CN107066442A - Detection method, device and the electronic equipment of mood value - Google Patents
Detection method, device and the electronic equipment of mood value Download PDFInfo
- Publication number
- CN107066442A CN107066442A CN201710081287.9A CN201710081287A CN107066442A CN 107066442 A CN107066442 A CN 107066442A CN 201710081287 A CN201710081287 A CN 201710081287A CN 107066442 A CN107066442 A CN 107066442A
- Authority
- CN
- China
- Prior art keywords
- mood
- classification results
- tested
- value
- text set
- 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
Classifications
-
- 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
- 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
- G06F16/353—Clustering; Classification into predefined classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/105—Human resources
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Databases & Information Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiments of the invention provide a kind of detection method, device and the electronic equipment of mood value, wherein, the detection method of mood value includes：Feature extraction is carried out to the multiple texts being tested in text set, multiple characteristic vectors are generated；The multiple characteristic vector is inputted into mood grader respectively, generate based on positive mood, negative emotions, neutral mood multiple first classification results；According to the multiple first classification results, the corresponding mood value of tested text set is calculated.Detection method, device and the electronic equipment of the mood value of the embodiment of the present invention, mood value is calculated by multiple first classification results based on positive mood, negative emotions, neutral mood, the calculation of the classification results for being based only on positive mood and negative emotions compared with prior art, results in more accurate mood value.
Description
Technical field
The present invention relates to field of computer technology, more particularly to a kind of detection method, device and the electronic equipment of mood value.
Background technology
The core resource of enterprise be people in itself, to being the management to people in the nature of management of enterprise, good manager needs
The affective state of each employee and the emotional state to working and living on corresponding post are grasped in time, by understanding employee
Last state, by employee and the effective and reasonable distribution of other ERMs, so as to reach Enterprise Efficiency optimization collocation.
Enterprise can produce substantial amounts of text information during office, the daily work report of such as employee, Weekly Work Report,
Work exchange content (work mail, work chat record) etc., can be by the extraction and analysis to these text informations, to detect
The mood value of employee, so as to preferably grasp the mood dynamic of employee.
In the prior art, the analysis of text based mood is main calculates detected personnel's using two methods classified
Mood value, is generally only divided into two classes for the emotional semantic classification for being detected personnel, i.e., positive mood and negative emotions, based on such
The mood value that model algorithm is obtained not is very accurate.Because often neutral mood also influences whether employee in real work life
The weights in attitude and temperature and then influence model algorithm on a certain event, finally will also have influence on company manager to member
The judgement and decision-making of work affective state.
The content of the invention
The embodiment of the present invention provides a kind of detection method, device and the electronic equipment of mood value, to improve the feelings detected
The accuracy of thread value.
An aspect of of the present present invention there is provided a kind of detection method of mood value, including：
Feature extraction is carried out to the multiple texts being tested in text set, multiple characteristic vectors are generated；
The multiple characteristic vector is inputted into mood grader respectively, generation is based on positive mood, negative emotions, neutral feelings
Multiple first classification results of thread；
According to the multiple first classification results, the corresponding mood value of tested text set is calculated.
The second aspect of the present invention there is provided a kind of detection means of mood value, including：
Feature vector generation module, for carrying out feature extraction to the multiple texts being tested in text set, generates multiple spies
Levy vector；
Mood grader, for obtaining the multiple characteristic vector as input, generation is based on positive mood, negative feelings
Multiple first classification results of thread, neutral mood；
Mood value computing module, for according to the multiple first classification results, calculating the corresponding feelings of tested text set
Thread value.
The third aspect of the present invention there is provided a kind of electronic equipment, including：
Memory, for storage program；
Processor, coupled to the memory, for performing described program, for：
Feature extraction is carried out to the multiple texts being tested in text set, multiple characteristic vectors are generated；
The multiple characteristic vector is inputted into mood grader respectively, generation is based on positive mood, negative emotions, neutral feelings
Multiple first classification results of thread；
According to the multiple first classification results, the corresponding mood value of tested text set is calculated.
Detection method, device and the electronic equipment of the mood value of the embodiment of the present invention, by based on positive mood, negative feelings
Thread, multiple first classification results of neutral mood calculate mood value, compared with prior art be based only on positive mood and negative
The calculation of the classification results of face mood, results in more accurate mood value.
Described above is only the general introduction of technical solution of the present invention, in order to better understand the technological means of the present invention,
And can be practiced according to the content of specification, and in order to allow above and other objects of the present invention, feature and advantage can
Become apparent, below especially exemplified by the embodiment of the present invention.
Brief description of the drawings
Fig. 1 is the technical principle schematic diagram of inventive embodiments；
The schematic flow sheet that Fig. 2 applies for the technical scheme of the embodiment of the present invention in enterprise；
Fig. 3 is the schematic flow sheet of the detection method of the mood value of the embodiment of the present invention one；
Fig. 4 is the structural representation of the mathematical modeling of the online awareness device grader of the embodiment of the present invention one；
Fig. 5 is the structural representation of the detection means of the mood value of the embodiment of the present invention two；
Fig. 6 is the structural representation of the electronic equipment of the embodiment of the present invention three.
Embodiment
The exemplary embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although showing the disclosure in accompanying drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure without should be by embodiments set forth here
Limited.On the contrary, these embodiments are provided to facilitate a more thoroughly understanding of the present invention, and can be by the scope of the present disclosure
Complete conveys to those skilled in the art.
Relational language explanation：
Mood mining analysis：Based on related data index, excavated using relevant way method and analyze tested personnel's
Mood, possible mood includes positive active mood, negative negative feeling, neutral mood.
Perceptron (Perception)：Perceptron is by american computer scientist's Rosenblat (Frank
Roseblatt) proposed in nineteen fifty-seven, be earliest artificial neural network, a kind of feedforward of simplest form can be regarded as
Formula artificial neural network and binary linearity grader.Perceptron principle comes from simplest neuron models adaptive characteristic,
It includes input layer and output layer, and input layer and output layer are joined directly together.Perceptron can be used in the embodiment of the present invention
It is used as mood grader.
Corpus：Refer to the extensive e-text storehouse through scientific sampling and processing.By computer analysis tool, researcher
Language theory and the application study of correlation can be carried out.Corpus is deposited in the basic resource of corpus linguistics research, corpus
What is put is the true linguistic data occurred in the actual use of language, is the main money of empiricism speech research method
Source.Applied to lexicography, language teaching, conventional language research, the research based on statistics or example in natural language processing etc.
Aspect.
TF-IDF(term frequency–inverse document frequency)：It is that one kind is used for information retrieval
With the conventional weighting technique of data mining.More directly for, TF-IDF is a kind of statistical method, for assess a word to N
The importance of wherein one article in article or a corpus.The number of times that one of word occurs in an article can not
Show the importance of the word, such as similar " everybody ", " ", " ", " obtaining " are although time that this common word occurs in article
Number is a lot, but can not represent that these words are just critically important, and now we are accomplished by TF-IDF statistical methods to calculate the weight of some word
The property wanted.The number of times that the importance of word occurs with it in an article is directly proportional increase, but simultaneously can be as it is in N articles
The frequency occurred in (article set) is inversely proportional decline.
Stop words：Stop words refers to during text-processing that if running into them then stopping immediately being handled, and is thrown away not
It is included in the word of statistical disposition.These words, which are thrown away, can increase recall precision and the degree of accuracy.Stop words mainly includes English character, number
Word, mathematical character, punctuation mark and the extra-high Chinese word character of frequency of use etc..
The technical principle to the present invention is illustrated below：
The basic fundamental thought of the embodiment of the present invention is that positive mood and negative emotions will be based only in the prior art
The methods of two classification calculate the mood value of detected personnel (employee of such as company), change into based on positive mood, negative
Mood calculates final mood value with the methods of three classification of neutral mood.In view of in real work life, often largely
Neutral mood also influence whether attitude and temperature of the employee to a certain event, therefore, during the determination of mood value, embody
The word or sentence of neutral mood are also particularly important.The present invention passes through based on positive mood, negative emotions and neutral mood
Three classification results calculate final mood value, the accuracy of mood value is substantially increased, so as to being preferably that enterprise manages
Judgement and decision-making of the reason person to employee's affective state provide reliable foundation.
Specifically, in embodiments of the present invention, by the sub-classifier of three two classification (in actual applications using perception
Device as the embodiment of the present invention mood grader) realize three classification to positive mood, negative emotions and neutral mood.Will
The value of the classification results of positive mood, negative emotions and neutral mood is respectively set as 1, -1 and 0 these three mood values.Such as
Shown in Fig. 1, it is the technical principle schematic diagram of the embodiment of the present invention, in embodiments of the present invention, constructs three perceptrons：Its
Middle perceptron 1 is only that 1 and -1 are classified to mood value；Perceptron 2 is only classified to mood value for 1 and 0；Perceptron
3 be only that 0 and -1 are classified to mood value, and final classification results are drawn final by three perceptrons with the mode voted
The classification results of mood value, i.e., finally belong to positive mood, negative emotions and neutral mood.
The technical scheme of the embodiment of the present invention can apply in enterprise, and enterprise can produce a large amount of during office
Text information, such as the daily work report of employee, Weekly Work Report, work exchange content (work mail, work chat record),
Analyzed by the mood to these texts, it can be deduced that the mood value of employee, so as to carry out the important of internal decision making as enterprise
Foundation.
As shown in Fig. 2 the schematic flow sheet that the technical scheme that it is the embodiment of the present invention is applied in enterprise, the technical side
Case can be realized on the server or business platform of OA (office automation) system of enterprise.First, server can be right
The textual resources such as weekly, daily paper, chat record and the contact mail of employee, which carry out data acquisition, (can gather certain time period
Interior textual resources, for example, can carry out the collection of textual resources with unit around), can be to all kinds of during collection
Textual resources are screened and integrated and simple data cleansing, then form (the i.e. pending mood analysis of tested text set
Text intersection), the tested text set can be formed by some employee, so that the mood value finally given is should
The mood value of employee, naturally it is also possible to form tested text set for some department or whole company, so as to obtain bigger
In the range of mood value.Then, each text in the tested text set based on formation carries out feature extraction, formed feature to
Amount, and progress mood classification in above-mentioned mood grader is input to, obtain based on positive mood, negative emotions and neutral mood
Mood classification results.Finally, the classification results based on each text, it is determined that the entirely mood value of tested text set, at this
During, can set different weights to each text according to actual needs, for example can according to the type of text it is different come
Weight (weight of such as weekly is higher than the weight of daily paper) is set, can also be set according to the difference of the employee of the generation of text
Weight (weight of the weight of the high employee of such as rank employee low higher than rank).
By such as Fig. 2 processing procedure, the mood value corresponding to tested text set can be obtained, so as to be used as business administration
The foundation of decision-making.Because tested text set corresponds generally to certain time period (with unit around), enterprise can also be based on the past
The Employees'Emotions value of period predicts the mood value of the employee of current or next period, so as to preferably to member
The mood tendency of work is tackled.
Technical scheme will be further illustrated by several specific embodiments below.
Embodiment one
As shown in figure 3, its schematic flow sheet for the detection method of the mood value of the embodiment of the present invention one, the present embodiment
Method includes：
S101：Feature extraction is carried out to the multiple texts being tested in text set, multiple characteristic vectors are generated.The step can be with
Performed by the OA system servers of enterprise or the business platform of enterprises, as previously mentioned, tested text set can come
From in textual resources such as the weekly of employee, daily paper, chat record and contact mails.
Specifically, in this step, characteristic vector can be produced in the following way：
S1011：It is each in text set for being tested according to the keyword dictionary that the keyword quantity pre-established is N
Text, calculates the corresponding TF-IDF values of each keyword in keyword dictionary.Keyword dictionary mentioned here is for above-mentioned
What tested text set was extracted, the dimension of characteristic vector can be greatly reduced by extracting keyword, so as to improve mood
The efficiency of classification, sets up process about keyword dictionary, will be described in detail later.
S1012：The corresponding characteristic vector of each text is generated according to the corresponding TF-IDF values of each keyword, wherein, it is special
The dimension for levying vector is that the component in N, each dimension of characteristic vector is the corresponding TF- of each keyword in keyword dictionary
IDF values.That is, it is the characteristic vector being made up of N number of TF-IDF values to finally enter the characteristic vector in mood grader.
S102：Multiple characteristic vectors are inputted into mood grader respectively, generation is based on positive mood, negative emotions, neutrality
Multiple first classification results of mood.The step can be held by the OA system servers of enterprise or the business platform of enterprises
OK, mood grader can be arranged on OA system servers or business platform.Specifically, as described previously, mood
Grader can be realized with the sub-classifier of three two classification, can specifically be included：For entering to positive mood and negative emotions
First sub-classifier of row classification, for positive mood and neutral mood are classified the second sub-classifier, for centering
The 3rd sub-classifier that vertical mood and negative emotions are classified.Based on the framework of such mood grader, this step can be with
Specifically include：
S1021：Each characteristic vector is inputted into each sub-classifier respectively, the second classification knot of each sub-classifier is generated
Really.In actual applications, each sub-classifier can use online awareness device (Perception) grader, the grader mathematics
The structural representation of model is as shown in figure 4, its specific formula form of expression is as follows：
Wherein, w=[w1,w2,w3,...,wn]T, be the weight vectors (or weight vector) of perceptron, be set-point or
Person's preset value, x=[x1,x2,x3,...,xn]TFor the n dimensional feature vectors of input, the x in the n dimensional feature vectors1To xnCorrespond to
P in Fig. 41To pn, b is bias, is set-point or preset value.
Function f (L) it is defined as：
Wherein, the independent variable of the X representative functions f (L) in formula (2), equivalent to the w in formula (1)TX, f (L) are excitation function
(or being activation primitive), t is scalar output (i.e. the mood value as classification results of perceptron final output).Perceptron
Output behavior is tried to achieve after the characteristic vector of input and the inner product of weight vector, through a scalar result obtained by an activation primitive.
In actual applications, the value of the first classification results is typically set as 1, -1 and 0, corresponds respectively to positive feelings
Thread, negative emotions and neutral mood, and in the mathematical modeling of said sensed device, the original output result of each perceptron
For -1 and+1, the difference of implication classified according to each perceptron mood is so also needed to, output result is converted to and first
The second unified classification results of the values of classification results, for example ,+1 and -1 can directly corresponding to of being exported of the first sub-classifier
In positive mood and negative emotions, and+the 1 of the second mood sub-classifier output corresponds to positive mood and neutral feelings with -1
Thread, that is, need to be converted to+1 and 0 unified with the first classification results, and similarly the 3rd mood sub-classifier is also required to carry out output knot
The conversion of fruit.In a word, it with the values of the first classification results is unified that the value of the second classification results of final output, which is,.
S1022：Ballot computing is carried out to the second classification results of each sub-classifier, each characteristic vector is obtained corresponding
First classification results.
As shown in fig. 1, according to the difference of each sub-classifier, same characteristic vector may go out second classification results
Existing the second different classification results.For example, some characteristic vector is positive mood in the classification results of the first sub-classifier, the
The classification results of two sub-classifiers are positive mood, and are neutral mood in the classification results of the 3rd sub-classifier.So need
By Voting Algorithm, to obtain the first final classification results.Specific ballot mode can be tied with the second classification occupied the majority
Fruit exports as the first classification results.Such as the example above, the second classification results that there is two sub-classifiers are positive mood,
Second classification results of one sub-classifier are neutral mood, then the first classification results exported are that positive mood, i.e. output are made
It is 1 for the mood value of classification results.
S103：According to multiple first classification results, the corresponding mood value of tested text set is calculated.The step can be by looking forward to
The OA system servers of industry or the business platform of enterprises are performed.In step S102 above, for tested text
The full text of this concentration has calculated the first classification results, in this step, and the first classification results of full text are carried out
After conformity calculation, it is possible to obtain the corresponding mood value of tested text set.In view of each text is for final Employees'Emotions
For analysis, significance level is probably differentiated.For example, tested text set from the daily paper of the same employee, weekly and toward
Carry out the documents such as mail to be constituted, in view of weekly is more regular, therefore, weekly is for the analysis of mood value, its significance level
Daily paper can be higher than.Considered based on such, can be in the corresponding mood value of the final tested text set of calculating determination, by each text
Originally processing is weighted, weights can be decided according to the actual requirements.For another example several employees of the tested text set by a department
Daily paper, weekly and the contact document such as mail constituted, in this case, can also be set according to the rank of different employees
The weights of text are determined, for example, the weights of the text of department manager are higher than the weights of the text of general employee.
Based on above-mentioned consideration, step S103 can be specially：According to the weighted value of each text set in advance, to each
First classification results of text are weighted summation operation, to obtain the corresponding mood value of tested text set.
In addition, as the linear two classification device of standard, can also further use PA (passive-aggressive)
When algorithm improves accuracy, i.e. repairing positive weights vector, a corrected parameter is added, when predicting correct, it is not necessary to adjustment power
Weight vector, during prediction error, active accommodation weight vectors, its advantage is to reduce the classification number of mistake, so as to improve correlation
Accuracy.
The tested corresponding mood value of text set embodies the mood value for the personnel to form tested text set, for example, the tested text
This collection comes from the daily paper within past one week, weekly and contact mail of same employee, then final mood value just embodies
The synthesis mood value that the employee was embodied past one week.Value setting based on the first classification results above, finally
The corresponding mood value of tested text set can value between+1 and-1, by final mood value, can understand and obtain exactly
The emotional status of employee is known, so as to provide important evidence for the internal control and decision-making of enterprise.
Above-mentioned text can be an envelope mail, a daily paper or weekly or division statement therein etc..
In practical application, the calculating of Employees'Emotions value is general in units of week, according to the mood value of employee's last week, prediction this week employee's
Mood is with respect to being last week more preferable or more bad.The calculation formula of mood value is：Assuming that i-th week mood tendency is+1, -1 and 0
The quantity of text be respectively xi、yi、zi, then the mood value of the prediction in this week is SiFor：
si=(zi*w1+w2)*(w3*xi+w4*yi+w5) ... ... ... ... ... ... formula (3)
Wherein, w1、w2、w3、w4The weighted value being inclined to for each different mood.For neutral text, it is not pair
The mood value of employee is not contributed, and it can reflect temperature of some employee to something, is inclined to by adjusting different moods
The weighted value of text cause the final mood value to be reasonable and accurate.
Elaborate the extraction process of keyword again below, specifically, before step S101, can also include：
S1001：Remove the stop words in tested text set.Stop words mainly includes auxiliary words of mood, adverbial word, preposition, connection
Word etc., these usual words have no clear and definite meaning, and only putting it into a complete sentence just has certain effect, such as common
" ", " " etc.The dimension of characteristic vector can be greatly reduced by removing stop words.
S1002：According to default corpus, it is tested in text set, extracts the word conduct that TF-IDF values are higher than predetermined threshold value
The keyword of the tested text set, and according to the whole keywords extracted, generation is tested the corresponding keyword of text set with this
Dictionary, keyword is included in corpus.Keyword can be specifically selected based on FudanNLP corpus, predetermined threshold value can
To set according to actual needs.
By two steps above, can remove a large amount of be inclined on text semantic influences and unconspicuous word, greatly
Reduce the dimension of characteristic vector.Counted according to actual experiment, when not extracting keyword, for carrying out Employees'Emotions detection
In text, containing more than 16000 word in 100000 training texts, and extract after keyword, 100000 training texts contain
There are 2000 keywords, so that the dimension of characteristic vector greatly reduces.
Embodiment two
As shown in figure 5, its for the embodiment of the present invention two mood value detection means structural representation, including feature to
Measure generation module 11, mood grader 12, mood value computing module 13：
Feature vector generation module 11, for carrying out feature extraction to the multiple texts being tested in text set, is generated multiple
Characteristic vector.In this feature vector generation module 11, feature extraction is carried out to the multiple texts being tested in text set, generated many
The processing of individual characteristic vector can be specifically included：
According to the keyword dictionary that the keyword quantity pre-established is N, for each text being tested in text set, meter
Calculate the corresponding TF-IDF values of each keyword in keyword dictionary；
The corresponding characteristic vector of each text is generated according to the corresponding TF-IDF values of each keyword, wherein, characteristic vector
Dimension be N, the component in each dimension of characteristic vector is the corresponding TF-IDF values of each keyword in keyword dictionary.
Mood grader 12, for obtaining multiple characteristic vectors as input, generation based on positive mood, negative emotions,
Multiple first classification results of neutral mood.Specifically, mood grader 12 can be specifically included：For to positive mood and negative
The first sub-classifier 121 that face mood is classified, the second subclassification for being classified to positive mood and neutral mood
Device 122, the 3rd sub-classifier 123 for being classified to neutral mood and negative emotions.
Based on the framework of above three sub-classifier, the multiple characteristic vectors of above-mentioned acquisition are as input, and generation is based on just
Face mood, negative emotions, the processing of multiple first classification results of neutral mood can be specifically included：
Each characteristic vector is inputted into each sub-classifier respectively, the second classification results of each sub-classifier are generated；
Ballot computing is carried out to the second classification results of each sub-classifier, each characteristic vector is obtained corresponding first point
Class result.
Mood value computing module 13, for according to multiple first classification results, calculating the corresponding mood of tested text set
Value.In mood value computing module 13, the first classification results using value as 1, -1 and 0 can correspond respectively to positive mood, it is negative
Mood and neutral mood, correspondingly, according to multiple first classification results, calculate the place of the corresponding mood value of tested text set
Reason can be specifically included：According to the weighted value of each text set in advance, the first classification results of each text are added
Summation operation is weighed, to obtain the corresponding mood value of tested text set.
In addition, the detection means of the present embodiment, can also include：
Stop words removes module 14, for removing the stop words in tested text set；
Keyword extracting module 15, for according to default corpus, extracting in multiple texts in tested text set,
TF-IDF values are tested the keyword of text set higher than the word of predetermined threshold value as this, and according to the whole keywords extracted, it is raw
Into keyword dictionary corresponding with the tested text set, keyword is included in corpus.
The detection means of the mood value of the present embodiment, passes through three points based on positive mood, negative emotions and neutral mood
Class result calculates final mood value, substantially increases the accuracy of mood value, so as to being preferably company manager couple
The judgement and decision-making of employee's affective state provide reliable foundation.
Embodiment three
The built-in function and structure of the detection means of mood value are the foregoing described, Fig. 6 is electricity provided in an embodiment of the present invention
The structural representation of sub- equipment, as shown in fig. 6, in practice, the detection means of the mood value in above-described embodiment can be realized as one
Electronic equipment is planted, can be included：Memory 111 and processor 112.
Memory 111, for storage program.
In addition to said procedure, memory 111 is also configured to the other various data of storage to support in electronic equipment
On operation.The example of these data includes the instruction of any application program or method for operating on an electronic device, example
Such as electric business platform, telephone book data, message, picture, video etc. on shopping platform on line, line.
Memory 111 can realize by any kind of volatibility or non-volatile memory device or combinations thereof,
Such as static RAM (SRAM), Electrically Erasable Read Only Memory (EEPROM), erasable programmable is read-only
Memory (EPROM), programmable read only memory (PROM), read-only storage (ROM), magnetic memory, flash memory, disk
Or CD.
Processor 112, coupled to memory 111, for performing the program in memory 111, for：
Feature extraction is carried out to the multiple texts being tested in text set, multiple characteristic vectors are generated；
Multiple characteristic vectors are inputted into mood grader respectively, generated based on positive mood, negative emotions, neutral mood
Multiple first classification results；
According to multiple first classification results, the corresponding mood value of tested text set is calculated.
Wherein, mood grader can include：For the first subclassification classified to positive mood and negative emotions
Device, for positive mood and neutral mood are classified the second sub-classifier, for entering to neutral mood and negative emotions
3rd sub-classifier of row classification,
Multiple characteristic vectors are then inputted into mood grader respectively, generation is based on positive mood, negative emotions, neutral mood
Multiple first classification results can include：
Each characteristic vector is inputted into each sub-classifier respectively, the second classification results of each sub-classifier are generated；
Ballot computing is carried out to the second classification results of each sub-classifier, each characteristic vector is obtained corresponding first point
Class result.
Above-mentioned specific processing operation is described in detail in preceding embodiment, will not be repeated here.
Further, as shown in fig. 6, electronic equipment can also include：Communication component 113, power supply module 114, audio-frequency assembly
115th, other components such as display 116.Members are only schematically provided in Fig. 6, are not meant to that electronic equipment only includes Fig. 6
Shown component.
Communication component 113 is configured to facilitate the communication of wired or wireless way between electronic equipment and other equipment.Electricity
Sub- equipment can access the wireless network based on communication standard, such as WiFi, 2G or 3G, or combinations thereof.It is exemplary at one
In embodiment, communication component 113 receives broadcast singal or broadcast correlation from external broadcasting management system via broadcast channel
Information.In one exemplary embodiment, communication component 113 also includes near-field communication (NFC) module, to promote junction service.
For example, radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra wide band (UWB) skill can be based in NFC module
Art, bluetooth (BT) technology and other technologies are realized.
Power supply module 114, electric power is provided for the various assemblies of electronic equipment.Power supply module 114 can include power management
System, one or more power supplys, and other generate, manage and distribute the component that electric power is associated with for electronic equipment.
Audio-frequency assembly 115 is configured as output and/or input audio signal.For example, audio-frequency assembly 115 includes a Mike
Wind (MIC), when electronic equipment be in operator scheme, when such as call model, logging mode and speech recognition mode, microphone by with
It is set to reception external audio signal.The audio signal received can be further stored in memory 111 or via communication set
Part 113 is sent.In certain embodiments, audio-frequency assembly 115 also includes a loudspeaker, for exports audio signal.
Display 116 includes screen, and its screen can include liquid crystal display (LCD) and touch panel (TP).If screen
Curtain includes touch panel, and screen may be implemented as touch-screen, to receive the input signal from user.Touch panel includes one
Individual or multiple touch sensors are with the gesture on sensing touch, slip and touch panel.Touch sensor can not only sense tactile
Touch or sliding action border, but also detection and touch or slide related duration and pressure.
One of ordinary skill in the art will appreciate that：Realizing all or part of step of above-mentioned each method embodiment can lead to
The related hardware of programmed instruction is crossed to complete.Foregoing program can be stored in a computer read/write memory medium.The journey
Sequence upon execution, performs the step of including above-mentioned each method embodiment；And foregoing storage medium includes：ROM, RAM, magnetic disc or
Person's CD etc. is various can be with the medium of store program codes.
Finally it should be noted that：Various embodiments above is merely illustrative of the technical solution of the present invention, rather than its limitations；To the greatest extent
The present invention is described in detail with reference to foregoing embodiments for pipe, it will be understood by those within the art that：Its according to
The technical scheme described in foregoing embodiments can so be modified, or which part or all technical characteristic are entered
Row equivalent substitution；And these modifications or replacement, the essence of appropriate technical solution is departed from various embodiments of the present invention technology
The scope of scheme.
Claims (12)
1. a kind of detection method of mood value, it is characterised in that including：
Feature extraction is carried out to the multiple texts being tested in text set, multiple characteristic vectors are generated；
The multiple characteristic vector is inputted into mood grader respectively, generated based on positive mood, negative emotions, neutral mood
Multiple first classification results；
According to the multiple first classification results, the corresponding mood value of tested text set is calculated.
2. detection method according to claim 1, it is characterised in that
The mood grader includes：For positive mood and negative emotions are classified the first sub-classifier, for pair
The second sub-classifier that positive mood and neutral mood are classified, for neutral mood and negative emotions are classified the
Three sub-classifiers,
The multiple characteristic vector is inputted into mood grader respectively, generated based on positive mood, negative emotions, neutral mood
Multiple first classification results include：
Each characteristic vector is inputted into each sub-classifier respectively, the second classification results of each sub-classifier are generated；
Ballot computing is carried out to the second classification results of each sub-classifier, the corresponding institute of each characteristic vector is obtained
State the first classification results.
3. detection method according to claim 1, it is characterised in that
Feature extraction is carried out to the multiple texts being tested in text set, generating multiple characteristic vectors includes：
According to the keyword dictionary that the keyword quantity pre-established is N, for each text being tested in text set, institute is calculated
State the corresponding TF-IDF values of each keyword in keyword dictionary；
The corresponding characteristic vector of each text is generated according to the corresponding TF-IDF values of each described keyword, wherein, the feature
The dimension of vector is that the component in N, each dimension of the characteristic vector is the corresponding TF- of each keyword in keyword dictionary
IDF values.
4. detection method according to claim 1, it is characterised in that the value of first classification results is 1, -1 and 0,
Positive mood, negative emotions and neutral mood are corresponded respectively to,
According to the multiple first classification results, calculating the tested corresponding mood value of text set includes：
According to the weighted value of each text set in advance, summation operation is weighted to the first classification results of each text,
To obtain the corresponding mood value of tested text set.
5. detection method according to claim 1, it is characterised in that carried out in multiple texts to being tested in text set special
Levy extraction, before generating multiple characteristic vectors, in addition to：
Remove the stop words in the tested text set；
According to default corpus, in tested text set, word of the TF-IDF values higher than predetermined threshold value is extracted as the tested text
The keyword of this collection, and according to the whole keywords extracted, generation is tested the corresponding keyword dictionary of text set with this, described
Keyword is included in the corpus.
6. a kind of detection means of mood value, it is characterised in that including：
Feature vector generation module, for carrying out feature extraction to the multiple texts being tested in text set, generate multiple features to
Amount；
Mood grader, for obtaining the multiple characteristic vector as input, generation based on positive mood, negative emotions, in
Multiple first classification results of vertical mood；
Mood value computing module, for according to the multiple first classification results, calculating the corresponding mood value of tested text set.
7. detection means according to claim 6, its characteristic value is,
The mood grader includes：For positive mood and negative emotions are classified the first sub-classifier, for pair
The second sub-classifier that positive mood and neutral mood are classified, for neutral mood and negative emotions are classified the
Three sub-classifiers,
The multiple characteristic vector is obtained as input, generate based on positive mood, negative emotions, neutral mood multiple first
Classification results include：
Each characteristic vector is inputted into each sub-classifier respectively, the second classification results of each sub-classifier are generated；
Ballot computing is carried out to the second classification results of each sub-classifier, the corresponding institute of each characteristic vector is obtained
State the first classification results.
8. detection means according to claim 6, it is characterised in that in the feature vector generation module, to tested
Multiple texts in text set carry out feature extraction, and generating multiple characteristic vectors includes：
According to the keyword dictionary that the keyword quantity pre-established is N, for each text being tested in text set, institute is calculated
State the corresponding TF-IDF values of each keyword in keyword dictionary；
The corresponding characteristic vector of each text is generated according to the corresponding TF-IDF values of each described keyword, wherein, the feature
The dimension of vector is that the component in N, each dimension of the characteristic vector is the corresponding TF- of each keyword in keyword dictionary
IDF values.
9. detection means according to claim 6, it is characterised in that the value of first classification results is 1, -1 and 0,
Positive mood, negative emotions and neutral mood are corresponded respectively to,
In the mood value computing module, according to the multiple first classification results, the corresponding feelings of tested text set are calculated
Thread value includes：
According to the weighted value of each text set in advance, summation operation is weighted to the first classification results of each text,
To obtain the corresponding mood value of tested text set.
10. detection means according to claim 6, it is characterised in that also include：
Stop words removes module, for removing the stop words in the tested text set；
Keyword extracting module, for according to default corpus, in tested text set, extracts TF-IDF values and is higher than default threshold
The word of value is tested the keyword of text set as this, and according to the whole keywords extracted, generation and the tested text set pair
The keyword dictionary answered, the keyword is included in the corpus.
11. a kind of electronic equipment, it is characterised in that including：
Memory, for storage program；
Processor, coupled to the memory, for performing described program, for：
Feature extraction is carried out to the multiple texts being tested in text set, multiple characteristic vectors are generated；
The multiple characteristic vector is inputted into mood grader respectively, generated based on positive mood, negative emotions, neutral mood
Multiple first classification results；
According to the multiple first classification results, the corresponding mood value of tested text set is calculated.
12. electronic equipment according to claim 11, it is characterised in that
The mood grader includes：For positive mood and negative emotions are classified the first sub-classifier, for pair
The second sub-classifier that positive mood and neutral mood are classified, for neutral mood and negative emotions are classified the
Three sub-classifiers,
The multiple characteristic vector is inputted into mood grader respectively, generated based on positive mood, negative emotions, neutral mood
Multiple first classification results include：
Each characteristic vector is inputted into each sub-classifier respectively, the second classification results of each sub-classifier are generated；
Ballot computing is carried out to the second classification results of each sub-classifier, the corresponding institute of each characteristic vector is obtained
State the first classification results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710081287.9A CN107066442A (en) | 2017-02-15 | 2017-02-15 | Detection method, device and the electronic equipment of mood value |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710081287.9A CN107066442A (en) | 2017-02-15 | 2017-02-15 | Detection method, device and the electronic equipment of mood value |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107066442A true CN107066442A (en) | 2017-08-18 |
Family
ID=59621362
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710081287.9A Pending CN107066442A (en) | 2017-02-15 | 2017-02-15 | Detection method, device and the electronic equipment of mood value |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107066442A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019096047A1 (en) * | 2017-11-16 | 2019-05-23 | 阿里巴巴集团控股有限公司 | Interface display method, interface data processing method, client and server |
| CN110245236A (en) * | 2019-06-25 | 2019-09-17 | 北京向上一心科技有限公司 | Information demonstrating method, device and electronic equipment |
| CN110609996A (en) * | 2018-06-15 | 2019-12-24 | 阿里巴巴集团控股有限公司 | Text emotion recognition method and device and electronic equipment |
| WO2020119533A1 (en) * | 2018-12-14 | 2020-06-18 | 深圳壹账通智能科技有限公司 | Public sentiment warning method and apparatus based on recurrent neural network algorithm, terminal and medium |
| TWI712948B (en) * | 2018-09-20 | 2020-12-11 | 大數據股份有限公司 | Method for document sentimental analysis, apparatus and computer program product thereof |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100217595A1 (en) * | 2009-02-24 | 2010-08-26 | Korea Institute Of Science And Technology | Method For Emotion Recognition Based On Minimum Classification Error |
| CN102929937A (en) * | 2012-09-28 | 2013-02-13 | 福州博远无线网络科技有限公司 | Text-subject-model-based data processing method for commodity classification |
| CN103365997A (en) * | 2013-07-12 | 2013-10-23 | 华东师范大学 | Opinion mining method based on ensemble learning |
| CN103761239A (en) * | 2013-12-09 | 2014-04-30 | 国家计算机网络与信息安全管理中心 | Method for performing emotional tendency classification to microblog by using emoticons |
| CN104331451A (en) * | 2014-10-30 | 2015-02-04 | 南京大学 | Recommendation level scoring method for theme-based network user comments |
| CN106202481A (en) * | 2016-07-18 | 2016-12-07 | 量子云未来（北京）信息科技有限公司 | The evaluation methodology of a kind of perception data and system |
-
2017
- 2017-02-15 CN CN201710081287.9A patent/CN107066442A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100217595A1 (en) * | 2009-02-24 | 2010-08-26 | Korea Institute Of Science And Technology | Method For Emotion Recognition Based On Minimum Classification Error |
| CN102929937A (en) * | 2012-09-28 | 2013-02-13 | 福州博远无线网络科技有限公司 | Text-subject-model-based data processing method for commodity classification |
| CN103365997A (en) * | 2013-07-12 | 2013-10-23 | 华东师范大学 | Opinion mining method based on ensemble learning |
| CN103761239A (en) * | 2013-12-09 | 2014-04-30 | 国家计算机网络与信息安全管理中心 | Method for performing emotional tendency classification to microblog by using emoticons |
| CN104331451A (en) * | 2014-10-30 | 2015-02-04 | 南京大学 | Recommendation level scoring method for theme-based network user comments |
| CN106202481A (en) * | 2016-07-18 | 2016-12-07 | 量子云未来（北京）信息科技有限公司 | The evaluation methodology of a kind of perception data and system |
Non-Patent Citations (1)
| Title |
|---|
| 杜亚楠 刘业政: "基于修正G2特征筛选的中文微博情感组合分类", 《情报学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019096047A1 (en) * | 2017-11-16 | 2019-05-23 | 阿里巴巴集团控股有限公司 | Interface display method, interface data processing method, client and server |
| CN110609996A (en) * | 2018-06-15 | 2019-12-24 | 阿里巴巴集团控股有限公司 | Text emotion recognition method and device and electronic equipment |
| CN110609996B (en) * | 2018-06-15 | 2023-05-02 | 阿里巴巴集团控股有限公司 | Text emotion recognition method and device and electronic equipment |
| TWI712948B (en) * | 2018-09-20 | 2020-12-11 | 大數據股份有限公司 | Method for document sentimental analysis, apparatus and computer program product thereof |
| WO2020119533A1 (en) * | 2018-12-14 | 2020-06-18 | 深圳壹账通智能科技有限公司 | Public sentiment warning method and apparatus based on recurrent neural network algorithm, terminal and medium |
| CN110245236A (en) * | 2019-06-25 | 2019-09-17 | 北京向上一心科技有限公司 | Information demonstrating method, device and electronic equipment |
| CN110245236B (en) * | 2019-06-25 | 2021-07-20 | 北京向上一心科技有限公司 | Information presentation method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109783632B (en) | Customer service information pushing method and device, computer equipment and storage medium | |
| CN107066442A (en) | Detection method, device and the electronic equipment of mood value | |
| US10387784B2 (en) | Technical and semantic signal processing in large, unstructured data fields | |
| Wesslen | Computer-assisted text analysis for social science: Topic models and beyond | |
| WO2018184518A1 (en) | Microblog data processing method and device, computer device and storage medium | |
| CN112396108A (en) | Service data evaluation method, device, equipment and computer readable storage medium | |
| CN112084334B (en) | Label classification method and device for corpus, computer equipment and storage medium | |
| CN107807968A (en) | Question and answer system, method and storage medium based on Bayesian network | |
| CN108228808A (en) | Determine the method, apparatus of focus incident and storage medium and electronic equipment | |
| CN110941953A (en) | Automatic identification method and system for network false comments considering interpretability | |
| Kathuria et al. | Real time sentiment analysis on twitter data using deep learning (Keras) | |
| US20220156862A1 (en) | System and method for analyzing grantability of a legal filing | |
| US11099107B2 (en) | Component testing plan considering distinguishable and undistinguishable components | |
| Khemani et al. | A review on reddit news headlines with nltk tool | |
| CN113392920A (en) | Method, apparatus, device, medium, and program product for generating cheating prediction model | |
| CN113761193A (en) | Log classification method and device, computer equipment and storage medium | |
| Richetti et al. | Analysis of knowledge-intensive processes focused on the communication perspective | |
| CN111859955A (en) | Public opinion data analysis model based on deep learning | |
| Pohl et al. | Online processing of social media data for emergency management | |
| KR101592756B1 (en) | Method for dialogue management using memory intensity of user knowledge and apparatus for performing the method | |
| CN114462742A (en) | Risk prompting method, device, equipment and computer storage medium | |
| CN112069807A (en) | Text data theme extraction method and device, computer equipment and storage medium | |
| CN111242519A (en) | User characteristic data generation method and device and electronic equipment | |
| Ulyanovska et al. | STUDY OF THE PROCESS OF IDENTIFYING THE AUTHORSHIP OF TEXTS WRITTEN IN NATURAL LANGUAGE. | |
| Gupta | A Directionality & Novel Function for Stock Price Modelling Prediction Dynamics based on Multiple Learning |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170818 |
|
| RJ01 | Rejection of invention patent application after publication |