CN113792126A - Conversation method and system for intelligent customer service for school enrollment - Google Patents
Conversation method and system for intelligent customer service for school enrollment Download PDFInfo
- Publication number
- CN113792126A CN113792126A CN202111069244.1A CN202111069244A CN113792126A CN 113792126 A CN113792126 A CN 113792126A CN 202111069244 A CN202111069244 A CN 202111069244A CN 113792126 A CN113792126 A CN 113792126A
- Authority
- CN
- China
- Prior art keywords
- main body
- query
- condition
- preset
- matching
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3343—Query execution using phonetics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/216—Parsing using statistical methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/237—Lexical tools
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L13/00—Speech synthesis; Text to speech systems
- G10L13/02—Methods for producing synthetic speech; Speech synthesisers
- G10L13/027—Concept to speech synthesisers; Generation of natural phrases from machine-based concepts
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/26—Speech to text systems
Abstract
The invention discloses a dialogue method and a dialogue system for intelligent customer service for school enrollment, wherein the method comprises the following steps: step one, receiving voice information of a user call and converting the voice information into an initial Chinese sentence through an ASR technology; selecting a corresponding response mode according to the obtained initial statement; step three, generating a corresponding reply text according to the obtained initial sentence and the selected response mode; and step four, carrying out conversation recording according to the acquired initial sentence and the acquired reply text, converting the reply text into voice through a TTS technology, and making a reply for the user. The method can quickly and accurately identify the incoming call consultation intention of the user, can deduce the possible incoming call intention of the user according to the hot spot problem counted in real time even if the abnormality occurs in the intention understanding link, improves the conversation fluency of the intelligent customer service, can truly solve a great part of incoming call consultation problems, and truly lightens the enrollment and handling pressure of schools.
Description
Technical Field
The invention relates to the technical field of voice interaction, in particular to a dialogue method and system for intelligent customer service for school enrollment.
Background
With the continuous advance of the comprehensive innovation of examination enrollment systems of advanced schools, in the affairs of examinee's volunteering and enrollment work, a system is needed to facilitate the communication of simple problems between examinee users and teacher customer service, so that the examinee can know general problems, the stress of school enrollment is relieved better, and a series of intelligent customer service platforms for school enrollment are brought to the students.
At present, the existing intelligent customer service main bodies are divided into two types, one type is a main body keyword matching reply form, and the other type is a reply generated by training a dialogue model through a neural network such as RNN (neural network), LSTM (least squares TM) and the like, wherein the keyword matching form is simple in structure, low in matching rate and too dependent on the richness of a keyword lexicon; although the semantic understanding ability of the intelligent customer service built through the neural network is strong, the intelligent customer service also depends on the accuracy and the richness of the trained corpus, and a model with strong semantic understanding ability can be trained only by needing a large amount of high-quality corpora, so that the workload of model building is large, the time consumption of model optimization is long, and under the business scene of school enrollment, the linguistic data are generally few, and the consultation topics are not wide, so that the cost performance of the scheme is not high, and the current intelligent customer service cannot well predict the intention when the semantics cannot be accurately identified.
Disclosure of Invention
The invention aims to solve the problems that in the prior art, the intelligent customer service has low service understanding capability, so that the conversation is not smooth, and the query efficiency is low due to troublesome operation and easy error during information query.
In order to achieve the purpose, the invention adopts the following technical scheme:
a conversation method for intelligent customer service for school enrollment, comprising the steps of:
step one, receiving voice information of a user call and converting the voice information into an initial Chinese sentence through an ASR technology;
selecting a corresponding response mode according to the obtained initial statement;
step three, generating a corresponding reply text according to the obtained initial sentence and the selected response mode;
and step four, carrying out conversation recording according to the acquired initial sentence and the acquired reply text, converting the reply text into voice through a TTS technology, and making a reply for the user.
Preferably, in the second step, selecting a corresponding response mode according to the obtained initial sentence specifically includes:
if the obtained initial sentence does not contain the preset query key words in the condition library, selecting a matching mode to respond;
and if the obtained initial sentence contains the preset query key words in the condition library, selecting a query mode to respond.
Preferably, when the matching mode is selected for response, the method specifically comprises the following steps:
s1, extracting a main body and content of the obtained initial sentence by adopting a regular matching method, respectively matching the extracted main body and content with a preset main body and preset content, adding one to a corresponding weight value every time the main body and content are matched with the preset main body and the preset content, and calculating a main body weight dictionary and a content weight dictionary of the initial sentence;
s2, extracting a main body and content of the obtained initial sentence by adopting a bidirectional word segmentation method, respectively matching the extracted main body and content with a preset main body and preset content, and adding one to a corresponding weight value every time the main body and the content are matched with the preset main body and the preset content, so as to calculate a main body weight dictionary and a content weight dictionary of the initial sentence;
s3, extracting a main body and contents of the obtained initial sentence by adopting a main body related word matching method, respectively matching the extracted main body and contents with a preset main body and preset contents, and adding one to a corresponding weight value every time the main body and contents are matched with the preset main body and the preset contents, so as to calculate a main body weight dictionary and a content weight dictionary of the initial sentence;
and S4, integrating the main body weight dictionary and the content weight dictionary calculated in S1, S2 and S3, if the integrated main body weight dictionary and the content weight dictionary are both empty, extracting the most frequent question from the hot questions as a reply text, and if the integrated main body weight dictionary and the content weight dictionary are not both empty, taking a group of main bodies and contents with the highest weight as a semantic understanding result and matching the reply text.
Preferably, the step of using the group of subjects and contents with the highest weight as semantic understanding results and matching reply texts specifically includes:
if a group of main bodies and contents with the highest weight do not contain preset query keywords in the condition library, matching a reply text from the dialogistic library;
and if a group of main bodies and contents with the highest weight contain preset query keywords in the condition library, selecting a query mode to respond and acquiring a reply text.
Preferably, when the query mode is selected for responding, the method specifically includes:
if the query conditions are complete, the query result is used as a reply text;
if the query condition is missing, sequentially acquiring the missing condition query text and querying the user, waiting for the user to reply, if the missing condition is successfully acquired, acquiring the reply text, otherwise, giving up the query.
The invention also comprises a dialogue system for intelligent customer service for school enrollment, which has two modes of a matching mode and a query mode, and the system comprises the following modules:
the voice analysis module is used for converting the voice of the user call into an initial Chinese sentence through an ASR technology;
the condition extraction module is used for sequentially searching all conditions in the condition library according to the initial sentence until a matching condition is extracted, continuing to be processed by the condition acquisition module, if the matching condition cannot be matched, issuing a question to the user again to acquire a required query condition, and if the required condition cannot be acquired, determining that the query fails and giving up the query;
the context word segmentation module is used for extracting a main body and contents of the initial sentence by sequentially utilizing a regular matching method, a bidirectional word segmentation method and a main body associated word matching method, and if the word segmentation is successful, the obtained main body and contents are used as output and transmitted to the intention extraction module; if the word segmentation fails, the problem recommendation module carries out the next processing;
the question recommending module is used for extracting a recommended question to inquire a user;
the intention extraction module is used for judging the consultation intention of the user;
the information query module is used for querying information;
the condition acquisition module is used for inquiring a preset condition acquisition dictionary to obtain a condition acquisition statement when the condition required to be inquired is absent, and outputting the condition acquisition statement to the voice synthesis module for the next processing;
and the voice synthesis module synthesizes the input dialect sentences, the condition acquisition sentences and the query result sentences into voice through a TTS technology, makes voice response to the user, and responds again for the user to enter the next cycle.
Preferably, the condition extraction module is only available when the system is in the query mode.
Preferably, the context segmentation module is only available when the system is in a matching mode.
Compared with the prior art, the invention provides a conversation method and a conversation system for intelligent customer service for school enrollment, which have the following beneficial effects:
1. the method extracts possible main bodies and contents in sentences through three-wheel matching of a regular matching method, a two-way matching word segmentation method and a context association matching method to obtain the weight of each main body and content, uses a group of main bodies and contents with the highest weight as semantic understanding results, and has high reliability, so that the comprehension of the obscure problems is realized in a convenient and fast mode, the service understanding capability of the intelligent customer service is improved, meanwhile, the frequency and the sequence of the identified main bodies and contents are counted, and under the condition that the intention of a user cannot be obtained, the user is actively inquired whether to consult related problems according to the result of reverse statistics, and the conversation fluency of the intelligent customer service is improved;
2. when the method and the device are used for inquiring information, only voice conversation is needed, the operation is simple, and the experience of a user in inquiring information is improved.
According to the invention, through mutual cooperation among all components, the incoming call consultation intention of students or parents can be quickly and accurately identified, even if abnormality occurs in the idea understanding link, the possible intention of the students or the parents for incoming calls can be inferred according to the hot spot problem counted in real time, the conversation fluency of intelligent customer service is improved, the incoming call experience is improved to a greater extent, a great part of incoming call consultation problems can be truly solved, and the enrollment pressure of schools is really relieved.
Drawings
Fig. 1 is a schematic flow chart of a conversation method for intelligent customer service for school enrollment according to the present invention;
fig. 2 is a schematic structural diagram of a dialogue system for intelligent customer service for school students according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
In the description of the present invention, it is to be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", and the like, indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, are merely for convenience in describing the present invention and simplifying the description, and do not indicate or imply that the device or element being referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, should not be construed as limiting the present invention.
Referring to fig. 1, a conversation method for intelligent customer service for school recruitment, comprising the steps of:
step one, receiving voice information of a user call and converting the voice information into an initial Chinese sentence through an ASR technology;
step two, selecting a corresponding response mode according to the obtained initial statement, and specifically comprising the following steps:
if the obtained initial sentence does not contain the preset query keyword in the condition library, selecting a matching mode for responding, and if the obtained initial sentence contains the preset query keyword in the condition library, selecting a query mode for responding;
step three, generating a corresponding reply text according to the obtained initial sentence and the selected reply mode, and when the matching mode is selected for replying, specifically comprising the following steps:
s1, extracting a main body and content of the obtained initial sentence by adopting a regular matching method, respectively matching the extracted main body and content with a preset main body and preset content, adding one to a corresponding weight value every time the main body and content are matched with the preset main body and the preset content, and calculating a main body weight dictionary and a content weight dictionary of the initial sentence;
s2, extracting a main body and content of the obtained initial sentence by adopting a bidirectional word segmentation method, respectively matching the extracted main body and content with a preset main body and preset content, and adding one to a corresponding weight value every time the main body and the content are matched with the preset main body and the preset content, so as to calculate a main body weight dictionary and a content weight dictionary of the initial sentence;
s3, extracting a main body and contents of the obtained initial sentence by adopting a main body related word matching method, respectively matching the extracted main body and contents with a preset main body and preset contents, and adding one to a corresponding weight value every time the main body and contents are matched with the preset main body and the preset contents, so as to calculate a main body weight dictionary and a content weight dictionary of the initial sentence;
s4, integrating the subject weight dictionary and the content weight dictionary calculated in S1, S2 and S3, if the integrated subject weight dictionary and the content weight dictionary are both empty, extracting the most frequent question from the hot questions as a reply text, and if the integrated subject weight dictionary and the content weight dictionary are not both empty, taking a group of subjects and contents with the highest weight as a semantic understanding result and matching the reply text, specifically including:
if a group of main bodies and contents with the highest weight do not contain preset query keywords in the condition library, matching a reply text from the dialogistic library;
if a group of main bodies and contents with the highest weight contain preset query keywords in the condition library, selecting a query mode to respond and acquiring a reply text, wherein the method specifically comprises the following steps:
if the query conditions are complete, the query result is used as a reply text;
if the query condition is missing, sequentially acquiring the missing condition query text and querying the user, waiting for the user to reply, if the missing condition is successfully acquired, acquiring the reply text, otherwise, giving up the query;
and step four, carrying out conversation recording according to the reply text, converting the reply text into voice through a TTS technology, and making a reply for the user.
In this embodiment, it is assumed that the speech text converted by the ASR technique is as follows:
q1 is you good, do i want to ask a few people at school dormitory?
Q2 is how many people he wants to ask about the school year A professional?
Taking Q1 as an example, the specific process analysis is as follows:
q1 is the original sentence obtained by ASR technique analyzed by the voice analysis module in the mentioned system, and is recorded as S1, and the current system state is the default matching mode;
2. the sentence S1 is transmitted into the context word segmentation module in the matching mode, and a preset main dictionary and a preset content dictionary are loaded;
3. the sentence S1 is to firstly obtain a main body weight dictionary and a content weight dictionary by a regular matching method, now, assuming that the preset regular matching rules of the main body are 'school' and 'canteen', etc., every time matching is hit, the weight value of the main body is added with 1, the initial value is 0, then the main body weight dictionary of the sentence S1 is { 'school': 1, 'dormitory': 1} by the regular matching method, and is recorded as Z1; presetting content matching rules such as 'people' and 'money', adding 1 to the weight value of the content when the same matching is hit once, and setting the initial value to be 0, and obtaining a content weight dictionary of a sentence S1 as { 'several people': 1} through a regular matching method, wherein the dictionary is recorded as N1;
4. the sentence S1 is further obtained by a bidirectional word segmentation method to obtain a main body weight dictionary and a content weight dictionary, now, assuming that a main body word segmentation list [ "school", "dormitory", "dining room". ] is preset and is recorded as K1, and a content word segmentation list [ "school fee", "number of people", "air conditioner" ] is recorded as K2, the forward main body word segmentation matching is firstly carried out on S to obtain the weight dictionary { "school": 1}, and then the reverse main body word segmentation matching is carried out on the main body word segmentation to obtain the weight dictionary { "school": 1}, so the main body word segmentation weight dictionary is { "school": 1, "dormitory": 1}, and is recorded as Z2; performing bidirectional word segmentation on the content without matching items, so that the content word segmentation weight dictionary is empty and is recorded as N2;
5. the sentence S1 obtains a main body weight dictionary and a content weight dictionary by a main body associated word matching method, wherein the main body associated dictionary structure is { main body p1: { content c1: [ "associated word w 1", "associated word w 2", "associated word w 3" ], content c2: [ "associated word w 4", "associated word w 5" ] }, main body p2: { content c3: [ "associated word w 6", "associated word w 7" ], content c4: [ "associated word w 8" ], };
now, assume that the preset subject associated dictionary D1 is:
the method comprises the steps of { "school" { "number of people", "number of people" ], "area" [ "large", "area", "square" ] }, and "dormitory" { "number of people" [ "one room", "live", "several people", "number of people" ], "environment" [ "air conditioner", "warm air", "individual guard", "upper and lower berths" ]. }, wherein the matching is performed by traversing the associated words, so that S1 is matched with "school-number of people-several people", "dormitory-number of people-live", so that a subject weight dictionary of { "school": 1, "dormitory": 2}, is written as Z3, and a content weight dictionary of { "number of people": 3}, which is written as N3;
6. the weight dictionary of the subject obtained by integrating the Z1, Z2 and Z3 is { "school": 3, "dormitory": 4}, so the possibility of the subject being "dormitory" is higher, the weight dictionary of the content obtained by integrating the N1, N2 and N3 is { "several persons": 1, "number of persons": 3}, so the possibility of the content being "number of persons" is higher, the comprehensive obtained S1 records the word segmentation result as "dormitory-number" as R1, if each round of matching is not hit, the weight dictionary is empty, and the word segmentation result is also empty and is recorded as R2;
7. after word segmentation is completed, if word segmentation is successful, transmitting a result R1 into a consultation statistical module, recording, counting the frequency of questioning of contents under each subject, storing the word segmentation result R1 into a word technique matching module according to the frequency reverse arrangement, if word segmentation fails, extracting a problem q1 with the highest frequency from the hot problems, and in an initial state, presetting a hot problem q, and transmitting q1 into a voice synthesis module;
8. after the R1 is introduced into the topic matching module, it will match the answer from the dialogies library, now assume that there is a dialogies dictionary: the method comprises the following steps of { "dormitory-number of people" { "res": i am for 6 people in a male dormitory, 4 people in a female dormitory, flag ": 0}, query-score {" res ": and flag": 1} }, wherein "res" is a reply sentence recorded as s, flag "is a system mode flag bit, default 0 indicates a matching mode, matched s is transmitted into a voice synthesis module, and" flag "1 indicates a query mode, and a word segmentation result R1 is transmitted into an information query module;
9. when R1 is transmitted into the information query module, the system compares whether each condition required by the query information is missing, and the query result can be obtained when the conditions are complete, and then the query result is transmitted into the speech synthesis module, otherwise, the missing condition query text is sequentially obtained and transmitted into the speech synthesis module; for example, in the original sentence Q2, it is assumed that the word segmentation result R3 obtained by the context word segmentation module is "query-a professional score", and then the system mode flag "obtained by the word matching module is 1, that is, the system enters the query mode, the missing condition" examinee province "is obtained by the information query module, and it is written as m1 and transmitted to the condition acquisition module;
10. after the condition acquisition module receives the missing condition m1 sent by the information receiving module, the condition acquisition module queries a missing condition query text, and assumes that a text "ask what province is you are? ", denoted t1, passed into the speech synthesis module;
11. after the t1 is transmitted into the speech synthesis module, the speech is stored in a dialogue record, and then a TTS technology is used for synthesizing a speech answer;
12. after the user replies, the new reply sentence enters the voice analysis module again, if the current user replies that the answer is 'I is an examinee in Anhui', the answer is recorded as S2, the condition matching module matches the missing condition, if the matching is unsuccessful, the last dialog is rescued from the dialog record, and if the query is still unsuccessful again, the query is abandoned, and the user is informed that the query is failed. If matching is successful, obtaining a missing condition m1 as 'Anhui', then transmitting the R3 and m1 into an information query module, obtaining a query result if the conditions are complete, and then transmitting the query result Rs1 into a voice synthesis module; if the query conditions are still missing, repeating the steps 9, 10, 11 and 12 to obtain the missing conditions in sequence until the conditions are complete to obtain a query result or abandon the query due to matching failure;
13. when the system is in a matching mode, assuming that the R1 obtains a reply text R1 in the dialect library, R1 is transmitted to the speech synthesis module;
14. after receiving various texts, the speech synthesis module firstly records the conversation, repeats the previous conversation when the abnormality occurs, and then converts the texts into speech by a TTS technology to make a response for a user;
15. after making the response via the speech synthesis module, the system will go to the next cycle for the user to make a new response.
As shown in fig. 2, the present invention further includes a dialogue system for intelligent customer service for school students, which has two modes, a matching mode and a query mode, and the system includes the following modules:
the voice analysis module is used for converting the voice of the user call into an initial Chinese sentence through an ASR technology;
the condition extraction module is used for sequentially searching all conditions in the condition library according to the initial sentence until a matching condition is extracted, continuing to be processed by the condition acquisition module, if the matching condition cannot be matched, issuing a question to the user again to acquire a required query condition, and if the required condition cannot be acquired, determining that the query fails and giving up the query;
the context word segmentation module is used for extracting a main body and contents of the obtained initial sentence by sequentially utilizing a regular matching method, a bidirectional word segmentation method and a main body associated word matching method, and if the word segmentation is successful, the obtained main body and contents are used as output and transmitted to the intention extraction module; if the word segmentation fails, the problem recommendation module carries out the next processing;
the question recommending module extracts a recommendation question from the question recommending module to ask the user if the sentence word segmentation fails, and the working mode is as follows: recording the next question of each question by using a dictionary and sequencing according to frequency reverse, when word segmentation fails, inquiring the dictionary to obtain the next question which is possibly concerned with the highest frequency, then submitting the next question to a voice synthesis module to synthesize voice, and actively inquiring whether a user consults the question;
the intention extraction module is used for obtaining a main body and contents after the original sentence is successfully participated by the word segmentation module, then obtaining the user consultation intention by the intention dictionary, if the intention is information such as query scores, enrollment number and the like, setting the system mode as a query mode, and if the intention is information such as query scores, enrollment number and the like, setting the system mode as a matching mode, and then handing the system mode to the word technology matching module for next processing;
the information query module is used for outputting the missing conditions X2 and X3 to the condition acquisition module for further processing if the queried problem needs to know the conditions X1, X2 and X3 and only X2 is known currently when the system is in a query mode, and outputting the query result to the voice synthesis module and setting the system mode as a matching mode when the required conditions are known;
the condition acquisition module is used for inquiring a preset condition acquisition dictionary to obtain a condition acquisition statement when the system is in an inquiry mode and the required inquiry condition is absent, and outputting the condition acquisition statement to the voice synthesis module for further processing;
and the voice synthesis module synthesizes the input dialect sentences, the condition acquisition sentences and the query result sentences into voice through a TTS technology, makes voice response to the user, and responds again for the user to enter the next cycle.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.
Claims (8)
1. A conversation method of intelligent customer service for school enrollment, characterized by comprising the steps of:
step one, receiving voice information of a user call and converting the voice information into an initial Chinese sentence through an ASR technology;
selecting a corresponding response mode according to the obtained initial statement;
step three, generating a corresponding reply text according to the obtained initial sentence and the selected response mode;
and step four, carrying out conversation recording according to the acquired initial sentence and the acquired reply text, converting the reply text into voice through a TTS technology, and making a reply for the user.
2. The conversation method for intelligent customer service for school recruitment according to claim 1, wherein in the second step, the corresponding response mode is selected according to the obtained initial sentence, and specifically comprises:
if the obtained initial sentence does not contain the preset query key words in the condition library, selecting a matching mode to respond;
and if the obtained initial sentence contains the preset query key words in the condition library, selecting a query mode to respond.
3. The conversation method for intelligent customer service for school students according to claim 2, wherein when a matching mode is selected for response, the conversation method specifically comprises the following steps:
s1, extracting a main body and content of the obtained initial sentence by adopting a regular matching method, respectively matching the extracted main body and content with a preset main body and preset content, adding one to a corresponding weight value every time the main body and content are matched with the preset main body and the preset content, and calculating a main body weight dictionary and a content weight dictionary of the initial sentence;
s2, extracting a main body and content of the obtained initial sentence by adopting a bidirectional word segmentation method, respectively matching the extracted main body and content with a preset main body and preset content, and adding one to a corresponding weight value every time the main body and the content are matched with the preset main body and the preset content, so as to calculate a main body weight dictionary and a content weight dictionary of the initial sentence;
s3, extracting a main body and contents of the obtained initial sentence by adopting a main body related word matching method, respectively matching the extracted main body and contents with a preset main body and preset contents, and adding one to a corresponding weight value every time the main body and contents are matched with the preset main body and the preset contents, so as to calculate a main body weight dictionary and a content weight dictionary of the initial sentence;
and S4, integrating the main body weight dictionary and the content weight dictionary calculated in S1, S2 and S3, if the integrated main body weight dictionary and the content weight dictionary are both empty, extracting the most frequent question from the hot questions as a reply text, and if the integrated main body weight dictionary and the content weight dictionary are not both empty, taking a group of main bodies and contents with the highest weight as a semantic understanding result and matching the reply text.
4. The conversation method for intelligent customer service for school enrollment as claimed in claim 3, wherein said using the group of subjects and contents with highest weight as semantic understanding result and matching reply text specifically comprises:
if a group of main bodies and contents with the highest weight do not contain preset query keywords in the condition library, matching a reply text from the dialogistic library;
and if a group of main bodies and contents with the highest weight contain preset query keywords in the condition library, selecting a query mode to respond and acquiring a reply text.
5. The conversation method for intelligent customer service for school students according to claim 2 or 4, wherein when the query mode is selected for response, the conversation method specifically comprises:
if the query conditions are complete, the query result is used as a reply text;
if the query condition is missing, sequentially acquiring the missing condition query text and querying the user, waiting for the user to reply, if the missing condition is successfully acquired, acquiring the reply text, otherwise, giving up the query.
6. A dialogue system for intelligent customer service for school enrollment is characterized in that the system has two modes of a matching mode and a query mode, and the system comprises the following modules:
the voice analysis module is used for converting the voice of the user call into an initial Chinese sentence through an ASR technology;
the condition extraction module is used for sequentially searching all conditions in the condition library according to the initial sentence until a matching condition is extracted, continuing to be processed by the condition acquisition module, if the matching condition cannot be matched, issuing a question to the user again to acquire a required query condition, and if the required condition cannot be acquired, determining that the query fails and giving up the query;
the context word segmentation module is used for extracting a main body and contents of the initial sentence by sequentially utilizing a regular matching method, a bidirectional word segmentation method and a main body associated word matching method, and if the word segmentation is successful, the obtained main body and contents are used as output and transmitted to the intention extraction module; if the word segmentation fails, the problem recommendation module carries out the next processing;
the question recommending module is used for extracting a recommended question to inquire a user;
the intention extraction module is used for judging the consultation intention of the user;
the information query module is used for querying information;
the condition acquisition module is used for inquiring a preset condition acquisition dictionary to obtain a condition acquisition statement when the condition required to be inquired is absent, and outputting the condition acquisition statement to the voice synthesis module for the next processing;
and the voice synthesis module synthesizes the input dialect sentences, the condition acquisition sentences and the query result sentences into voice through a TTS technology, makes voice response to the user, and responds again for the user to enter the next cycle.
7. The dialog system for an intelligent customer service for school recruitment according to claim 1, wherein the condition extraction module is only available when the system is in query mode.
8. The dialog system for intelligent customer service for school students according to claim 1, wherein said context segmentation module is only available when the system is in a matching mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111069244.1A CN113792126A (en) | 2021-09-13 | 2021-09-13 | Conversation method and system for intelligent customer service for school enrollment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111069244.1A CN113792126A (en) | 2021-09-13 | 2021-09-13 | Conversation method and system for intelligent customer service for school enrollment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113792126A true CN113792126A (en) | 2021-12-14 |
Family
ID=78880073
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111069244.1A Pending CN113792126A (en) | 2021-09-13 | 2021-09-13 | Conversation method and system for intelligent customer service for school enrollment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113792126A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116757203A (en) * | 2023-08-16 | 2023-09-15 | 杭州北冥星火科技有限公司 | Natural language matching method, device, computer equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107315766A (en) * | 2017-05-16 | 2017-11-03 | 广东电网有限责任公司江门供电局 | A kind of voice response method and its device for gathering intelligence and artificial question and answer |
| CN110019739A (en) * | 2017-11-30 | 2019-07-16 | 上海智臻智能网络科技股份有限公司 | Answering method and device, computer equipment and storage medium based on necessary condition |
| CN112651750A (en) * | 2020-09-11 | 2021-04-13 | 安徽九广全景智慧科技有限公司 | Intelligent customer service system for college student recruitment telephone |
-
2021
- 2021-09-13 CN CN202111069244.1A patent/CN113792126A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107315766A (en) * | 2017-05-16 | 2017-11-03 | 广东电网有限责任公司江门供电局 | A kind of voice response method and its device for gathering intelligence and artificial question and answer |
| CN110019739A (en) * | 2017-11-30 | 2019-07-16 | 上海智臻智能网络科技股份有限公司 | Answering method and device, computer equipment and storage medium based on necessary condition |
| CN112651750A (en) * | 2020-09-11 | 2021-04-13 | 安徽九广全景智慧科技有限公司 | Intelligent customer service system for college student recruitment telephone |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116757203A (en) * | 2023-08-16 | 2023-09-15 | 杭州北冥星火科技有限公司 | Natural language matching method, device, computer equipment and storage medium |
| CN116757203B (en) * | 2023-08-16 | 2023-11-10 | 杭州北冥星火科技有限公司 | Natural language matching method, device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP3472194B2 (en) | Automatic response method and device, and medium recording the program | |
| US7216080B2 (en) | Natural-language voice-activated personal assistant | |
| CN101566998B (en) | Chinese question-answering system based on neural network | |
| CN101309327B (en) | Sound chat system, information processing device, speech recognition and key words detection | |
| WO2008128423A1 (en) | An intelligent dialog system and a method for realization thereof | |
| CN108345690A (en) | Intelligent answer method and system | |
| CN110164447B (en) | Spoken language scoring method and device | |
| CN109543020B (en) | Query processing method and system | |
| US20100114684A1 (en) | Chat rooms search engine queryer | |
| CN108536807B (en) | Information processing method and device | |
| CN111477231B (en) | Man-machine interaction method, device and storage medium | |
| US20070136067A1 (en) | Audio dialogue system and voice browsing method | |
| CN110266900A (en) | Recognition methods, device and the customer service system that client is intended to | |
| CN113569023A (en) | Chinese medicine question-answering system and method based on knowledge graph | |
| Lickley | Dialogue moves and disfluency rates | |
| CN107562911A (en) | More wheel interaction probabilistic model training methods and auto-answer method | |
| CN113792126A (en) | Conversation method and system for intelligent customer service for school enrollment | |
| CN112287082A (en) | Data processing method, device, equipment and storage medium combining RPA and AI | |
| Nishimura et al. | Automatic n-gram language model creation from web resources | |
| Atef et al. | AQAD: 17,000+ arabic questions for machine comprehension of text | |
| CN110675292A (en) | Child language ability evaluation method based on artificial intelligence | |
| Galibert et al. | Ritel: an open-domain, human-computer dialog system. | |
| Fletcher et al. | LARSPing by numbers | |
| CN109086372A (en) | It is a kind of based on artificial intelligence application in the more wheel session methods and system of legal advice | |
| van Schooten et al. | Handling speech input in the ritel QA dialogue system. |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |