CN113239164B - Multi-round dialogue flow construction method and device, computer equipment and storage medium - Google Patents
Multi-round dialogue flow construction method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN113239164B CN113239164B CN202110520914.0A CN202110520914A CN113239164B CN 113239164 B CN113239164 B CN 113239164B CN 202110520914 A CN202110520914 A CN 202110520914A CN 113239164 B CN113239164 B CN 113239164B
- Authority
- CN
- China
- Prior art keywords
- question
- dialogue
- answer
- constructing
- user
- 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.)
- Active
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
- 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
- G06F16/35—Clustering; Classification
-
- 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
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/231—Hierarchical techniques, i.e. dividing or merging pattern sets so as to obtain a dendrogram
-
- 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
- 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
- G06F40/295—Named entity recognition
-
- 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
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The embodiment of the invention discloses a method and a device for constructing a multi-round dialogue flow, computer equipment and a storage medium. The method comprises the following steps: collecting dialogue corpus; word segmentation and recognition are carried out on the dialogue corpus so as to obtain a recognition result; constructing a question answer set according to the identification result; and constructing a dialogue flow tree by adopting a hierarchical clustering mode according to the answer set of the questions. By implementing the method of the embodiment of the invention, the on-line data can be effectively and automatically constructed into a higher-quality dialogue flow, and the conventional question-answer set is mined by adopting a question-answer pair mining mode.
Description
Technical Field
The invention relates to the technical field of intelligent question and answer, in particular to a method and a device for constructing a multi-round dialogue flow, computer equipment and a storage medium.
Background
In the intelligent dialogue system, the correct understanding and processing of questions are the key points of the whole dialogue interaction, and the key points are needed to be established on the basis of a huge knowledge base, the current knowledge base construction comprises entity vocabularies, entity relations, QA (question and answer, question Answering) pairs, similar questions, dialogue flows and the like, the method of combining manual auditing and machine statistics is basically adopted, the mining of the entity vocabularies, the entity relations and the questions is mainly focused, and the automatic mining construction of the dialogue flows is very limited in practical application.
The actual production environment has a huge amount of dialogue corpora, for example, in a banking system, centralized outbound or customer service agents are provided, massive dialogue corpora are collected through accumulation of accumulated sediment in daily and monthly periods, and the corpora have very strong bank field characteristics, but at present, no method is available for better utilizing dialogue knowledge corpora, automatically enriching and perfecting an intelligent dialogue system knowledge base and automatically constructing a dialogue flow with field characteristics.
Therefore, a new method is needed to be designed, so that on-line data can be effectively and automatically constructed into a higher-quality dialogue flow, and a conventional question-answer set is mined through a question-answer pair mining mode.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a multi-round dialogue flow construction method, a multi-round dialogue flow construction device, computer equipment and a storage medium.
In order to achieve the above purpose, the present invention adopts the following technical scheme: the multi-round dialogue flow construction method comprises the following steps:
collecting dialogue corpus;
word segmentation and recognition are carried out on the dialogue corpus so as to obtain a recognition result;
constructing a question answer set according to the identification result;
and constructing a dialogue flow tree by adopting a hierarchical clustering mode according to the answer set of the questions.
The further technical scheme is as follows: the collecting dialogue corpus includes:
acquiring a dialogue recording file;
performing voice recognition on the dialogue record file to obtain dialogue stream data;
capturing question-answer pairs through a crawler technology;
and integrating the dialogue stream data and the question-answer pairs to obtain dialogue corpus.
The further technical scheme is as follows: the word segmentation and recognition are carried out on the dialogue corpus to obtain a recognition result, which comprises the following steps:
and performing word segmentation and entity recognition on the dialogue corpus by using a Jieba word segmentation tool to obtain a recognition result.
The further technical scheme is as follows: the construction of the question answer set according to the recognition result comprises the following steps:
constructing a user problem set according to the identification result;
constructing a robot response set corresponding to the user problem set by adopting a cluster analysis mode;
the problem answer set comprises a user problem set and a robot answer set corresponding to the user problem set.
The further technical scheme is as follows: the method for constructing the robot response set corresponding to the user problem set by adopting the cluster analysis mode comprises the following steps:
and constructing a robot response set corresponding to the user problem set by adopting a Canopy clustering algorithm.
The further technical scheme is as follows: the constructing a robot response set corresponding to the user problem set by adopting a Canopy clustering algorithm comprises the following steps:
determining a first similarity threshold and a second similarity threshold;
initializing a user problem set and initializing a clustering result set;
randomly selecting a question and creating a cluster centered on the question;
traversing each answer in the user question set, and calculating the distance from the answer to each cluster;
classifying answers with the distance smaller than a first similarity threshold value into the clusters, deleting the answers with the user distance smaller than a second similarity threshold value, and determining the questions as new cluster center points when the distances from the answers corresponding to the questions to each cluster are larger than the first similarity threshold value;
and determining the cluster as a robot response set corresponding to the user problem set.
The further technical scheme is as follows: the construction of the dialogue flow tree by hierarchical clustering mode according to the question answer set comprises the following steps:
acquiring a robot question and an answer of a starting node;
obtaining answers of subsequent robots, and constructing clustering branches for each robot by adopting a Canopy clustering algorithm;
and acquiring response branch response nodes of the robot, and marking response nodes which belong to question nodes and have no answer.
The invention also provides a multi-round dialogue flow construction device, which comprises:
the corpus collection unit is used for collecting dialogue corpus;
the recognition unit is used for word segmentation and recognition of the dialogue corpus to obtain a recognition result;
the set construction unit is used for constructing a question answer set according to the recognition result;
and the flow tree construction unit is used for constructing a dialogue flow tree in a hierarchical clustering mode according to the answer set of the questions.
The invention also provides a computer device which comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the method when executing the computer program.
The present invention also provides a storage medium storing a computer program which, when executed by a processor, performs the above-described method.
Compared with the prior art, the invention has the beneficial effects that: according to the invention, the existing dialogue corpus is collected, word segmentation and recognition are carried out on the dialogue corpus, a question answer set is constructed on the basis, a dialogue flow tree is constructed on the basis of the question answer set in a hierarchical clustering mode, so that on-line data can be effectively and automatically constructed into a higher-quality dialogue flow, and a conventional question answer set is mined in a question answer pair mining mode.
The invention is further described below with reference to the drawings and specific embodiments.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is an application scenario schematic diagram of a multi-round dialog flow construction method according to an embodiment of the present invention;
FIG. 2 is a flow diagram of a multi-round dialogue flow construction method according to an embodiment of the present invention;
FIG. 3 is a schematic sub-flowchart of a multi-round dialog flow construction method according to an embodiment of the present invention;
FIG. 4 is a schematic sub-flowchart of a multi-round dialog flow construction method according to an embodiment of the present invention;
FIG. 5 is a schematic sub-flowchart of a multi-round dialog flow construction method according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a sub-flow of a multi-round dialog flow construction method according to an embodiment of the present invention;
FIG. 7 is a schematic flow chart of a Canopy clustering algorithm provided by an embodiment of the invention;
FIG. 8 is a schematic flow diagram of a dialogue flow tree construction according to an embodiment of the present invention;
FIG. 9 is a schematic block diagram of a multi-round dialog flow construction device according to an embodiment of the present invention;
FIG. 10 is a schematic block diagram of a corpus collection unit of a multi-round dialog flow construction device according to an embodiment of the present invention;
FIG. 11 is a schematic block diagram of a set building unit of the multi-round dialog flow building device provided by an embodiment of the present invention;
FIG. 12 is a schematic block diagram of a response set construction subunit of the multi-round dialog flow construction device provided by an embodiment of the present invention;
FIG. 13 is a schematic block diagram of a process tree construction unit of the multi-round dialog process construction device according to an embodiment of the present invention;
fig. 14 is a schematic block diagram of a computer device according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be understood that the terms "comprises" and "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It is also to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.
Referring to fig. 1 and fig. 2, fig. 1 is a schematic application scenario diagram of a multi-round dialog flow construction method according to an embodiment of the present invention. Fig. 2 is a schematic flow chart of a multi-round dialog flow construction method according to an embodiment of the present invention. The multi-round dialogue flow construction method is applied to a server, the server performs data interaction with a robot and a terminal, after dialogue corpus is obtained from the robot and the terminal, the dialogue corpus is identified, a question answer set is constructed, and a dialogue flow tree is constructed by the question answer set.
Fig. 2 is a flow diagram of a multi-round dialog flow construction method according to an embodiment of the present invention. As shown in fig. 2, the method includes the following steps S110 to S140.
S110, collecting dialogue corpus.
In this embodiment, the dialogue corpus refers to dialogue contents between a robot and a customer or between customer service and a customer in a banking system.
In one embodiment, referring to fig. 3, the step S110 may include steps S111 to S114.
S111, acquiring a dialogue record file.
In this embodiment, the dialogue recording file refers to a recording file of a phone dialogue of a manual call or customer service of a banking system, where the phone dialogue of the manual call or customer service of the banking system generally exists in a dialogue recording file manner.
S112, carrying out voice recognition on the dialogue record file to obtain dialogue stream data.
In this embodiment, the dialogue stream data is data for distinguishing information such as dialogue order and time between the two parties after the dialogue record file is subjected to speech recognition.
Specifically, a voice recognition technology is adopted to convert dialogue record into dialogue stream data in text form, and the dialogue sequence, time and other information of the two parties of the speaker are distinguished. The common methods of speech recognition technology are linguistic and acoustic based methods; a random model method; a method using an artificial neural network; analyzing the probability grammar; details of converting the dialogue record into dialogue stream data in text form by these four methods belong to the prior art and are not described here.
S113, capturing question-answer pairs through a crawler technology.
In this embodiment, the question-answer and comment data in the bank community type question-answer system can obtain the question-answer pair by the crawler specifically grasping out the message board of hundred degrees, knowing the hope and the bank homepage website.
S114, integrating the dialogue stream data and the question-answer pairs to obtain dialogue corpus.
In this embodiment, the dialogue data of the robot outbound system, which are the dialogues between the telephone robot and the clients, while there may be some shortages in the dialogue response of the telephone robot, the other party of the dialogues is a natural person, which can give reasonably correct feedback according to the current robot response, even including complaints or even emotional responses where the other party cannot understand his own response. This is also the question that the building knowledge needs to solve, i.e. the better answer to the customer questions.
After the corpus of questions and answers is arranged, the knowledge range of banking business can be basically covered, and the knowledge questions and answers comprise account passwords, property inquiry, credit cards, deposit financing, loan repayment, legal policies and the like. The following are examples of corpus data for a one-way conversation in a telephone outbound system:
And (3) a robot: you good, i are the customer manager of the bank, you are the good customers of my, there is a good deposit product worth paying attention to you, the demand for deposit is low, but the yield rate is similar to that of a large deposit bill, and you can draw at any time, i introduce you to go down the bar? (R131)
The user: i go to the bar on the first day. (Y3)
And (3) a robot: the Bohai fixed deposit is innovative deposit which is pushed out by the inventor, can be drawn in advance at any time, has the highest annual income ratio of 4.125%, ensures that you obtain high income without affecting urgent use of your funds, and ensures that a customer manager responsible for the Bohai fixed deposit contacts your later and introduces a bar in detail? (R134)
The user: that mani he comes back but me is now at the outside head. (Q32)
And (3) a robot: you can transact through an internet banking, a mobile phone banking or to a nearby website. (R156)
The user: the mobile phone bank of that mobile phone bank can also find, is it a bar? (Q29)
And (3) a robot: can be handled through a mobile phone bank. Do me let me get in touch with you without later going to give you a detailed description of what can be? (R157)
The user: row by row. (Y3)
And (3) a robot: good, thank you for support, i walk customers later to contact you via a conversation, please answer, get you happy, see-! (R135)
Converting according to the sound recording file, and marking conversation contents of the two parties of the conversation, for example Q, R, Y, Q is a problem; r is a response, Y is a positive response, and the corresponding relation between the questions and the answers can be obtained.
S120, word segmentation and recognition are carried out on the dialogue corpus so as to obtain a recognition result.
In this embodiment, the recognition result refers to the meaning of each word, such as adjective, in each dialogue content of the dialogue corpus.
Specifically, the dialog corpus is subjected to word segmentation and entity recognition by using a Jieba word segmentation tool so as to obtain a recognition result.
The words are basic units of the semantics, firstly, the open-source Jieba word segmentation tool is utilized to segment words and identify entities, and the word parts of nouns, verbs, adjectives, adverbs, auxiliary words and the like are mainly identified for each text in the dialogue corpus, and entity phrase words and the like are identified. Since individual words and phrases are converted into homophones or wrongly written words in telephone speech, word spelling information and the like need to be analyzed together when performing lexical analysis. In addition, potential ngram phrases are identified according to some rules.
The following is a word segmentation result example:
We/r this/r version/q '/wp Bo/j deposit/n'/wp is/v me line/n push/v/b deposit/n,/wp may/v advance/v pick up/v at any time/d advance/v,/wp maximum/a annual/v profitability/n may/v up to/v 4.125%/m,/wp lets/v you/r get/u simultaneous/n of/v high/a earn/n,/wp does not yet/d affect/v you/r funds/n urgent/v,/wp me/r let/v be responsible/v this block/v/u customer/n manager/n later/d contact/v you/r,/wp detailed/a introduction/v under/m bar/u? Wp
After splitting each word of each dialogue content, the meaning is divided, so that each dialogue content of the dialogue corpus can be identified.
S130, constructing a question answer set according to the recognition result.
In this embodiment, the answer set of questions refers to a set formed by one-to-one correspondence between questions and answers.
The historical dialogue corpus contains a large number of question-answer dialogues. The following examples contain question-answer pairs.
User (B): the mobile phone bank of that mobile phone bank can also find, is it a bar? (Q29)
Robot (a): can be handled through a mobile phone bank. Do me let me get in touch with you without later going to give you a detailed description of what can be? (R157)
In one embodiment, referring to fig. 4, the step S130 may include steps S131 to S132.
S131, constructing a user problem set according to the identification result.
In this embodiment, the user question set refers to all questions initiated by the client.
Specifically, a question classifier needs to be constructed to judge whether the user text in the recognition result is an effective question, and identify and screen out the potential effective questions in all dialogue corpora. The training corpus of the question recognition classifier is arranged in a mode of combining rules and manual inspection, and the rules for judging the questions mainly judge whether the questions contain question marks, whether the questions contain query words, whether the questions contain verb words, whether the specific rule sentence patterns are met or not. Classification category y= {0,1}, each user question X is converted into a question word based TF-IDF (common weighting technique for information retrieval data mining, term frequ)ency-inverse document frequency), prepared sample t= { (X) 1 ,Y 1 ),(X 2 ,Y 2 )(X 3 ,Y 3 ) ... N samples of the sample number, the SVM classification hyperplane (split line) is wx+b=0; the geometric distance from the sample point to the hyperplane is:
the goal of the SVM (support vector machine ) training is to find the hyperplane corresponding to the largest value of all the margins. The mathematical language description is therefore to determine w, b such that the outer edge distance is maximized: 
Thereby determining a user problem set for the user.
S132, constructing a robot response set corresponding to the user problem set in a cluster analysis mode.
In this embodiment, the robot answer set refers to an answer set corresponding to the answer user question set.
The problem answer set comprises a user problem set and a robot answer set corresponding to the user problem set.
In this embodiment, as shown in fig. 7, a Canopy clustering algorithm is used to construct a robot response set corresponding to the user problem set.
In one embodiment, referring to fig. 5, the step S132 may include steps S1321 to S1326.
S1321, determining a first similarity threshold and a second similarity threshold.
In this embodiment, after the question is identified, the response to the question is mined. Considering that the effective question-answer pair is in the form of one question-answer, the original position of the truly distinguished question sentence is traced back, the text word immediately below is used as the potential answer of the question sentence, and one question is a truly effective answer, the frequency of occurrence is relatively high, so that the potential question answers are directly gathered to the same category by adopting a clustering analysis mode. I treat the following answers as basic question answer pairs sample X. In this embodiment, a Canopy clustering algorithm is adopted to perform clustering, and a first similarity threshold t1 and a second similarity threshold t2, t1> t2 need to be determined.
S1322, initializing a user problem set and initializing a clustering result set.
In this embodiment, firstly, question a in the user question set is initialized, the text a is vectorized and then placed in a, and the clustering result set Canopy is initialized.
S1323, randomly selecting a problem and creating a cluster centering on the problem.
In this embodiment, a question qa is randomly selected, and a cluster canopy centered on the question qa is created.
S1324, traversing each answer in the user question set, and calculating the distance from the answer to each cluster.
In this embodiment, each response a in the remaining set of user questions is traversed, and the distance d of the answer to each cluster is calculated.
S1325, classifying answers with the distance smaller than a first similarity threshold value into the clusters, deleting the answers with the user distance smaller than a second similarity threshold value, and determining that the question is a new cluster center point when the distance from the answer corresponding to the question to each cluster is larger than the first similarity threshold value.
Specifically, if d < t1, categorize the problem qa into the cluster; if d < t2, the question is deleted from the user question set; if each d is d > t1 for a question QA, the question QA is the center point of a new cluster, i.e. the question QA is a potential new question QA. The above steps S1324 and S1325 are repeated until the user problem set is empty. Thus each cluster is a set of similar questions and answers.
S1326, determining the clusters as a robot response set corresponding to the user problem set.
After the clustering is completed, the potential high-quality answer can be obtained if the average value of the degrees of similarity in the clustering is high or the number of the aggregation categories is large. Each cluster large question or answer text represents the size of that cluster, which can then be used in the calculation of transition probabilities.
And S140, constructing a dialogue flow tree by adopting a hierarchical clustering mode according to the answer set of the questions.
In this embodiment, the dialogue flow tree refers to a result formed by constructing questions and answers in a tree shape.
As shown in fig. 8, a dialogue knowledge base is in the form of a tree hierarchy, and a question of a user corresponds to an answer of a customer service; the user can further ask related business problems, so that the dialogue becomes a dialogue tree hierarchy expansion, the answering technique of the robot is regarded as a node, the answering technique of the user is regarded as a transfer edge, and one answering technique of the robot can lead out a plurality of different user answering interactions. Thus, the establishment of a dialogue hierarchical tree is to construct a transition relationship between nodes, one node having a plurality of child nodes. In addition, the constructed dialog logic graph transition tree is probabilistic because the probability of a robot's transition from one phone to another is not exactly consistent.
In one embodiment, referring to fig. 6, the step S140 may include steps S141 to S143.
S141, acquiring the robot questions and answers of the starting node.
Specifically, the first sentence questions and answers of the robots in all dialogue records are clustered, so that the subsequent branches of the starting node are obtained.
S142, obtaining answers of subsequent robots, and constructing clustering branches for each robot by adopting a Canopy clustering algorithm.
Specifically, for each clustering branch of the robot, different responses of the subsequent users are firstly obtained, and a Canopy clustering algorithm is adopted by a specific clustering method. If the user hangs up here, such a state is classified as a special state alone. And adopting a Canopy clustering method to further mine the speech of different users so as to obtain the questions or the questions of the users.
S143, acquiring response branch response nodes of the robot, and marking response nodes which belong to question nodes and have no answers.
For the response of the user, after the response branch response node of the subsequent robot is obtained, whether the response branch response node is a question node is judged according to a question classification discriminator, if the response branch response node is a question node but no answer is found, the question node is required to be marked with a key mark to indicate that the current user question does not respond correctly. After hierarchical clustering is completed, transition probabilities between the clusters Q1 and the question clusters Q2 can be calculated, and the transition probabilities are based on the cluster sizes.
According to the multi-round dialogue flow construction method, the existing dialogue corpus is collected, word segmentation and recognition are carried out on the dialogue corpus, the question answer set is constructed on the basis, the dialogue flow tree is constructed on the basis of the question answer set in a hierarchical clustering mode, high-quality dialogue flow is effectively constructed through automatic on-line data, and a conventional question answer set is mined through a question answer pair mining mode.
Fig. 9 is a schematic block diagram of a multi-round dialog flow construction device 300 according to an embodiment of the present invention. As shown in fig. 9, the present invention further provides a multi-round dialog flow construction device 300 corresponding to the above multi-round dialog flow construction method. The multi-round dialog flow construction device 300, which may be configured in a server, comprises means for performing the multi-round dialog flow construction method described above. Specifically, referring to fig. 9, the multi-turn dialog flow construction device 300 includes a corpus collection unit 301, a recognition unit 302, a set construction unit 303, and a flow tree construction unit 304.
A corpus collection unit 301, configured to collect dialogue corpus; the recognition unit 302 is configured to segment and recognize the dialogue corpus to obtain a recognition result; a set construction unit 303, configured to construct a question answer set according to the recognition result; the process tree construction unit 304 is configured to construct a dialogue process tree according to the answer set of the question in a hierarchical clustering manner.
In one embodiment, as shown in fig. 10, the corpus collection unit 301 includes a sound recording file acquisition subunit 13011, a speech recognition subunit 3012, a crawling subunit 3013, and an integration subunit 3014.
A recording file obtaining subunit 13011, configured to obtain a dialogue recording file; a voice recognition subunit 3012, configured to perform voice recognition on the dialogue record file to obtain dialogue stream data; a crawling subunit 3013, configured to grab the question-answer pair through a crawler technology; and the integration subunit 3014 is configured to integrate the dialog flow data and the question-answer pair to obtain a dialog corpus.
In an embodiment, the recognition unit 302 is configured to perform word segmentation and entity recognition on the dialog corpus by using a Jieba word segmentation tool to obtain a recognition result.
In one embodiment, as shown in fig. 11, the set construction unit 303 includes a question set construction subunit 3031 and a answer set construction subunit 3032.
A question set construction subunit 3031, configured to construct a user question set according to the identification result; a response set construction subunit 3032, configured to construct a robot response set corresponding to the user problem set by adopting a cluster analysis manner; the problem answer set comprises a user problem set and a robot answer set corresponding to the user problem set.
In an embodiment, the answer set construction subunit 3032 is configured to construct a robot answer set corresponding to the user question set by using a Canopy clustering algorithm.
In one embodiment, as shown in fig. 12, the answer set construction subunit 3032 includes a threshold determination module 30321, an initialization module 30322, a cluster construction module 30323, a traversal module 30324, a categorization module 30325, and a determination module 30326.
A threshold determining module 30321 for determining a first similarity threshold and a second similarity threshold; an initialization module 30322, configured to initialize a user problem set and initialize a clustering result set; a cluster construction module 30323, configured to randomly select a problem and create a cluster centered on the problem; a traversing module 30324, configured to traverse each answer in the user question set, and calculate a distance from the answer to each cluster; a classifying module 30325, configured to classify answers with the distance smaller than a first similarity threshold into the clusters, delete answers with the user distance smaller than a second similarity threshold, and determine that the question is a new cluster center point when the distance from the answer corresponding to the question to each cluster is greater than the first similarity threshold; and the determining module 30326 is used for determining the cluster as a robot response set corresponding to the user problem set.
In one embodiment, as shown in fig. 13, the flow tree construction unit 304 includes an acquisition subunit 3041, a branch construction subunit 3042, and a labeling subunit 3043.
An acquiring subunit 3041, configured to acquire a robot question and an answer of the start node; a branch construction subunit 3042, configured to obtain an answer of a subsequent robot, and construct a clustering branch for each robot by adopting a Canopy clustering algorithm; the labeling subunit 3043 is configured to obtain a response branch answer node of the robot, and label an answer node that belongs to a question node and has no answer.
It should be noted that, as will be clearly understood by those skilled in the art, the specific implementation process of the multi-round conversation process constructing apparatus 300 and each unit may refer to the corresponding descriptions in the foregoing method embodiments, and for convenience and brevity of description, the description is omitted here.
The multi-round dialog flow construction device 300 described above may be implemented in the form of a computer program which can be run on a computer arrangement as shown in fig. 14.
Referring to fig. 14, fig. 14 is a schematic block diagram of a computer device according to an embodiment of the present application. The computer device 500 is a server, where the server may be a stand-alone server or may be a server cluster formed by a plurality of servers.
With reference to FIG. 14, the computer device 500 includes a processor 502, memory, and a network interface 505 connected by a system bus 501, where the memory may include a non-volatile storage medium 503 and an internal memory 504.
The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform a multi-round dialog flow construction method.
The processor 502 is used to provide computing and control capabilities to support the operation of the overall computer device 500.
The internal memory 504 provides an environment for the execution of a computer program 5032 in the non-volatile storage medium 503, which computer program 5032, when executed by the processor 502, causes the processor 502 to perform a multi-round dialog flow construction method.
The network interface 505 is used for network communication with other devices. It will be appreciated by those skilled in the art that the structure shown in fig. 14 is merely a block diagram of a portion of the structure associated with the present application and does not constitute a limitation of the computer device 500 to which the present application is applied, and that a particular computer device 500 may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
Wherein the processor 502 is configured to execute a computer program 5032 stored in a memory to implement the steps of:
collecting dialogue corpus; word segmentation and recognition are carried out on the dialogue corpus so as to obtain a recognition result; constructing a question answer set according to the identification result; and constructing a dialogue flow tree by adopting a hierarchical clustering mode according to the answer set of the questions.
In one embodiment, when implementing the step of collecting the dialogue corpus, the processor 502 specifically implements the following steps:
acquiring a dialogue recording file; performing voice recognition on the dialogue record file to obtain dialogue stream data; capturing question-answer pairs through a crawler technology; and integrating the dialogue stream data and the question-answer pairs to obtain dialogue corpus.
In an embodiment, when implementing the step of word segmentation and recognition on the dialog corpus to obtain a recognition result, the processor 502 specifically implements the following steps:
and performing word segmentation and entity recognition on the dialogue corpus by using a Jieba word segmentation tool to obtain a recognition result.
In one embodiment, when the step of building the answer set of the question according to the recognition result is implemented by the processor 502, the following steps are specifically implemented:
constructing a user problem set according to the identification result; constructing a robot response set corresponding to the user problem set by adopting a cluster analysis mode; the problem answer set comprises a user problem set and a robot answer set corresponding to the user problem set.
In an embodiment, when implementing the step of constructing the robot answer set corresponding to the user question set by using the cluster analysis method, the processor 502 specifically implements the following steps:
and constructing a robot response set corresponding to the user problem set by adopting a Canopy clustering algorithm.
In an embodiment, when implementing the step of constructing the robot answer set corresponding to the user question set by using the Canopy clustering algorithm, the processor 502 specifically implements the following steps:
determining a first similarity threshold and a second similarity threshold; initializing a user problem set and initializing a clustering result set; randomly selecting a question and creating a cluster centered on the question; traversing each answer in the user question set, and calculating the distance from the answer to each cluster; classifying answers with the distance smaller than a first similarity threshold value into the clusters, deleting the answers with the user distance smaller than a second similarity threshold value, and determining the questions as new cluster center points when the distances from the answers corresponding to the questions to each cluster are larger than the first similarity threshold value; and determining the cluster as a robot response set corresponding to the user problem set.
In one embodiment, when implementing the step of constructing the dialogue flow tree according to the answer set of the question in a hierarchical clustering manner, the processor 502 specifically implements the following steps:
Acquiring a robot question and an answer of a starting node; obtaining answers of subsequent robots, and constructing clustering branches for each robot by adopting a Canopy clustering algorithm; and acquiring response branch response nodes of the robot, and marking response nodes which belong to question nodes and have no answer.
It should be appreciated that in embodiments of the present application, the processor 502 may be a central processing unit (Central Processing Unit, CPU), the processor 502 may also be other general purpose processors, digital signal processors (Digital Signal Processor, DSPs), application specific integrated circuits (Application Specific Integrated Circuit, ASICs), off-the-shelf programmable gate arrays (Field-Programmable Gate Array, FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or the like. Wherein the general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
Those skilled in the art will appreciate that all or part of the flow in a method embodying the above described embodiments may be accomplished by computer programs instructing the relevant hardware. The computer program comprises program instructions, and the computer program can be stored in a storage medium, which is a computer readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the flow steps of the embodiments of the method described above.
Accordingly, the present invention also provides a storage medium. The storage medium may be a computer readable storage medium. The storage medium stores a computer program which, when executed by a processor, causes the processor to perform the steps of:
collecting dialogue corpus; word segmentation and recognition are carried out on the dialogue corpus so as to obtain a recognition result; constructing a question answer set according to the identification result; and constructing a dialogue flow tree by adopting a hierarchical clustering mode according to the answer set of the questions.
In one embodiment, when the processor executes the computer program to implement the step of collecting the dialogue corpus, the following steps are specifically implemented:
acquiring a dialogue recording file; performing voice recognition on the dialogue record file to obtain dialogue stream data; capturing question-answer pairs through a crawler technology; and integrating the dialogue stream data and the question-answer pairs to obtain dialogue corpus.
In one embodiment, when the processor executes the computer program to perform the step of word segmentation and recognition on the dialogue corpus to obtain a recognition result, the following steps are specifically implemented:
and performing word segmentation and entity recognition on the dialogue corpus by using a Jieba word segmentation tool to obtain a recognition result.
In one embodiment, when the processor executes the computer program to implement the step of constructing the answer set of questions according to the recognition result, the following steps are specifically implemented:
constructing a user problem set according to the identification result; constructing a robot response set corresponding to the user problem set by adopting a cluster analysis mode; the problem answer set comprises a user problem set and a robot answer set corresponding to the user problem set.
In an embodiment, when the processor executes the computer program to implement the step of constructing the robot response set corresponding to the user problem set by adopting a cluster analysis mode, the method specifically includes the following steps:
and constructing a robot response set corresponding to the user problem set by adopting a Canopy clustering algorithm.
In an embodiment, when the processor executes the computer program to implement the step of constructing the robot answer set corresponding to the user question set by adopting a Canopy clustering algorithm, the following steps are specifically implemented:
determining a first similarity threshold and a second similarity threshold; initializing a user problem set and initializing a clustering result set; randomly selecting a question and creating a cluster centered on the question; traversing each answer in the user question set, and calculating the distance from the answer to each cluster; classifying answers with the distance smaller than a first similarity threshold value into the clusters, deleting the answers with the user distance smaller than a second similarity threshold value, and determining the questions as new cluster center points when the distances from the answers corresponding to the questions to each cluster are larger than the first similarity threshold value; and determining the cluster as a robot response set corresponding to the user problem set.
In one embodiment, when the processor executes the computer program to implement the step of constructing the dialogue flow tree according to the answer set of the question in a hierarchical clustering manner, the method specifically includes the following steps:
acquiring a robot question and an answer of a starting node; obtaining answers of subsequent robots, and constructing clustering branches for each robot by adopting a Canopy clustering algorithm; and acquiring response branch response nodes of the robot, and marking response nodes which belong to question nodes and have no answer.
The storage medium may be a U-disk, a removable hard disk, a Read-Only Memory (ROM), a magnetic disk, or an optical disk, or other various computer-readable storage media that can store program codes.
Those of ordinary skill in the art will appreciate that the elements and algorithm steps described in connection with the embodiments disclosed herein may be embodied in electronic hardware, in computer software, or in a combination of the two, and that the elements and steps of the examples have been generally described in terms of function in the foregoing description to clearly illustrate the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the several embodiments provided by the present invention, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the device embodiments described above are merely illustrative. For example, the division of each unit is only one logic function division, and there may be another division manner in actual implementation. For example, multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed.
The steps in the method of the embodiment of the invention can be sequentially adjusted, combined and deleted according to actual needs. The units in the device of the embodiment of the invention can be combined, divided and deleted according to actual needs. In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The integrated unit may be stored in a storage medium if implemented in the form of a software functional unit and sold or used as a stand-alone product. Based on such understanding, the technical solution of the present invention is essentially or a part contributing to the prior art, or all or part of the technical solution may be embodied in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a terminal, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention.
While the invention has been described with reference to certain preferred embodiments, it will be understood by those skilled in the art that various changes and substitutions of equivalents may be made and equivalents will be apparent to those skilled in the art without departing from the scope of the invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (7)
1. The multi-round dialogue flow construction method is characterized by comprising the following steps:
collecting dialogue corpus;
word segmentation and recognition are carried out on the dialogue corpus so as to obtain a recognition result;
constructing a question answer set according to the identification result;
constructing a dialogue flow tree by adopting a hierarchical clustering mode according to the question answer set;
the step of constructing a question answer set according to the recognition result comprises the following steps:
constructing a user problem set according to the identification result;
constructing a robot response set corresponding to the user problem set by adopting a cluster analysis mode;
the problem answer set comprises a user problem set and a robot answer set corresponding to the user problem set;
the step of constructing a user problem set according to the identification result comprises the following steps:
Constructing a question classifier, judging whether the user text in the recognition result is an effective question, and recognizing and screening out potential effective questions in all dialogue corpora; training corpus of the question recognition classifier is sorted by adopting a mode of combining rules with manual inspection;
classification category y= {0,1}, each user question X is converted into TF-IDF based on question words for feature representation, and prepared sample t= { (X) 1 ,Y 1 ) , (X 2 ,Y 2 ) (X 3 ,Y 3 ) ... N samples of the sample number, the SVM classification hyperplane is wx+b=0; the geometric distance from the sample point to the hyperplane is:
the goal of SVM training is to find out the hyper-plane corresponding to the largest value in all the outer margins; the mathematical language description is therefore to determine w, b such that the outer edge distance is maximized:
，
thereby determining a user problem set for the user;
the collecting dialogue corpus includes:
acquiring a dialogue recording file;
performing voice recognition on the dialogue record file to obtain dialogue stream data;
capturing question-answer pairs through a crawler technology;
integrating the dialogue stream data and question-answer pairs to obtain dialogue corpus;
the construction of the dialogue flow tree by hierarchical clustering mode according to the question answer set comprises the following steps:
Acquiring a robot question and an answer of a starting node;
obtaining answers of subsequent robots, and constructing clustering branches for each robot by adopting a Canopy clustering algorithm; namely, for each clustering branch of the robots, firstly, different responses of subsequent users are obtained, and a Canopy clustering algorithm is adopted by a specific clustering method; if the user hangs up the phone here, such a state is classified as a special state separately; adopting a Canopy clustering method to further mine the speech of different users so as to obtain questions or answers of the users;
acquiring response branch response nodes of the robot, and marking response nodes belonging to question nodes but without answers; that is, for the response of the user, after obtaining the response branch response node of the subsequent robot, judging whether the response branch response node is a question node according to a question classification discriminator, if the response branch response node is a question node but has no answer, the response branch response node needs to be used for marking the key points to indicate that the current user question does not respond correctly; after hierarchical clustering is completed, calculating transition probability between the clusters Q1 and the question clusters Q2, wherein the transition probability is obtained based on the cluster size.
2. The method for constructing a multi-turn dialog flow according to claim 1, wherein the word segmentation and recognition of the dialog corpus to obtain a recognition result includes:
And performing word segmentation and entity recognition on the dialogue corpus by using a Jieba word segmentation tool to obtain a recognition result.
3. The method for constructing a multi-turn conversation process according to claim 1, wherein the constructing a robot response set corresponding to the user question set by using a cluster analysis method includes:
and constructing a robot response set corresponding to the user problem set by adopting a Canopy clustering algorithm.
4. The method for constructing a multi-round dialog flow according to claim 3, wherein the constructing a robot response set corresponding to the user question set by using a Canopy clustering algorithm includes:
determining a first similarity threshold and a second similarity threshold;
initializing a user problem set and initializing a clustering result set;
randomly selecting a question and creating a cluster centered on the question;
traversing each answer in the user question set, and calculating the distance from the answer to each cluster;
classifying answers with the distance smaller than a first similarity threshold value into the clusters, deleting the answers with the user distance smaller than a second similarity threshold value, and determining the questions as new cluster center points when the distances from the answers corresponding to the questions to each cluster are larger than the first similarity threshold value;
And determining the cluster as a robot response set corresponding to the user problem set.
5. The multi-round dialogue flow constructing device is characterized by comprising:
the corpus collection unit is used for collecting dialogue corpus;
the recognition unit is used for word segmentation and recognition of the dialogue corpus to obtain a recognition result;
the set construction unit is used for constructing a question answer set according to the recognition result;
the flow tree construction unit is used for constructing a dialogue flow tree in a hierarchical clustering mode according to the question answer set;
the step of constructing a question answer set according to the recognition result comprises the following steps:
constructing a user problem set according to the identification result;
constructing a robot response set corresponding to the user problem set by adopting a cluster analysis mode;
the problem answer set comprises a user problem set and a robot answer set corresponding to the user problem set;
the step of constructing a user problem set according to the identification result comprises the following steps:
constructing a question classifier, judging whether the user text in the recognition result is an effective question, and recognizing and screening out potential effective questions in all dialogue corpora; training corpus of the question recognition classifier is sorted by adopting a mode of combining rules with manual inspection;
Classification category y= {0,1}, each user question X is converted into TF-IDF based on question words for feature representation, and prepared sample t= { (X) 1 ,Y 1 ) , (X 2 ,Y 2 ) (X 3 ,Y 3 ) ... N samples of the sample number, the SVM classification hyperplane is wx+b=0; the geometric distance from the sample point to the hyperplane is:
；
the goal of SVM training is to find out the hyper-plane corresponding to the largest value in all the outer margins; the mathematical language description is therefore to determine w, b such that the outer edge distance is maximized:
thereby determining a user problem set for the user;
the collecting dialogue corpus includes:
acquiring a dialogue recording file;
performing voice recognition on the dialogue record file to obtain dialogue stream data;
capturing question-answer pairs through a crawler technology;
integrating the dialogue stream data and question-answer pairs to obtain dialogue corpus;
the construction of the dialogue flow tree by hierarchical clustering mode according to the question answer set comprises the following steps:
acquiring a robot question and an answer of a starting node;
obtaining answers of subsequent robots, and constructing clustering branches for each robot by adopting a Canopy clustering algorithm; namely, for each clustering branch of the robots, firstly, different responses of subsequent users are obtained, and a Canopy clustering algorithm is adopted by a specific clustering method; if the user hangs up the phone here, such a state is classified as a special state separately; adopting a Canopy clustering method to further mine the speech of different users so as to obtain questions or answers of the users;
Acquiring response branch response nodes of the robot, and marking response nodes belonging to question nodes but without answers; that is, for the response of the user, after obtaining the response branch response node of the subsequent robot, judging whether the response branch response node is a question node according to a question classification discriminator, if the response branch response node is a question node but has no answer, the response branch response node needs to be used for marking the key points to indicate that the current user question does not respond correctly; after hierarchical clustering is completed, calculating transition probability between the clusters Q1 and the question clusters Q2, wherein the transition probability is obtained based on the cluster size.
6. A computer device, characterized in that it comprises a memory on which a computer program is stored and a processor which, when executing the computer program, implements the method according to any of claims 1-4.
7. A storage medium storing a computer program which, when executed by a processor, performs the method of any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110520914.0A CN113239164B (en) | 2021-05-13 | 2021-05-13 | Multi-round dialogue flow construction method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110520914.0A CN113239164B (en) | 2021-05-13 | 2021-05-13 | Multi-round dialogue flow construction method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113239164A CN113239164A (en) | 2021-08-10 |
| CN113239164B true CN113239164B (en) | 2023-07-04 |
Family
ID=77134175
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110520914.0A Active CN113239164B (en) | 2021-05-13 | 2021-05-13 | Multi-round dialogue flow construction method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113239164B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015058604A1 (en) * | 2013-10-21 | 2015-04-30 | 北京奇虎科技有限公司 | Apparatus and method for obtaining degree of association of question and answer pair and for search ranking optimization |
| CN111767366A (en) * | 2019-04-01 | 2020-10-13 | 北京百度网讯科技有限公司 | Question and answer resource mining method and device, computer equipment and storage medium |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105335400B (en) * | 2014-07-22 | 2018-11-23 | 阿里巴巴集团控股有限公司 | Enquirement for user is intended to obtain the method and device of answer information |
| CN104408639A (en) * | 2014-10-22 | 2015-03-11 | 百度在线网络技术（北京）有限公司 | Multi-round conversation interaction method and system |
| CN106294321B (en) * | 2016-08-04 | 2019-05-31 | 北京儒博科技有限公司 | A kind of the dialogue method for digging and device of specific area |
| US11373632B2 (en) * | 2017-05-10 | 2022-06-28 | Oracle International Corporation | Using communicative discourse trees to create a virtual persuasive dialogue |
| US20180341870A1 (en) * | 2017-05-23 | 2018-11-29 | International Business Machines Corporation | Managing Indecisive Responses During a Decision Tree Based User Dialog Session |
| US10498898B2 (en) * | 2017-12-13 | 2019-12-03 | Genesys Telecommunications Laboratories, Inc. | Systems and methods for chatbot generation |
| US11321536B2 (en) * | 2019-02-13 | 2022-05-03 | Oracle International Corporation | Chatbot conducting a virtual social dialogue |
| CN111274378B (en) * | 2020-02-13 | 2021-09-24 | 南京云问网络技术有限公司 | Data processing method and device for question answering, equipment and storage medium |
-
2021
- 2021-05-13 CN CN202110520914.0A patent/CN113239164B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015058604A1 (en) * | 2013-10-21 | 2015-04-30 | 北京奇虎科技有限公司 | Apparatus and method for obtaining degree of association of question and answer pair and for search ranking optimization |
| CN111767366A (en) * | 2019-04-01 | 2020-10-13 | 北京百度网讯科技有限公司 | Question and answer resource mining method and device, computer equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 交互式问答的关系结构体系及标注;周小强;王晓龙;陈清财;;中文信息学报(05);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113239164A (en) | 2021-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11216510B2 (en) | Processing an incomplete message with a neural network to generate suggested messages | |
| WO2020244073A1 (en) | Speech-based user classification method and device, computer apparatus, and storage medium | |
| US20220337538A1 (en) | Customized message suggestion with user embedding vectors | |
| US7912714B2 (en) | Method for segmenting communication transcripts using unsupervised and semi-supervised techniques | |
| CN110414004B (en) | Method and system for extracting core information | |
| CN112735383A (en) | Voice signal processing method, device, equipment and storage medium | |
| CN107229627B (en) | Text processing method and device and computing equipment | |
| WO2017091985A1 (en) | Method and device for recognizing stop word | |
| CN113254643B (en) | Text classification method and device, electronic equipment and text classification program | |
| CN112818109B (en) | Intelligent reply method, medium, device and computing equipment for mail | |
| WO2021036439A1 (en) | Method for responding to complaint, and device | |
| CN108287848B (en) | Method and system for semantic parsing | |
| CN115274086B (en) | Intelligent diagnosis guiding method and system | |
| CN115238688B (en) | Method, device, equipment and storage medium for analyzing association relation of electronic information data | |
| CN111782793A (en) | Intelligent customer service processing method, system and equipment | |
| CN111429157A (en) | Method, device and equipment for evaluating and processing complaint work order and storage medium | |
| CN111611358A (en) | Information interaction method and device, electronic equipment and storage medium | |
| CN111143508B (en) | Event detection and tracking method and system based on communication type short text | |
| CN114756675A (en) | Text classification method, related equipment and readable storage medium | |
| CN116150651A (en) | AI-based depth synthesis detection method and system | |
| CN113392205A (en) | User portrait construction method, device and equipment and storage medium | |
| Ali et al. | K-means clustering to improve the accuracy of decision tree response classification | |
| CN112487186A (en) | Human-human conversation log analysis method, system, equipment and storage medium | |
| CN113239164B (en) | Multi-round dialogue flow construction method and device, computer equipment and storage medium | |
| CN116305257A (en) | Privacy information monitoring device and privacy information monitoring method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |