Consultation element extraction method, electronic device and storage medium for consultation dialogue

Abstract

The present invention discloses a method for extracting consulting elements from consulting conversations, an electronic device, and a storage medium. The method includes: obtaining a topic grouping according to a topic field and its application field. Using a set segmented deep neural network, identifying consulting key words from conversation content data. Dividing the conversation content data into multiple groups of conversation content data according to the consulting key words and set categories. Clustering the question sentences in each group of conversation content, and obtaining consulting elements through the key sentences in each cluster. The present invention is mainly based on a clustering method to obtain semantically abstracted consulting elements, and at the same time obtains existing questions for the consulting elements. The recognition rate is high and the main consulting elements can be obtained from consulting conversations, which is more convenient for subsequent answers or response matching needs.

Description

Technical Field

[0001] The present invention belongs to the technical field of speech processing, and in particular relates to a method for extracting consulting elements of consulting dialogues, an electronic device and a storage medium. Background Art

[0002] Current patents focus on how to identify intent categories in sentences or how to mine categories, or how to get answers to inquiries, but do not focus on how to initiate questions.

[0003] The patent method and device for obtaining answer information based on the user's question intention (CN105335400A) is most relevant to the method of this article. The method has a similar process, but the patent focuses on how to obtain the intended answer information rather than how to ask questions about the elements of specific intent.

[0004] Patent A method for identifying user query intention (CN103268348A) solves the problem of how to determine the user's query intention when the intention is known. Patent Method and Device for Mining Intent (CN103942204B) solves the problem of mining the user input's intention and obtains all the user's candidate intentions. Patent A method and device for generating questions (CN106909677A) focuses on asking questions based on unobtained answer parameters. Questions are retrieved based on unobtained answer parameters (similar to the consultation elements in the present invention) and returned to the user. In specific fields, such as the medical field, a disease complication mining method based on the FP-GROWTH algorithm (CN106709248) focuses only on the mining of this single element in the consultation scenario.

[0005] The patent most similar to the present invention is a method and device for obtaining answer information based on a user's question intention (CN105335400A), which is characterized by comprising: obtaining historical conversation records during instant communication between a first user and multiple second users; identifying user intentions for question corpora raised by each second user from the historical conversation records, obtaining each question corpus included under the same user intention, and obtaining reply corpora given by the first user to each question corpus under the same user intention; clustering each reply corpus for the same user intention to obtain multiple categories, and in a target category containing more reply corpora than a preset threshold, respectively calculating the weight of each reply corpus in the target category as the category center; and determining the central answer to the user intention based on the weight.

[0006] Secondly, a similar patent is a method and device for generating questions (CN106909677A). The method includes: determining an unobtained answer parameter; obtaining a question set for the unobtained answer parameter from a question library; obtaining a question for the unobtained answer parameter based on the question set for the unobtained answer parameter in the question library; and sending the question to the user.

[0007] A method and device for obtaining answer information based on the user's question intention (CN105335400A) only focuses on how to answer questions, but does not propose a method for asking questions. A method and device for generating questions (CN106909677A) is a method for question retrieval based on unavailable answer parameters. The defects of this method include: 1. All required answer parameters are known or manually defined 2. The main focus is on how to ask questions about unavailable answer parameters.

[0008] The above patent CN105335400A only focuses on the answer. The answer parameters of CN106909677A are known, and only the answer parameters that have not been obtained are targeted. It is a difficult task to let AI understand how to ask questions, and it is difficult to achieve overnight. Summary of the invention

[0009] The embodiments of the present invention aim to solve at least one of the above technical problems.

[0010] In a first aspect, an embodiment of the present invention provides a method for extracting consulting elements from consulting conversations, comprising: obtaining consulting conversation data. The consulting conversation data includes a subject field and conversation content data. The conversation content data includes: question sentences and answer sentences. Topic grouping is obtained according to the subject field and its application field. Consultation key words are identified from the conversation content data through a set fragment-based deep neural network. The conversation content data is divided into multiple groups of conversation content data according to the consultation key words and set categories. Question sentences are extracted from the multiple groups of conversation content data. The question sentences in each group of conversation content are clustered to obtain the key sentences in each cluster. Cluster keywords are extracted according to the key sentences in each cluster. Consultation elements are obtained according to the cluster keywords.

[0011] In a second aspect, an embodiment of the present invention provides an electronic device comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any one of the above-mentioned consulting element extraction methods for consulting conversations of the present invention.

[0012] In a third aspect, an embodiment of the present invention provides a storage medium, in which one or more programs including execution instructions are stored, and the execution instructions can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to execute any of the above-mentioned consulting element extraction methods of the consulting dialogue of the present invention.

[0013] In a fourth aspect, an embodiment of the present invention further provides a computer program product, comprising a computer program stored on a storage medium, wherein the computer program comprises program instructions, and when the program instructions are executed by a computer, the computer executes any one of the above-mentioned consulting element extraction methods for consulting conversations.

[0014] The embodiment of the present invention obtains topic groupings according to the subject field and its corresponding application field. The consulting key words are identified from the conversation content data through a set segmented deep neural network. The conversation content data is divided into multiple groups of conversation content data according to the consulting key words and set categories. The question sentences in each group of conversation content are clustered, and the consulting elements are obtained through the key sentences in each cluster. The present invention is mainly based on the clustering method to obtain the consulting elements after semantic abstraction, and at the same time obtains the existing questions for the consulting elements. The recognition rate is high and the main consulting elements can be obtained from the consulting conversation, which is more convenient for the subsequent answer or response matching needs.

[0015] ` BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0017] Figure 1 A flowchart of an embodiment of a method for extracting consulting elements from consulting dialogues of the present invention;

[0018] Figure 2 A flowchart of another embodiment of the method for extracting consulting elements from consulting dialogues of the present invention;

[0019] Figure 3 It is a schematic structural diagram of an embodiment of an electronic device of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application may be combined with each other.

[0022] The present invention may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0023] In the present invention, "module", "device", "system" and the like refer to related entities applied to computers, such as hardware, a combination of hardware and software, software or software in execution, etc. In detail, for example, an element can be, but is not limited to, a process, a processor, an object, an executable element, an execution thread, a program and / or a computer running on a processor. In addition, an application or script program running on a server, a server can all be an element. One or more elements can be in an execution process and / or thread, and an element can be localized on a computer and / or distributed between two or more computers, and can be operated by various computer-readable media. An element can also communicate through local and / or remote processes according to a signal with one or more data packets, for example, a signal from a data that interacts with another element in a local system, a distributed system, and / or a network on the Internet through a signal to interact with other systems.

[0024] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include" and "comprise" include not only those elements, but also other elements that are not explicitly listed, or also include elements inherent to such processes, methods, articles or equipment. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the existence of other identical elements in the process, method, article or equipment that includes the elements.

[0025] The embodiment of the present invention provides a method for extracting consulting elements from consulting dialogues, which can be applied to electronic devices. The electronic devices can be computers, servers, or other electronic products, etc., which are not limited in the present invention.

[0026] Please refer to Figure 1 , which shows a method for extracting consulting elements from consulting dialogues provided by an embodiment of the present invention.

[0027] like Figure 1 As shown, a method for extracting consulting elements from consulting dialogues of the present invention includes:

[0028] Step S101, obtaining consultation dialogue data.

[0029] In this step, consulting dialogue data is obtained. The consulting dialogue data includes a subject field and dialogue content data. The dialogue content data includes: question sentences and answer sentences.

[0030] Step S102, obtaining subject groups.

[0031] In this step, the subject groups are obtained based on the subject fields and their application fields.

[0032] Step S103, obtaining consultation key words.

[0033] In this step, the consultation key words are identified from the conversation content data through the set fragment-based deep neural network.

[0034] Step S104, dividing the conversation content data into multiple groups.

[0035] In this step, the conversation content data is divided into a plurality of groups of conversation content data according to the consultation key words and the set categories.

[0036] Step S105, extracting question sentences from the multiple groups of dialogue content data groups.

[0037] Step S106, obtaining key sentences in each cluster.

[0038] In this step, the question sentences in each group of dialogue content are clustered to obtain the key sentences in each cluster.

[0039] Step S107: extract cluster keywords based on the key sentences in each cluster.

[0040] Step S108, obtaining consulting elements according to the cluster keywords.

[0041] In some optional embodiments, the consulting key words include: proper nouns and non-entity words belonging to the central word of the sentence.

[0042] In some optional embodiments, step S106 includes:

[0043] The question sentences in each group of dialogue content are clustered using the k-means clustering algorithm.

[0044] In some optional embodiments, the K value in the k-means clustering algorithm is obtained by the Elbow Method. The value interval of k is set to [10,50]. The k value is an integer.

[0045] In some optional embodiments, if the Elbow method has no obvious elbow point, a value is selected where the slope is greater than the set slope value and k is greater than the set K value. The slope is set to the average value of the conventional slope. The K value is set to the average value of the conventional K.

[0046] In some optional embodiments, the word vector of each word in the question sentence is trained by FastText. The question sentence includes a question sentence and its corresponding answer sentence. The sentence vector is obtained by averaging the word vectors through the BOW bag of words model.

[0047] The sentence vectors of the question sentences in each group of dialogue content are clustered using the k-means clustering algorithm to obtain the key sentences in each cluster.

[0048] In some optional embodiments, step S107 includes merging the cluster keywords whose cosine similarity of the keywords is greater than a set threshold value. The threshold value is set to 0.8.

[0049] In some optional embodiments, step S107 further includes: merging synonymous cluster keywords. If the cluster keyword is a medical inquiry word, then removing the symptom word in the cluster keyword. Step S108 further includes: sorting the cluster keywords from large to small according to their weights, obtaining the cluster keyword corresponding to the maximum weight, and using the cluster keyword as the consultation element word.

[0050] In one embodiment of the present invention, the present invention generally adopts a rule or template matching method to solve the problem of obtaining consulting elements. Or the consulting elements are manually defined. The present invention mainly obtains the consulting elements after semantic abstraction based on the clustering method, and obtains the existing questions for the consulting elements. The main difficulty lies in obtaining the consulting elements from the dialogue data. The dialogue data has the difficulties of missing components and serious colloquialism.

[0051] The existing solutions for obtaining consultation elements of specific intent are not general enough and the labor cost is high. The solution and principle of the present invention are introduced in detail below, taking the example of seeing a doctor in a hospital.

[0052] First, if Figure 2 The following is a basic flow chart of the present invention in one embodiment. Each step is described in detail below, using two consultation dialogues under two topics as examples:

[0053] Scenario Example 1: Medicine - Respiratory Medicine - Pre-consultation:

[0054] Dialogue scenario 1 is as follows:

[0055] Patient: I have an itchy throat, cough and headache. What medicine should I take? (Male, 27 years old)

[0056] Doctor: Hello, how long has this situation lasted?

[0057] Patient: Two days.

[0058] Doctor: I caught a cold.

[0059] Patient: Yesterday.

[0060] Patient: What medicine should I take?

[0061] Doctor: It's probably due to overwork.

[0062] Doctor: Is there phlegm in your cough?

[0063] Patient: He is a delivery man. He was sweating yesterday and it was cold in the afternoon. It was windy and he was blown by the wind.

[0064] Patient: No sputum, mainly itchy throat and dry cough

[0065] Doctor: Do you have any history of allergies?

[0066] Patient: No

[0067] Doctor: Lianhua Qingwen Granules and Suhuang Cough Capsules

[0068] Doctor: Drink more water

[0069] Doctor: Take some rest.

[0070] Patient: Just drink these two, right?

[0071] Doctor: Yes.

[0072] Patient: Okay.

[0073] Dialogue 2 is as follows:

[0074] Patient: What is the reason for my persistent cough? (Female, 18 years old)

[0075] Doctor: Do you have phlegm? What color is it?

[0076] Patient: No

[0077] Doctor: How long have you been coughing?

[0078] Patients: 5 to 6 days

[0079] Doctor: Do you have symptoms such as fever, nasal congestion, runny nose, sore throat, headache, and aches in the limbs?

[0080] Patient: No

[0081] Doctor: Yes, have you had any recent contact with people from area A?

[0082] Patient: No

[0083] Doctor: Yes, have you used any medicine?

[0084] Patient: No

[0085] Doctor: Well, I suggest you take compound pholcodine orally to relieve cough and treat the symptoms.

[0086] Patient: OK

[0087] Patient: Thank you

[0088] Doctor: You’re welcome. Happy New Year!

[0089] Scenario 2: Car-Test Drive

[0090] Session 1:

[0091] Consumer: I would like to ask if your store can test drive a Toyota Camry.

[0092] Sales: Hello, the 2021 models currently on sale include 2.0T and 2.5T. Which one would you like to try?

[0093] Consumer: 2.5T, maybe?

[0094] Sales: 2.5T cars include gasoline cars and hybrid cars, that is, dual-engine cars. Do you want to try both?

[0095] Consumer: Yes, let me take a look at them all.

[0096] Consumer: Are both cars available for test drive now?

[0097] Sales: You can test drive it now. We can arrange a test drive for you if you come over.

[0098] Sales: Do you have a driver’s license?

[0099] Consumer: I have a driver’s license but it’s not with me. Can I show you the electronic one?

[0100] Sales: Sure. Please send it to me for review. After I confirm it, I can sign the agreement on the spot and start the test drive.

[0101] Sales: When will you come for a test drive?

[0102] Consumer: Uh, let’s talk about it later. I’ll contact you then, okay.

[0103] Sales: OK.

[0104] Session 2:

[0105] Consumer: Um, I would like to ask if you have a Cadillac CT4 that I can test drive in your store?

[0106] Sales: Yes, you can test drive it. Do you have a driver's license?

[0107] Consumer: Yes, can I just go to the store with my driver’s license?

[0108] Sales: Uh-huh, yes.

[0109] Consumer: No problem, I’ll bring my driver’s license with me when the time comes.

[0110] Sales: Do you drive a lot?

[0111] Consumer: It’s okay, I’m driving my family car.

[0112] Sales: Yes. When do you plan to come?

[0113] Consumer: I went there around 3pm.

[0114] Sales: Yes, please call me when you arrive and I will arrange it for you.

[0115] Consumer: Ok, bye.

[0116] Sales: Goodbye.

[0117] A method for extracting consulting elements from consulting dialogues provided by an embodiment of the present invention comprises the following steps:

[0118] 1. Group by topic

[0119] Group all counseling sessions into themes.

[0120] As shown in the above example, the subject category of Sample 1 is: Medicine-Respiratory Medicine-Pre-consultation.

[0121] The subject category of Example 2 is: Car-Test Drive

[0122] 2. Extraction of consultation points

[0123] Extract the key points of the consultation, which include entity words (i.e. meaningful proper nouns) in the consultation party's description and non-entity words that are the central words of the sentence.

[0124] The present invention uses an entity extraction method based on a fragment-based deep neural network.

[0125] In Sample 1, the symptoms contained in the patient's main statement were extracted, such as the main statement in Session 1 "Itchy throat, cough, headache, what medicine should I take? (Male, 27 years old)" and the main statement in Session 2 "Why can't I cough all the time (Female, 18 years old)", and the consultation points "itchy throat, cough, headache, medicine" and "cough" were extracted respectively.

[0126] In Sample 2, the inquiry sentences of consumers are extracted. For example, the sentence in Session 1 is "I would like to ask if you can test drive Toyota Camry." The sentence in Session 2 is "Um, I would like to ask if you have Cadillac CT4 for test drive?" The inquiry points are "test drive, Toyota, Camry" and "Cadillac, CT4, test drive" respectively.

[0127] Methods for extracting consulting key points include, but are not limited to, using dictionaries or template matching, using sequence labeling models using deep neural networks, using fragment-based deep neural network models and other information extraction methods.

[0128] 3. Group all session data by consulting points

[0129] The sessions are grouped according to the extracted symptoms, such as:

[0130] Example 1:

[0131] "cough": {session 1, session 2}

[0132] "Headache": {Session 1}

[0133] "Itchy Throat": {Session 1}

[0134] Example 2:

[0135] "Test Drive": {Session 1, Session 2}

[0136] "Toyota": {Session 1}

[0137] "Camry": {Session 1}

[0138] "Cadillac": {Session 2}

[0139] "CT4": {Session 2}

[0140] 4. Extract consulting questions

[0141] Extract all sentences in which the party being consulted asks questions in the conversation.

[0142] In Example 1, the doctor asks the patient a question.

[0143] In Example 2, the salesperson asks the consumer a question.

[0144] For example, the inquiry sentences in Session 2 of Example 1 are: {“Do you have phlegm? What color is it?”, “How long have you been coughing?”, “Do you have symptoms such as fever, runny nose, sore throat, headache, and limb pain?”, “Yes, have you had any contact with people from Area A recently?”, “Yes, have you used any medicine?”}

[0145] Inquiry sentences in Session 2 of Example 2: {“Do you have a driver’s license?”, “Do you drive a lot?”, “When do you plan to come over?”}

[0146] The present invention adopts a method based on regular matching to match questions.

[0147] The query sentence extraction methods include but are not limited to text classification based on deep neural networks, rule-based or template-based matching and other methods.

[0148] 5. Cluster the problems.

[0149] That is, cluster the query sentences in the same consulting point group

[0150] The present invention uses the k-means clustering method. For the selection of K, the present invention uses the Elbow Method to select, and the value interval of k is set to [10,50], where k is an integer.

[0151] After k is determined, the doctor's inquiry sentence and the patient's answer are concatenated as the vector of the inquiry sentence. This paper uses the Bag-of-Words model (BOW) to average and obtain the sentence vector. The word vector in BOW is based on the word vector obtained by FastText training.

[0152] If the Elbow method has no obvious elbow point, select a larger slope and a larger value of k.

[0153] The steps to determine k in k-means using the Elbow method include: 5.1. Set all the value sets of k. 5.2. For each k, calculate the Euclidean distance from each point in each cluster to the cluster center, and sum them up as WSSk. 5.3. Plot with WSSk as the y-axis and k as the x-axis. The turning point along the way is the desired k value.

[0154] After clustering, keywords are extracted from each cluster, and clusters whose keyword cosine similarity is greater than a threshold (set to 0.8) are merged.

[0155] For example, in Example 1, the query sentences in the consultation point group "cough" are clustered to obtain the corresponding clusters.

[0156] Cluster 1: "Do you have phlegm? What color is it?", "Do you have phlegm when you cough?"

[0157] Cluster 2: "How long have you been coughing?", "Hello, how long has this been going on?"

[0158] Cluster 3: "Any history of allergies?"

[0159] Cluster 4: "Do you have symptoms of fever, nasal congestion, runny nose, sore throat, headache, and limb aches?"

[0160] Cluster 5: "Hmm, have you had any recent contact with people in area A?"

[0161] Cluster 6: "Well, have you used any medicine?"

[0162] For example, in Example 2, the query sentence about the consulting point "test drive" is clustered:

[0163] Cluster 1: "Excuse me, do you have a driver's license?", "Do you have a driver's license?"

[0164] Cluster 2: "When will you come for a test drive?", "When do you plan to come?"

[0165] Cluster 3: "Which one do you want to try?", "Do you want to try both?"

[0166] Cluster 4: "Do you drive a lot?"

[0167] Clustering methods include, but are not limited to, k-means clustering, density clustering, hierarchical clustering, and their variants.

[0168] Sentence vector representation methods in clustering methods, including but not limited to BOW, TF-IDF, sentenc-transformer and other methods.

[0169] 6. Perform semantic abstraction on the clusters obtained by clustering

[0170] Keyword extraction is performed on all sentences in a cluster. The present invention uses the keyword extraction method of Textrank to obtain all keywords of a cluster.

[0171] Then, synonymous keywords are merged, the symptom words of the cluster are removed, and the keywords are sorted by their weights. The top 1 keyword is taken as the semantic abstract word of the cluster.

[0172] Keyword extraction methods include but are not limited to textrank, tf-idf, deep neural network (pointer network) keyword extraction, etc. Synonym methods include but are not limited to synonym dictionaries, word vector cosine similarity, etc.

[0173] For example, in Example 1, after removing the symptom words of the cluster and sorting them, the cluster keywords are as follows:

[0174] Cluster 1: Phlegm, color

[0175] Cluster 2: How long

[0176] Cluster 3: History of allergies

[0177] Cluster 4: Fever, runny nose, sore throat, headache, and limb pain

[0178] Cluster 5: Contact history, Region A

[0179] Cluster 6: Medicine

[0180] The final cluster keywords obtained in Example 2 are as follows:

[0181] Cluster 1: Driver’s license

[0182] Cluster 2: When

[0183] Cluster 3:

[0184] Cluster 4: Driving

[0185] 7. Obtain the elements of the consultation corresponding to the symptoms

[0186] According to the result of semantic abstraction of the keywords of the cluster, the top 1 word is selected as the consulting element of the consulting point to which the cluster belongs.

[0187] As in Sample 1, the consultation elements for “cough” include: sputum, duration, allergy history, fever, contact history, and medicine.

[0188] As shown in Sample 2, the consultation elements of “test drive” include: driver’s license, when, and drive.

[0189] The above-mentioned other patents have different solutions from the present invention. The present invention mainly focuses on the extraction of consulting elements.

[0190] The detailed process of the present invention has been described in the previous question.

[0191] The innovation of the present invention is to propose a method for extracting consulting elements from consulting dialogue data. The method follows certain steps, and the main steps protected are:

[0192] 1.1. Cluster the extracted query sentences and perform semantic abstraction.

[0193] 1.2. In clustering, the method of representing the query sentence vector is to concatenate the query sentence and the answer as a sentence to obtain the query sentence vector. This is because sentences in dialogue data often have omitted components, so a certain context needs to be concatenated.

[0194] 1.3. The method of cluster semantic abstraction is to extract cluster keywords and combine the synonyms to obtain the first word as the cluster semantic word.

[0195] The present invention focuses on the extraction of consulting elements, and the ultimate task is to let AI learn to ask questions. Therefore, patents related to generating questions, extracting question points, question parameters, question core, etc. can be verified for infringement.

[0196] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of combined actions, but those skilled in the art should be aware that the present invention is not limited by the described order of actions, because according to the present invention, certain 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 present invention. 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, please refer to the relevant description of other embodiments.

[0197] In some embodiments, an embodiment of the present invention provides a non-volatile computer-readable storage medium, in which one or more programs including execution instructions are stored. The execution instructions can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to execute any of the above-mentioned consulting element extraction methods for consulting conversations of the present invention.

[0198] In some embodiments, an embodiment of the present invention further provides a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes any one of the above-mentioned consulting element extraction methods for consulting conversations.

[0199] In some embodiments, an embodiment of the present invention also provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute a consulting element extraction method for a consulting conversation.

[0200] Figure 3 FIG. 1 is a schematic diagram of the hardware structure of an electronic device for executing a method for extracting consulting elements from consulting dialogues provided by another embodiment of the present application. Figure 3 As shown, the device includes:

[0201] One or more processors 310 and memory 320, Figure 3 A processor 310 is taken as an example.

[0202] The device for executing the method for extracting consulting elements in consulting dialogues may further include: an input device 330 and an output device 340 .

[0203] The processor 310, the memory 320, the input device 330 and the output device 340 may be connected via a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0204] The memory 320, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules corresponding to the method for extracting consulting elements from consulting conversations in the embodiment of the present application. The processor 310 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 320, that is, implements the method for extracting consulting elements from consulting conversations in the above method embodiment.

[0205] The memory 320 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created according to the use of a consulting element extraction device for consulting conversations, etc. In addition, the memory 320 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 320 may optionally include a memory remotely disposed relative to the processor 310, and these remote memories may be connected to the consulting element extraction device for consulting conversations via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0206] The input device 330 can receive input digital or character information and generate signals related to user settings and function control of the consultation element extraction device of the consultation dialogue. The output device 340 can include display devices such as display screens.

[0207] The one or more modules are stored in the memory 320, and when executed by the one or more processors 310, the consulting element extraction method for consulting dialogue in any of the above method embodiments is executed.

[0208] The above product can execute the extraction method provided in the embodiment of the present application, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the extraction method provided in the embodiment of the present application.

[0209] The electronic devices of the embodiments of the present application exist in various forms, including but not limited to:

[0210] (1) Mobile communication equipment: This type of equipment is characterized by having mobile communication functions and its main purpose is to provide voice and data communications. This type of terminal includes: smart phones, multimedia phones, functional phones, and low-end phones.

[0211] (2) Ultra-mobile personal computer devices: These devices belong to the category of personal computers, have computing and processing functions, and generally also have mobile Internet access features. These terminals include: PDA, MID and UMPC devices, etc.

[0212] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players, handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0213] (4) Other onboard electronic devices with data interaction functions, such as on-board devices installed in vehicles.

[0214] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0215] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the extraction method described in each embodiment or some parts of the embodiment.

[0216] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for extracting consulting elements in consulting dialogues. include: Get consulting conversation data; The consultation dialogue data includes a subject field and dialogue content data; The conversation content data includes: question sentences and answer sentences; obtaining topic groups according to the topic fields and the application fields to which they belong; Identifying consulting key words from the conversation content data through a set segment-based deep neural network; dividing the conversation content data into multiple groups of conversation content data according to the consulting key words and set categories; extracting question sentences from the multiple groups of conversation content data; Clustering the question sentences in each group of dialogue content through the k-means clustering algorithm, including: obtaining the k value in the k-means clustering algorithm through the Elbow Method, the value range of k is set to [10,50], the k value is an integer, if the Elbow method has no obvious elbow point, then selecting a value whose slope is greater than a set slope value and k is greater than a set K value, the set slope is the average value of the regular slope, the set K value is the average value of the regular K, obtaining the key sentences in each cluster; extracting cluster keywords according to the key sentences in each cluster; obtaining the consulting elements according to the cluster keywords.

2. The extraction method according to claim 1, in, The consulting key words include: proper nouns and non-entity words belonging to the central words of the sentence.

3. The extraction method according to claim 1, in, The step of clustering the question sentences in each group of dialogue content by the k-means clustering algorithm includes: training the word vector of each word in the question sentence by FastText; the question sentence includes a question sentence and its corresponding answer sentence; obtaining a sentence vector by averaging the word vectors by the BOW bag of words model; clustering the sentence vectors of the question sentences in each group of dialogue content by the k-means clustering algorithm, and obtaining the key sentences in each cluster.

4. The extraction method according to claim 1, in, The step of extracting cluster keywords according to the key sentences in each cluster includes merging cluster keywords whose keyword cosine similarity is greater than a set threshold; the set threshold is 0.

8.

5. The extraction method according to claim 1, in, The step of extracting cluster keywords according to the key sentences in each cluster also includes: Merge synonymous cluster keywords; if the cluster keyword is a medical inquiry word, remove the symptom word in the cluster keyword; the step of obtaining the consultation element based on the cluster keyword also includes sorting the cluster keywords from large to small according to their weights, obtaining the cluster keyword corresponding to the maximum weight, and using the cluster keyword as the consultation element word.

6. An electronic device, wherein include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the method described in any one of claims 1 to 5.

7. A storage medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images