CN108153727A - Utilize the method for semantic mining algorithm mark sales calls and the system of improvement sales calls - Google Patents
Utilize the method for semantic mining algorithm mark sales calls and the system of improvement sales calls Download PDFInfo
- Publication number
- CN108153727A CN108153727A CN201711363955.3A CN201711363955A CN108153727A CN 108153727 A CN108153727 A CN 108153727A CN 201711363955 A CN201711363955 A CN 201711363955A CN 108153727 A CN108153727 A CN 108153727A
- Authority
- CN
- China
- Prior art keywords
- sales calls
- label
- type
- mining algorithm
- term vector
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—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
- 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
- G06F16/355—Class or cluster creation or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/374—Thesaurus
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2216/00—Indexing scheme relating to additional aspects of information retrieval not explicitly covered by G06F16/00 and subgroups
- G06F2216/03—Data mining
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses the systems that the method for sales calls is identified using semantic mining algorithm and administers sales calls.The present invention is included using the method specific steps that semantic mining algorithm identifies sales calls:S1, divide phone label be different types;S2, dictionary is established, dictionary includes the corresponding term vector of label;Belong to same kind of label in S3, extraction dictionary and form one layer of training sample;S4, it is trained using multilayer training sample, obtains disaggregated model;S5, according to disaggregated model, identify the type belonging to the term vector in dictionary;The invention also discloses a kind of systems for administering sales calls using semantic mining algorithm.The system that the present invention is identified the method for sales calls and administered sales calls using semantic mining algorithm, can accurately classify to sales calls, and user can independently select the classification of access phone, achieve the purpose that precisely to intercept sales calls.
Description
Technical field
The invention belongs to technical field of telecommunications more particularly to a kind of sides that sales calls are identified using semantic mining algorithm
Method and the system for administering sales calls.
Background technology
It is reported that telemarketing has become the main marketing mould of the industries such as current real estate, finance and money management, educational training
Formula, the portfolio that telemarketing is brought account for nearly 2/3rds, that is to say, that telemarketing has its existing soil, part citizen
These information are also required to, sales calls are as a special type in harassing call, it is impossible to simply using interception or pass
The mode stopped is disposed.According to tactile precious telephone statistics, sales calls surmounted other types since 2016, became mobile phone user
Most important harassing and wrecking type.
Have many softwares at present to mark available for phone, as 360 mobile guards, Tencent mobile phone manager, Baidu's mobile phone are defended
Scholar etc., the mobile phones such as Huawei, millet, vivo also provide phone mark function to the user.Phone label can be helped to a certain extent
User is helped to avoid economic loss, if for example, party A-subscriber is connected to marketing fraudulent call, and has been identified for label, the phone
When removing harassing and wrecking user B again, since existing swindle is reminded, the defence psychology of user B is improved.But in practical applications, user couple
The label of phone is multifarious.For example, belonging to the phone of marketing swindle type, user may be labeled as marketing swindle, ring one
Sound, doubtful marketing swindle, swindle fishing, harassing call, illegal marketing etc..Such case not only can accurately identify phone to user
Type belt carrys out certain difficulty, while also analyzes phone treatment status for operator and cause to perplex.
Therefore, according to the different labels of phone, to phone into the understanding and classification of row label, needs are independently set by user
The classification of interception, has important practical significance.
The present invention according in above-mentioned technology to the label of sales calls and processing there are the defects of, the present inventor is based on being engaged in
Such product design manufactures abundant for many years practical experience and professional knowledge, and coordinates the utilization of scientific principle, is actively studied wound
Newly, the method for identifying and administering to propose a kind of to sales calls based on label semantic understanding, integration and sorting algorithm, is led to
It crosses and establishes label dictionary, divide structure disaggregated model using history tab, identify known label major class and unknown classification, it is right
Model training is reused in unknown classification.
It so, it is possible to improve the general existing method for identifying sales calls and administering, and then phone is recognized accurately
Type, make its have more practicability.By constantly studying, designing, and after studying sample repeatedly and improving, create finally
Set out the present invention having practical value.
Invention content
For above-mentioned technical problem of the existing technology, semantic mining algorithm mark battalion is utilized the present invention provides a kind of
The system sold the method for phone and administer sales calls, the present invention is using machine learning algorithm to the sales calls mark of user's mark
Label carry out semantic understanding and excavate, and label is integrated and is classified, and then realizes the exact classification to sales calls and disposition,
The defects of to solve to prepare to identify to phone type in the prior art, to reach the effect precisely intercepted to sales calls
Fruit.
To reach above-mentioned technical purpose, the present invention adopts the following technical scheme that:
A kind of method for identifying sales calls using semantic mining algorithm, including:
S1, divide phone label be different types;
S2, dictionary is established, the dictionary includes the corresponding term vector of label;
Belong to same kind of label in S3, the extraction dictionary and form one layer of training sample;
S4, it is trained using training sample described in multilayer, obtains disaggregated model;
S5, according to the disaggregated model, identify the type belonging to label in the sales calls.
It is specifically included as a preference of the present invention, step S2 establishes dictionary:
S21, acquisition language material storage preparation training text;
S22, the training text is segmented to obtain word using stammerer tool;
S23, institute's predicate is converted into term vector using Word2Vec algorithms, is made of institute's predicate and the term vector described
Dictionary.
As a preference of the present invention, the training sample is divided into training set and test set, divided with training set training
Class model;The disaggregated model tested with test set, and using test set to disaggregated model tuning.
As a preference of the present invention, a default threshold value, calculates the term vector according to disaggregated model and belongs to each type
Probability value;If the probability value that the term vector belongs to a kind of type in mentioned kind is more than the threshold value, maximum is exported
The corresponding type of probability value, the type as the corresponding label of the term vector.
If as a preference of the present invention, the term vector belongs to the probability value of any type all no more than the threshold value,
The corresponding label of the term vector is then added to unknown class library, and it is Unknown Label to mark label.
As a preference of the present invention, a default similarity threshold, if the Unknown Label correspond to term vector with it is another unknown
The similarity that label corresponds to term vector is more than the similarity threshold, then uniformly extracts similar Unknown Label and be divided into new
Type.
The system that a kind of semantic mining algorithm of utilization administers sales calls, based on the method for above-mentioned mark sales calls, sheet
Invention is included using the system that semantic mining algorithm administers sales calls:
Setup module sets the type of call interceptor;
Transmission module transmits the information of setting to communication network;
Judgment module judges whether the type of the phone of access to communication networks is identical with the type of setting;
Interrupt module, if the type of the phone of access to communication networks is identical with the type set, by the phone
It hangs up.
As a preference of the present invention, further including sales calls classification intercepts platform, the sales calls classification intercepts platform
For control the communication network by the call block and hang up and for record intercept phone information.
As a preference of the present invention, sales calls classification intercept platform at least by the number of the phone of interception,
Turn-on time, marketing type Periodic Notice user.
Technical solution provided by the invention can include the following benefits:
The present invention is based on current existing sales calls tag along sort, the sales calls tag along sort of enterprises registration and use
Self-defined label or description of the family for number, using semantic understanding and the label integration excavated and sorting algorithm technology, to battalion
Pin phone is accurately classified, and is selected according to user individual, some users is allowed to wish the phone incoming call for the type answered,
Others are intercepted, have achieved the purpose that precisely to intercept sales calls.
Description of the drawings
Fig. 1 is the flow diagram that the present invention identifies sales calls using semantic mining algorithm;
Fig. 2 is the formation of 1 dictionary of the embodiment of the present invention and source schematic diagram;
Fig. 3 builds flow diagram for 1 disaggregated model of the embodiment of the present invention;
Fig. 4 is work flow diagram of the present invention to known class output;
Fig. 5 is the present invention to unknown class library source schematic diagram;
Fig. 6 is work flow diagram of the present invention to the output of unknown classification;
Fig. 7 is Word2Vec algorithm arrangements schematic diagram of the present invention;
Fig. 8 intercepts platform schematic diagram for sales calls of the present invention classification;
Fig. 9 is 3 specific service handling flow of the embodiment of the present invention;
Figure 10 is 3 operating diagram of the embodiment of the present invention.
Specific embodiment
The following is specific embodiments of the present invention and with reference to attached drawing, technical scheme of the present invention is further described,
But the present invention is not limited to these embodiments.
Embodiment 1:
The present embodiment is as shown in Figure 1, Figure 2, shown in Fig. 3, Fig. 4, Fig. 7, semantic mining algorithm is utilized present embodiment describes a kind of
The method for identifying sales calls, specific operating process are as follows:
S1, divide phone label be different types;
The label of phone is divided by the big types of N according to business and daily life habit, class of such as calling a taxi, swindle class, finance
Class promotes class etc., literary style or pronunciation or pronunciation difference but label equivalent in meaning is divided into same class, to literary style or pronunciation
Or pronunciation is different, the also different label that looks like is divided into other classifications.
The source that type foundation is divided to label is divided into following three classes:
1. existing sales calls classification, from internet works software and mobile phone terminal from tape sorting.
Such as:The type of sales calls is divided by 360 mobile guards:Harassing call, ad promotions, recruitment hunter etc.;Hundred
The type of sales calls is divided by degree mobile phone bodyguard:Swindle, advertisement, harassing and wrecking etc.;The type of sales calls is divided by Huawei's mobile phone:
Harassing call, fraudulent call, ad promotions phone etc.
2. the type of number that the enterprise for carrying out sale sales calls registers in platform;
3. unknown classification, this is unknown to be classified as:Other than existing classification, the unknown classification phone of user's experience, it may be possible to
A kind of classification type of phone, it is also possible to which passage describes, such as:Whether dialog context is needs invoice.
S2, dictionary is established, the dictionary includes the corresponding term vector of label;
The main purpose for establishing dictionary is that label text is converted into term vector, and computer is facilitated to understand the meaning of word,
It realizes and semantic understanding is carried out to word.
As shown in Fig. 2, specifically include following steps:
Step A1:Corpus acquires, and crawls with the relevant corpus of each type as training text;
Corpus can be news, article etc.;
Step A2:Text segments, and text is segmented using stammerer tool to obtain single word or word;
Step A3:Word is converted by term vector using Word2Vec algorithms, i.e., text segmented after each word
It is corresponding that there are one vectors;
Step A4:The set of all words and the corresponding term vector of word is dictionary.
Word2Vec is a kind of term vector Core Generator for deep learning.Substantially it is that neural network language is utilized
Model simultaneously simplifies it, not only ensure that effect but also has improved computation complexity.Word2Vec, with high dimension vector (word to
Amount, Word Embedding) represent word, and the word of similar import is placed on similar position, and be real number to
It measures (being not limited to integer).We only need the language material of certain a large amount of language, it is possible to it come training pattern, obtain word to
Amount.There are two types of the common algorithms of the model:CBOW (Continuous Bag-of-Words Model) and Skip-gram
(Continuous Skip-gram Model).Wherein CBOW models go prediction current word using each k word before and after word W (t);And
Skip-gram models are just the opposite, it goes to predict its front and rear each k word using word W (t), and the present invention uses Skip-gram moulds
Type.
The problem of natural language understanding, will be converted into the problem of machine can be handled, and the first step must be by these sign digits
Change, i.e., be mapped to the expression of text in the vector space of k dimensions.Word2Vec algorithms are by the Chinese word in the corpus segmented
Language is converted to term vector, and the term vector trained by Word2Vec is as follows
vi=(a0,a1,L,ad) (1)
(1) in formula, d is the dimension of term vector.
As shown in fig. 7, Word2Vec algorithms specific implementation process is:
Step A31:The keyword being collected about in phone label text feature library is counted, it is assumed that have m pass
Keyword;
Step A32:One word first with one-hot-vector is converted into the vector x of a n dimension, is with " swindle "
Example:
" swindle " → [0,0,0,0,1..., 0,0]
Step A33:There is m neuron in hidden layer, it is known that input layer is a n-dimensional vector and connect entirely with hidden layer,
So it needs in the hidden neuron that the weight matrix w of a n*m is 1*m the DUAL PROBLEMS OF VECTOR MAPPING that n is tieed up to latitude;
Step A34:Also with full connection from hidden layer to output layer, added in when output unit is calculated
Softmax graders, can be in the hope of final vectorial w by back transfer, and by being multiplied with initial term vector, i.e. x*w can be with
Acquire final term vector, that is, the vectorial W (i) of 1*m;
X*w=W (i)=[Wi1 Wi2 … Wim]
Step A35:The corresponding term vector of each label taken on the telephone is added, obtains belonging to the label taken on the telephone
Vectorial d.
Belong to same kind of label in S3, the extraction dictionary and form one layer of training sample;
S4, it is trained using training sample described in multilayer, obtains disaggregated model;
By existing known label and the correspondence of classification, the present embodiment carries out disaggregated model using softmax algorithms
Structure, passes through softmax algorithm train classification models.
As shown in figure 3, the wherein implementation procedure of S3-S4 specifically includes:
The correspondence of B1, the term vector that a variety of labels are obtained in step S1-S2 and type, are corresponded to by multiple term vectors
A kind of one layer of training sample of Plant composition;
B2, by multilayer training sample with 3:1 ratio cut partition is training set and test set;
B3, training set is trained data by softmax recurrence graders, obtains preliminary classification model;
B4, test set is inputted into preliminary classification model, calculates accuracy rate, the recall rate of disaggregated model;
B5, according to the obtained accuracys rate of B4 and recall rate to preliminary classification model tuning, adjust the parameter of disaggregated model, weight
New training preliminary classification model is expected to have arrived disaggregated model until accuracy rate reaches with recall rate.
Softmax algorithms:
Softmax regression models are popularization of the logistic regression models in more classification problems, in more classification problems,
Class label y can take more than two values.Softmax recurrence has supervision.In this patent, phone label is divided into
N (N2) a major class, belongs to supervised learning, therefore can build Softmax regression models and carry out category division to phone label.
In softmax recurrence, solution is more classification problems, and category y can take N number of different value.For what is given
Test input x, with assuming that function estimates probability value p (y=j | x) for each classification j, that is, each for estimating x is classified
As a result the probability occurred.Assuming that function will export the vector of a N-dimensional to represent the probability value of this N number of estimation.Assuming that function
hθ(x) form is as follows:
Wherein θ1,θ2,L,θNIt is the parameter of model.Model exports the probability value that label to be measured belongs to each classification, chooses most
Classification of the corresponding classification of greatest as the phone label.
S5, according to the disaggregated model, identify the type belonging to label in the sales calls.
As shown in figure 4, step S5 is specifically included:
C1, label to be predicted in sales calls is inputted into trained disaggregated model;
C2, the corresponding term vector of the label in A4 step dictionarys is found;
C3, the term vector is input to trained Softmax graders, exports the corresponding term vector of the label and belong to
The probability value of each type;
C4, a default threshold value, if the probability value that term vector belongs to certain type in mentioned kind is more than the threshold value;
C5, if it has, then exporting classification of the corresponding type of most probable value as the label.
In conclusion establishing dictionary in S2 steps in the present embodiment, word is converted into using Word2Vec algorithms
Term vector realized to semantic understanding, the disaggregated model gone out in S3-S5 by known classification based training, can be with by disaggregated model
The classification belonging to arbitrary label is exported, said program is semantic mining algorithm, can be to arbitrarily seeking by semantic mining algorithm
Phone identification is sold, convenient for user accurately to the identification of phone type.
Embodiment 2:
As shown in Figure 5, Figure 6, the present embodiment further includes the side for creating new Genre categories on the basis of above-described embodiment 1
Case, present embodiment describes new classification is established to belonging to the label in unknown class library using semantic mining algorithm, specifically just
Case is as follows:
As shown in figure 5, C5, the disaggregated model obtained according to embodiment 1, it is existing that disaggregated model output label belongs to each
If type probability value both less than or equal to the threshold value, which is added to unknown class library.
As described in Figure 6, the processing specially to the label in unknown class library.
D1, the label for arranging unknown class library and corresponding term vector;
D2, the term vector of labels all in library is subjected to similarity analysis;
D3, be more than by the similarity of multiple labels similarity threshold extraction, wherein similarity threshold independently sets by user
It is fixed;
D4, by the label universal formulation with certain similarity be new classification, be added in the classification of S1 steps.
So disaggregated model output category includes:Output to known class and output two parts to unknown classification.
In this programme to similarity calculation used by method be:
Cosine similarity:
Carrying out similarity analysis to label can be by cosine similarity algorithm.Cosine similarity is in vector space
Measurement of two vectorial angle cosine values as the size for weighing two inter-individual differences.For the label vector a of two n dimensions
(x1,x2,L,xn),b(y1,y2,L,yn), the included angle cosine of a and b are equal to:
Cosine value indicates that angle closer to 0 degree, that is, two vectors are more similar, angle is equal to 0, i.e., two closer to 1
A vector is equal.
The present embodiment is based on label semantic understanding, integration and sorting algorithm, by establishing label dictionary, utilizes history mark
Label divide structure disaggregated model, identify known label major class and unknown classification, model training is reused for for unknown classification.
The processing of unknown classification can not only be helped to improve dynamic sensing ability of the operator to harassing call type, moreover it is possible to further
Improve the recognition accuracy of disaggregated model.
Embodiment 3
As shown in Fig. 8, Fig. 9, Figure 10, the present embodiment is based on the method for identifying sales calls using semantic mining algorithm, root
The type belonging to each label is obtained according to above-described embodiment, and the present embodiment selects to block the phone of certain type, provides one
The system that kind administers sales calls using semantic mining algorithm.
It is registered, and the phone that needs is set to intercept current embodiment require that user intercepts platform in sales calls classification
Type is shown in Fig. 8.The acceptance range of business includes all telecommunications, unicom and cell phone, and specific service handling flow is shown in Fig. 9.
Communication network intercepts particular kind of sales calls, and concrete scheme is as follows:
Setup module sets the type of call interceptor;
101:Called subscriber intercepts platform in sales calls classification and is registered, and the telephone type that needs is set to intercept
Type;
Transmission module transmits the information of setting to communication network;
102:User information is transmitted to NGIN (Next Generation Intelligent Network, communication by platform
Network) carry out Call- Control1 subscription;
103:NGIN returns to user and subscribes to success notification;
104:Sales calls classification intercepts platform notice user registration success.
Judgment module judges whether the type of the phone of access to communication networks is identical with the type of setting;
Sales calls intercept principle:
It is by the way that the rule of calling number type and user setting is compared the side being filtered that sales calls classification, which intercepts,
Method.
Interrupt module, if the type of the phone of access to communication networks is identical with the type set, by the phone
It hangs up.
As the preferred embodiment of the present embodiment, platform is intercepted by sales calls classification and is hung up phone and by marketing electricity
Words classification intercepts the relevant information of phone that platform record intercepts.
As shown in Figure 10, after communication network interception, interception situation is sent to sales calls classification and intercepts platform, by seeking
Pin call classifier interception platform is by intercept information by short message sending to user, for example, specific flow is as follows:
201:Initiate user of the calling to registered business at sales calls center, and calling is linked into the NGIN of operator;
202:The calling and called of calling are sent to sales calls classification and intercept platform by NGIN;
203:Sales calls classification intercepts platform and the type of sales calls and the situation of user setting is compared and judged, such as seeks
The classification that pin number is intercepted for the needs of user setting, then returning to NGIN needs to intercept;
204:NGIN hangs up calling, while will intercept result and return to sales calls classification interception platform;
205:Sales calls classification, which intercepts platform, will intercept record with short message mode Periodic Notice user, number including interception
Code intercepts time, the marketing type intercepted.
In conclusion the present embodiment can have needle according to the type of call interceptor set by user to sales calls
It intercepts to property and accurately and effectively, solves the puzzlement that operator's analysis electricity administers phone, also avoid sales calls and get through use
Interference of the phone at family to user.
Specific embodiment described herein is only an example for the spirit of the invention.Technology belonging to the present invention is led
The technical staff in domain can do various modifications or additions to described specific embodiment or replace in a similar way
In generation, however, it does not deviate from the spirit of the invention or beyond the scope of the appended claims.
Claims (9)
- A kind of 1. method for identifying sales calls using semantic mining algorithm, which is characterized in that including:S1, divide phone label be different types;S2, dictionary is established, the dictionary includes the corresponding term vector of label;Belong to same kind of label in S3, the extraction dictionary and form one layer of training sample;S4, it is trained using training sample described in multilayer, obtains disaggregated model;S5, according to the disaggregated model, identify the type belonging to label in the sales calls.
- 2. the method according to claim 1 for identifying sales calls using semantic mining algorithm, which is characterized in that step S2 Dictionary is established to specifically include:S21, acquisition language material storage preparation training text;S22, the training text is segmented to obtain word using stammerer tool;S23, institute's predicate is converted into term vector using Word2Vec algorithms, the dictionary is formed by institute's predicate and the term vector Library.
- 3. the method according to claim 1 for identifying sales calls using semantic mining algorithm, which is characterized in that the instruction Practice sample and be divided into training set and test set, disaggregated model is obtained with training set training;The classification mould tested with test set Type, and using test set to disaggregated model tuning.
- 4. the method according to claim 1 for identifying sales calls using semantic mining algorithm, which is characterized in that the step Rapid S5 is specifically included:A default threshold value belongs to the probability value of each type according to the disaggregated model calculating term vector;If the probability value that the term vector belongs to a kind of type in mentioned kind is more than the threshold value, most probable value is exported Corresponding type, the type as the corresponding label of the term vector.
- 5. the method according to claim 4 for identifying sales calls using semantic mining algorithm, which is characterized in that if described Term vector belongs to the probability value of any type all no more than the threshold value, then is added to the corresponding label of the term vector not Know class library, and it is Unknown Label to mark label.
- 6. the method according to claim 5 for identifying sales calls using semantic mining algorithm, which is characterized in that default one Similarity threshold, if the similarity that the Unknown Label corresponds to term vector term vector corresponding with another Unknown Label is more than the phase Like degree threshold value, then similar Unknown Label is uniformly extracted and be divided into new type.
- 7. a kind of system for administering sales calls using semantic mining algorithm, which is characterized in that based on any one of claim 1-6 The method for identifying sales calls using semantic mining algorithm, which is characterized in that the system comprises:Setup module sets the type of call interceptor;Transmission module transmits the information of setting to communication network;Judgment module judges whether the type of the phone of access to communication networks is identical with the type of setting;If the type of the phone of access to communication networks is identical with the type set, the phone is hung up for interrupt module.
- 8. the system for administering sales calls using semantic mining algorithm according to claim 7, which is characterized in that further include battalion It sells call classifier and intercepts platform, the sales calls classification intercepts platform for controlling the communication network by the call block And the information of phone hung up and intercepted for record.
- 9. the system for administering sales calls using semantic mining algorithm according to claim 8, which is characterized in that the marketing Call classifier intercepts platform at least by the number of the phone of interception, turn-on time, marketing type Periodic Notice user.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711363955.3A CN108153727B (en) | 2017-12-18 | 2017-12-18 | Method for identifying marketing call by semantic mining algorithm and system for managing marketing call |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711363955.3A CN108153727B (en) | 2017-12-18 | 2017-12-18 | Method for identifying marketing call by semantic mining algorithm and system for managing marketing call |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108153727A true CN108153727A (en) | 2018-06-12 |
CN108153727B CN108153727B (en) | 2020-09-08 |
Family
ID=62467493
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711363955.3A Active CN108153727B (en) | 2017-12-18 | 2017-12-18 | Method for identifying marketing call by semantic mining algorithm and system for managing marketing call |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108153727B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109120612A (en) * | 2018-08-06 | 2019-01-01 | 浙江衣拿智能科技有限公司 | A kind of packet filtering method, system and application program |
CN111131593A (en) * | 2018-11-01 | 2020-05-08 | 百度在线网络技术(北京)有限公司 | Crank call identification method and device |
CN111465021A (en) * | 2020-04-01 | 2020-07-28 | 北京中亦安图科技股份有限公司 | Graph-based crank call identification model construction method |
CN111669757A (en) * | 2020-06-15 | 2020-09-15 | 国家计算机网络与信息安全管理中心 | Terminal fraud call identification method based on conversation text word vector |
CN111985901A (en) * | 2020-08-24 | 2020-11-24 | 北京思特奇信息技术股份有限公司 | Marketing product configuration method, device, equipment and storage medium in telecommunication industry |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150187353A1 (en) * | 2013-12-31 | 2015-07-02 | Abbyy Infopoisk Llc | Audio tagging |
US20160232452A1 (en) * | 2013-09-17 | 2016-08-11 | Zte Corporation | Method and device for recognizing spam short messages |
CN106534596A (en) * | 2016-10-31 | 2017-03-22 | 努比亚技术有限公司 | Anti-harassment call filtering method and filtering system thereof |
CN107301246A (en) * | 2017-07-14 | 2017-10-27 | 河北工业大学 | Chinese Text Categorization based on ultra-deep convolutional neural networks structural model |
CN107360300A (en) * | 2017-08-01 | 2017-11-17 | 中国联合网络通信集团有限公司 | Harassing call hold-up interception method and device |
-
2017
- 2017-12-18 CN CN201711363955.3A patent/CN108153727B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160232452A1 (en) * | 2013-09-17 | 2016-08-11 | Zte Corporation | Method and device for recognizing spam short messages |
US20150187353A1 (en) * | 2013-12-31 | 2015-07-02 | Abbyy Infopoisk Llc | Audio tagging |
CN106534596A (en) * | 2016-10-31 | 2017-03-22 | 努比亚技术有限公司 | Anti-harassment call filtering method and filtering system thereof |
CN107301246A (en) * | 2017-07-14 | 2017-10-27 | 河北工业大学 | Chinese Text Categorization based on ultra-deep convolutional neural networks structural model |
CN107360300A (en) * | 2017-08-01 | 2017-11-17 | 中国联合网络通信集团有限公司 | Harassing call hold-up interception method and device |
Non-Patent Citations (1)
Title |
---|
胡玮: "基于语义的垃圾邮件过滤技术的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109120612A (en) * | 2018-08-06 | 2019-01-01 | 浙江衣拿智能科技有限公司 | A kind of packet filtering method, system and application program |
CN109120612B (en) * | 2018-08-06 | 2021-04-30 | 浙江衣拿智能科技股份有限公司 | Data packet filtering method, system and application program |
CN111131593A (en) * | 2018-11-01 | 2020-05-08 | 百度在线网络技术(北京)有限公司 | Crank call identification method and device |
CN111131593B (en) * | 2018-11-01 | 2021-04-13 | 百度在线网络技术(北京)有限公司 | Crank call identification method and device |
CN111465021A (en) * | 2020-04-01 | 2020-07-28 | 北京中亦安图科技股份有限公司 | Graph-based crank call identification model construction method |
CN111465021B (en) * | 2020-04-01 | 2023-06-09 | 北京中亦安图科技股份有限公司 | Graph-based crank call identification model construction method |
CN111669757A (en) * | 2020-06-15 | 2020-09-15 | 国家计算机网络与信息安全管理中心 | Terminal fraud call identification method based on conversation text word vector |
CN111669757B (en) * | 2020-06-15 | 2023-03-14 | 国家计算机网络与信息安全管理中心 | Terminal fraud call identification method based on conversation text word vector |
CN111985901A (en) * | 2020-08-24 | 2020-11-24 | 北京思特奇信息技术股份有限公司 | Marketing product configuration method, device, equipment and storage medium in telecommunication industry |
CN111985901B (en) * | 2020-08-24 | 2024-02-02 | 北京思特奇信息技术股份有限公司 | Marketing product configuration method, device, equipment and storage medium in telecom industry |
Also Published As
Publication number | Publication date |
---|---|
CN108153727B (en) | 2020-09-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108153727A (en) | Utilize the method for semantic mining algorithm mark sales calls and the system of improvement sales calls | |
CN107861942B (en) | Suspected power complaint work order identification method based on deep learning | |
CN107678845A (en) | Application program management-control method, device, storage medium and electronic equipment | |
CN108776671A (en) | A kind of network public sentiment monitoring system and method | |
WO2020038100A1 (en) | Feature relationship recommendation method and apparatus, computing device and storage medium | |
CN107517463A (en) | A kind of recognition methods of telephone number and device | |
CN107704868B (en) | User clustering method based on mobile application use behaviors | |
CN107527240A (en) | A kind of operator's industry product Praise effect identification system and method | |
Ghosh et al. | Empirical analysis of ensemble methods for the classification of robocalls in telecommunications | |
Peng et al. | Fraud phone calls analysis based on label propagation community detection algorithm | |
CN112184484A (en) | Differentiated service method and system for power users | |
CN114693317A (en) | Telecommunication fraud security federation detection method fusing homogeneous graph and bipartite graph | |
Rehman et al. | Customer churn prediction, segmentation and fraud detection in telecommunication industry | |
Yang et al. | K-means based clustering on mobile usage for social network analysis purpose | |
CN109474755A (en) | Abnormal phone active predicting method and system based on sequence study and integrated study | |
CN109982272A (en) | A kind of fraud text message recognition methods and device | |
CN109377436A (en) | A kind of accurate monitoring and managing method of environment and device, terminal device and storage medium | |
Jamil et al. | Churn comprehension analysis for telecommunication industry using ALBA | |
CN112734425A (en) | Identification method for phishing users in Ether house platform | |
Droftina et al. | A diffusion model for churn prediction based on sociometric theory | |
Wang et al. | A Comparative Study on Contract Recommendation Model: Using Macao Mobile Phone Datasets | |
Moudani et al. | Fraud detection in mobile telecommunication | |
Anjum et al. | Optimizing coverage of churn prediction in telecommunication industry | |
Xu et al. | Fraud detection in telecommunication: a rough fuzzy set based approach | |
CN113726963A (en) | Intelligent outbound harassment prevention method, device, equipment and medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |