CN111090726A - NLP-based electric power industry character customer service interaction method - Google Patents
NLP-based electric power industry character customer service interaction method Download PDFInfo
- Publication number
- CN111090726A CN111090726A CN201911226062.3A CN201911226062A CN111090726A CN 111090726 A CN111090726 A CN 111090726A CN 201911226062 A CN201911226062 A CN 201911226062A CN 111090726 A CN111090726 A CN 111090726A
- Authority
- CN
- China
- Prior art keywords
- information
- semantic
- voice
- database
- customer service
- 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 20
- 230000003993 interaction Effects 0.000 title claims abstract description 11
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 230000006403 short-term memory Effects 0.000 claims description 9
- 230000002457 bidirectional effect Effects 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 8
- 230000007246 mechanism Effects 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 2
- 238000000605 extraction Methods 0.000 claims description 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 238000003058 natural language processing Methods 0.000 abstract 2
- 238000004458 analytical method Methods 0.000 abstract 1
- 238000004891 communication Methods 0.000 abstract 1
- 230000007787 long-term memory Effects 0.000 description 8
- 238000011161 development Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 239000000203 mixture Substances 0.000 description 3
- 230000002452 interceptive effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241000238558 Eucarida Species 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000003786 synthesis reaction Methods 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
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/02—Feature extraction for speech recognition; Selection of recognition unit
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/08—Speech classification or search
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
- G10L2015/225—Feedback of the input speech
Abstract
The invention relates to the technical field of artificial intelligence customer service and intelligent voice analysis, and particularly discloses a character customer service interaction method in the power industry based on NLP, which specifically comprises the following steps: firstly, preprocessing voice data; step two, voice recognition and a natural generation type question-answering system; step three, generating a semantic database; step four, semantic matching; step five, generating characters by a natural language processing method; step six, outputting the characters; by adopting intelligent voice text interaction, the communication efficiency and experience of customers can be improved, the labor cost is reduced, and the method has higher research value.
Description
Technical Field
The invention relates to the technical field of voice recognition interaction, in particular to a character customer service interaction method based on NLP in the power industry.
Background
The electric power industry is one of the most important basic industries in China, and great progress and development are achieved in recent years. With the increase of the user base number, the manual service pressure is huge in the field of power grid customer service. With the improvement of the technical level of artificial intelligence, intelligent customer service has become an important development direction in the customer service field. The method is continuously promoted in the fields of interactive experience, function perception and scene service, and can gradually strengthen and exceed the traditional service. Due to the complexity of the industry in the field of power grids, the application of intelligent customer service is still in the exploration and starting stage and is not mature.
The intelligent voice technology is started from the publication of Siri of apple corporation in 2010, the development of the intelligent voice technology in China is basically and internationally synchronous, and with the successive establishment and development of various large voice enterprises, the intelligent voice technology represented by voice recognition and voice synthesis is continuously broken through, and the intelligent voice industry in China is rapidly developed. In the electric power electric wire netting field, also need follow the development of technique, fully combine technologies such As Speech Recognition (ASR), characters change pronunciation (TTS), natural semantic processing (NLP), be applied to electric power electric wire netting customer service field and realize intellectuality, the high efficiency alleviates the manpower burden to reduce the human cost, promote customer viscidity and customer experience.
Disclosure of Invention
In order to solve the above problems, the present invention provides an NLP-based method for interacting text and customer service in the power industry, which specifically comprises the following steps:
the method comprises the following steps: preprocessing input general voice signals and voice signals with region information to form training data, wherein the preprocessing comprises denoising processing and feature extraction processing;
step two: carrying out voice recognition on the preprocessed general voice signal to obtain semantic information, and generating a semantic library by all the semantic information; meanwhile, the preprocessed voice signal with the regional information is subjected to a natural generation type question-answering system consisting of sequence-to-sequence models to obtain syntax information and keyword information of sentences;
step three: fusing the syntactic information, the keyword information and the semantic library to form a database with syntactic information; step four: extracting key words in the client voice sentences by the voice recognition model through semantic understanding, and matching the key words with key words in a database;
step five: generating reply characters corresponding to the replies according to the answers matched with the database and by combining the syntactic information and the keyword information;
step six: and displaying the corresponding characters according to the reply voice to complete the interaction with the client.
Preferably, the database includes a plurality of customer question sentences, a keyword corresponding to each of the plurality of customer question sentences, a reply voice corresponding to each of the plurality of customer question sentences, and region information of the voices of the plurality of customers.
Preferably, the natural-generating question-answering system is composed of a Sequence-to-Sequence (Sequence-to-Sequence, Sequence 2Seq) model, and is trained by customer service standard questions and extended questions in the power industry.
Compared with the prior art, the invention has the beneficial effects that: the intelligent interaction in the customer service field of the power industry is realized, the problem of high repetition rate in the customer service field can be completely completed by the method, the semantic library is perfected by using the naturally-generated question-answering system, and the output interactive characters are more accurate and are more like sentences spoken by natural people. Not only improves the efficiency and the customer satisfaction, but also greatly reduces the labor cost.
Drawings
FIG. 1 is a block diagram showing the structure of the present invention.
FIG. 2 is a flow chart of speech processing according to the present invention.
FIG. 3 is a schematic diagram of a model for a natural-occurring question-answering system according to the present invention.
FIG. 4 is a schematic diagram of a general long-term and short-term memory network.
FIG. 5 is a schematic diagram of a bidirectional long term and short term memory network according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1 to 5, the present invention provides a customer service intelligent navigation voice data processing method based on artificial intelligence, which specifically includes the following steps:
step one, voice preprocessing, namely, processing modes such as pre-emphasis, windowing, discrete Fourier transform, filter filtering and the like are carried out on an input voice signal to obtain a voice characteristic vector.
Step two, in voice recognition; the speech feature vector is input to a bidirectional long and short term memory network (BLSTM) to train a speech recognition model, the network not only has the data storage capacity of a common long and short term memory network (LSTM), but also takes the influence of reverse timing sequence information into consideration, the network is more advantageous in speech data recognition, and a Gaussian Mixture Model (GMM) is added behind the network to replace the traditional full-connection network, so that the recognition capacity of the network is improved. The output of the Gaussian mixture model is the posterior probability of the feature vector, and the speech features are obtained by decoding. The data set is used as 300-hour voice data collected in the field of power grids, and a two-channel wav format of 8000Hz, 16bit and 128kpbs is adopted. The BLSTM output information is calculated as follows:
it=sigmoid(Wxixt+Wrirt-1+Wcict-1+bi)
ft=sigmoid(Wxfxt+Wrfrt-1+Wcfct-1+bf)
gt=tanh(Wxgxt+|Wrgrt-1+bg)
ht=sigmoid(Wxhxt+Wrhrt-1+Wchct-1+bh)
i is an input gate which controls the amount of information input to the BLSTM; f is a forgetting gate and controls the operation of the excitation signal; h is an output gate which controls the output information quantity of the BLSTM; c is information of BLSTM after the gate is forgotten; r is the information output by BLSTM; w is the weight matrix and b is the offset. In the speech recognition, a client speech statement is input into a Gaussian mixture model for recognition, the final activation function of the bidirectional long-short term memory network is a Sigmoid function, data is mapped between 0 and 1, whether recognition is successful or not is judged, and a specific calculation function is as follows:
and obtaining a client voice statement through voice recognition, and preparing for semantic matching in the next step.
The naturally-occurring question-answering uses a Sequence-to-Sequence (Seq 2Seq) model, and obtains syntactic information through training standard question-answering and extended question-answering predictions in customer service in the power industry. Wherein the system is formulated as:
p represents the probability of the model output sequence; x represents an input sequence; y represents an output sequence; i denotes the sequence index.
Step three, merging the syntax information and the keyword information obtained by the natural generation type question-answering system with a semantic base into a database with syntax information, wherein the database is called a syntax information database;
step four, semantic matching; and obtaining the intention of inputting the user voice through the processing of semantic hierarchy, and extracting keywords in the sentence and corresponding keywords in the syntax confidence database for matching. The matching method is as follows:
1) in the bidirectional long and short term memory network, the speech sentences of the clients are read in two directions, so that the semantic information in the sentences can be better obtained, and the reading process is as follows:
where c represents the customer speech statement read from both directions and h is the resulting implicit semantic information.
2) And dividing the obtained semantic information h by the characters of the whole client voice sentence to obtain the final semantic representation W at the character level. The following formula:
where Q is the number of characters in the entire customer speech statement.
3) Judging whether the answer speech is matched with the reply speech in the database through a matching algorithm, wherein the judgment formula is as follows:
r=σ(Wrq+br)
Score=Wscorer+bscore
where r is the data weighted information, b is the bias, and w is each weight matrix.
4) On the basis of the bidirectional long and short term memory network framework, a bidirectional long and short term memory network of an attention mechanism is added, the attention network adopts Softmax to carry out importance ranking and normalization on semantic units, and in combination with the bidirectional long and short term memory network used in the invention, a final formula to a matched text is as follows:
wherein T is a text semantic unit; w is a weight matrix of the attention network; z represents a high-dimensional semantic text matching vector output by the layered convolutional neural network; omega is the weight of text semantic learning; gamma is the bias of the attention mechanism in the text semantics. After the attention mechanism is added, the voice recognition capability and the text matching capability of the network are effectively improved.
5) After obtaining the semantic information, the error rate is judged by the following formula:
where N is the number of total words identified, S is a replacement error, I is an insertion error, and D is a deletion error. And judging whether the output results are matched or not through a preset threshold value.
Step five, generating a text; and after the semantic matching is completed, displaying matched answer generation texts in the syntactic information database to a display to complete the interaction with the user.
To illustrate the effectiveness and adaptability of the present invention, the conventional method without a natural-generating question-answering system was used for comparison in the experiment. In the test of using 50 test voices and actual scenes, the recognition rate of the method is improved from 80 percent to 91 percent. Therefore, the method can effectively improve the customer service capacity and level in the field of power grids.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (4)
1. An NLP-based electric power industry character customer service interaction method is characterized by comprising the following steps:
the method comprises the following steps: preprocessing input general voice signals and voice signals with region information to form training data, wherein the preprocessing comprises denoising processing and feature extraction processing;
step two: carrying out voice recognition on the preprocessed general voice signal to obtain semantic information, and generating a semantic library by all the semantic information; meanwhile, the preprocessed voice signal with the region information is processed through a natural generation type question-answering system composed of sequence-to-sequence models to obtain syntax information and keyword information of sentences;
step three: fusing the syntactic information, the keyword information and the semantic library to form a database with syntactic information;
step four: extracting key words in the client voice sentences by the voice recognition model through semantic understanding, and matching the key words with key words in a database;
step five: generating reply characters corresponding to the replies according to the answers matched with the database and by combining the syntactic information and the keyword information;
step six: and displaying the corresponding characters according to the reply voice to complete the interaction with the client.
2. The method of claim 1, wherein the speech recognition model uses a bidirectional long short term memory network (BLSTM) with attention to extract the keywords and converts the keywords into corresponding textual information for matching against the database.
3. The method of claim 1, wherein the database comprises a plurality of customer question statements, a keyword associated with each of the plurality of customer question statements, and a reply utterance associated with each of the plurality of customer question statements.
4. The method of claim 1, wherein the matching of the speech recognition model to the database in step four is formulated as follows:
wherein T is a text semantic unit; w is a weight matrix of the attention network; z represents a high-dimensional semantic text matching vector output by the layered convolutional neural network; omega is the weight of text semantic learning; gamma is the bias of an attention mechanism in text semantics, the voice recognition capability and the text matching capability of the network are effectively improved after the attention mechanism is added, h is an output gate, and b is the bias.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911226062.3A CN111090726A (en) | 2019-12-04 | 2019-12-04 | NLP-based electric power industry character customer service interaction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911226062.3A CN111090726A (en) | 2019-12-04 | 2019-12-04 | NLP-based electric power industry character customer service interaction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111090726A true CN111090726A (en) | 2020-05-01 |
Family
ID=70394124
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911226062.3A Pending CN111090726A (en) | 2019-12-04 | 2019-12-04 | NLP-based electric power industry character customer service interaction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111090726A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111965531A (en) * | 2020-07-15 | 2020-11-20 | 广西电网有限责任公司电力科学研究院 | High-voltage circuit breaker on-line monitoring and evaluating system based on multi-signal acquisition |
CN112653798A (en) * | 2020-12-22 | 2021-04-13 | 平安普惠企业管理有限公司 | Intelligent customer service voice response method and device, computer equipment and storage medium |
CN117234341A (en) * | 2023-11-15 | 2023-12-15 | 中影年年(北京)文化传媒有限公司 | Virtual reality man-machine interaction method and system based on artificial intelligence |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108491517A (en) * | 2018-03-22 | 2018-09-04 | 青岛农业大学 | A kind of region agricultural information service speech polling terminal |
CN109413286A (en) * | 2018-10-22 | 2019-03-01 | 北京移数通电讯有限公司 | A kind of intelligent customer service voice response system and method |
CN109658271A (en) * | 2018-12-19 | 2019-04-19 | 前海企保科技(深圳)有限公司 | A kind of intelligent customer service system and method based on the professional scene of insurance |
CN110209791A (en) * | 2019-06-12 | 2019-09-06 | 百融云创科技股份有限公司 | It is a kind of to take turns dialogue intelligent speech interactive system and device more |
-
2019
- 2019-12-04 CN CN201911226062.3A patent/CN111090726A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108491517A (en) * | 2018-03-22 | 2018-09-04 | 青岛农业大学 | A kind of region agricultural information service speech polling terminal |
CN109413286A (en) * | 2018-10-22 | 2019-03-01 | 北京移数通电讯有限公司 | A kind of intelligent customer service voice response system and method |
CN109658271A (en) * | 2018-12-19 | 2019-04-19 | 前海企保科技(深圳)有限公司 | A kind of intelligent customer service system and method based on the professional scene of insurance |
CN110209791A (en) * | 2019-06-12 | 2019-09-06 | 百融云创科技股份有限公司 | It is a kind of to take turns dialogue intelligent speech interactive system and device more |
Non-Patent Citations (1)
Title |
---|
刘书齐等: "基于CNN-ATTBiLSTM的文本情感分析", 《贵州大学计算机科学与技术学院》, vol. 36, no. 2, pages 85 - 88 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111965531A (en) * | 2020-07-15 | 2020-11-20 | 广西电网有限责任公司电力科学研究院 | High-voltage circuit breaker on-line monitoring and evaluating system based on multi-signal acquisition |
CN112653798A (en) * | 2020-12-22 | 2021-04-13 | 平安普惠企业管理有限公司 | Intelligent customer service voice response method and device, computer equipment and storage medium |
CN117234341A (en) * | 2023-11-15 | 2023-12-15 | 中影年年(北京)文化传媒有限公司 | Virtual reality man-machine interaction method and system based on artificial intelligence |
CN117234341B (en) * | 2023-11-15 | 2024-03-05 | 中影年年(北京)科技有限公司 | Virtual reality man-machine interaction method and system based on artificial intelligence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11908451B2 (en) | Text-based virtual object animation generation method, apparatus, storage medium, and terminal | |
CN108564942A (en) | One kind being based on the adjustable speech-emotion recognition method of susceptibility and system | |
CN113205817B (en) | Speech semantic recognition method, system, device and medium | |
CN109119072A (en) | Civil aviaton's land sky call acoustic model construction method based on DNN-HMM | |
CN114116994A (en) | Welcome robot dialogue method | |
CN111400469A (en) | Intelligent generation system and method for voice question answering | |
CN109949799B (en) | Semantic parsing method and system | |
CN111090726A (en) | NLP-based electric power industry character customer service interaction method | |
CN112581964B (en) | Multi-domain oriented intelligent voice interaction method | |
CN112233646A (en) | Voice cloning method, system, device and storage medium based on neural network | |
CN115206293B (en) | Multi-task air traffic control voice recognition method and device based on pre-training | |
CN116110405B (en) | Land-air conversation speaker identification method and equipment based on semi-supervised learning | |
CN111524535A (en) | Feature fusion method for speech emotion recognition based on attention mechanism | |
CN114203177A (en) | Intelligent voice question-answering method and system based on deep learning and emotion recognition | |
CN115394287A (en) | Mixed language voice recognition method, device, system and storage medium | |
CN111653270A (en) | Voice processing method and device, computer readable storage medium and electronic equipment | |
Ling | An acoustic model for English speech recognition based on deep learning | |
Wu et al. | Speech synthesis with face embeddings | |
CN113393841A (en) | Training method, device and equipment of speech recognition model and storage medium | |
CN115497455B (en) | Intelligent evaluating method, system and device for oral English examination voice | |
CN116911869A (en) | Intelligent customer service system based on artificial intelligence and method thereof | |
CN110956859A (en) | VR intelligent voice interaction English method based on deep learning | |
CN110738989A (en) | method for solving automatic recognition task of location-based voice by using end-to-end network learning of multiple language models | |
CN114626424B (en) | Data enhancement-based silent speech recognition method and device | |
Ihori et al. | Parallel corpus for Japanese spoken-to-written style conversion |
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 |