CN111858875A - Intelligent interaction method, device, equipment and storage medium - Google Patents
Intelligent interaction method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN111858875A CN111858875A CN202010386988.5A CN202010386988A CN111858875A CN 111858875 A CN111858875 A CN 111858875A CN 202010386988 A CN202010386988 A CN 202010386988A CN 111858875 A CN111858875 A CN 111858875A
- Authority
- CN
- China
- Prior art keywords
- question
- information
- classification
- user
- model
- 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
- 238000000034 method Methods 0.000 title claims abstract description 86
- 230000003993 interaction Effects 0.000 title claims abstract description 62
- 238000013145 classification model Methods 0.000 claims abstract description 60
- 239000013598 vector Substances 0.000 claims description 123
- 238000012549 training Methods 0.000 claims description 65
- 238000004590 computer program Methods 0.000 claims description 13
- 230000015654 memory Effects 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 7
- 238000013136 deep learning model Methods 0.000 claims description 6
- 230000001502 supplementing effect Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 24
- 230000004927 fusion Effects 0.000 abstract description 6
- 238000013461 design Methods 0.000 description 18
- 238000012545 processing Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000013519 translation Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
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
- G06F16/35—Clustering; Classification
-
- 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
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/04—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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
Abstract
The embodiment of the invention provides an intelligent interaction method, an intelligent interaction device, intelligent interaction equipment and a storage medium, wherein after user problem information is obtained, problem classification is carried out on the user problem information according to a problem classification model to obtain a classification result, and whether the classification result meets a preset condition is judged; if the preset conditions are met, acquiring corresponding answer information from a preset question-answer library according to the classification result; if the preset conditions are not met, inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information; and outputting the acquired answer information to respond to the user instruction. The method comprises the steps of carrying out semantic classification on user question information through a question classification model to obtain a classification result, obtaining answer information from a preset question-answer library according to the classification result when the classification result is determined to be relatively accurate, obtaining the answer information through an answer generation model when the classification result is determined to be inaccurate, improving the accuracy of obtaining answers through fusion of two modes, and improving user experience in an interaction process.
Description
Technical Field
The embodiment of the invention relates to the technical field of intelligent voice, in particular to an intelligent interaction method, an intelligent interaction device, intelligent interaction equipment and a storage medium.
Background
The intelligent interaction is an important subject in the field of artificial intelligence, is generally applied to equipment such as intelligent sound boxes and intelligent robots, can realize voice interaction between human and machines, and can answer user questions or chat with users.
The existing intelligent interaction method usually adopts a retrieval-based method, needs to collect corresponding answers according to common questions of a user in advance, constructs a question-answer library, loads the question-answer library into a search engine, searches related questions from the question-answer library through the search engine according to the user questions when in online use, and returns the answers corresponding to the questions with the highest matching degree to the user.
The existing retrieval-based method adopts text similarity for retrieval, has poor retrieval effect, and often cannot search matched problems in a question-answer library for some long question sentences or complex question sentences, so that answers cannot be accurately obtained.
Disclosure of Invention
The embodiment of the invention provides an intelligent interaction method, an intelligent interaction device, intelligent interaction equipment and a storage medium, so that a more accurate answer can be obtained and returned to a user aiming at a user problem, and the user experience in an interaction process is improved.
A first aspect of an embodiment of the present invention provides an intelligent interaction method, including:
Acquiring a user instruction, and acquiring user problem information according to the user instruction;
performing problem category classification on the user problem information according to a pre-trained problem classification model to obtain a classification result, and judging whether the classification result meets a preset condition;
if the classification result meets the preset condition, acquiring corresponding answer information from a preset question-answer library according to the classification result, wherein the preset question-answer library comprises different question categories and answer information corresponding to the question categories;
if the classification result is determined not to meet the preset condition, inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information;
and outputting the acquired answer information to respond to the user instruction.
In one possible design, the performing problem category classification on the user problem information according to a pre-trained problem classification model to obtain a classification result, and determining whether the classification result meets a predetermined condition includes:
inputting the user question information into the question classification model, and acquiring the question category and the corresponding confidence of the user question information through the question classification model;
And judging whether the confidence coefficient meets a preset confidence coefficient threshold value.
In one possible design, the problem classification model includes a word vector acquisition submodel and a word vector classification submodel;
the obtaining of the question category and the corresponding confidence of the user question information through the question classification model includes:
obtaining a sub-model through the word vector, segmenting words of the user problem information, and obtaining the word vector of the user problem information according to a word segmentation result;
and classifying the word vectors of the user problem information through the word vector classification submodel to obtain the problem category and the corresponding confidence of the user problem information.
In a possible design, the classifying the word vector of the user question information by the word vector classification submodel to obtain the question category and the corresponding confidence of the user question information includes:
and performing similarity matching on the word vector of the user problem information and a preset word vector corresponding to a preset problem category through the word vector classification submodel, acquiring the preset problem category with the highest similarity as the problem category of the user problem information, and taking the similarity as the confidence.
In one possible design, the word vector classification submodel is a deep learning model.
In one possible design, the answer generation model is a Seq2Seq model;
the inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information includes:
and encoding the user question information through an encoder of the Seq2Seq model, and decoding through a decoder of the Seq2Seq model to obtain answer information corresponding to the user question information.
In one possible design, the encoder of the Seq2Seq model employs an LSTM model, a GRU model, or a Transformer model;
the decoder of the Seq2Seq model adopts an LSTM model, a GRU model or a Transformer model.
In one possible design, the method further includes:
taking a question sample included in each question category in the preset question-answering library as first training data, and training the word vector acquisition submodel;
and acquiring a word vector of each first training data according to the trained word vector acquisition submodel, taking the word vector as second training data, and training the word vector classification submodel until the word vector classification submodel is converged.
In one possible design, the method further includes:
and taking the question sample included in each question category in the preset question-answering library and the corresponding answer information as third training data, and training the Seq2Seq model.
In one possible design, the method further includes:
acquiring historical problems, classifying the historical problems, and acquiring related approximate problems of each problem category and supplementing the related approximate problems to the corresponding problem categories;
acquiring corresponding answer information for each question category;
and constructing the preset question-answer library according to the questions included in each question category and the corresponding answer information.
A second aspect of an embodiment of the present invention provides an intelligent interaction apparatus, including:
the acquisition module is used for acquiring a user instruction and acquiring user problem information according to the user instruction;
the classification module is used for performing problem category classification on the user problem information according to a pre-trained problem classification model to obtain a classification result;
the judging module is used for judging whether the classification result meets a preset condition or not;
a first answer obtaining module, configured to, if it is determined that the classification result meets the predetermined condition, obtain corresponding answer information from a preset question-and-answer library according to the classification result, where the preset question-and-answer library includes different question categories and answer information corresponding to each question category;
The second answer obtaining module is used for inputting the user question information into a pre-trained answer generating model to obtain corresponding answer information if the classification result is determined not to meet the preset condition;
and the output module is used for outputting the acquired answer information so as to respond to the user instruction.
In one possible design, the classification module is configured to perform problem category classification on the user problem information according to a pre-trained problem classification model, and when obtaining a classification result, is configured to:
inputting the user question information into the question classification model, and acquiring the question category and the corresponding confidence of the user question information through the question classification model;
the judging module is used for judging whether the classification result meets the preset condition or not:
and judging whether the confidence coefficient meets a preset confidence coefficient threshold value.
In one possible design, the problem classification model includes a word vector acquisition submodel and a word vector classification submodel;
the classification module is configured to, when the problem classification and the corresponding confidence of the user problem information are obtained through the problem classification model:
obtaining a sub-model through the word vector, segmenting words of the user problem information, and obtaining the word vector of the user problem information according to a word segmentation result;
And classifying the word vectors of the user problem information through the word vector classification submodel to obtain the problem category and the corresponding confidence of the user problem information.
In one possible design, the classification module is configured to, when classifying the word vectors of the user question information through the word vector classification submodel and obtaining the question categories and the corresponding confidence degrees of the user question information,:
and performing similarity matching on the word vector of the user problem information and a preset word vector corresponding to a preset problem category through the word vector classification submodel, acquiring the preset problem category with the highest similarity as the problem category of the user problem information, and taking the similarity as the confidence.
In one possible design, the word vector classification submodel is a deep learning model.
In one possible design, the answer generation model is a Seq2Seq model;
the second answer obtaining module is configured to, when inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information:
and encoding the user question information through an encoder of the Seq2Seq model, and decoding through a decoder of the Seq2Seq model to obtain answer information corresponding to the user question information.
In one possible design, the encoder of the Seq2Seq model employs an LSTM model, a GRU model, or a Transformer model;
the decoder of the Seq2Seq model adopts an LSTM model, a GRU model or a Transformer model.
In one possible design, the apparatus further includes a first training module to:
taking a question sample included in each question category in the preset question-answering library as first training data, and training the word vector acquisition submodel;
and acquiring a word vector of each first training data according to the trained word vector acquisition submodel, taking the word vector as second training data, and training the word vector classification submodel until the word vector classification submodel is converged.
In one possible design, the apparatus further includes a second training module to:
and taking the question sample included in each question category in the preset question-answering library and the corresponding answer information as third training data, and training the Seq2Seq model.
In one possible design, the apparatus further includes a question-and-answer library construction module configured to:
acquiring historical problems, classifying the historical problems, and acquiring related approximate problems of each problem category and supplementing the related approximate problems to the corresponding problem categories;
Acquiring corresponding answer information for each question category;
and constructing the preset question-answer library according to the questions included in each question category and the corresponding answer information.
A third aspect of an embodiment of the present invention provides an intelligent interaction device, including:
a memory;
a processor; and
a computer program;
wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method of the first aspect.
A fourth aspect of embodiments of the present invention is to provide a computer-readable storage medium having stored thereon a computer program;
which when executed by a processor implements the method according to the first aspect.
A fifth aspect of embodiments of the present invention provides a computer program comprising program code for performing the method according to the first aspect when the computer program is run by a computer.
According to the intelligent interaction method, the intelligent interaction device, the intelligent interaction equipment and the intelligent interaction storage medium, the user problem information is obtained according to the user instruction by acquiring the user instruction; performing problem category classification on user problem information according to a pre-trained problem classification model to obtain a classification result, and judging whether the classification result meets a preset condition; if the classification result meets the preset condition, acquiring corresponding answer information from a preset question-answer library according to the classification result; if the classification result is determined not to meet the preset condition, inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information; and outputting the acquired answer information to respond to the user instruction. According to the method, firstly, the user question information is subjected to semantic classification through the question classification model to obtain the classification result, when the classification result is determined to be relatively accurate, the answer information corresponding to the question category to which the user question information belongs is retrieved from the preset question-answer library according to the classification result, the accuracy of the retrieval process can be improved, when the classification result is determined to be inaccurate, the answer generation model is adopted to obtain the answer information, the accuracy of obtaining the answer can be improved through the fusion of the two modes, the problem that the accurate answer cannot be obtained when the problem cannot be accurately matched based on the retrieval mode is avoided, and further the user experience in the interaction process can be improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a system architecture diagram of an intelligent interaction method according to an embodiment of the present invention;
FIG. 2 is a flowchart of an intelligent interaction method according to an embodiment of the present invention;
FIG. 3 is a flowchart of an intelligent interaction method according to another embodiment of the present invention;
FIG. 4 is a flowchart of an intelligent interaction method according to another embodiment of the present invention;
FIG. 5 is a flowchart of an intelligent interaction method according to another embodiment of the present invention;
FIG. 6 is a block diagram of an intelligent interaction device according to an embodiment of the present invention;
fig. 7 is a structural diagram of an intelligent interactive device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments of the present invention without any creative efforts shall fall within the protection scope of the embodiments of the present invention.
Existing intelligent interaction methods typically employ a search-based approach. The existing retrieval-based method adopts text similarity for retrieval, has poor retrieval effect, and often cannot search matched questions in a question-answer library for some long question sentences or complex question sentences, so that answers cannot be accurately acquired. In order to obtain more accurate answers to user questions and return the answers to the user, a method generated through a model is fused on the basis of a retrieval-based method, firstly, semantic classification is carried out on user question information to obtain a classification result, questions with the same (or similar) semantics but different expression modes are classified into a question category, answer information corresponding to the question category to which the user question information belongs is retrieved from a preset question-answer library according to the classification result when the classification result is determined to be relatively accurate, and the accuracy of the retrieval process can be improved; and when the classification result is determined to be inaccurate, the answer generation model is adopted to obtain the answer information, the accuracy of obtaining the answer can be improved through the fusion of two modes, the problem that the accurate answer cannot be obtained when the problem cannot be accurately matched based on a retrieval mode is avoided, and the user experience in the interaction process can be further improved.
The method provided by the embodiment of the invention can be applied to the communication system shown in figure 1. As shown in fig. 1, the communication system includes a collecting device 10 and a processing device 11, where the collecting device 10 may be configured to collect a user instruction and send the user instruction to the processing device 11, and the processing device 11 may be any device capable of executing a flow of the intelligent interaction method provided in the embodiment of the present invention, and after receiving the user instruction sent by the collecting device 10, the processing device obtains answer information by the method of the present invention and outputs the answer information, so as to respond to the user instruction.
The intelligent interaction process is described in detail below with reference to specific embodiments.
Fig. 2 is a flowchart of an intelligent interaction method according to an embodiment of the present invention. The embodiment provides an intelligent interaction method, an execution subject of which is the processing device in fig. 1, as shown in fig. 2, the intelligent interaction method includes the following specific steps:
s201, collecting a user instruction, and acquiring user problem information according to the user instruction.
In this embodiment, the user instruction may be a voice instruction, which may be collected by a voice collecting device such as a microphone, and the voice instruction is converted into a text by using technologies such as voice recognition, so that the user problem information may be obtained, and the specific process of converting the voice instruction into the text is not described here again. In addition, the user command may also be a text command, for example, the user may input the user command through an input device such as a keyboard, and the user question information may be acquired from the text command.
S202, performing problem category classification on the user problem information according to a pre-trained problem classification model to obtain a classification result, and judging whether the classification result meets a preset condition.
In this embodiment, the question classification model is used for classifying question categories, and may classify questions with the same (or similar) semantics but different expression modes into one question category, and then subsequently retrieve answer information corresponding to the question category from a preset question-and-answer library. The problem classification model can be any classification model and can be trained offline through training data in advance.
After the user question information is obtained, the user question information is firstly input into a pre-trained question classification model, and a classification result is obtained through the question classification model.
Considering the error of the question classification model, some user questions may not be accurately classified into a certain question category, for example, a certain user question may be similar to more than two question categories at the same time, or the confidence of the certain user question classified into a certain question category is low, and then the method of retrieving answers from the preset question-answer library through the question categories still is still adopted at this time, which may cause the obtained answer information to be inaccurate, so in this embodiment, after the classification result output by the question classification model is obtained, the classification result is judged, whether the classification result meets the predetermined condition is judged, if the classification result meets the predetermined condition, the classification result is relatively accurate, and at this time, it is enough to retrieve the answer information from the preset question-answer library by using S203; if the classification result does not meet the predetermined condition, it is indicated that the classification result may be inaccurate, at this time, answer information is obtained by adopting S204 through the pre-trained answer generation model, and at this time, the obtained answer information is more accurate. When it needs to be explained, the classification result is relatively accurate in practical application, that is, in most cases, answer information can be retrieved from the preset question-answer library through S203, and in some cases, answer information is obtained through the answer generation model through S204.
S203, if the classification result meets the preset condition, acquiring corresponding answer information from a preset question-answer library according to the classification result;
the preset question-answer library comprises different question categories and answer information corresponding to the question categories.
In this embodiment, the preset question-answer library may include different question categories, each question category may include question samples having the same (or similar) semantics but different expression manners, and each question category corresponds to answer information, where the answer information may be in a text form, and certainly the answer information may also be in a multimedia form such as audio, video, picture, and the like, and one question category may correspond to one answer information, and certainly also may correspond to multiple answer information, and when obtaining the answer information, one of the answer information may be selected to increase the diversity of answers. Optionally, in this embodiment, the search process may apply a search engine, load the preset question-and-answer library into the search engine in advance, and search from the preset question-and-answer library through the search engine.
In this embodiment, when the classification result meets the predetermined condition, it is described that the classification result is relatively accurate, and at this time, answer information corresponding to the question category to which the user question information belongs is directly retrieved from a preset question-answer library according to the classification result, and at this time, the obtained answer information is relatively accurate. In the embodiment, the question category classification is performed on the user question information according to semantics, and then the answers are retrieved from the question-answer library according to the question categories, so that the answer retrieval effect is better compared with the answer retrieval effect in the prior art by directly retrieving answers from the question-answer library through text similarity, and particularly, the problem that a long question or a complex question cannot be accurately retrieved can be effectively solved.
Optionally, the preset question-answer library in this embodiment may be obtained in advance through the following processes:
firstly, acquiring historical problems, classifying the historical problems, and acquiring related approximate problems of each problem category and supplementing the related approximate problems into the corresponding problem category; acquiring corresponding answer information for each question category; and constructing the preset question-answer library according to the questions included in each question category and the corresponding answer information.
In this embodiment, historical problems (high-frequency historical problems may be obtained) may be obtained from a historical voice interaction log, the historical problems are classified by a human or a machine, problem categories are labeled, and related approximate problems of each problem category are obtained by a human or a machine and supplemented to the corresponding problem category, so that the problem samples in each problem category are sufficiently abundant, for example, for the historical problems "hello" and "hello", the problem categories are "greetings (greetings)", and the related approximate problems such as "hello" and "hello" may be obtained as supplementary problem samples of the "greetings (greetings)" problem category.
And S204, if the classification result is determined not to meet the preset condition, inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information.
In this embodiment, when the classification result does not satisfy the predetermined condition, it indicates that the classification result may be inaccurate, and then, the question-answer library is retrieved according to the classification result, which may cause the obtained answer information to be inaccurate.
In an alternative embodiment, the answer generation model may be a Sequence-to-Sequence (Seq 2Seq) model or other model capable of automatically generating answers. Wherein the Seq2Seq model is a network of an Encoder-Decoder structure, the input of which is a sequence and the output of which is a sequence. In an encoder, a sequence is converted into a vector with a fixed length, and then the vector is converted into a required sequence through a decoder and output, and the Seq2Seq model is widely applied to the fields of machine translation, question answering systems and the like.
In this embodiment, the user question information may be encoded by the encoder of the Seq2Seq model, and then decoded by the decoder of the Seq2Seq model, so as to obtain answer information corresponding to the user question information, and the specific process is not described in detail here.
The encoder and decoder of the Seq2Seq model may adopt a general RNN (recurrent neural network) model. In order to reduce the gradient decay problem existing in the RNN and the problem that the accuracy of generating the answer is low due to too Long input user question information, an LSTM (Long Short-Term Memory), a GRU (Gated recursive Units) or a Transformer machine translation model may be used to replace the RNN model of the encoder and/or decoder, or to optimally adjust the RNN model.
And S205, outputting the acquired answer information to respond to the user instruction.
In this embodiment, after the answer information is obtained through S203 or S204, the answer to the question may be output, so as to correspond to the user instruction, specifically, for example, the answer information is in a text form, and may be converted into a voice before performing voice broadcast, and if the answer information is in a multimedia form, the answer information may be directly played, for example, the audio may be directly subjected to voice broadcast, a picture or a video may be displayed through a display screen, and of course, the answer information in the text form may be displayed through the display screen.
According to the intelligent interaction method provided by the embodiment, the user problem information is obtained according to the user instruction by acquiring the user instruction; performing problem category classification on user problem information according to a pre-trained problem classification model to obtain a classification result, and judging whether the classification result meets a preset condition; if the classification result meets the preset condition, acquiring corresponding answer information from a preset question-answer library according to the classification result; if the classification result is determined not to meet the preset condition, inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information; and outputting the acquired answer information to respond to the user instruction. According to the method, firstly, the user question information is subjected to semantic classification through the question classification model to obtain the classification result, when the classification result is determined to be relatively accurate, the answer information corresponding to the question category to which the user question information belongs is retrieved from the preset question-answer library according to the classification result, the accuracy of the retrieval process can be improved, when the classification result is determined to be inaccurate, the answer generation model is adopted to obtain the answer information, the accuracy of obtaining the answer can be improved through the fusion of the two modes, the problem that the accurate answer cannot be obtained when the problem cannot be accurately matched based on the retrieval mode is avoided, and further the user experience in the interaction process can be improved.
On the basis of the foregoing embodiment, as shown in fig. 3, the step S202 of performing problem category classification on the user problem information according to a pre-trained problem classification model to obtain a classification result, and determining whether the classification result meets a predetermined condition may specifically include:
s301, inputting the user question information into the question classification model, and obtaining the question category and the corresponding confidence of the user question information through the question classification model.
In this embodiment, the input of the question classification model is user question information, and the output classification result is a question category and a corresponding confidence of the user question information.
Specifically, the problem classification model includes a Word vector obtaining sub-model and a Word vector classification sub-model, optionally, the Word vector obtaining sub-model may be a Word2vec (Word to vector) model, and may segment the input user problem information and obtain Word vectors; the word vector classification submodel is a deep learning model, such as a CNN (convolutional neural Networks) classification model, an LSTM classification model, a transform-based classification model, and the like.
Further, as shown in fig. 4, the S301 may specifically include:
S3011, obtaining a sub-model through the word vector, segmenting words of the user question information, and obtaining the word vector of the user question information according to a word segmentation result;
s3012, classifying the word vectors of the user problem information through the word vector classification submodel to obtain the problem category and the corresponding confidence of the user problem information.
More specifically, in this embodiment, the word vector of the user question information and the preset word vector corresponding to the preset question category are subjected to similarity matching through the word vector classification submodel, the preset question category with the highest similarity is obtained as the question category of the user question information, and the similarity is used as the confidence.
S302, judging whether the confidence coefficient meets a preset confidence coefficient threshold value.
In this embodiment, the confidence obtained in the above process is compared with a predetermined confidence threshold, and if the confidence is greater than the predetermined confidence threshold, corresponding answer information may be obtained from a preset question-answer library according to the question category of the user question information; if the confidence is not greater than the preset confidence threshold, the user question information can be input into an answer generation model to obtain corresponding answer information.
On the basis of any of the above embodiments, the intelligent interaction method may also perform offline training of the problem classification model in advance, as shown in fig. 5, the specific training process may be as follows:
s401, taking a question sample included in each question category in the preset question-answering library as first training data, and training the word vector acquisition submodel;
s402, obtaining word vectors of the first training data according to the trained word vector obtaining submodel, using the word vectors as second training data, and training the word vector classification submodel until the word vector classification submodel is converged.
In this embodiment, since the preset question-answer library includes different question classifications, and each question classification includes a plurality of question samples (including historical questions obtained from logs and supplementary related approximate questions), the question samples classified into question categories in the preset question-answer library may be used as training data (i.e., first training data), and first, a Word vector obtaining sub-model may be trained, i.e., a Word2vec model may be trained, so that the Word2vec model may accurately perform Word segmentation and obtain Word vectors; further, after the training of the word vector obtaining submodel is completed, the word vector of the first training data is obtained through the word vector obtaining submodel, the word vector of the first training data is used as the training data of the word vector classification submodel, the word vector of the first training data is input into the word vector classification submodel (a CNN classification model, an LSTM classification model or a transformation-based classification model) to be trained until the word vector classification submodel is converged, namely a loss function of a past model in the training process, when the loss function is converged, the word vector classification submodel is converged, and at this time, the training of the word vector classification submodel can be completed.
On the basis of any of the above embodiments, the intelligent interaction method may also perform offline training of an answer generation model in advance, and the specific training process may be as follows:
and taking the question sample included in each question category in the preset question-answering library and the corresponding answer information as third training data, and training the Seq2Seq model.
In this embodiment, each question type in the preset question-answer library includes a question sample and corresponding answer information, the question sample and the corresponding answer information may be used as training data of an answer generation model, and a specific training process for the Seq2Seq model may adopt an existing training method, which is not described herein again.
Fig. 6 is a structural diagram of an intelligent interaction device according to an embodiment of the present invention. The intelligent interaction device provided in this embodiment may execute the processing flow provided in the intelligent interaction method embodiment, as shown in fig. 3, the intelligent interaction device 600 includes an acquisition module 601, a classification module 602, a judgment module 603, a first answer obtaining module 604, a second answer obtaining module 605, and an output module 606.
The acquisition module 601 is used for acquiring a user instruction and acquiring user problem information according to the user instruction;
A classification module 602, configured to perform problem category classification on the user problem information according to a pre-trained problem classification model to obtain a classification result;
a judging module 603, configured to judge whether the classification result meets a predetermined condition;
a first answer obtaining module 604, configured to, if it is determined that the classification result meets the predetermined condition, obtain corresponding answer information from a preset question-and-answer library according to the classification result, where the preset question-and-answer library includes different question categories and answer information corresponding to each question category;
a second answer obtaining module 605, configured to, if it is determined that the classification result does not satisfy the predetermined condition, input the user question information into a pre-trained answer generation model to obtain corresponding answer information;
and an output module 606, configured to output the obtained answer information to respond to the user instruction.
On the basis of the foregoing embodiment, the classification module 602 performs problem category classification on the user problem information according to a pre-trained problem classification model, and when obtaining a classification result, is configured to:
inputting the user question information into the question classification model, and acquiring the question category and the corresponding confidence of the user question information through the question classification model;
The determining module 603, when determining whether the classification result satisfies a predetermined condition, is configured to:
and judging whether the confidence coefficient meets a preset confidence coefficient threshold value.
On the basis of any one of the above embodiments, the problem classification model includes a word vector acquisition submodel and a word vector classification submodel;
the classification module 602, when obtaining the question category and the corresponding confidence of the user question information through the question classification model, is configured to:
obtaining a sub-model through the word vector, segmenting words of the user problem information, and obtaining the word vector of the user problem information according to a word segmentation result;
and classifying the word vectors of the user problem information through the word vector classification submodel to obtain the problem category and the corresponding confidence of the user problem information.
On the basis of any of the above embodiments, the classification module 602, when classifying the word vector of the user question information through the word vector classification submodel and obtaining the question category and the corresponding confidence of the user question information, is configured to:
and performing similarity matching on the word vector of the user problem information and a preset word vector corresponding to a preset problem category through the word vector classification submodel, acquiring the preset problem category with the highest similarity as the problem category of the user problem information, and taking the similarity as the confidence.
On the basis of any one of the above embodiments, the word vector classification submodel is a deep learning model.
On the basis of any one of the above embodiments, the answer generation model is a Seq2Seq model;
when the second answer obtaining module 605 inputs the user question information into a pre-trained answer generation model to obtain corresponding answer information, the second answer obtaining module is configured to:
and encoding the user question information through an encoder of the Seq2Seq model, and decoding through a decoder of the Seq2Seq model to obtain answer information corresponding to the user question information.
On the basis of any one of the above embodiments, the encoder of the Seq2Seq model adopts an LSTM model, a GRU model or a Transformer model;
the decoder of the Seq2Seq model adopts an LSTM model, a GRU model or a Transformer model.
On the basis of any of the above embodiments, the apparatus further includes a first training module configured to:
taking a question sample included in each question category in the preset question-answering library as first training data, and training the word vector acquisition submodel;
and acquiring a word vector of each first training data according to the trained word vector acquisition submodel, taking the word vector as second training data, and training the word vector classification submodel until the word vector classification submodel is converged.
On the basis of any of the above embodiments, the apparatus further includes a second training module configured to:
and taking the question sample included in each question category in the preset question-answering library and the corresponding answer information as third training data, and training the Seq2Seq model.
On the basis of any of the above embodiments, the apparatus further includes a question-answer library construction module, configured to:
acquiring historical problems, classifying the historical problems, and acquiring related approximate problems of each problem category and supplementing the related approximate problems to the corresponding problem categories;
acquiring corresponding answer information for each question category;
and constructing the preset question-answer library according to the questions included in each question category and the corresponding answer information.
The intelligent interaction device provided in the embodiment of the present invention may be specifically configured to execute the embodiment of the intelligent interaction method provided in fig. 1, and specific functions are not described herein again.
According to the intelligent interaction device provided by the embodiment of the invention, user problem information is obtained according to a user instruction by acquiring the user instruction; performing problem category classification on user problem information according to a pre-trained problem classification model to obtain a classification result, and judging whether the classification result meets a preset condition; if the classification result meets the preset condition, acquiring corresponding answer information from a preset question-answer library according to the classification result; if the classification result is determined not to meet the preset condition, inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information; and outputting the acquired answer information to respond to the user instruction. According to the method, firstly, the user question information is subjected to semantic classification through the question classification model to obtain the classification result, when the classification result is determined to be relatively accurate, the answer information corresponding to the question category to which the user question information belongs is retrieved from the preset question-answer library according to the classification result, the accuracy of the retrieval process can be improved, when the classification result is determined to be inaccurate, the answer generation model is adopted to obtain the answer information, the accuracy of obtaining the answer can be improved through the fusion of the two modes, the problem that the accurate answer cannot be obtained when the problem cannot be accurately matched based on the retrieval mode is avoided, and further the user experience in the interaction process can be improved.
Fig. 7 is a schematic structural diagram of an intelligent interaction device according to an embodiment of the present invention. The intelligent interaction device provided by the embodiment of the present invention may execute the processing flow provided by the intelligent interaction method embodiment, as shown in fig. 7, the intelligent interaction device 70 includes a memory 71, a processor 72, a computer program, and a communication interface 73; wherein the computer program is stored in the memory 71 and is configured to be executed by the processor 72 for performing the intelligent interaction method as described in the above embodiments.
The intelligent interaction device in the embodiment shown in fig. 7 may be used to implement the technical solution of the above method embodiment, and the implementation principle and technical effect are similar, which are not described herein again.
According to the intelligent interaction device provided by the embodiment, the user problem information is obtained according to the user instruction by acquiring the user instruction; performing problem category classification on user problem information according to a pre-trained problem classification model to obtain a classification result, and judging whether the classification result meets a preset condition; if the classification result meets the preset condition, acquiring corresponding answer information from a preset question-answer library according to the classification result; if the classification result is determined not to meet the preset condition, inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information; and outputting the acquired answer information to respond to the user instruction. According to the method, firstly, the user question information is subjected to semantic classification through the question classification model to obtain the classification result, when the classification result is determined to be relatively accurate, the answer information corresponding to the question category to which the user question information belongs is retrieved from the preset question-answer library according to the classification result, the accuracy of the retrieval process can be improved, when the classification result is determined to be inaccurate, the answer generation model is adopted to obtain the answer information, the accuracy of obtaining the answer can be improved through the fusion of the two modes, the problem that the accurate answer cannot be obtained when the problem cannot be accurately matched based on the retrieval mode is avoided, and further the user experience in the interaction process can be improved.
In addition, the present embodiment also provides a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to implement the intelligent interaction method described in the above embodiments.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional unit.
The integrated unit implemented in the form of a software functional unit may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute some steps of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
It is obvious to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional modules is merely used as an example, and in practical applications, the above function distribution may be performed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to perform all or part of the above described functions. For the specific working process of the device described above, reference may be made to the corresponding process in the foregoing method embodiment, which is not described herein again.
Finally, it should be noted that: the above embodiments are only used for illustrating the technical solutions of the embodiments of the present invention, and are not limited thereto; although embodiments of the present invention have been described in detail with reference to the foregoing embodiments, those skilled in the art will understand that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (22)
1. An intelligent interaction method, comprising:
acquiring a user instruction, and acquiring user problem information according to the user instruction;
performing problem category classification on the user problem information according to a pre-trained problem classification model to obtain a classification result, and judging whether the classification result meets a preset condition;
if the classification result meets the preset condition, acquiring corresponding answer information from a preset question-answer library according to the classification result, wherein the preset question-answer library comprises different question categories and answer information corresponding to the question categories;
If the classification result is determined not to meet the preset condition, inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information;
and outputting the acquired answer information to respond to the user instruction.
2. The method according to claim 1, wherein the performing problem category classification on the user problem information according to a pre-trained problem classification model to obtain a classification result, and determining whether the classification result satisfies a predetermined condition includes:
inputting the user question information into the question classification model, and acquiring the question category and the corresponding confidence of the user question information through the question classification model;
and judging whether the confidence coefficient meets a preset confidence coefficient threshold value.
3. The method of claim 2, wherein the problem classification model comprises a word vector acquisition submodel and a word vector classification submodel;
the obtaining of the question category and the corresponding confidence of the user question information through the question classification model includes:
obtaining a sub-model through the word vector, segmenting words of the user problem information, and obtaining the word vector of the user problem information according to a word segmentation result;
And classifying the word vectors of the user problem information through the word vector classification submodel to obtain the problem category and the corresponding confidence of the user problem information.
4. The method of claim 3, wherein the classifying the word vectors of the user question information by the word vector classification submodel to obtain the question categories and the corresponding confidences of the user question information comprises:
and performing similarity matching on the word vector of the user problem information and a preset word vector corresponding to a preset problem category through the word vector classification submodel, acquiring the preset problem category with the highest similarity as the problem category of the user problem information, and taking the similarity as the confidence.
5. The method of claim 3 or 4, wherein the word vector classification submodel is a deep learning model.
6. The method according to any one of claims 1-4, wherein the answer generation model is a Seq2Seq model;
the inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information includes:
And encoding the user question information through an encoder of the Seq2Seq model, and decoding through a decoder of the Seq2Seq model to obtain answer information corresponding to the user question information.
7. The method of claim 6,
an encoder of the Seq2Seq model adopts an LSTM model, a GRU model or a Transformer model;
the decoder of the Seq2Seq model adopts an LSTM model, a GRU model or a Transformer model.
8. The method of claim 3 or 4, further comprising:
taking a question sample included in each question category in the preset question-answering library as first training data, and training the word vector acquisition submodel;
and acquiring a word vector of each first training data according to the trained word vector acquisition submodel, taking the word vector as second training data, and training the word vector classification submodel until the word vector classification submodel is converged.
9. The method of claim 6, further comprising:
and taking the question sample included in each question category in the preset question-answering library and the corresponding answer information as third training data, and training the Seq2Seq model.
10. The method of claim 1, further comprising:
acquiring historical problems, classifying the historical problems, and acquiring related approximate problems of each problem category and supplementing the related approximate problems to the corresponding problem categories;
acquiring corresponding answer information for each question category;
and constructing the preset question-answer library according to the questions included in each question category and the corresponding answer information.
11. An intelligent interaction device, comprising:
the acquisition module is used for acquiring a user instruction and acquiring user problem information according to the user instruction;
the classification module is used for performing problem category classification on the user problem information according to a pre-trained problem classification model to obtain a classification result;
the judging module is used for judging whether the classification result meets a preset condition or not;
a first answer obtaining module, configured to, if it is determined that the classification result meets the predetermined condition, obtain corresponding answer information from a preset question-and-answer library according to the classification result, where the preset question-and-answer library includes different question categories and answer information corresponding to each question category;
the second answer obtaining module is used for inputting the user question information into a pre-trained answer generating model to obtain corresponding answer information if the classification result is determined not to meet the preset condition;
And the output module is used for outputting the acquired answer information so as to respond to the user instruction.
12. The apparatus of claim 11, wherein the classification module, when performing problem category classification on the user problem information according to a pre-trained problem classification model to obtain a classification result, is configured to:
inputting the user question information into the question classification model, and acquiring the question category and the corresponding confidence of the user question information through the question classification model;
the judging module is used for judging whether the classification result meets the preset condition or not:
and judging whether the confidence coefficient meets a preset confidence coefficient threshold value.
13. The apparatus of claim 12, wherein the problem classification model comprises a word vector acquisition submodel and a word vector classification submodel;
the classification module is configured to, when the problem classification and the corresponding confidence of the user problem information are obtained through the problem classification model:
obtaining a sub-model through the word vector, segmenting words of the user problem information, and obtaining the word vector of the user problem information according to a word segmentation result;
And classifying the word vectors of the user problem information through the word vector classification submodel to obtain the problem category and the corresponding confidence of the user problem information.
14. The apparatus of claim 13, wherein the classification module, when classifying the word vector of the user question information through the word vector classification submodel to obtain the question category and the corresponding confidence of the user question information, is configured to:
and performing similarity matching on the word vector of the user problem information and a preset word vector corresponding to a preset problem category through the word vector classification submodel, acquiring the preset problem category with the highest similarity as the problem category of the user problem information, and taking the similarity as the confidence.
15. The apparatus of claim 13 or 14, wherein the word vector classification submodel is a deep learning model.
16. The apparatus according to any one of claims 11-14, wherein the answer generation model is a Seq2Seq model;
the second answer obtaining module is configured to, when inputting the user question information into a pre-trained answer generation model to obtain corresponding answer information:
And encoding the user question information through an encoder of the Seq2Seq model, and decoding through a decoder of the Seq2Seq model to obtain answer information corresponding to the user question information.
17. The apparatus of claim 16,
an encoder of the Seq2Seq model adopts an LSTM model, a GRU model or a Transformer model;
the decoder of the Seq2Seq model adopts an LSTM model, a GRU model or a Transformer model.
18. The apparatus of claim 13 or 14, further comprising a first training module to:
taking a question sample included in each question category in the preset question-answering library as first training data, and training the word vector acquisition submodel;
and acquiring a word vector of each first training data according to the trained word vector acquisition submodel, taking the word vector as second training data, and training the word vector classification submodel until the word vector classification submodel is converged.
19. The apparatus of claim 16, further comprising a second training module to:
and taking the question sample included in each question category in the preset question-answering library and the corresponding answer information as third training data, and training the Seq2Seq model.
20. The apparatus of claim 11, further comprising a question-and-answer library construction module configured to:
acquiring historical problems, classifying the historical problems, and acquiring related approximate problems of each problem category and supplementing the related approximate problems to the corresponding problem categories;
acquiring corresponding answer information for each question category;
and constructing the preset question-answer library according to the questions included in each question category and the corresponding answer information.
21. An intelligent interaction device, comprising:
a memory;
a processor; and
a computer program;
wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method of any one of claims 1-10.
22. A computer-readable storage medium, having stored thereon a computer program;
the computer program, when executed by a processor, implementing the method of any one of claims 1-10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010386988.5A CN111858875A (en) | 2020-05-09 | 2020-05-09 | Intelligent interaction method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010386988.5A CN111858875A (en) | 2020-05-09 | 2020-05-09 | Intelligent interaction method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111858875A true CN111858875A (en) | 2020-10-30 |
Family
ID=72985736
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010386988.5A Pending CN111858875A (en) | 2020-05-09 | 2020-05-09 | Intelligent interaction method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111858875A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112364128A (en) * | 2020-11-06 | 2021-02-12 | 北京乐学帮网络技术有限公司 | Information processing method and device, computer equipment and storage medium |
| CN112883173A (en) * | 2021-02-08 | 2021-06-01 | 联想(北京)有限公司 | Text response method and device |
| JP7370115B1 (en) | 2023-04-26 | 2023-10-27 | 堺財経電算合同会社 | Answering device and answering method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107329967A (en) * | 2017-05-12 | 2017-11-07 | 北京邮电大学 | Question answering system and method based on deep learning |
| KR20170127882A (en) * | 2016-05-13 | 2017-11-22 | 한국전자통신연구원 | Question answering apparatus and method for processing of truth or false type question |
| US20190087408A1 (en) * | 2017-09-15 | 2019-03-21 | International Business Machines Corporation | Training data update |
| CN109783622A (en) * | 2018-12-20 | 2019-05-21 | 出门问问信息科技有限公司 | One kind determining problem answers method, apparatus and electronic equipment based on Question Classification |
| CN109947909A (en) * | 2018-06-19 | 2019-06-28 | 平安科技（深圳）有限公司 | Intelligent customer service answer method, equipment, storage medium and device |
| JP6537211B1 (en) * | 2018-07-06 | 2019-07-03 | Ｊｅインターナショナル株式会社 | Search device and program |
-
2020
- 2020-05-09 CN CN202010386988.5A patent/CN111858875A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20170127882A (en) * | 2016-05-13 | 2017-11-22 | 한국전자통신연구원 | Question answering apparatus and method for processing of truth or false type question |
| CN107329967A (en) * | 2017-05-12 | 2017-11-07 | 北京邮电大学 | Question answering system and method based on deep learning |
| US20190087408A1 (en) * | 2017-09-15 | 2019-03-21 | International Business Machines Corporation | Training data update |
| CN109947909A (en) * | 2018-06-19 | 2019-06-28 | 平安科技（深圳）有限公司 | Intelligent customer service answer method, equipment, storage medium and device |
| JP6537211B1 (en) * | 2018-07-06 | 2019-07-03 | Ｊｅインターナショナル株式会社 | Search device and program |
| CN109783622A (en) * | 2018-12-20 | 2019-05-21 | 出门问问信息科技有限公司 | One kind determining problem answers method, apparatus and electronic equipment based on Question Classification |
Non-Patent Citations (2)
| Title |
|---|
| 张小艳;: "中文自动答疑系统的研究与实现", 微计算机信息, no. 36 * |
| 徐健;张栋;李寿山;王红玲;: "基于双语信息的问题分类方法研究", 中文信息学报, no. 05 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112364128A (en) * | 2020-11-06 | 2021-02-12 | 北京乐学帮网络技术有限公司 | Information processing method and device, computer equipment and storage medium |
| CN112883173A (en) * | 2021-02-08 | 2021-06-01 | 联想(北京)有限公司 | Text response method and device |
| JP7370115B1 (en) | 2023-04-26 | 2023-10-27 | 堺財経電算合同会社 | Answering device and answering method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108829757B (en) | Intelligent service method, server and storage medium for chat robot | |
| CN111339283B (en) | Method and device for providing customer service answers aiming at user questions | |
| CN111858875A (en) | Intelligent interaction method, device, equipment and storage medium | |
| CN109992669B (en) | Keyword question-answering method based on language model and reinforcement learning | |
| CN112818109B (en) | Intelligent reply method, medium, device and computing equipment for mail | |
| CN113505198B (en) | Keyword-driven generation type dialogue reply method and device and electronic equipment | |
| CN111898369B (en) | Article title generation method, model training method and device and electronic equipment | |
| CN113392265A (en) | Multimedia processing method, device and equipment | |
| CN111026840A (en) | Text processing method, device, server and storage medium | |
| CN112069781A (en) | Comment generation method and device, terminal device and storage medium | |
| CN115146124A (en) | Question-answering system response method and device, equipment, medium and product thereof | |
| CN117332789A (en) | Semantic analysis method and system for dialogue scene | |
| TWI734085B (en) | Dialogue system using intention detection ensemble learning and method thereof | |
| CN112084788A (en) | Automatic marking method and system for implicit emotional tendency of image captions | |
| CN116978367A (en) | Speech recognition method, device, electronic equipment and storage medium | |
| CN113408287B (en) | Entity identification method and device, electronic equipment and storage medium | |
| CN115168558A (en) | Method for realizing multi-round man-machine conversation | |
| CN112035648B (en) | User data processing method and device and electronic equipment | |
| CN114330701A (en) | Model training method, device, computer equipment, storage medium and program product | |
| CN112749554A (en) | Method, device and equipment for determining text matching degree and storage medium | |
| CN116775848B (en) | Control method, device, computing equipment and storage medium for generating dialogue information | |
| CN117453895B (en) | Intelligent customer service response method, device, equipment and readable storage medium | |
| CN114048319B (en) | Humor text classification method, device, equipment and medium based on attention mechanism | |
| CN114117034B (en) | Method and device for pushing texts of different styles based on intelligent model | |
| CN116844530A (en) | Voice information processing method and device, readable storage medium and electronic device |
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 |