CN114118059A - Sample statement processing method and device, computer equipment and storage medium - Google Patents
Sample statement processing method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN114118059A CN114118059A CN202111302435.8A CN202111302435A CN114118059A CN 114118059 A CN114118059 A CN 114118059A CN 202111302435 A CN202111302435 A CN 202111302435A CN 114118059 A CN114118059 A CN 114118059A
- Authority
- CN
- China
- Prior art keywords
- sample
- sentence
- statement
- similar
- target
- 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
- 238000003672 processing method Methods 0.000 title abstract description 17
- 238000001514 detection method Methods 0.000 claims abstract description 84
- 238000000034 method Methods 0.000 claims abstract description 41
- 238000013145 classification model Methods 0.000 claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000012545 processing Methods 0.000 claims abstract description 24
- 238000012937 correction Methods 0.000 claims abstract description 13
- 238000004590 computer program Methods 0.000 claims description 23
- 238000012217 deletion Methods 0.000 claims description 10
- 230000037430 deletion Effects 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 9
- 238000012216 screening Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 description 12
- 238000013528 artificial neural network Methods 0.000 description 8
- 238000013473 artificial intelligence Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 4
- 238000012512 characterization method Methods 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000001960 triggered effect Effects 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- 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
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/23—Updating
-
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application relates to a sample statement processing method, a sample statement processing device, computer equipment and a storage medium. The method comprises the following steps: acquiring a target statement with an intention mark error; inputting the target sentence marked with the error intention label into a trained similar sample detection model; based on a similar sample detection model, comparing the similarity of the target sentence marked with the wrong intention label with each original sample sentence marked with the intention label to detect a wrong marked sample sentence similar to the target sentence from the sample sentence library; carrying out sample sentence correction processing on the sample sentence library based on the error marking sample sentence so as to update the sample sentence library; the updated sample sentence library is used for training the intention classification model; the intent classification model is used to identify the intent of a user input sentence in a robot conversation scenario and to instruct the chat robot to respond based on the identified intent. By adopting the method, the updating efficiency of the sample sentence library can be improved.
Description
Technical Field
The present application relates to the field of artificial intelligence technologies, and in particular, to a method and an apparatus for processing sample sentences, a computer device, and a storage medium.
Background
With the development of artificial intelligence technology, a chat robot technology appears, and the chat robot can be used for customer service, can capture keywords input by a user on a conversation page, and then searches for the most appropriate answer sentence from a sample library. The chat robot can improve the service quality and reduce the service cost.
Obviously, for the application of the chat robot, the most important is the accuracy of the sample library. In some cases, there may be confounding samples in the sample library. The traditional method mainly depends on manpower, and the time and labor are too long and the sample database is not updated efficiently because the confusing samples are manually deleted from the sample database.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a sample statement processing method, device, computer device, and storage medium for improving the efficiency of updating a sample library.
In a first aspect, the present application provides a sample statement processing method. The method comprises the following steps:
acquiring a target statement with an intention mark error;
inputting the target sentence marked with the wrong intention label into a trained similar sample detection model; the similar sample detection model is trained on the basis of a set of original sample sentences marked with intentional drawing labels in a sample sentence library;
comparing the similarity of the target sentence marked with the wrong intention label with each original sample sentence marked with the intention label based on the similar sample detection model so as to detect a wrong marked sample sentence similar to the target sentence from the sample sentence library;
carrying out sample statement correction processing on the sample statement library based on the error marking sample statement so as to update the sample statement library; the updated sample sentence library is used for training the intention classification model; the intent classification model is used to identify an intent of a user input sentence in a robot conversation scenario and to instruct the chat robot to respond based on the identified intent.
In one embodiment, the method further comprises a training step of a similar sample detection model; the training step of the similar sample detection model comprises the following steps:
selecting a base sample statement from the set of original sample statements;
acquiring an original sample statement which meets a preset similar condition with the basic sample statement from the set of the original sample statements to serve as a similar sample statement corresponding to the basic sample statement;
screening similar sample sentences with the same intention labels as the basic sample sentences from the similar sample sentences corresponding to each basic sample sentence, and taking the basic sample sentences and the screened similar sample sentences as similar sample pairs;
and training the similar sample detection model according to the similar sample pair.
In one embodiment, the performing a sample sentence correction process on the sample sentence library based on the error marked sample sentence to update the sample sentence library includes:
if the revision operation aiming at the error marking sample statement is detected, triggering and revising the intention label of the error marking sample statement in the sample statement library;
and if the deletion operation aiming at the error marking sample statement is detected, triggering to delete the error marking sample statement from the sample statement library.
In one embodiment, after the triggering of revising the intention label of the error marked sample statement in the sample statement library if the revision operation for the error marked sample statement is detected, the method further includes:
identifying a history sentence similar to the target sentence from a history conversation record of the chat robot and the user;
adding a revised intention label corresponding to the error marking sample statement aiming at the historical statement;
and adding the historical statement added with the revised intention label into the sample statement library as a sample statement.
In one embodiment, the target sentence is a user input sentence generated in a historical conversation of the chat robot with the user;
the method further comprises the following steps:
determining a first context statement corresponding to the target statement in the target history session;
identifying a second contextual statement from a session record of the non-targeted historical session that is similar to the first contextual statement;
determining reference sentences which are positioned between the second context sentences and marked with correct intention labels from the conversation records of the non-target historical conversation;
and if the reference statement and the target statement meet preset similar conditions, adding a correct intention label of the reference statement to the target statement, and adding the target statement added with the correct intention label to the sample statement library.
In one embodiment, the comparing the target sentence marked with the wrong intention label with the similarity of each original sample sentence marked with the intention label based on the similar sample detection model to detect the wrong marked sample sentence similar to the target sentence from the sample sentence library includes:
comparing the intention labels respectively marked for the original sample sentences with the error intention labels marked for the target sentences;
taking an original sample statement corresponding to an intention label which is the same as the error intention label as a target original sample statement;
and comparing the similarity of the target statement with each target original sample statement based on the similar sample detection model so as to detect an error marked sample statement similar to the target statement from the sample statement library.
In a second aspect, the present application further provides a sample sentence processing apparatus. The device comprises:
the acquisition module is used for acquiring a target statement with an intention mark error;
the detection module is used for inputting the target sentence marked with the wrong intention label into a trained similar sample detection model; the similar sample detection model is trained on the basis of a set of original sample sentences marked with intentional drawing labels in a sample sentence library; comparing the similarity of the target sentence marked with the wrong intention label with each original sample sentence marked with the intention label based on the similar sample detection model so as to detect a wrong marked sample sentence similar to the target sentence from the sample sentence library;
the correcting module is used for carrying out sample statement correcting processing on the sample statement library based on the error marking sample statement so as to update the sample statement library; the updated sample sentence library is used for training the intention classification model; the intention classification model is used for identifying the intention of a user input sentence in a robot dialogue scene and instructing the robot to respond based on the identified intention.
In a third aspect, the present application also provides a computer device. The computer device comprises a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the method according to embodiments of the present application.
In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium has stored thereon a computer program which, when executed by a processor, causes the processor to perform the steps of the method according to embodiments of the application.
In a fifth aspect, the present application further provides a computer program product. The computer program product comprises a computer program which, when executed by a processor, causes the processor to perform the steps of the method according to embodiments of the present application.
The sample statement processing method, the sample statement processing device, the computer equipment and the storage medium obtain the target statement with the intention of marking errors; inputting the target sentence marked with the error intention label into a trained similar sample detection model; based on the similar sample detection model, the target sentences marked with the wrong intention labels are compared with the original sample sentences marked with the intention labels in similarity, so that the wrong marked sample sentences similar to the target sentences are detected from the sample sentence library. Sample sentence correction processing is carried out on the sample sentence library based on the error marked sample sentence so as to update the sample sentence library, and the sample sentence library can be quickly updated and corrected based on the error marked sample sentence; the updated sample sentence library is used for training the intention classification model; the intention classification model is used for identifying the intention of the user input sentence in the robot dialogue scene, so that the sample sentence library can be automatically updated, and the updating efficiency of the sample sentence library is improved.
Drawings
FIG. 1 is a diagram of an application environment of a sample statement processing method in one embodiment;
FIG. 2 is a flow diagram illustrating a sample statement processing method according to one embodiment;
FIG. 3a is a schematic diagram of a similar sample detection model in one embodiment;
FIG. 3b is a schematic diagram of the infrastructure network in one embodiment;
FIG. 4 is a flowchart illustrating a sample statement processing method according to another embodiment;
FIG. 5 is a block diagram showing the structure of a sample sentence processing means in one embodiment;
FIG. 6 is a diagram illustrating an internal structure of a computer device according to an embodiment;
FIG. 7 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The sample statement processing method provided by the embodiment of the application can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104, or may be located on the cloud or other network server. The server 104 can acquire the target sentence with the intention of marking error through the terminal 102; the server 104 may input the target sentence marked with the incorrect intention label to the trained similar sample detection model, and compare the similarity between the target sentence marked with the incorrect intention label and each original sample sentence marked with the intention label based on the similar sample detection model, so as to detect the incorrectly marked sample sentence similar to the target sentence from the sample sentence library; the server 104 may perform a sample sentence correction process on the sample sentence library through the terminal 102 based on the error marked sample sentence to update the sample sentence library. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things devices and portable wearable devices, and the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart car-mounted devices, and the like. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, and the like. The server 104 may be implemented as a stand-alone server or as a server cluster comprised of multiple servers.
In one embodiment, as shown in fig. 2, a sample statement processing method is provided, which is described by taking the application of the method to the server in fig. 1 as an example, and includes the following steps:
s202, obtaining the target sentence marked with the error intention, and inputting the target sentence marked with the error intention label into the trained similar sample detection model.
The intention label is a label for identifying the intention of the sentence. It is to be appreciated that in a scenario where the chat robot is conversing with the user, the intent of the user can be identified based on the matched intent tag of the user input sentence. An intent tag error means that the intent of the statement itself does not correspond to the intent tag of the statement. It will be appreciated that a false intent tag actually refers to an intent tag of a statement that does not correspond to the intent of the statement itself. The similar sample detection model is trained on a set of original sample sentences marked with intentional drawing labels in the sample sentence library. And the sample sentence library is used for storing the sample sentences. It is understood that the sample sentence library is a collection of all sample sentences. The original sample sentence refers to a sample sentence marked with a conscious graph label in the sample sentence library before correction. The sample sentence is a sample formed by combining the sentence and the corresponding intention label.
Specifically, the server can obtain the target sentence marked with the error intention through the terminal, and input the target sentence marked with the error intention label into the trained similar sample detection model.
In one embodiment, the terminal may obtain a target sentence with an intention to mark an error, and send the target sentence to the server.
In one embodiment, the terminal may store the trained similar sample detection models. The terminal can obtain the target sentence marked with the error intention, and directly input the target sentence marked with the error intention label into the trained similar sample detection model.
In one embodiment, the trainee can judge the sentence with the wrong label mark by the terminal and mark the sentence with the wrong label. It is understood that the terminal may take the sentence marked as the intention marking error by the trainer as the target sentence.
The machine trainer is an artificial intelligence trainer, and is an enterprise, wherein the machine trainer is an operator who knows services and also knows the principle of artificial intelligence, can be combined with a service scene of practical application according to the characteristics of the artificial intelligence, can independently deploy and configure an intelligent product, can specifically train, improve and optimize certain functions of the intelligent product, and finally makes the artificial intelligence exert a specific service value in the application process.
In one embodiment, the target statement may include at least one of a sample statement in a sample statement library and a statement in a historical conversation record.
In one embodiment, the similar sample detection model may include at least one of a trigram network model, a twin neural network model, and a Sentence-transformer-based bidirectional encoder characterization (sequence-BERT) model.
S204, based on the similar sample detection model, the target sentence marked with the wrong intention label is compared with each original sample sentence marked with the intention label in similarity, so that the wrong marked sample sentence similar to the target sentence is detected from the sample sentence library.
The similar sample detection model refers to a model for detecting similar sample statements.
Specifically, the server may compare the similarity between the target sentence marked with the incorrect intention label and each of the original sample sentences marked with the intention labels based on the similar sample detection model to detect an incorrectly marked sample sentence similar to the target sentence from the sample sentence library. The server can obtain the error marking sample statement and send the error marking sample statement to the terminal.
In one embodiment, the terminal may store a similar sample detection model, and directly detect an error marked sample sentence similar to the target sentence from the sample sentence library after comparing the similarity between the target sentence marked with the error intention tag and each original sample sentence marked with the intention tag through the similar sample detection model.
In one embodiment, the similar sample detection model has a preset detection condition for the similarity comparison result. It can be understood that if the similarity comparison result between the target statement and the original sample statement satisfies the preset detection condition, the original sample statement is an error marked sample statement similar to the target statement.
The preset detection condition may include that the similarity between the target sentence and the original sample sentence is greater than or equal to a similarity threshold and at least one of the ranks of the similarities is located at a preset top rank. It is to be understood that if the similarity between the target sentence and the original sample sentence is greater than or equal to the similarity threshold, the original sample sentence is an error marked sample sentence similar to the target sentence. If the rank of the similarity is in the preset order after the similarity between the target sentence and the original sample sentence is arranged in a descending order, the original sample sentence is an error marked sample sentence similar to the target sentence. It is understood that the pre-set number of bits refers to the number of bits that are arranged before the similarity is arranged in descending order. For example, if the previous preset bit number is 20, the original sample sentence with the similarity degree to the target sentence being located at the top 20 of the rank is the error marked sample sentence similar to the target sentence.
And S206, carrying out sample statement correction processing on the sample statement library based on the error marking sample statement so as to update the sample statement library.
And the updated sample sentence library is used for training the intention classification model. The intent classification model is used to identify the intent of a user input sentence in a robot conversation scenario and to instruct the chat robot to respond based on the identified intent.
Specifically, the terminal may perform sample sentence correction processing on the sample sentence library based on the error marking sample sentence, and the server may update the sample sentence library in response to the sample sentence correction processing.
In one embodiment, the server may delete the error marked sample statement from the sample corpus.
In one embodiment, the server may modify the intent tags of the error marked sample sentences from a sample corpus of sentences.
In one embodiment, the server may obtain a statement similar to the target statement from the historical conversation record based on the target statement, and update the statement with the correct intent tag as a sample statement to the sample statement library.
The sample statement processing method obtains the target statement with the wrong intention mark; inputting the target sentence marked with the error intention label into a trained similar sample detection model; based on the similar sample detection model, the target sentences marked with the wrong intention labels are compared with the original sample sentences marked with the intention labels in similarity, so that the wrong marked sample sentences similar to the target sentences are detected from the sample sentence library. Sample sentence correction processing is carried out on the sample sentence library based on the error marked sample sentence so as to update the sample sentence library, and the sample sentence library can be quickly updated and corrected based on the error marked sample sentence; the updated sample sentence library is used for training the intention classification model; the intention classification model is used for identifying the intention of a sentence input by a user in a robot dialogue scene, so that the sample sentence library can be automatically updated, the updating efficiency of the sample sentence library is improved, the intention classification model can be trained on the basis of the updated sample sentence library in time, and the classification accuracy of the intention classification model is improved.
In one embodiment, the method further comprises a training step of a similar sample detection model; the training step of the similar sample detection model comprises the following steps: selecting a basic sample statement from a set of original sample statements; acquiring original sample sentences meeting preset similar conditions with basic sample sentences from a set of the original sample sentences as similar sample sentences corresponding to the basic sample sentences; screening similar sample sentences with the same intention labels as the basic sample sentences from the similar sample sentences corresponding to each basic sample sentence, and taking the basic sample sentences and the screened similar sample sentences as similar sample pairs; and training a similar sample detection model according to the similar sample pair.
Wherein, the basic sample statement is used for constructing similar sample pairs. It can be understood that the base sample statement is essentially a reference statement, and only the original sample statement similar to the base sample statement and having the same intent tag can be obtained to construct a similar sample pair. And presetting a similar condition for judging similar sample sentences of the basic sample sentences from the set of the original sample sentences. The similar sample sentence refers to a sample sentence similar to the basic sample sentence.
Specifically, the server may select a base sample sentence from the set of original sample sentences, obtain, from the set of original sample sentences and by a text similarity algorithm, an original sample sentence that satisfies a preset similarity condition with the base sample sentence as a similar sample sentence corresponding to the base sample sentence, compare intentions of the base sample sentence and the similar sample sentence, screen a similar sample sentence having the same intention label as the base sample sentence, and use the base sample sentence and the screened similar sample sentence as a similar sample pair. The server may train a similar sample detection model based on the similar sample pairs.
In one embodiment, the server may select all of the original sample sentences from the sample corpus as base sample sentences.
In one embodiment, the text similarity algorithm may be the BM25 algorithm. The server may obtain the basic sample statement, and search, through the BM25 algorithm, an original sample statement that satisfies a preset similar condition with the basic sample statement from the sample statement library, as a similar sample statement corresponding to the basic sample statement. For example, the preset similarity condition may be that the similarity value output by the BM25 algorithm is located at the previous preset bit number. It is understood that the larger the similarity value output by the BM25 algorithm, the higher the rank, and the more similar between the similar sample statement and the base sample statement. When the current preset bit number is 15, the server may obtain 15 original sample statements that are most similar to the base sample statement, that is, 15 similar sample statements.
The BM25 algorithm is an algorithm for calculating similarity between a sentence and a document, and can perform word segmentation on an input sentence, then calculate similarity between each word in the sentence and the document, perform weighted summation, and finally obtain similarity score between the sentence and the document.
In one embodiment, the server may compare the intention labels of the similar sample sentence and the basic sample sentence, wherein the same intention label is 1, and the different intention labels are 0, so as to output the comparison result of the intention labels. It is understood that the base sample statement with an output of 1 and the similar sample statement are consistent in intention label, and can be used as a similar sample pair for training a similar sample detection model.
In one embodiment, fig. 3a is a schematic structural diagram of a similar sample detection model. The similar sample detection model is a twin neural network model. The server can respectively input the two sample sentences into the two basic networks, the similarity is evaluated by using a loss function after the two basic networks are processed, and finally the similarity between the two sample sentences is input by using cosine similarity comparison. It is understood that the server may use the similar sample statements and the basic sample statements in the similar sample pairs as inputs to the two basic networks, respectively, to train the similar sample detection model.
The twin neural network model is a coupling framework built based on two artificial neural networks. The twin neural network takes two samples as input and outputs the characterization of embedding high-dimensional space of the two samples so as to compare the similarity degree of the two samples. The narrowly defined twin neural network is formed by splicing two neural networks which have the same structure and share the weight. A generalized twin neural network, or a pseudo-twin neural network (pseudo-twin neural network), may be formed by splicing any two neural networks. The loss function of the similar sample detection model adopts a contrast loss function, and the optimizer adopts an RMSpro (root mean square (back) propagation) optimizer. It is to be appreciated that in training similar sample detection models, the server can use an optimizer to minimize the loss function based on the gradient.
In one embodiment, the server may use the target sentence and each sample sentence in the sample sentence library as inputs of two basic networks of the similar sample detection model in fig. 3a, and evaluate similarity of the two inputs through computation of contrast loss function to detect the sample sentence similar to the target sentence. After detecting the sample sentences similar to the target sentences by using the similar sample detection model, the server can perform statistics by using cosine similarity, and after two times of similarity evaluation, error marked sample sentences are obtained.
Fig. 3b is a schematic diagram of the structure of the basic network. It is to be understood that the similar sample detection model may include an underlying network. The base network in fig. 3b corresponds to the base network in fig. 3 a. The base network includes Inputs (Inputs), an embedding layer (InputEmbedding), a flattening layer (Flatten), three fully connected and disposable layers (Dense & Dropout), and Outputs (Outputs).
When a sample statement is input, the embedding layer can realize mapping from a semantic space to a vector space, and simultaneously, the relation of an original sample in the semantic space is kept in the vector space as much as possible, for example, the positions of two words with similar semantics in the vector space are also relatively close. The flattening layer may one-dimensionally input the multiple dimensions processed by the embedding layer. A fully connected layer (density) can map the distributed feature representation to the sample label space, essentially linearly transforming from one feature space to another. The random discard (Dropout) can make the activation value of a certain neuron stop working with a certain probability during forward propagation, so that the model is more generalized, because it is not too dependent on some local features.
In this embodiment, a base sample statement is selected from a set of original sample statements; acquiring original sample sentences meeting preset similar conditions with basic sample sentences from a set of the original sample sentences as similar sample sentences corresponding to the basic sample sentences; screening similar sample sentences with the same intention labels as the basic sample sentences from the similar sample sentences corresponding to each basic sample sentence, and taking the basic sample sentences and the screened similar sample sentences as similar sample pairs; and training the similar sample detection model according to the similar sample pair, so that the trained similar sample detection model is matched with the sample sentence library, and when a confusing sample in the sample sentence library is searched subsequently, the confusing sample similar to the target sentence can be quickly and accurately positioned based on the target sentence, thereby reducing the labor cost.
In one embodiment, performing a sample statement modification process on the sample statement base based on the error marked sample statement to update the sample statement base comprises: if the revision operation aiming at the error marking sample statement is detected, triggering the intention label of the error marking sample statement in the revision sample statement library; and if the deletion operation aiming at the error marked sample statement is detected, triggering to delete the error marked sample statement from the sample statement library.
The revision operation refers to an operation of revising the intent tag of the error marked sample sentence from the sample sentence library. The delete operation is an operation of deleting the intent tag of the error marked sample sentence from the sample sentence library.
Specifically, if the server detects that the trainee revises the error marked sample sentence through the terminal, the server may receive the revision information sent by the terminal, and revise the intention label of the error marked sample sentence in the sample sentence library according to the revision information. If the server detects that the trainee deletes the error marked sample sentence through the terminal, the server can receive the deletion information sent by the terminal and delete the error marked sample sentence from the sample sentence library according to the deletion information.
The revision information includes information that the trainee revises the intention label of the error marking sample sentence through the terminal. For example, what kind of revision is specifically made on which intent label of the error marked sample sentence is. The deletion information includes information that the trainee deletes the error marked sample sentence through the terminal. For example, it is specifically which error marked sample sentence is deleted.
In one embodiment, the trainees can modify the intention labels of the error marked sample sentences in batch and delete the error marked sample sentences in batch at the terminal.
In this embodiment, if a revision operation for an error marked sample statement is detected, an intention tag of the error marked sample statement in a revision sample statement library is triggered; if the deletion operation aiming at the error marking sample sentence is detected, the error marking sample sentence is triggered to be deleted from the sample sentence library, the intention label of the error marking sample sentence is revised, and the error marking sample sentence is deleted, so that the sample sentence library can be updated in time, the quality of the sample sentence library is improved, and the identification accuracy of the intention classification model obtained based on the training of the sample sentence library is improved.
In one embodiment, after triggering the intent tag of the error marked sample statement in the revised sample statement library if the revision operation for the error marked sample statement is detected, the method further comprises: identifying historical sentences similar to the target sentences from historical conversation records of the chat robot and the user; adding a revised intention label corresponding to the error marking sample statement aiming at the historical statement; and adding the historical sentences added with the revised intention labels into the sample sentence library as sample sentences.
The historical conversation record refers to the historical conversation record of the chat robot and the user in the conversation system. It can be understood that the chat robot can be applied to a conversation system to realize chatting with users. History statements refer to statements in the history session record.
Specifically, the server can obtain historical conversation records of the chat robot and the user from the conversation system, and identify historical sentences similar to the target sentences from the historical conversation records by using a similar sample detection model. The server may add, to the historical statement, the revised intention tag corresponding to the error marking sample statement, and add, as the sample statement, the historical statement to which the revised intention tag is added to the sample statement library.
In the embodiment, historical sentences similar to the target sentences are identified from historical conversation records of the chat robot and the user; adding a revised intention label corresponding to the error marking sample statement aiming at the historical statement; and adding the revised historical sentences with the added intention labels as sample sentences into a sample sentence library, revising the error marked sample sentences, efficiently supplementing similar samples, and improving the richness of the sample sentence library so as to improve the identification generalization capability of the intention classification model.
In one embodiment, the target sentence is a user input sentence generated in a historical conversation of the chat robot with the user; the method further comprises the following steps: determining a first context statement corresponding to a target statement in a target history session; identifying a second contextual statement from the session record of the non-targeted historical session that is similar to the first contextual statement; determining reference sentences which are positioned between the second context sentences and marked with correct intention labels from the conversation records of the non-target historical conversation; and if the preset similar conditions are met between the reference statement and the target statement, adding the correct intention label of the reference statement to the target statement, and adding the target statement added with the correct intention label to the sample statement library.
The first context statement refers to a context statement of the target statement. The second context sentence is a context sentence of the reference sentence. The reference sentence refers to a sentence corresponding to the target sentence, and the target sentence may be labeled with an intention tag with reference to the sentence. It is understood that in different historical conversations, if the contextual statements are intended to be the same, then the statements between the contextual statements are intended to be the same with a high probability.
Specifically, the server may determine an identifier of a target history session to which the target statement belongs, locate the target history session corresponding to the identifier, and search for a first context statement corresponding to the target statement according to a time sequence of the target statement in the target history session. The server may compare the relevance between the statements in the session record of the non-targeted historical session and the first contextual statement to identify a second contextual statement that is similar to the first contextual statement. The server may determine the identifier of the history session in the second context statement, and determine, from the session record of the non-target history session corresponding to the identifier, a reference statement that is located between the second context statements and is marked with a correct intention tag according to the time sequence of the second context statement. And if the preset similar conditions are met between the reference statement and the target statement, adding the correct intention label of the reference statement to the target statement, and adding the target statement added with the correct intention label to the sample statement library.
In one embodiment, the server may target a user input sentence for which the chat robot cannot recognize the intent. It can be understood that the user may have a misstatement during the process of interacting with the chat robot, the chat robot cannot accurately identify the wrong sentence, and if the wrong sentence occurs many times, the server may mark the wrong sentence with an intention tag by comparing the context of the correct sentence with the context of the wrong sentence, and add the intention tag to the sample sentence library.
In one embodiment, the server may compare the relevance between the statement in the session record of the non-targeted historical session and the first context statement through a similar sample detection model.
In one embodiment, the preset similarity condition may include at least one of a similarity greater than or equal to a similarity threshold and a similarity descending ranking in a preset order.
In the embodiment, a first context statement corresponding to a target statement in a target history session is determined; identifying a second contextual statement from the session record of the non-targeted historical session that is similar to the first contextual statement; determining reference sentences which are positioned between the second context sentences and marked with correct intention labels from the conversation records of the non-target historical conversation; if preset similar conditions are met between the reference sentences and the target sentences, correct intention labels of the reference sentences are added to the target sentences, the target sentences added with the correct intention labels are added to the sample sentence library, correct intention labels can be marked for error sentences which are frequently wrongly spoken by a user, the error labels are added to the sample sentence library, and the sample sentence library is enriched, so that the recognition generalization capability of the intention classification model is improved.
In one embodiment, the comparing the similarity between the target sentence marked with the incorrect intention label and each original sample sentence marked with the intention label based on the similar sample detection model to detect the incorrectly marked sample sentence similar to the target sentence from the sample sentence library comprises: comparing the intention labels respectively marked for all original sample sentences with the error intention labels marked for the target sentences; taking an original sample statement corresponding to an intention label which is the same as the error intention label as a target original sample statement; and comparing the similarity of the target sentences with each target original sample sentence based on a similar sample detection model so as to detect error marked sample sentences similar to the target sentences from the sample sentence library.
Specifically, the server may compare the intention label marked for each original sample sentence with the error intention label marked for the target sentence through the classification model, and use the original sample sentence corresponding to the intention label identical to the error intention label as the target original sample sentence. The server may compare the similarity between the target sentence and each target original sample sentence based on the similar sample detection model to detect an error marked sample sentence similar to the target sentence from the sample sentence library.
The classification model may include at least one of a multi-class classification model, a multi-label classification model, an unbalanced classification model, and the like.
In this embodiment, the intention labels respectively marked for each original sample sentence are compared with the wrong intention labels marked for the target sentence; taking an original sample statement corresponding to an intention label which is the same as the error intention label as a target original sample statement; based on the similar sample detection model, the target sentences and all the target original sample sentences are compared in similarity, so that the error marked sample sentences similar to the target sentences are detected from the sample sentence library, the error marked sample sentences can be detected based on the similar sample detection model after the original sample sentences with the same intention labels are screened out, and the accuracy of detecting the error marked sample sentences is improved.
FIG. 4 is a flowchart illustrating a sample statement processing method according to another embodiment;
s402, marking by the trainer, and acquiring the target sentence marked as the error.
The trainees can find the target sentences marked with the wrong intention labels through the terminal, mark the target sentences as the wrong sentences, and the terminal can send the target sentences marked with the errors to the server. For example, when a user says a certain sentence, the chat robot considers the intention A, the follow-up trainer manually judges whether the answer is correct, and if the answer is incorrect, the language of the user is marked. It will be appreciated that marking is simply marking a statement as an error. Or the training engineer may mark the sample in the sample sentence library, where the target sentence refers to a sentence with an incorrect intention label in the sample sentence library.
And S404, recalling candidate error marking sample sentences similar to the target sentence in the sample sentence library through the similar sample detection model.
And the server inputs the received target statement into the similar sample detection model. The similar sample detection model can compare the similarity of the target statement and the sample statement in the sample statement library, and directly output the sample statement similar to the target statement as a candidate error marking sample statement. It is to be understood that the similar sample detection model may be a twin neural network model. It should be noted that the candidate error marked sample statement may also be directly used as an error marked sample statement.
S406, counting the similarity between the candidate error marking sample sentences and the target sentences to screen the error marking sample sentences from the candidate error marking sample sentences.
After detecting the candidate error marked sample sentences output by the similar sample detection model, the server may compare the cosine similarity between the candidate error marked sample sentences and the target sentences again, and perform descending order sorting according to the cosine similarity, and take the candidate error marked sample sentences with the cosine similarity at the previous preset order as error marked samples. For example, if the previous preset bit number is 20, and the candidate error sample sentences with descending ranks before 20 and 20 are considered to be more similar to the target sentence, the server finally sends 20 error marked sample sentences similar to the target sentence to the terminal. It is understood that S406 is essentially a further filtering of the error marked sample statements based on S404.
S408, the trainee judges whether to delete the error marking sample sentence and the target sentence.
The terminal can show the error marking sample to the trainee, and the trainee judges whether to delete the target sentence and the error marking sample. If not, go to step S410. If deleted, S412 is executed. It will be appreciated that the target sentence is essentially a confusing sample. The confusing sample is a sample sentence marked with a wrong intention label, namely a confusing sample sentence, which can cause the chat robot to mistake one thing for another. For example, if the intention label associated with the sample sentence a in the sample sentence library is intention B, but it is manually confirmed that the intention a should be associated with the sample sentence a, the sample a is called a confusing sample.
And S410, revising the intention label, simultaneously acquiring a sentence similar to the target sentence from the historical conversation record through the similar sample detection model, and marking the intention label to be used as a sample sentence to be supplemented into the sample sentence library.
The trainees can revise the intention labels of the target sentences and the error marking sentences through the terminal and store the revised target sentences and error marking sentences into the sample sentence library. Meanwhile, the server can obtain sentences similar to the target sentences from the historical conversation records on the customer service conversation system by using a similar sample detection model, mark the revised intention labels of the target sentences and store the intention labels into the sample sentence library. It can be understood that the server can directly revise the intention labels of the target sentence and the error marked sentence, and does not need a trainer to manually revise the intention labels at the terminal.
S412, deleting the error sample statement and the target statement.
The trainee may perform a deletion operation for the target sentence and the error markup sentence at the terminal, and the server may delete the target sentence and the error markup sentence from the sample sentence library in response to the deletion operation.
After performing the above step of modifying the sample corpus, the server may retrain the intent classification model according to the modified sample corpus.
It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially displayed as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the flowcharts related to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the execution order of the steps or stages is not necessarily sequential, but may be rotated or alternated with other steps or at least a part of the steps or stages in other steps.
Based on the same inventive concept, the embodiment of the present application further provides a sample sentence processing apparatus for implementing the above-mentioned sample sentence processing method. The implementation scheme for solving the problem provided by the apparatus is similar to the implementation scheme described in the above method, so specific limitations in one or more embodiments of the sample sentence processing apparatus provided below can be referred to the limitations of the sample sentence processing method in the foregoing, and details are not described here.
In one embodiment, as shown in fig. 5, there is provided a sample sentence processing apparatus 500, including: an acquisition module 502, a detection module 504, and a correction module 506, wherein:
an obtaining module 502, configured to obtain a target statement that is intended to mark an error.
A detection module 504, configured to input the target sentence labeled with the error intention label to a trained similar sample detection model; the similar sample detection model is trained on the basis of a set of original sample sentences marked with intentional drawing labels in a sample sentence library; and comparing the similarity of the target sentence marked with the wrong intention label with each original sample sentence marked with the intention label based on the similar sample detection model so as to detect the wrong marked sample sentence similar to the target sentence from the sample sentence library.
A correcting module 506, configured to perform sample statement correction processing on the sample statement library based on the error marked sample statement, so as to update the sample statement library; the updated sample sentence library is used for training the intention classification model; the intent classification model is used to identify an intent of a user input sentence in a robot conversation scenario and to instruct the chat robot to respond based on the identified intent.
In one embodiment, the apparatus further comprises a training module. An obtaining module 502, further configured to select a basic sample statement from the set of original sample statements; acquiring an original sample statement which meets a preset similar condition with the basic sample statement from the set of the original sample statements to serve as a similar sample statement corresponding to the basic sample statement; screening similar sample sentences with the same intention labels as the basic sample sentences from the similar sample sentences corresponding to each basic sample sentence, and taking the basic sample sentences and the screened similar sample sentences as similar sample pairs; and the training module is used for training the similar sample detection model according to the similar sample pair.
In one embodiment, the modifying module 506 is further configured to trigger the modifying of the intention label of the error marked sample sentence in the sample sentence library if the modifying operation for the error marked sample sentence is detected; and if the deletion operation aiming at the error marking sample statement is detected, triggering to delete the error marking sample statement from the sample statement library.
In one embodiment, the modification module 506 is further configured to identify a history sentence similar to the target sentence from the history conversation record of the chat robot and the user; adding a revised intention label corresponding to the error marking sample statement aiming at the historical statement; and adding the historical statement added with the revised intention label into the sample statement library as a sample statement.
In one embodiment, the target sentence is a user input sentence generated in a historical conversation of the chat robot with the user; the modification module 506 is further configured to determine a first context statement corresponding to the target statement in the target history session; identifying a second contextual statement from a session record of the non-targeted historical session that is similar to the first contextual statement; determining reference sentences which are positioned between the second context sentences and marked with correct intention labels from the conversation records of the non-target historical conversation; and if the reference statement and the target statement meet preset similar conditions, adding a correct intention label of the reference statement to the target statement, and adding the target statement added with the correct intention label to the sample statement library.
In one embodiment, the detecting module 504 is further configured to compare the intent tag marked for each original sample sentence with the incorrect intent tag marked for the target sentence; taking an original sample statement corresponding to an intention label which is the same as the error intention label as a target original sample statement; and comparing the similarity of the target statement with each target original sample statement based on the similar sample detection model so as to detect an error marked sample statement similar to the target statement from the sample statement library.
The respective modules in the sample sentence processing apparatus described above may be wholly or partially implemented by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 6. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used to store the sample statements. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a sample statement processing method.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 7. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a sample statement processing method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 7 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the above-described method embodiments when executing the computer program.
In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.
In an embodiment, a computer program product is provided, comprising a computer program which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.
It should be noted that, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, presented data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile memory may include a Read-only memory (ROM), a magnetic tape, a floppy disk, a flash memory, an optical memory, a high-density embedded nonvolatile memory, a resistive random access memory (ReRAM), a Magnetic Random Access Memory (MRAM), a Ferroelectric Random Access Memory (FRAM), a Phase Change Memory (PCM), a graphene memory, and the like. Volatile memory can include Random Access Memory (RAM), external cache memory, or the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.
Claims (10)
1. A method for processing sample statements, the method comprising:
acquiring a target statement with an intention mark error;
inputting the target sentence marked with the wrong intention label into a trained similar sample detection model; the similar sample detection model is trained on the basis of a set of original sample sentences marked with intentional drawing labels in a sample sentence library;
comparing the similarity of the target sentence marked with the wrong intention label with each original sample sentence marked with the intention label based on the similar sample detection model so as to detect a wrong marked sample sentence similar to the target sentence from the sample sentence library;
carrying out sample statement correction processing on the sample statement library based on the error marking sample statement so as to update the sample statement library; the updated sample sentence library is used for training the intention classification model; the intent classification model is used to identify an intent of a user input sentence in a robot conversation scenario and to instruct the chat robot to respond based on the identified intent.
2. The method of claim 1, further comprising the step of training a similar sample detection model; the training step of the similar sample detection model comprises the following steps:
selecting a base sample statement from the set of original sample statements;
acquiring an original sample statement which meets a preset similar condition with the basic sample statement from the set of the original sample statements to serve as a similar sample statement corresponding to the basic sample statement;
screening similar sample sentences with the same intention labels as the basic sample sentences from the similar sample sentences corresponding to each basic sample sentence, and taking the basic sample sentences and the screened similar sample sentences as similar sample pairs;
and training the similar sample detection model according to the similar sample pair.
3. The method of claim 1, wherein said performing a sample sentence revision process on said sample sentence library based on said error marked sample sentence to update said sample sentence library comprises:
if the revision operation aiming at the error marking sample statement is detected, triggering and revising the intention label of the error marking sample statement in the sample statement library;
and if the deletion operation aiming at the error marking sample statement is detected, triggering to delete the error marking sample statement from the sample statement library.
4. The method according to claim 3, wherein after said triggering revision of the intention label of the error marked sample sentence in the sample sentence library if a revision operation for the error marked sample sentence is detected, further comprising:
identifying a history sentence similar to the target sentence from a history conversation record of the chat robot and the user;
adding a revised intention label corresponding to the error marking sample statement aiming at the historical statement;
and adding the historical statement added with the revised intention label into the sample statement library as a sample statement.
5. The method of claim 1, wherein the target sentence is a user input sentence generated in a historical conversation of the chat robot with the user;
the method further comprises the following steps:
determining a first context statement corresponding to the target statement in the target history session;
identifying a second contextual statement from a session record of the non-targeted historical session that is similar to the first contextual statement;
determining reference sentences which are positioned between the second context sentences and marked with correct intention labels from the conversation records of the non-target historical conversation;
and if the reference statement and the target statement meet preset similar conditions, adding a correct intention label of the reference statement to the target statement, and adding the target statement added with the correct intention label to the sample statement library.
6. The method according to any one of claims 1 to 5, wherein said comparing, based on the similar sample detection model, the target sentence labeled with an incorrect intention label with the similarity of each of the original sample sentences labeled with intention labels to detect an incorrectly labeled sample sentence similar to the target sentence from the sample sentence library comprises:
comparing the intention labels respectively marked for the original sample sentences with the error intention labels marked for the target sentences;
taking an original sample statement corresponding to an intention label which is the same as the error intention label as a target original sample statement;
and comparing the similarity of the target statement with each target original sample statement based on the similar sample detection model so as to detect an error marked sample statement similar to the target statement from the sample statement library.
7. A sample sentence processing apparatus, characterized in that the apparatus comprises:
the acquisition module is used for acquiring a target statement with an intention mark error;
the detection module is used for inputting the target sentence marked with the wrong intention label into a trained similar sample detection model; the similar sample detection model is trained on the basis of a set of original sample sentences marked with intentional drawing labels in a sample sentence library; comparing the similarity of the target sentence marked with the wrong intention label with each original sample sentence marked with the intention label based on the similar sample detection model so as to detect a wrong marked sample sentence similar to the target sentence from the sample sentence library;
the correcting module is used for carrying out sample statement correcting processing on the sample statement library based on the error marking sample statement so as to update the sample statement library; the updated sample sentence library is used for training the intention classification model; the intention classification model is used for identifying the intention of a user input sentence in a robot dialogue scene and instructing the robot to respond based on the identified intention.
8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 6.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program realizes the steps of the method of any one of claims 1 to 6 when executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111302435.8A CN114118059A (en) | 2021-11-04 | 2021-11-04 | Sample statement processing method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111302435.8A CN114118059A (en) | 2021-11-04 | 2021-11-04 | Sample statement processing method and device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114118059A true CN114118059A (en) | 2022-03-01 |
Family
ID=80380708
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111302435.8A Pending CN114118059A (en) | 2021-11-04 | 2021-11-04 | Sample statement processing method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114118059A (en) |
-
2021
- 2021-11-04 CN CN202111302435.8A patent/CN114118059A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112632385B (en) | Course recommendation method, course recommendation device, computer equipment and medium | |
| CN111753060B (en) | Information retrieval method, apparatus, device and computer readable storage medium | |
| US20240046043A1 (en) | Multi-turn Dialogue Response Generation with Template Generation | |
| CN109376222B (en) | Question-answer matching degree calculation method, question-answer automatic matching method and device | |
| JP2018517959A (en) | Selecting a representative video frame for the video | |
| CN110968695A (en) | Intelligent labeling method, device and platform based on active learning of weak supervision technology | |
| CN110825956A (en) | Information flow recommendation method and device, computer equipment and storage medium | |
| CN110362798B (en) | Method, apparatus, computer device and storage medium for judging information retrieval analysis | |
| US11586838B2 (en) | End-to-end fuzzy entity matching | |
| CN110619050A (en) | Intention recognition method and equipment | |
| CN113343091A (en) | Industrial and enterprise oriented science and technology service recommendation calculation method, medium and program | |
| CN110377618B (en) | Method, device, computer equipment and storage medium for analyzing decision result | |
| KR102532216B1 (en) | Method for establishing ESG database with structured ESG data using ESG auxiliary tool and ESG service providing system performing the same | |
| CN111597336A (en) | Processing method and device of training text, electronic equipment and readable storage medium | |
| CN115758271A (en) | Data processing method, data processing device, computer equipment and storage medium | |
| CN114118059A (en) | Sample statement processing method and device, computer equipment and storage medium | |
| JP2022111020A (en) | Transfer learning method of deep learning model based on document similarity learning and computer device | |
| CN112632246A (en) | Robot dialogue method and device based on deep learning and computer equipment | |
| CN114490996B (en) | Intention recognition method and device, computer equipment and storage medium | |
| CN117151247B (en) | Method, apparatus, computer device and storage medium for modeling machine learning task | |
| CN117938951B (en) | Information pushing method, device, computer equipment and storage medium | |
| US20240004888A1 (en) | Sorting method, apparatus and device, and computer storage medium | |
| CN117150114A (en) | Information recommendation method, apparatus, computer device, storage medium, and program product | |
| CN116861071A (en) | Information pushing method, information pushing device, computer equipment, storage medium and program product | |
| CN116383508A (en) | Searching method, searching device, computer equipment and storage medium |
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 | ||
| CB02 | Change of applicant information |
Country or region after: China Address after: 518000 Room 201, building A, No. 1, Qian Wan Road, Qianhai Shenzhen Hong Kong cooperation zone, Shenzhen, Guangdong (Shenzhen Qianhai business secretary Co., Ltd.) Applicant after: Zhaolian Consumer Finance Co.,Ltd. Address before: 518000 Room 201, building A, No. 1, Qian Wan Road, Qianhai Shenzhen Hong Kong cooperation zone, Shenzhen, Guangdong (Shenzhen Qianhai business secretary Co., Ltd.) Applicant before: MERCHANTS UNION CONSUMER FINANCE Co.,Ltd. Country or region before: China |
|
| CB02 | Change of applicant information |