CN111984769A - Information processing method and device of response system - Google Patents
Information processing method and device of response system Download PDFInfo
- Publication number
- CN111984769A CN111984769A CN202010620827.8A CN202010620827A CN111984769A CN 111984769 A CN111984769 A CN 111984769A CN 202010620827 A CN202010620827 A CN 202010620827A CN 111984769 A CN111984769 A CN 111984769A
- Authority
- CN
- China
- Prior art keywords
- emotion
- theme
- topic
- polarity
- classification
- 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
- 230000010365 information processing Effects 0.000 title claims abstract description 20
- 230000004044 response Effects 0.000 title claims abstract description 20
- 238000003672 processing method Methods 0.000 title claims abstract description 12
- 230000008451 emotion Effects 0.000 claims abstract description 428
- 238000000034 method Methods 0.000 claims abstract description 48
- 238000013145 classification model Methods 0.000 claims abstract description 37
- 238000004458 analytical method Methods 0.000 claims abstract description 32
- 230000002996 emotional effect Effects 0.000 claims abstract description 30
- 238000007781 pre-processing Methods 0.000 claims description 44
- 238000000611 regression analysis Methods 0.000 claims description 23
- 238000011156 evaluation Methods 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 13
- 238000010801 machine learning Methods 0.000 claims description 7
- 230000001419 dependent effect Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 description 42
- 230000008569 process Effects 0.000 description 14
- 230000008901 benefit Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 208000027534 Emotional disease Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000009193 crawling Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008909 emotion recognition Effects 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 230000036651 mood Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Databases & Information Systems (AREA)
- Human Computer Interaction (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the application discloses an information processing method and device of a response system. The method comprises the following steps: firstly, carrying out emotion analysis on a received statement to obtain a theme corresponding to the statement and an emotion score representing the emotion polarity and the emotion intensity of each theme; secondly, determining a theme classification and an emotion weight value corresponding to each theme according to each theme and the emotion polarity and theme classification model aiming at each theme, wherein under the condition that the emotion polarity of different theme classifications is positive or negative, the emotion weight values of the different theme classifications are different due to different degrees of influence on the whole emotional tendency; then, carrying out weighted summation according to the emotion scores and the emotion weight values corresponding to the themes to obtain an overall emotion score; and then, determining the matched sentences according to the overall emotion scores and responding. In the answering system, the overall emotion score obtained by the method is closer to the real emotion of the user, so that answering sentences which are more matched with the emotion of the user can be output.
Description
Technical Field
The present application relates to the field of information processing technologies, and in particular, to an information processing method and apparatus for an answering system.
Background
In answering systems, there is often a need for emotion analysis of the current utterance of a user. However, there are some topics that the user has related to a plurality of aspects at a time, and the emotion of each topic is different, which results in inaccurate overall emotion recognition for the user.
For example: the words input by the user are that "the mobile phone has good appearance and good pixels, and the NFC (Near Field Communication) function is very good, i.e. the mobile phone cannot be normally turned on". The user mentions four topics, and the emotion score of each topic is also different: the sentiment score for the appearance design is positive, the sentiment score for the quality of the camera is positive, the sentiment score for NFC is positive, and the sentiment score for power-on is negative. In this case, how to accurately calculate whether the emotion score of the whole user is positive or negative, and find a matching answer sentence according to the emotion score becomes a technical problem to be solved in the answer system.
Disclosure of Invention
In order to solve the above problem, embodiments of the present application provide an information processing method and apparatus for an answering system.
According to a first aspect of embodiments of the present application, there is provided an information processing method of an answering system, the method including: acquiring a first statement; obtaining a theme corresponding to the first sentence and an emotion score aiming at each theme according to the first sentence, wherein the emotion score comprises emotion polarity and emotion intensity, the emotion polarity comprises positive and negative, the positive represents positive emotion, the negative represents negative emotion, and the emotion intensity is a numerical value representing the emotion intensity; determining a theme classification and an emotion weight value corresponding to each theme according to each theme and an emotion polarity and theme classification model for each theme, wherein the theme classification comprises a first theme classification in which the emotion weight values are all lower than a median value when the emotion polarity is positive or negative, a second theme classification in which the emotion weight values are lower than the median value when the emotion polarity is positive and higher than the median value when the emotion polarity is negative, a third theme classification in which the emotion weight values are greater than or equal to the median value when the emotion polarity is positive or negative, and a fourth theme classification in which the emotion weight values are higher than the median value when the emotion polarity is positive and lower than the median value when the emotion polarity is negative; carrying out weighted summation according to the theme corresponding to the first statement, the emotion score of each theme and the emotion weight value corresponding to each theme to obtain the overall emotion score of the first statement; and determining a second sentence matched with the overall emotion score according to the overall emotion score so as to answer the first sentence.
According to an embodiment of the present application, obtaining a topic corresponding to a first sentence and an emotion score for each topic according to the first sentence includes: performing language preprocessing on the first sentence to obtain a first language preprocessing result, wherein the preprocessing comprises removing stop words and removing special characters; and obtaining the theme corresponding to the first sentence and the emotion score aiming at each theme according to the first language preprocessing result.
According to an embodiment of the present application, obtaining a topic corresponding to a first sentence and an emotion score for each topic according to a first language preprocessing result includes: obtaining a theme corresponding to the first sentence and emotion polarity aiming at each theme according to the first language preprocessing result; and determining the emotional intensity of each theme under the corresponding emotional polarity according to the first language preprocessing result, the theme corresponding to the first sentence and the emotional polarity aiming at each theme.
According to an embodiment of the present application, obtaining a topic corresponding to a first sentence and an emotion polarity for each topic according to a first language preprocessing result includes: and inputting the first language preprocessing result into an emotion polarity analysis system to obtain a theme corresponding to the first statement and emotion polarity for each theme, wherein the emotion polarity analysis system comprises a combined model established based on Bert, LSTM and CRF.
According to an embodiment of the present application, determining, according to the first language preprocessing result, the topic corresponding to the first sentence, and the emotion polarity for each topic, the emotion intensity for each topic under the corresponding emotion polarity includes: and inputting the first language preprocessing result, each theme in the theme corresponding to the first sentence and the emotion polarity aiming at each theme into an emotion polarity intensity analysis system to obtain the emotion intensity of each theme under the corresponding emotion polarity, wherein the emotion polarity intensity analysis system comprises a model based on a machine learning regression algorithm.
According to an embodiment of the present application, before determining the topic classification and the emotion weight value corresponding to each topic according to each topic, the emotion polarity for each topic, and the topic classification model, the method further includes: determining a regression analysis algorithm used by the topic classification model; training the topic classification model according to the user evaluation data to obtain a topic library corresponding to different topic classifications; correspondingly, determining a theme classification and an emotion weight value corresponding to each theme according to each theme, the emotion score for each theme and the theme classification model, comprising: determining a theme library where each theme is located according to each theme and the theme library; determining a theme classification corresponding to each theme according to the theme library; and determining the emotion weight value of each theme according to the theme classification, the emotion polarity of each theme and the emotion weight value of the theme classification under the corresponding emotion polarity.
According to an embodiment of the present application, the user evaluation data includes user scores and user comment sentences, the regression analysis algorithm includes a regression analysis algorithm using the user scores as dependent variables and using emotion polarities and emotion intensities of the user comment sentences as independent variables, and accordingly, the topic classification model is trained according to the user evaluation data to obtain topic libraries corresponding to different topic classifications, including: acquiring user scores and user comment sentences; obtaining themes corresponding to the user comment sentences and emotion polarity and emotion intensity aiming at each theme according to the user comment sentences; determining a theme classification corresponding to each theme according to the user score, the theme corresponding to the user comment statement, the emotion polarity and the emotion intensity aiming at each theme and a regression analysis algorithm; and storing each theme into a corresponding theme library according to the theme classification.
According to an embodiment of the present application, before determining the emotion weight value of each topic according to the topic classification, the emotion polarity of each topic, and the emotion weight value of the topic classification under the corresponding emotion polarity, the method further includes: and setting emotion weight values of different theme classifications under different emotion polarities.
According to a second aspect of embodiments of the present application, an information processing apparatus of a response system, the apparatus comprising: the first statement acquisition module is used for acquiring a first statement; the emotion analysis module is used for obtaining the corresponding theme of the first sentence and the emotion score aiming at each theme according to the first sentence, wherein the emotion score comprises emotion polarity and emotion intensity, the emotion polarity comprises positive and negative, positive represents positive emotion, negative represents negative emotion, and the emotion intensity is a numerical value representing the emotion intensity; the emotion weight value determining module is used for determining a theme classification and an emotion weight value corresponding to each theme according to each theme, the emotion polarity of each theme and a theme classification model, wherein the theme classification comprises a first theme classification with the emotion weight value lower than a median value under the condition that the emotion polarity is positive or negative, a second theme classification with the emotion weight value higher than the median value under the condition that the emotion polarity is positive and lower than the median value under the condition that the emotion polarity is negative, a third theme classification with the emotion weight value larger than or equal to the median value under the condition that the emotion polarity is positive or negative, and a fourth theme classification with the emotion weight value lower than the median value under the condition that the emotion polarity is positive and higher than the median value under the condition that the emotion polarity is negative; the overall emotion score calculation module is used for carrying out weighted summation according to the theme corresponding to the first statement, the emotion score of each theme and the emotion weight value corresponding to each theme to obtain the overall emotion score of the first statement; and the second sentence answering determination module is used for determining a second sentence matched with the overall emotion score according to the overall emotion score so as to answer the first sentence.
According to an embodiment of the present application, the apparatus further includes: the regression analysis algorithm determining module is used for determining a regression analysis algorithm used by the topic classification model; the theme classification model training module is used for training the theme classification model according to the user evaluation data to obtain a theme base corresponding to different theme classifications; accordingly, the emotion weight value determination module comprises: the theme base determining submodule is used for determining a theme base in which each theme is located according to each theme and the theme base; the theme class determining submodule is used for determining a theme class corresponding to each theme according to the theme library; and the emotion weight value determining module is used for determining the emotion weight value of each theme according to the theme classification, the emotion polarity of each theme and the emotion weight value of each theme under the corresponding emotion polarity of the theme classification.
The embodiment of the application discloses an information processing method and device of a response system. The method comprises the following steps: firstly, carrying out emotion analysis on a received statement to obtain a theme corresponding to the statement and an emotion score representing the emotion polarity and the emotion intensity of each theme; secondly, determining a theme classification and an emotion weight value corresponding to each theme according to each theme and the emotion polarity and theme classification model aiming at each theme, wherein under the condition that the emotion polarity of different theme classifications is positive or negative, the emotion weight values of the different theme classifications are different due to different degrees of influence on the whole emotional tendency; then, carrying out weighted summation according to the emotion scores and the emotion weight values corresponding to the themes to obtain an overall emotion score; and then, determining the matched sentences according to the overall emotion scores and responding. In the response system, the overall emotion score obtained by the method is closer to the real emotion of the user, so that response sentences which are more matched with the emotion of the user can be output, and the use satisfaction of the user can be greatly improved.
It is to be understood that the teachings of this application need not achieve all of the above-described benefits, but rather that specific embodiments may achieve specific technical results, and that other embodiments of this application may achieve benefits not mentioned above.
Drawings
The above and other objects, features and advantages of exemplary embodiments of the present application will become readily apparent from the following detailed description read in conjunction with the accompanying drawings. Several embodiments of the present application are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which:
in the drawings, the same or corresponding reference numerals indicate the same or corresponding parts.
Fig. 1 is a schematic diagram illustrating an implementation flow of an information processing method of a response system according to an embodiment of the present application;
FIG. 2 is a schematic diagram illustrating a topic classification interval determined by an information processing method of an answering system according to an embodiment of the application;
fig. 3 is a schematic diagram of a configuration of an information processing apparatus of a response system according to an embodiment of the present application.
Detailed Description
In order to make the objects, features and advantages of the present application more obvious and understandable, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
In the description herein, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present application, "a plurality" means two or more unless specifically limited otherwise.
According to a first aspect of the embodiments of the present application, there is provided an information processing method of an answering system, as shown in fig. 1, the method including: operation 110, obtain a first statement; an operation 120 of obtaining corresponding topics of the first sentence and an emotion score for each topic according to the first sentence, wherein the emotion score includes emotion polarities and emotion intensities, the emotion polarities include positive and negative, the positive represents a positive emotion, the negative represents a negative emotion, and the emotion intensity is a numerical value representing the degree of emotion; an operation 130 of determining a theme classification and an emotion weight value corresponding to each theme according to each theme and an emotion polarity and theme classification model for each theme, wherein the theme classification includes a first theme classification in which the emotion weight value is lower than a median value in a case where the emotion polarity is positive or negative, a second theme classification in which the emotion weight value is lower than the median value in a case where the emotion polarity is positive and the emotion weight value is higher than the median value in a case where the emotion polarity is negative, a third theme classification in which the emotion weight value is greater than or equal to the median value in a case where the emotion polarity is positive or negative, and a fourth theme classification in which the emotion weight value is higher than the median value in a case where the emotion polarity is negative; operation 140, performing weighted summation according to the theme corresponding to the first sentence, the emotion score of each theme, and the emotion weight value corresponding to each theme to obtain the overall emotion score of the first sentence; at operation 150, a second sentence that matches the overall sentiment score is determined from the overall sentiment score to answer the first sentence.
In operation 110, the first sentence refers to a sentence that needs to be answered and is received during the dialog, and is used for inquiring information or advancing the dialog. The embodiment of the present application does not limit the way of obtaining the first sentence, and may be: a first sentence in a voice form received by the voice receiving means; a first sentence received by the text entry system in text form; a first sentence received by the image acquisition system in the form of an image, and so on.
In operation 120, the topic corresponding to the first sentence refers to the topic related to the first sentence, and is a real word that can be used to express semantics, highlighting intent, or distinguishing classification, and is usually from a preset topic relation system. For example, assume that a topic relation system preset by the response system takes an evaluated object as a topic, wherein the topics related to the mobile phone include: the "mobile phone appearance", "photographing function", "boot process", "NFC function", "mobile phone gift", "storage capacity", and the like. The first statement received in operation 110 is "the mobile phone has good appearance, the pixels are also good, the NFC function is very good, that is, the mobile phone cannot be normally turned on" often, wherein "the mobile phone has good appearance" mainly expresses the evaluation of the mobile phone appearance, and the corresponding theme is "the mobile phone appearance"; the 'good pixel' is the evaluation of the photographing function of the mobile phone, and the corresponding theme is the 'photographing function'; the 'NFC function is also excellent' and is an evaluation on the NFC function, and a corresponding theme of the 'NFC function' is supposed to be the 'NFC function'; the 'normal startup is not always performed' is the evaluation of the startup process, and the corresponding subject is the 'startup process'. It can be seen that the first sentence corresponds to a plurality of themes, which are "mobile phone appearance", "photographing function", "booting process", and "NFC function", respectively.
The emotion score is a numerical value obtained by predicting the emotional tendency and the emotional intensity that a sentence wants to express, wherein the emotional polarity represents whether the emotional tendency is positive or negative, and the emotional intensity represents whether the emotional tendency is strong or weak. If the value range of the emotion score of the response system is real number of (1, 1), the emotion score of 0.1 represents a weak negative emotion. And the emotion scores aiming at each theme represent that the user wants to express emotional tendency and emotional intensity aiming at each theme.
In the embodiment of the present application, a specific implementation manner of how to obtain the topic corresponding to the first sentence and the emotion score for each topic according to the first sentence is not limited, and an implementer may adopt any applicable implementation manner. For example, a deep learning model can be established, features in a text are automatically learned through a machine, semantic relation between hidden words in the text is increased, and a topic corresponding to a first sentence is determined by a sentence classification method based on a convolutional neural network and a topic distribution method based on a cyclic neural network; then, the first sentence and each theme are input into the emotion analysis model to obtain emotion scores for each theme.
For example, for the first received sentence "the mobile phone has a good appearance, the pixels are also good, the NFC function is very good, that is, the mobile phone cannot be normally turned on, and after operation 120, the following themes and the emotion scores corresponding to each theme can be obtained:
the subject is as follows: the mobile phone appearance, the emotion score for the mobile phone appearance is + 0.3;
theme two: a photographing function, an emotion score for the photographing function being + 0.4;
theme three: an NFC function for which the sentiment score is + 0.3;
subject four: and in the starting process, the emotion score aiming at the starting process is-0.5.
The method comprises the steps that a theme corresponding to a first sentence is obtained according to the first sentence, the emotion score of each theme is an important data basis for determining the whole emotion score of the first sentence, whether the determined theme is correct or not is judged, and the accuracy of the whole emotion score of the first sentence is directly influenced if the emotion score of each theme is predicted to be accurate or not.
In operation 130, determining that the topic classification and the emotion weight value corresponding to each topic are an important link for determining the overall emotion score of the first sentence by using the topic classification model, which is also a main feature of the information processing method of the response system of the present application, different from the prior art.
Generally, the satisfaction degree of a user for a certain commodity is often determined by certain functions or characteristics of the commodity and whether the experience of the user brought by the functions or characteristics is pleasant, and how the influence of the experience of different functions on the satisfaction degree of the user is different. For example, for a smart phone, the main purpose is data connection, photographing and media playing, when the user experiences poor functions, the satisfaction evaluation is particularly affected, and very angry negative emotion is generated, while the user experiences gift on the mobile phone has relatively weak influence on the user satisfaction, and it is difficult for the user to generate strong emotional reaction. Therefore, the functions and the characteristics of the products are used as themes, the themes are classified, the emotion scores expressed by the themes are weighted and summed by different emotion weight values to obtain the overall emotion score of the user, and the overall emotion score can be closer to the real emotion feeling of the user. Based on the above inventive thought, the inventor of the present application provides a method for classifying themes and setting different emotion weight values for different theme classifications under different emotion polarities to obtain a theme classification and an emotion weight value corresponding to each theme.
In operation 130, the topic classification and emotion weight value corresponding to each topic are obtained according to a topic classification model, which sets some topic classifications in advance and sets emotion weight values of each topic classification under different emotion polarities.
It should be noted that the median in operation 130 refers to the middle value of the range of emotion weight values. Assuming that the weighted value ranges from a real number between 0 and 1, the median value is 0.5. Furthermore, in practice it is common to take a value about equal to or close to the median value, e.g. 0.4, rather than necessarily exactly equal to the median value.
The first category of topics among the categories of topics, e.g. undifferentiated topics, mainly means that the user satisfaction does not change much, or the user does not at all, regardless of the user experience with such functions or features. For example: when the mobile phone gift film is purchased, the user can not generate too much positive emotion when the requirement is met, and the user can not generate too much negative emotion when the requirement is not met.
A second category of topics among the categories of topics, e.g. required topics, mainly means that the user is not satisfied if the user experiences such functions or features poorly; the user may not exhibit a particularly positive mood if the user experiences such a function or feature well. For example: the voice call quality of the mobile phone can cause great dissatisfaction of a user if the call quality is poor, so that the passive emotion weight is high; if the call quality is normal, the user thinks that this is due, and the positive weight is not too high.
A third topic category in the topic categories, for example, the expectation topic, mainly means that if the user experiences better with such functions or characteristics, the user satisfaction will be improved; i.e., user satisfaction may be reduced if the user experiences no good with such functions or features. For example: the mobile phone pixels, the face recognition function and the like, when the user experiences the functions well, the positive emotion weight of the user is increased, and when the user does not meet the functions, the negative emotion weight of the user is increased.
The fourth topic classification in the topic classifications, for example, the excitement-type topic, mainly refers to some functions or features that are unexpected for the user, and if the functions or features are not available or the experience of the part of functions or features is not good, the satisfaction of the user is not reduced, but when the functions or features are available or the experience of the part of functions or features is good, the satisfaction of the user is greatly improved. For example: the mobile phone screen can be folded, when the mobile phone screen has the function, the positive emotion of a user can be increased sharply, and when the mobile phone screen does not have the function, the user can not generate too much negative emotion.
It is assumed that, through analysis of the topic classification model, the topics "mobile phone appearance", "photographing function", "booting process", and "NFC function" corresponding to the first sentence in the above example belong to the "no-difference special topic", "expected topic", and "required topic", respectively.
And the emotion weight of each topic classification under different emotion polarities is shown in the following table:
TABLE 1
Then, the emotion weight corresponding to each topic is as follows:
the subject is as follows: the emotion weight value corresponding to the appearance of the mobile phone is 0.1;
theme two: the emotion weight value corresponding to the photographing function is 0.4;
theme three: the emotion weight value corresponding to the NFC function is 0.4;
subject four: the emotion weight value corresponding to the boot process is 0.8.
In operation 140, the overall sentiment score may be calculated using the following formula:
wherein S is the overall sentiment score; n is the number of topics; i represents each topic; siIs the sentiment score for each topic; q. q.siIs the emotion weight value corresponding to the theme.
For example, the overall emotion score of the first sentence "the mobile phone has good appearance, good pixels and excellent NFC function, that is, the mobile phone cannot be normally turned on" can be calculated by the following equation:
0.3*0.1+0.4*0.4+0.3*0.4-0.5*0.8＝-0.09
from the above-mentioned overall emotion score-0.09 of the first sentence, it can be seen that the emotional tendency of the first sentence is a negative emotion, and the result is closer to the real emotional feeling that the user wants to express through the sentence.
In operation 150, the second sentence identified herein is not only semantically responsive to the first sentence, but also emotionally compatible with the first sentence, i.e., the emotion tag or emotion score of the second sentence matches the overall emotion score of the first sentence. For example, when the emotion of the user is positive, the user can respond with a response sentence with a positive emotion score, so that the user's interest is higher; when the emotion of the user is negative, the user can respond with an answer sentence with an emotion tag for understanding, commenting or soothing, so that the emotion of the user can be soothed. Therefore, the user experience of the response system can be greatly improved.
For example, for the first statement above, the system determines the following two alternative answer statements:
alternative statement one: "happy you like our cell phone, please always support our brand. "the sentiment score of the sentence is positive, the sentiment tags of the sentence are" and "," thank you "and" happy ";
alternative statement two: sorry, our product brought a bad experience for you, i recorded your precious opinion, we continued to improve the quality of the product, the emotional score of the statement was negative, and the emotional labels of the statement were "sorry", "pacify" and "fellow".
Then according to the overall emotion score-0.09 of the first sentence, the emotion score and the alternative label of the alternative sentence II can be determined to be matched with the overall emotion score and the alternative label, and the alternative sentence II is returned to respond. At this point, the user's experience may be much better than getting a response as shown in alternative sentence one.
According to an embodiment of the present application, obtaining a topic corresponding to a first sentence and an emotion score for each topic according to the first sentence includes: performing language preprocessing on the first sentence to obtain a first language preprocessing result, wherein the preprocessing comprises removing stop words and removing special characters; and obtaining the theme corresponding to the first sentence and the emotion score aiming at each theme according to the first language preprocessing result.
Here, Stop Words (Stop Words) mean Words having no specific meaning in information retrieval, for example, "hey", "ha", "do", and the like; or words that have too many meanings to determine a particular meaning, such as "about," "than," "from," "typing," etc. Most of the stop words are manually judged, continuously summarized and enriched to form a stop word list. When the stop words are removed, the stop words can be determined and removed according to a preset stop word list. The removal of stop words can greatly improve the search efficiency. In addition, the stop words also have the characteristic of extremely high occurrence frequency, and the removal of the stop words can greatly save the storage space.
Special characters are symbols that are less frequently used and difficult to directly enter than traditional or commonly used symbols. Such as mathematical symbols, unit symbols, tabs, etc. Typically, no special character has any semantics. Therefore, the preservation of these special characters does not substantially help emotion analysis or semantic analysis, but increases the amount of information processing, so that the removal of the special characters in the process of preprocessing the first sentence is beneficial to shortening the processing time.
According to an embodiment of the present application, obtaining a topic corresponding to a first sentence and an emotion score for each topic according to a first language preprocessing result includes: obtaining a theme corresponding to the first sentence and emotion polarity aiming at each theme according to the first language preprocessing result; and determining the emotional intensity of each theme under the corresponding emotional polarity according to the first language preprocessing result, the theme corresponding to the first sentence and the emotional polarity aiming at each theme.
In this embodiment, the process of obtaining emotion scores is divided into two stages: the first stage is mainly used for acquiring the theme corresponding to the first sentence and the emotion polarity aiming at each theme; the second stage is mainly used to obtain the emotional intensity of each topic under the corresponding emotional polarity. The division decomposes a more complex task into two tasks which are relatively easy to implement, and because the technology for acquiring the theme corresponding to the first sentence and the emotion polarity of each theme is relatively mature, the existing technology or model can be reused, so that the implementation difficulty can be greatly reduced, and the best implementation effect can be obtained.
According to an embodiment of the present application, obtaining a topic corresponding to a first sentence and an emotion polarity for each topic according to a first language preprocessing result includes: and inputting the first language preprocessing result into an emotion polarity analysis system to obtain a theme corresponding to the first statement and emotion polarity for each theme, wherein the emotion polarity analysis system comprises a combined model established based on Bert, LSTM and CRF.
In the existing solution, when obtaining the topic corresponding to the first sentence and the emotion polarity for each topic according to the first language preprocessing result, two independent models and two stages are usually used for implementation, that is, the topic corresponding to the first sentence is obtained according to the topic determination model, and then the emotion polarity for each topic is obtained according to the emotion analysis model. In the embodiment, one emotion polarity analysis system implemented by the combined model obtains the topic corresponding to the first sentence and the emotion polarity for each topic at one time. The emotion polarity analysis system realized by the combined model established based on Bert, LSTM and CRF can combine semantic analysis, theme determination and emotion polarity determination, and utilizes the result feedback of the combined training to adjust all parameters of the whole combined model, so that the combined model has the advantages of higher accuracy and low possibility of errors when being used for actual prediction.
According to an embodiment of the present application, determining, according to the first language preprocessing result, the topic corresponding to the first sentence, and the emotion polarity for each topic, the emotion intensity for each topic under the corresponding emotion polarity includes: and inputting the first language preprocessing result, each theme in the theme corresponding to the first sentence and the emotion polarity aiming at each theme into an emotion polarity intensity analysis system to obtain the emotion intensity of each theme under the corresponding emotion polarity, wherein the emotion polarity intensity analysis system comprises a model based on a machine learning regression algorithm.
The machine learning regression algorithm may be Logistic regression, GBDT regression, SVR regression, or other regression algorithms. After the regression algorithm is determined, a machine learning model is established according to the regression algorithm, and the model is trained by utilizing training data with emotion intensity labels acquired through various ways, so that a model for predicting the intensity value of emotion polarity can be obtained.
Here, the value to identify mild intensity is generally set between [ -1,1], the closer the intensity value is to 1, the stronger the positive emotion is represented; the closer the intensity value is to-1, the stronger the negative emotion is represented.
According to an embodiment of the present application, before determining the topic classification and the emotion weight value corresponding to each topic according to each topic, the emotion polarity for each topic, and the topic classification model, the method further includes: determining a regression analysis algorithm used by the topic classification model; training the topic classification model according to the user evaluation data to obtain a topic library corresponding to different topic classifications; correspondingly, determining a theme classification and an emotion weight value corresponding to each theme according to each theme, the emotion score for each theme and the theme classification model, comprising: determining a theme library where each theme is located according to each theme and the theme library; determining a theme classification corresponding to each theme according to the theme library; and determining the emotion weight value of each theme according to the theme classification, the emotion polarity of each theme and the emotion weight value of the theme classification under the corresponding emotion polarity.
In this embodiment, the topic classification model is a machine learning model using a regression analysis algorithm, and the topic classification model is trained through user evaluation data to obtain a topic library corresponding to different topic classifications. That is, after the training process is finished, most of the topics in the topic relation system preset by the response system are allocated to the topic libraries corresponding to different topic classifications. During actual prediction, only each topic corresponding to the first sentence needs to be searched in the topic library corresponding to different topic classifications, and if the topic can be searched in the topic library corresponding to a certain topic classification, the topic classification to which the topic belongs can be determined to be the topic classification corresponding to the topic library.
According to an embodiment of the present application, the user evaluation data includes user scores and user comment sentences, the regression analysis algorithm includes a regression analysis algorithm using the user scores as dependent variables and using emotion polarities and emotion intensities of the user comment sentences as independent variables, and accordingly, the topic classification model is trained according to the user evaluation data to obtain topic libraries corresponding to different topic classifications, including: acquiring user scores and user comment sentences; obtaining themes corresponding to the user comment sentences and emotion polarity and emotion intensity aiming at each theme according to the user comment sentences; determining a theme classification corresponding to each theme according to the user score, the theme corresponding to the user comment statement, the emotion polarity and the emotion intensity aiming at each theme and a regression analysis algorithm; and storing each theme into a corresponding theme library according to the theme classification.
The user evaluation data can be obtained by crawling comment information of a user on a certain product on a webpage, and can also be purchased from a third-party data provider. Wherein the user score is the user's overall score for the product. For example, in evaluating a product, five stars lit for a score of 5, 4 stars lit for a score of 4, and so on. The comment sentence of the user is the character input by the user in the comment box.
According to an embodiment of the present application, a regression analysis algorithm using a user score as a dependent variable and using emotion polarity and emotion intensity of a user comment sentence as independent variables includes a regression analysis algorithm using the following regression equation:
y＝∑k(βk，pos×xk，pos+βk，neg×xk，neg)+
wherein y is the user score; x is the number ofk，[osPositive emotional intensity of the kth topic; x is the number ofk，negThe negation of being the kth topic is the intercept term; k is the number of topics; beta is ak，posThe influence amplitude of the negative emotion of the user on the kth theme on the average value of the whole emotion is shown; beta is ak，negThe influence amplitude of the negative emotion of the user on the kth theme on the average value of the whole emotion is shown; correspondingly, obtaining the topic classification corresponding to the comment sentences according to the user score, the topics corresponding to the comment sentences, the emotional polarity of each topic, the emotional intensity of each topic and a regression analysis algorithm, wherein the topic classification comprises the following steps: substituting the user scores, the topics corresponding to the comment sentences, the emotion polarity of each topic and the emotion intensity of each topic into a regression equation to obtain betak，posAnd betak，neg(ii) a According to betak，posAnd betak，negAnd obtaining the topic classification corresponding to the comment sentences.
To obtain betak，posAnd betak，negThen, according to the topic classification interval shown in fig. 2, the topic classification corresponding to each topic is determined according to the interval in which the coordinates representing the topic fall; or root of Chinese thorowaxAccording to beta, Table 2k，posAnd betak，negDirectly determining the topic classification corresponding to the topic:
TABLE 2
According to an embodiment of the present application, before determining the emotion weight value of each topic according to the topic classification, the emotion polarity of each topic, and the emotion weight value of the topic classification under the corresponding emotion polarity, the method further includes: and setting emotion weight values of different theme classifications under different emotion polarities.
In the present embodiment, the set weight values are emotion weight values of different theme classes under different emotion polarities, which are the weight values shown in table 1. The specific value of the weight value can be obtained by manual experience summary, or can be obtained by some expert systems through continuous verification and correction of historical data and experimental data.
According to a second aspect of the embodiment of the present application, an information processing apparatus of a response system, as shown in fig. 3, the apparatus 30 includes: a first sentence acquisition module 301, configured to acquire a first sentence; the emotion analysis module 302 is used for obtaining the corresponding theme of the first sentence and an emotion score aiming at each theme according to the first sentence, wherein the emotion score comprises emotion polarity and emotion intensity, the emotion polarity comprises positive and negative, positive represents positive emotion, negative represents negative emotion, and the emotion intensity is a numerical value representing the emotion intensity; an emotion weight value determination module 303, configured to determine, according to each topic, an emotion polarity and a topic classification model for each topic, a topic classification and an emotion weight value corresponding to each topic, where the topic classification includes a first topic classification in which the emotion weight values are all lower than a median value when the emotion polarity is positive or negative, a second topic classification in which the emotion weight values are higher than the median value when the emotion polarity is positive and lower than the median value when the emotion polarity is negative, a third topic classification in which the emotion weight values are all greater than or equal to the median value when the emotion polarity is positive or negative, and a fourth topic classification in which the emotion weight values are lower than the median value when the emotion polarity is positive and higher than the median value when the emotion polarity is negative; the overall emotion score calculation module 304 is configured to perform weighted summation according to the topics corresponding to the first sentence, the emotion score of each topic, and the emotion weight value corresponding to each topic to obtain an overall emotion score of the first sentence; a second sentence answer determination module 305 for determining a second sentence matching the overall emotion score according to the overall emotion score to answer the first sentence.
According to an embodiment of the present application, the emotion analysis module 302 includes: the preprocessing submodule is used for carrying out language preprocessing on the first statement to obtain a first language preprocessing result, and the preprocessing comprises the steps of removing stop words and removing special characters; and the emotion score acquisition submodule is used for acquiring the theme corresponding to the first statement and the emotion score aiming at each theme according to the first language preprocessing result.
According to an embodiment of the present application, the emotion score obtaining sub-module includes: the emotion polarity acquisition unit is used for acquiring a theme corresponding to the first statement and emotion polarity aiming at each theme according to the first language preprocessing result; and the emotion intensity acquisition unit is used for determining the emotion intensity of each theme under the corresponding emotion polarity according to the first language preprocessing result, the theme corresponding to the first sentence and the emotion polarity of each theme.
According to an embodiment of the application, the emotion polarity acquisition unit is specifically configured to input the first language preprocessing result to an emotion polarity analysis system to obtain a topic corresponding to the first sentence and an emotion polarity for each topic, where the emotion polarity analysis system includes a combined model established based on Bert, LSTM, and CRF.
According to an embodiment of the application, the emotion intensity obtaining unit is specifically configured to input the first language preprocessing result, each topic in the topics corresponding to the first sentence, and the emotion polarity for each topic into the emotion polarity intensity analysis system to obtain the emotion intensity of each topic under the corresponding emotion polarity, where the emotion polarity intensity analysis system includes a model based on a machine learning regression algorithm.
According to an embodiment of the present application, the apparatus 30 further includes: the regression analysis algorithm determining module is used for determining a regression analysis algorithm used by the topic classification model; the theme classification model training module is used for training the theme classification model according to the user evaluation data to obtain a theme base corresponding to different theme classifications; accordingly, the emotion weight value determination module comprises: the theme base determining submodule is used for determining a theme base in which each theme is located according to each theme and the theme base; the theme class determining submodule is used for determining a theme class corresponding to each theme according to the theme library; and the emotion weight value determining module is used for determining the emotion weight value of each theme according to the theme classification, the emotion polarity of each theme and the emotion weight value of each theme under the corresponding emotion polarity of the theme classification.
According to an embodiment of the present application, the theme base determination sub-module includes: the comment data acquisition unit is used for acquiring user scores and user comment sentences; the comment data sentiment analysis unit is used for obtaining topics corresponding to the user comment sentences and sentiment polarity and sentiment intensity aiming at each topic according to the user comment sentences; the comment data topic classification unit is used for determining topic classification corresponding to each topic according to user scores, topics corresponding to user comment sentences, emotion polarity and emotion intensity aiming at each topic and a regression analysis algorithm; and the theme base storage unit is used for storing each theme into the corresponding theme base according to the theme classification.
According to an embodiment of the present application, the apparatus 30 further includes an emotion weight value setting module, configured to set emotion weight values of different theme classifications under different emotion polarities.
Here, it should be noted that: the above description of the embodiment of the information processing apparatus of the response system is similar to the description of the foregoing method embodiment, and has similar beneficial effects to the foregoing method embodiment, and therefore, the description is omitted here for brevity. For technical details that have not been disclosed in the description of the information processing apparatus of the answering system, please refer to the description of the foregoing method embodiments of the present application for understanding, and therefore will not be described again for brevity.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above-described device embodiments are merely illustrative, for example, the division of a unit is only one logical function division, and there may be other division ways in actual implementation, such as: multiple units or components may be combined, or may be integrated into another device, or some features may be omitted, or not implemented. In addition, the coupling, direct coupling or communication connection between the components shown or discussed may be through some interfaces, and the indirect coupling or communication connection between the devices or units may be electrical, mechanical or other forms.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units; can be located in one place or distributed on a plurality of network units; some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, all functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may be separately regarded as one unit, or two or more units may be integrated into one unit; the integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional unit.
Those of ordinary skill in the art will understand that: all or part of the steps for realizing the method embodiments can be completed by hardware related to program instructions, the program can be stored in a computer readable storage medium, and the program executes the steps comprising the method embodiments when executed; and the aforementioned storage medium includes: various media capable of storing program codes, such as a removable storage medium, a Read Only Memory (ROM), a magnetic disk, and an optical disk.
Alternatively, the integrated units described above in the present application may be stored in a computer-readable storage medium if they are implemented in the form of software functional modules and sold or used as independent products. Based on such understanding, the technical solutions of the embodiments of the present application may be essentially implemented or portions thereof that contribute to the prior art may be embodied in the form of a software product stored in a storage medium, and including several instructions for enabling a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the methods of the embodiments of the present application. And the aforementioned storage medium includes: a removable storage medium, a ROM, a magnetic disk, an optical disk, or the like, which can store the program code.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (10)
1. An information processing method of a response system, the method comprising:
acquiring a first statement;
obtaining corresponding themes of the first sentence and an emotion score aiming at each theme according to the first sentence, wherein the emotion score comprises emotion polarity and emotion intensity, the emotion polarity comprises positive and negative, positive represents positive emotion, negative represents negative emotion, and the emotion intensity is a numerical value representing the emotion intensity;
determining a theme classification and an emotion weight value corresponding to each theme according to each theme and an emotion polarity and theme classification model for each theme, wherein the theme classification comprises a first theme classification which is lower than a median value when the emotion polarity is positive or negative, a second theme classification which is lower than the median value when the emotion polarity is positive and higher than the median value when the emotion polarity is negative, a third theme classification which is higher than or equal to the median value when the emotion polarity is positive or negative, and a fourth theme classification which is higher than the median value when the emotion polarity is positive and lower than the median value when the emotion polarity is negative;
according to the theme corresponding to the first statement, the emotion score aiming at each theme and the emotion weight value corresponding to each theme, carrying out weighted summation to obtain the whole emotion score of the first statement;
and determining a second sentence matched with the overall emotion score according to the overall emotion score so as to answer the first sentence.
2. The method of claim 1, wherein the obtaining the corresponding topic and emotion score for each topic of the first sentence according to the first sentence comprises:
performing language preprocessing on the first statement to obtain a first language preprocessing result, wherein the preprocessing comprises removing stop words and removing special characters;
and obtaining the theme corresponding to the first sentence and the emotion score aiming at each theme according to the first language preprocessing result.
3. The method of claim 2, wherein the obtaining the topic corresponding to the first sentence and the emotion score for each topic according to the first language preprocessing result comprises:
obtaining a theme corresponding to the first sentence and emotion polarity aiming at each theme according to the first language preprocessing result;
and determining the emotional intensity of each theme under the corresponding emotional polarity according to the first language preprocessing result, the theme corresponding to the first sentence and the emotional polarity aiming at each theme.
4. The method of claim 3, wherein the obtaining of the corresponding topic of the first sentence and the emotion polarity for each topic according to the first language preprocessing result comprises:
and inputting the first language preprocessing result into an emotion polarity analysis system to obtain a theme corresponding to the first statement and emotion polarity for each theme, wherein the emotion polarity analysis system comprises a combined model established based on Bert, LSTM and CRF.
5. The method of claim 3, the determining, from the first language preprocessing result, the topic corresponding to the first sentence, and the emotion polarity for each topic, the emotion intensity for each topic at the respective emotion polarity, comprising:
inputting the first language preprocessing result, each theme in the theme corresponding to the first sentence and the emotion polarity aiming at each theme into an emotion polarity intensity analysis system to obtain the emotion intensity of each theme under the corresponding emotion polarity, wherein the emotion polarity intensity analysis system comprises a model based on a machine learning regression algorithm.
6. The method of claim 1, prior to said determining a topic classification and emotion weight value for each topic from each topic, emotion polarity for each topic, and topic classification model, the method further comprising:
determining a regression analysis algorithm used by the topic classification model;
training the theme classification model according to the user evaluation data to obtain a theme base corresponding to different theme classifications;
correspondingly, the determining the theme classification and the emotion weight value corresponding to each theme according to each theme, the emotion score for each theme and the theme classification model comprises the following steps:
determining a theme library in which each theme is located according to each theme and the theme library;
determining a theme classification corresponding to each theme according to the theme library;
and determining the emotion weight value of each theme according to the theme classification, the emotion polarity of each theme and the emotion weight value of the theme classification under the corresponding emotion polarity.
7. The method of claim 6, the user evaluation data comprising a user score and a user comment sentence, the regression analysis algorithm comprising a regression analysis algorithm with the user score as a dependent variable and the emotional polarity and emotional intensity of the user comment sentence as independent variables,
correspondingly, the training the topic classification model according to the user evaluation data to obtain a topic library corresponding to different topic classifications includes:
acquiring user scores and user comment sentences;
obtaining a theme corresponding to the user comment statement and emotion polarity and emotion intensity aiming at each theme according to the user comment statement;
determining a topic classification corresponding to each topic according to the user score, the topics corresponding to the user comment sentences, the emotion polarity and the emotion intensity aiming at each topic and the regression analysis algorithm;
and storing each theme into a corresponding theme library according to the theme classification.
8. The method of claim 6, prior to said determining an emotion weight value for each topic from the topic classification, the emotion polarity for each topic, and the emotion weight value for the topic classification at the respective emotion polarity, the method further comprising:
and setting emotion weight values of different theme classifications under different emotion polarities.
9. An information processing apparatus of a response system, the apparatus comprising:
the first statement acquisition module is used for acquiring a first statement;
the emotion analysis module is used for obtaining the corresponding theme of the first sentence and an emotion score aiming at each theme according to the first sentence, wherein the emotion score comprises emotion polarity and emotion intensity, the emotion polarity comprises positive and negative, positive represents positive emotion, negative represents negative emotion, and the emotion intensity is a numerical value representing the emotion intensity degree;
the emotion weight value determining module is used for determining a theme classification and an emotion weight value corresponding to each theme according to each theme, the emotion polarity of each theme and a theme classification model, wherein the theme classification comprises a first theme classification of which the emotion weight values are lower than a median value under the condition that the emotion polarity is positive or negative, a second theme classification of which the emotion weight values are higher than the median value under the condition that the emotion polarity is positive and lower than the median value under the condition that the emotion polarity is negative, a third theme classification of which the emotion weight values are greater than or equal to the median value under the condition that the emotion polarity is positive or negative, and a fourth theme classification of which the emotion weight values are lower than the median value under the condition that the emotion polarity is positive and higher than the median value under the condition that the emotion polarity is negative;
the overall emotion score calculation module is used for carrying out weighted summation according to the theme corresponding to the first statement, the emotion score of each theme and the emotion weight value corresponding to each theme to obtain the overall emotion score of the first statement;
and the second sentence answering determination module is used for determining a second sentence matched with the overall emotion score according to the overall emotion score so as to answer the first sentence.
10. The apparatus of claim 9, the apparatus further comprising:
the regression analysis algorithm determining module is used for determining a regression analysis algorithm used by the topic classification model;
the theme classification model training module is used for training the theme classification model according to the user evaluation data to obtain a theme base corresponding to different theme classifications;
accordingly, the emotion weight value determination module comprises:
the theme bank determining submodule is used for determining a theme bank where each theme is located according to each theme and the theme bank;
the theme class determining submodule determines a theme class corresponding to each theme according to the theme library;
and the emotion weight value determining module is used for determining the emotion weight value of each theme according to the theme classification, the emotion polarity of each theme and the emotion weight value of the theme classification under the corresponding emotion polarity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010620827.8A CN111984769B (en) | 2020-06-30 | 2020-06-30 | Information processing method and device of response system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010620827.8A CN111984769B (en) | 2020-06-30 | 2020-06-30 | Information processing method and device of response system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111984769A true CN111984769A (en) | 2020-11-24 |
| CN111984769B CN111984769B (en) | 2024-04-26 |
Family
ID=73437714
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010620827.8A Active CN111984769B (en) | 2020-06-30 | 2020-06-30 | Information processing method and device of response system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111984769B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116630633A (en) * | 2023-07-26 | 2023-08-22 | 上海蜜度信息技术有限公司 | Automatic labeling method and system for semantic segmentation, storage medium and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015122026A1 (en) * | 2014-02-12 | 2015-08-20 | 哲祥 志村 | Disease detection system and disease detection method |
| CN107656917A (en) * | 2016-07-26 | 2018-02-02 | 深圳联友科技有限公司 | A kind of Chinese sentiment analysis method and system |
| CN107908715A (en) * | 2017-11-10 | 2018-04-13 | 中国民航大学 | Microblog emotional polarity discriminating method based on Adaboost and grader Weighted Fusion |
| CN108255803A (en) * | 2016-12-29 | 2018-07-06 | 北京国双科技有限公司 | The judgment method and device of document emotion |
| CN108733838A (en) * | 2018-05-29 | 2018-11-02 | 东北电力大学 | User's behavior prediction system and method based on multipole sentiment analysis |
-
2020
- 2020-06-30 CN CN202010620827.8A patent/CN111984769B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015122026A1 (en) * | 2014-02-12 | 2015-08-20 | 哲祥 志村 | Disease detection system and disease detection method |
| CN107656917A (en) * | 2016-07-26 | 2018-02-02 | 深圳联友科技有限公司 | A kind of Chinese sentiment analysis method and system |
| CN108255803A (en) * | 2016-12-29 | 2018-07-06 | 北京国双科技有限公司 | The judgment method and device of document emotion |
| CN107908715A (en) * | 2017-11-10 | 2018-04-13 | 中国民航大学 | Microblog emotional polarity discriminating method based on Adaboost and grader Weighted Fusion |
| CN108733838A (en) * | 2018-05-29 | 2018-11-02 | 东北电力大学 | User's behavior prediction system and method based on multipole sentiment analysis |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116630633A (en) * | 2023-07-26 | 2023-08-22 | 上海蜜度信息技术有限公司 | Automatic labeling method and system for semantic segmentation, storage medium and electronic equipment |
| CN116630633B (en) * | 2023-07-26 | 2023-11-07 | 上海蜜度信息技术有限公司 | Automatic labeling method and system for semantic segmentation, storage medium and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111984769B (en) | 2024-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109918676B (en) | Method and device for detecting intention regular expression and terminal equipment | |
| CN107391493B (en) | Public opinion information extraction method and device, terminal equipment and storage medium | |
| CN111767403B (en) | Text classification method and device | |
| US20180349781A1 (en) | Method and device for judging news quality and storage medium | |
| KR20210083706A (en) | Apparatus and method for classifying a category of data | |
| CN110827797B (en) | Voice response event classification processing method and device | |
| CN106776566B (en) | Method and device for recognizing emotion vocabulary | |
| CN106610990B (en) | Method and device for analyzing emotional tendency | |
| CN112686051B (en) | Semantic recognition model training method, recognition method, electronic device and storage medium | |
| CN110633464A (en) | Semantic recognition method, device, medium and electronic equipment | |
| CN114625858A (en) | Intelligent government affair question-answer replying method and device based on neural network | |
| CN112860896A (en) | Corpus generalization method and man-machine conversation emotion analysis method for industrial field | |
| CN109284389A (en) | A kind of information processing method of text data, device | |
| Kumar et al. | " Having 2 hours to write a paper is fun!": Detecting Sarcasm in Numerical Portions of Text | |
| CN115374259A (en) | Question and answer data mining method and device and electronic equipment | |
| CN107797981B (en) | Target text recognition method and device | |
| CN108536671B (en) | Method and system for recognizing emotion index of text data | |
| CN114443846A (en) | Classification method and device based on multi-level text abnormal composition and electronic equipment | |
| CN112527963B (en) | Dictionary-based multi-label emotion classification method and device, equipment and storage medium | |
| CN111984769A (en) | Information processing method and device of response system | |
| CN111428487B (en) | Model training method, lyric generation method, device, electronic equipment and medium | |
| CN117195046A (en) | Abnormal text recognition method and related equipment | |
| JP4054046B2 (en) | Opinion determination database creation method and apparatus and program, opinion determination method and apparatus and program, and computer-readable recording medium | |
| CN115659961B (en) | Method, apparatus and computer storage medium for extracting text views | |
| CN111126038A (en) | Information acquisition model generation method and device and information acquisition method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |