.

[0074] Step 202: Form a text file according to the interaction information.

[0075] In this embodiment, after obtaining the interaction information, the interaction information can be processed, such as being recognized and optimized, to form a corresponding text file. Taking the interaction information as the voice conversation record information as an example, first, the voice conversation record information can be subjected to speech recognition using the Hidden Markov Model (HMMLEARN, abbreviated as HMM) framework to obtain a text file; then, the obtained text file can be optimized using a Convolutional Neural Network (CNN), for example, correcting incorrect sentences in the text file, removing useless information in the text file, etc., so as to form the final text file.

[0076] Step 203: Generate a plurality of key-value pairs according to the text file; the key-value pairs include questions and answers.

[0077] In this embodiment, the key-value pairs can be in the form of key-value, where key corresponds to the question and value corresponds to the answer.

[0078] Specifically, after forming the text file, first, the sentences in the text file can be segmented to obtain the segmentation corresponding to each sentence; secondly, the part-of-speech of each segmentation can be determined to determine the interrogative sentences in the text file according to the part-of-speech of the segmentation; then, for each determined interrogative sentence, according to the context of each interrogative sentence in the text file, the answer corresponding to each interrogative sentence can be determined; finally, according to each interrogative sentence and the answer corresponding to each interrogative sentence, a plurality of key-value pairs are formed, where each key-value pair includes a question (i.e., an interrogative sentence) and an answer.

[0079] Step 204: Determine a plurality of question-and-answer categories according to the plurality of key-value pairs, where at least one key-value pair among the plurality of key-value pairs corresponds to each question-and-answer category, and each key-value pair corresponds to one question-and-answer category.

[0080] In this embodiment, after generating a plurality of key-value pairs, a plurality of question-and-answer categories can be determined according to the plurality of key-value pairs. Exemplarily, a plurality of question-and-answer categories can be determined according to the questions in each key-value pair. For example, in the first key-value pair and the second key-value pair, both are questions related to fee payment, so the determined question-and-answer category can be fee payment. By determining a plurality of question-and-answer categories, the classification processing of the plurality of key-value pairs is also realized, where at least one key-value pair among the plurality of key-value pairs corresponds to each question-and-answer category, and each key-value pair corresponds to one question-and-answer category.

[0081] Step 205: Perform duplicate removal processing on each key-value pair under each question-and-answer category to obtain the remaining key-value pairs under each question-and-answer category.

[0082] In this embodiment, usually at least two key-value pairs correspond to one Q&A category. Moreover, among the at least two key-value pairs corresponding to one Q&A category, there may be duplicate questions, or questions with the same meaning but different ways of asking. Therefore, it is necessary to remove duplicates from the key-value pairs under each Q&A category to obtain the remaining key-value pairs under each Q&A category. For example, two key-value pairs correspond to the Q&A category of "fee payment". Among them, the question corresponding to one key-value pair is "How to pay", and the other key-value pair corresponds to "How to make a payment". It can be seen that the two have the same meaning, so one of the key-value pairs can be deleted. In this way, after removing duplicates, only one key-value pair remains under the Q&A category of "fee payment".

[0083] Step 206: Generate question-answer information according to the remaining key-value pairs under each Q&A category.

[0084] In this embodiment, after obtaining the remaining key-value pairs under each Q&A category, question-answer information can be generated according to the remaining key-value pairs under each Q&A category. Among them, the question-answer information can specifically be Frequently Asked Questions (FAQ). In this way, after generating the question-answer information, for the human customer service, when being asked a question, the generated question-answer information can be used as a reference to answer the user's question better and faster, thereby improving the customer service efficiency; for the intelligent customer service, when being asked a question, the generated question-answer information corresponding to the asked question can be directly called, thereby greatly improving the customer service efficiency.

[0085] The embodiment of the present application provides an interactive information processing method. First, obtain the interactive information between the customer service and the user, and form a text file according to the interactive information. Secondly, generate multiple key-value pairs including questions and answers according to the text file, and each key-value pair includes a question and an answer. Then, determine multiple Q&A categories according to the generated multiple key-value pairs, where at least one key-value pair among the multiple key-value pairs corresponds to each Q&A category. Then, remove duplicates from the key-value pairs under each Q&A category to obtain the remaining key-value pairs under each Q&A category. Finally, generate question-answer information according to the remaining key-value pairs under each Q&A category. By processing the existing interactive information between the customer service and the user in this case, question-answer information is generated so that when the customer service receives the question asked by the user, the generated question-answer information can be used as a reference to quickly answer the user's question, thus greatly improving the customer service efficiency.

[0086] Figure 3 It is a schematic flowchart of the interactive information processing method provided in the second embodiment of the present application, as Figure 3 shown, this method may include:

[0087] Step 301: Obtain the interaction information between the customer service and the user, and form a text file according to the interaction information.

[0088] Step 302: Generate multiple key-value pairs according to the text file; the key-value pairs include questions and answers.

[0089] In this embodiment, regarding Step 301 and Step 302, for specific explanations, reference can be made to the explanations in Steps 201 - 203, which will not be elaborated here.

[0090] Step 303: Select a key-value pair from the multiple key-value pairs, determine the selected key-value pair as the initial centroid, and determine a question-and-answer category according to the selected key-value pair.

[0091] In this embodiment, in order to improve the processing efficiency, the MapReduce computing model can be introduced, that is, it is processed in a distributed manner. For example, this step can be executed at the Mapper end. Specifically, at the Mapper end, the question-and-answer category and the centroid can be determined based on the Kmeans clustering algorithm.

[0092] Among them, a key-value pair can be randomly selected from the multiple key-value pairs; when determining the question-and-answer category, the question-and-answer category corresponding to the selected key-value pair can be determined according to the question in the selected key-value pair.

[0093] Step 304: Select an unselected key-value pair from the multiple key-value pairs, calculate the first similarity between the currently selected key-value pair and each determined centroid. When it is determined that each first similarity is less than the preset first threshold, execute Step 305. When it is determined that there is at least one first similarity greater than or equal to the first threshold among the first similarities, execute Step 306.

[0094] In this embodiment, Step 304 will be repeatedly executed until there are no unselected key-value pairs.

[0095] For example, if a total of 100 key-value pairs are generated, step 304 needs to be executed 99 times. When step 304 is executed for the first time, that is, when selecting the second key-value pair from 100 key-value pairs, a non-selected key-value pair can be randomly selected, and the first similarity between the currently selected key-value pair and each determined centroid (at this time there is only the initial centroid) is calculated. If the first similarity is less than the preset first threshold, step 305 is executed, that is, the second selected key-value pair is also used as a centroid, and a new Q&A category is determined according to the second selected key-value pair; if the first similarity is greater than or equal to the first threshold, step 306 is executed, that is, it is determined that the currently selected key-value pair belongs to the Q&A category corresponding to the initial centroid, that is, the currently selected key-value pair and the first selected key-value pair belong to the same Q&A category. Next, the third key-value pair is selected from 100 key-value pairs, that is, a non-selected key-value pair is selected, and the first similarity between the currently selected key-value pair and each determined centroid (at this time there is the initial centroid, and there may also be a centroid determined for the second selected key-value pair) is calculated. If each first similarity is less than the first threshold, step 305 is executed, that is, a centroid and a Q&A category are determined for the currently selected key-value pair; if at least one of the first similarities is greater than or equal to the first threshold, step 305 is executed, that is, it is determined that the currently selected key-value pair belongs to the Q&A category corresponding to the centroid corresponding to the maximum first similarity. Subsequently, the 4th, 5th... 100th times are continued until there are no non-selected key-value pairs among the 100 key-value pairs.

[0096] In this embodiment, the first similarity between the key-value pair and the centroid can be calculated based on the Kmeans clustering algorithm; in addition, the first threshold can be set according to actual needs.

[0097] Step 305: Determine that the currently selected key-value pair is a new centroid, determine a new Q&A category according to the currently selected key-value pair, and execute step 304; when there are no non-selected key-value pairs among multiple key-value pairs, execute step 307.

[0098] In this embodiment, when each first similarity is less than the first threshold, it means that the currently selected key-value pair does not belong to the same Q&A category as each of the previously selected key-value pairs. Therefore, the currently selected key-value pair is determined as a new centroid, a new Q&A category is determined according to the currently selected key-value pair, and then it is determined whether there are still non-selected key-value pairs among the generated multiple key-value pairs. If there are, step 304 is continued to be executed. If not, step 307 is executed.

[0099] Step 306: Determine that the currently selected key-value pair belongs to the Q&A category corresponding to the centroid with the highest first similarity, and execute Step 304; when there are no unselected key-value pairs among multiple key-value pairs, execute Step 307.

[0100] In this embodiment, when determining each first similarity, if there is at least one first similarity greater than or equal to the first threshold, it indicates that the currently selected key-value pair belongs to the same Q&A category as one or some of the previously selected key-value pairs. At this time, determine that the currently selected key-value pair belongs to the Q&A category corresponding to the centroid with the highest first similarity.

[0101] Step 307: Perform duplicate removal processing on the key-value pairs under each Q&A category to obtain the remaining key-value pairs under each Q&A category.

[0102] In this embodiment, when there are at least two key-value pairs corresponding to each Q&A category, Step 307 may include the following steps: calculate the second similarity between every two key-value pairs under each Q&A category; if the second similarity is less than the preset second threshold, retain the corresponding two key-value pairs; if the second similarity is greater than or equal to the second threshold, remove one key-value pair from the corresponding two key-value pairs to obtain the remaining key-value pairs under each Q&A category; where, when there are at least two remaining key-value pairs under a Q&A category, the second similarity between any two remaining key-value pairs under the Q&A category is greater than the second threshold. Now, taking an example where there are 4 key-value pairs A, B, C, and D corresponding to a Q&A category, calculate the second similarity between A and B, A and C, A and D, B and C, and B and D respectively. For example, if the second similarity between A and B is less than the second threshold, then A and B can be retained first, and if the second similarity between A and C is greater than or equal to the second threshold, then one of A and C needs to be removed and only one is retained. The retention or removal of each key-value pair can be determined in turn.

[0103] Among them, this step can also be executed on the Mapper side; in addition, the second threshold can be set according to actual requirements; in addition, the second similarity between two key-value pairs can be calculated based on the prior art. [[ID=

[13] ]] [[ID=

[14] ]]

[0104] In this embodiment, after performing step 307, it further includes: according to a preset clustering algorithm, determining a core key-value pair from each of the remaining key-value pairs in each Q&A category, and determining each core key-value pair as a new centroid, and storing each centroid. The preset clustering algorithm may be the DBSCAN algorithm. Specifically, at the Reducer side, according to the DBSCAN algorithm, calculate the number of elements in the e domain of each Q&A category, and take the one with the largest number of elements in the e domain category, that is, the one with the largest minPts number, as the new centroid, that is, the one with the highest density as the new centroid, that is, determine a centroid for each Q&A category. In addition, to speed up the program access rate, multiple Reducer sides can be used to determine a centroid for each Q&A category. By using multiple Reducer sides, the rate of data reduction can be improved. Compared with single-threaded processing, it can not only save storage space, but also be simple and convenient to process.

[0105] In this embodiment, after determining a centroid for each Q&A category and storing each centroid, the following steps may further be included:

[0106] The first step: Obtain other interaction information, and form other text files according to the other interaction information;

[0107] The second step: Generate multiple other key-value pairs according to the other text files; the other key-value pairs include questions and answers;

[0108] The third step: Repeat the following steps until there are no unselected other key-value pairs: Select an unselected other key-value pair from the multiple other key-value pairs, calculate the third similarity between the currently selected other key-value pair and each stored centroid; when it is determined that each third similarity is less than a preset third threshold, determine the currently selected other key-value pair as a new centroid, and determine a new Q&A category according to the currently selected other key-value pair; when it is determined that at least one of the third similarities is greater than or equal to the third threshold, determine that the currently selected other key-value pair belongs to the Q&A category corresponding to the centroid corresponding to the largest third similarity;

[0109] The third step: Perform duplicate removal processing on each of the other key-value pairs under each Q&A category to obtain the remaining other key-value pairs under each Q&A category;

[0110] The fourth step: Generate other question-answering information according to the remaining other key-value pairs under each Q&A category.

[0111] When other interaction information is obtained subsequently, the other interaction information is processed to form other text files. Based on the other text files, multiple other key-value pairs are generated. Then, when determining the Q&A category, the stored centroid can be used to determine the Q&A category without determining the initial centroid and some other centroids, thereby improving the efficiency of data processing and further improving the customer service efficiency. Among them, the process of forming other text files for other interaction information and generating multiple other key-value pairs for the other text files can refer to steps 201 - 203 and will not be elaborated here.

[0112] Step 308: Generate question and answer information based on the remaining key-value pairs under each Q&A category.

[0113] In this embodiment, FAQs can be generated according to the remaining key-value pairs under each Q&A category.

[0114] Step 309: Determine whether the question and answer information exists in the preset knowledge base. If not, store the question and answer information in the knowledge base.

[0115] In this embodiment, the generated question and answer information can be stored in the preset knowledge base. Moreover, to avoid duplicate storage and reduce memory occupancy, after generating the question and answer information, it can be determined whether the question and answer information exists in the knowledge base. If it does not exist, it is stored; if it exists, it is not stored.

[0116] Step 310: Obtain the question information, determine the answer corresponding to the question information from the question and answer information, and provide the determined answer to the corresponding user.

[0117] In this embodiment, after the customer service obtains the user's question information, it can use the generated question and answer information as a reference to quickly and accurately answer the user's question. As Figure 4 shown, Figure 4 is a schematic diagram of the interface of the user-side device provided in the second embodiment of the present application. After the user sends the question information "How to pay" to the interaction information processing device, the interaction information processing device can use the generated question and answer information as a reference, quickly determine the corresponding answer from it, and send the answer to the user-side device, so that the user can view the answer through the user-side device.

[0118] In the embodiment of the present application, by calculating the first similarity between the currently selected key-value pairs and each determined centroid, the Q&A category to which each key-value pair belongs can be quickly determined. Then, duplicate removal processing is performed separately for each Q&A category, which can greatly improve the processing efficiency. In addition, by using multiple Reducer ends, the rate of data reduction can be improved. Compared with single-threaded processing, it can not only save storage space, but also be simple and convenient to process. In addition, according to the preset clustering algorithm, new centroids are determined and saved for each Q&A category. When other interaction information is obtained later, the other interaction information is processed to form other text files. According to the other text files, multiple other key-value pairs are generated. Then, when determining the Q&A category, the Q&A category can be determined based on the stored centroids without determining the initial centroids and some other centroids, thereby improving the efficiency of data processing and further improving the customer service efficiency.

[0119] Figure 5 FIG. is a schematic structural diagram of an interaction processing device provided in Embodiment 3 of the present application, as Figure 5 shown, including:

[0120] A first acquisition unit 501, configured to acquire interaction information between a customer service and a user, and form a text file according to the interaction information;

[0121] A first generation unit 502, configured to generate a plurality of key-value pairs according to the text file; the key-value pairs include questions and answers;

[0122] A first determination unit 503, configured to determine a plurality of Q&A categories according to the plurality of key-value pairs, where at least one key-value pair in the plurality of key-value pairs corresponds to each Q&A category, and each key-value pair corresponds to one Q&A category;

[0123] A first processing unit 504, configured to perform duplicate removal processing on each key-value pair under each Q&A category to obtain the remaining key-value pairs under each Q&A category;

[0124] A second generation unit 505, configured to generate question-answering information according to the remaining key-value pairs under each Q&A category.

[0125] The interaction information processing device provided in this embodiment is the same as the technical solution in the interaction information processing method provided in any of the foregoing embodiments, and its implementation principle and technical effects are similar, and will not be described in detail.

[0126] Figure 6 FIG. is a schematic structural diagram of an interaction information processing device provided in Embodiment 4 of the present application. On the basis of Embodiment 3, as Figure 6 shown,

[0127] The first determining unit 503 is specifically configured to:

[0128] Select a key-value pair from the multiple key-value pairs, determine the selected key-value pair as the initial centroid, and determine a question-and-answer category according to the selected key-value pair;

[0129] Repeat the following steps until there are no unselected key-value pairs: Select an unselected key-value pair from the multiple key-value pairs, calculate the first similarity between the currently selected key-value pair and each determined centroid; when it is determined that each first similarity is less than a preset first threshold, determine the currently selected key-value pair as a new centroid, and determine a new question-and-answer category according to the currently selected key-value pair; when it is determined that at least one of the first similarities is greater than the first threshold, determine that the currently selected key-value pair belongs to the question-and-answer category corresponding to the centroid corresponding to the maximum first similarity.

[0130] Further, when there are at least two key-value pairs corresponding to each question-and-answer category, the first processing unit 504 is specifically configured to:

[0131] Calculate the second similarity between every two key-value pairs under each question-and-answer category; if the second similarity is less than a preset second threshold, retain the corresponding two key-value pairs; if the second similarity is greater than or equal to the second threshold, remove one key-value pair from the corresponding two key-value pairs to obtain the remaining key-value pairs under each question-and-answer category; wherein, when there are at least two remaining key-value pairs under the question-and-answer category, the second similarity between any two remaining key-value pairs under the question-and-answer category is greater than the second threshold.

[0132] The device further includes:

[0133] A second determining unit 601, configured to, after the first processing unit 504 obtains the remaining key-value pairs under each question-and-answer category, determine a core key-value pair from the remaining key-value pairs under each question-and-answer category according to a preset clustering algorithm, determine each core key-value pair as a new centroid, and store each centroid.

[0134] The device further includes:

[0135] A second obtaining unit 602, configured to, after the second determining unit 601 stores each centroid, obtain other interaction information, and form other text files according to the other interaction information;

[0136] A third generating unit 603, configured to generate multiple other key-value pairs according to the other text files; the other key-value pairs include questions and answers.

[0137] The second processing unit 604 is configured to repeatedly execute the following steps until there are no other key-value pairs that have not been selected: select an unselected other key-value pair from the multiple other key-value pairs, and calculate the third similarity between the currently selected other key-value pair and each stored centroid; when it is determined that each third similarity is less than a preset third threshold, determine the currently selected other key-value pair as a new centroid, and determine a new question-and-answer category according to the currently selected other key-value pair; when it is determined that at least one of the third similarities is greater than the third threshold, determine that the currently selected other key-value pair belongs to the question-and-answer category corresponding to the centroid corresponding to the maximum third similarity.

[0138] The third processing unit 605 is configured to perform deduplication processing on each other key-value pair under each question-and-answer category to obtain the remaining other key-value pairs under each question-and-answer category.

[0139] The fourth generation unit 606 is configured to generate other question answer information according to the remaining other key-value pairs under each question-and-answer category.

[0140] Further, the second generation unit 505 is further configured to, after generating the question answer information, determine whether the question answer information exists in a preset knowledge base. If not, store the question answer information in the knowledge base.

[0141] The device further includes:

[0142] The fourth processing unit 607 is configured to, after the second generation unit 505 generates question answer information according to the remaining key-value pairs under each question-and-answer category, obtain question information; determine an answer corresponding to the question information from the question answer information, and provide the determined answer to the corresponding user.

[0143] The interactive information processing device provided in this embodiment is the same as the technical solution in the interactive information processing method provided in any of the foregoing embodiments, and its implementation principle and technical effects are similar, and will not be described in detail.

[0144] Figure 7 The structural schematic diagram of the interactive information processing device provided in the fifth embodiment of the present application is shown in Figure 7 as follows, and includes: a memory 701 and a processor 702;

[0145] The memory 701 is used to store a computer program.

[0146] Wherein, the processor 702 executes the computer program in the memory 701 to implement the method of any of the embodiments .

[0147] Embodiment 6 of the present application provides a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to implement the method of any embodiment.

[0148] After considering the specification and practicing the invention disclosed herein, those skilled in the art will readily conceive of other embodiments of the present disclosure. This application is intended to cover any variations, uses, or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common general knowledge or conventional technical means in the technical field not disclosed in the present disclosure. The specification and examples are only illustrative, and the true scope and spirit of the present disclosure are pointed out by the following claims.

[0149] It should be understood that the present disclosure is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is only limited by the appended claims.

Claims

1. An interactive information processing method, characterized in that, Including: Obtain the interaction information between the customer service and the user, and form a text file according to the interaction information, where the interaction information is the conversation record obtained within a preset time period; Generate a plurality of key-value pairs according to each question sentence in the text file and the answer corresponding to each question sentence; The key-value pair includes a question and an answer; Determine a plurality of question-and-answer categories according to the plurality of key-value pairs, including: select a key-value pair from the plurality of key-value pairs, determine the selected key-value pair as the initial centroid, and determine a question-and-answer category according to the selected key-value pair; repeat the following steps until there are no unselected key-value pairs: select an unselected key-value pair from the plurality of key-value pairs, calculate the first similarity between the currently selected key-value pair and each determined centroid; when it is determined that each first similarity is less than a preset first threshold, determine the currently selected key-value pair as a new centroid, and determine a new question-and-answer category according to the currently selected key-value pair; when it is determined that there is at least one first similarity greater than or equal to the first threshold among the first similarities, determine that the currently selected key-value pair belongs to the question-and-answer category corresponding to the centroid corresponding to the maximum first similarity, where each question-and-answer category corresponds to at least one key-value pair among the plurality of key-value pairs, and each key-value pair corresponds to a question-and-answer category; Perform deduplication processing on each key-value pair under each question-and-answer category to obtain the remaining key-value pairs under each question-and-answer category; Generate question-answering information according to the remaining key-value pairs under each question-and-answer category.

2. The method according to claim 1, wherein When there are at least two key-value pairs corresponding to each question-and-answer category, perform deduplication processing on each key-value pair under each question-and-answer category to obtain the remaining key-value pairs under each question-and-answer category, including: Calculate the second similarity between every two key-value pairs under each question-and-answer category; If the second similarity is less than a preset second threshold, retain the corresponding two key-value pairs; if the second similarity is greater than or equal to the second threshold, remove one key-value pair from the corresponding two key-value pairs to obtain the remaining key-value pairs under each question-and-answer category; Wherein, when there are at least two remaining key-value pairs under the question-and-answer category, the second similarity between any two remaining key-value pairs under the question-and-answer category is greater than or equal to the second threshold.

3. The method according to any one of claims 1-2, characterized in that, After obtaining the remaining key-value pairs under each question-and-answer category, it further includes: Determine a core key-value pair from each of the remaining key-value pairs under each question-and-answer category according to a preset clustering algorithm, determine each core key-value pair as a new centroid, and store each centroid.

4. The method according to claim 3, characterized in that After storing each centroid, it further includes: Obtain other interaction information, and form other text files according to the other interaction information; Generate a plurality of other key-value pairs according to the other text files; the other key-value pairs include questions and answers; Repeat the following steps until there are no other key-value pairs that have not been selected: Select an other key-value pair that has not been selected from the multiple other key-value pairs, and calculate the third similarity between the currently selected other key-value pair and each stored centroid; When it is determined that each third similarity is less than a preset third threshold, determine the currently selected other key-value pair as a new centroid, and determine a new Q&A category according to the currently selected other key-value pair; When it is determined that there is at least one third similarity greater than or equal to the third threshold among the third similarities, determine that the currently selected other key-value pair belongs to the Q&A category corresponding to the centroid corresponding to the maximum third similarity; For each other key-value pair under each Q&A category, perform a deduplication process to obtain the remaining other key-value pairs under each Q&A category; Generate other question answer information according to the remaining other key-value pairs under each Q&A category.

5. The method according to any one of claims 1-2, characterized in that, After generating the question answer information, it further includes: Judge whether the question answer information exists in the preset knowledge base. If not, store the question answer information in the knowledge base.

6. The method according to any one of claims 1-2, characterized in that, After generating the question answer information according to the remaining key-value pairs under each Q&A category, it includes: Obtain question information; Determine the answer corresponding to the question information from the question answer information, and provide the determined answer to the corresponding user.

7. An interactive information processing device, characterized in that, It includes: A first acquisition unit, configured to acquire the interaction information between the customer service and the user, and form a text file according to the interaction information, where the interaction information is the conversation record obtained within a preset time period; A first generation unit, configured to generate multiple key-value pairs according to each question sentence in the text file and the answer corresponding to each question sentence; The key-value pair includes a question and an answer; A first determination unit, configured to determine multiple Q&A categories according to the multiple key-value pairs, including: Select a key-value pair from the multiple key-value pairs, determine the selected key-value pair as the initial centroid, and determine a Q&A category according to the selected key-value pair; Repeat the following steps until there are no unselected key-value pairs: Select an unselected key-value pair from the multiple key-value pairs, and calculate the first similarity between the currently selected key-value pair and each determined centroid; When it is determined that each first similarity is less than a preset first threshold, determine the currently selected key-value pair as a new centroid, and determine a new Q&A category according to the currently selected key-value pair; When it is determined that there is at least one first similarity greater than or equal to the first threshold among the first similarities, then determine that the currently selected key-value pair belongs to the Q&A category corresponding to the centroid corresponding to the maximum first similarity, where at least one key-value pair in the multiple key-value pairs corresponds to each Q&A category, and each key-value pair corresponds to a Q&A category; A first processing unit, configured to perform a deduplication process on each key-value pair under each Q&A category to obtain the remaining key-value pairs under each Q&A category; A second generation unit, configured to generate question answer information according to the remaining key-value pairs under each Q&A category.

8. The device according to claim 7, characterized in that, When there are at least two key-value pairs corresponding to each Q&A category, the first processing unit is specifically configured to: Calculate the second similarity between every two key-value pairs under each Q&A category; if the second similarity is less than a preset second threshold, then retain the corresponding two key-value pairs; if the second similarity is greater than or equal to the second threshold, then remove one key-value pair from the corresponding two key-value pairs, to obtain the remaining key-value pairs under each Q&A category; wherein, when there are at least two remaining key-value pairs under the Q&A category, the second similarity between any two of the remaining key-value pairs under the Q&A category is greater than or equal to the second threshold.

9. The device according to any one of claims 7-8, characterized in that, The device further includes: A second determination unit, configured to, after the processing unit obtains the remaining key-value pairs under each Q&A category, determine a core key-value pair from each of the remaining key-value pairs under each Q&A category according to a preset clustering algorithm, determine each core key-value pair as a new centroid, and store each centroid.

10. The device according to claim 9, characterized in that, The device further includes: A second acquisition unit, configured to, after the second determination unit stores each centroid, acquire other interaction information, and form other text files according to the other interaction information. A third generation unit, configured to generate a plurality of other key-value pairs according to the other text files; the other key-value pairs include questions and answers. A second processing unit, configured to repeatedly execute the following steps until there are no unselected other key-value pairs: select an unselected other key-value pair from the plurality of other key-value pairs, calculate the third similarity between the currently selected other key-value pair and each stored centroid; when it is determined that each third similarity is less than a preset third threshold, determine the currently selected other key-value pair as a new centroid, and determine a new Q&A category according to the currently selected other key-value pair; when it is determined that there is at least one third similarity greater than or equal to the third threshold among each third similarity, determine that the currently selected other key-value pair belongs to the Q&A category corresponding to the centroid corresponding to the maximum third similarity. A third processing unit, configured to perform duplicate removal processing on the other key-value pairs under each Q&A category, to obtain the remaining other key-value pairs under each Q&A category. A fourth generation unit, configured to generate other question answering information according to the remaining other key-value pairs under each Q&A category.

11. An interactive information processing device, characterized in that, Comprising: A memory and a processor; The memory is configured to store a computer program; Wherein, the processor executes the computer program in the memory to implement the method according to any one of claims 1-6.

12. A computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to implement the method according to any one of claims 1-6.

Images