CN113590744B - Expandable emotion tracing method - Google Patents
Expandable emotion tracing method Download PDFInfo
- Publication number
- CN113590744B CN113590744B CN202110743688.2A CN202110743688A CN113590744B CN 113590744 B CN113590744 B CN 113590744B CN 202110743688 A CN202110743688 A CN 202110743688A CN 113590744 B CN113590744 B CN 113590744B
- Authority
- CN
- China
- Prior art keywords
- emotion
- model
- tracing
- training
- text
- 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.)
- Active
Links
- 230000008451 emotion Effects 0.000 title claims abstract description 86
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000004458 analytical method Methods 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims description 40
- 238000012360 testing method Methods 0.000 claims description 21
- 238000000605 extraction Methods 0.000 claims description 12
- 238000002372 labelling Methods 0.000 claims description 9
- 230000009193 crawling Effects 0.000 claims description 8
- 238000011156 evaluation Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000004140 cleaning Methods 0.000 claims description 5
- 230000007935 neutral effect Effects 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000006467 substitution reaction Methods 0.000 claims 1
- 238000011161 development Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 3
- 238000003058 natural language processing Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000000873 masking effect Effects 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
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
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/103—Formatting, i.e. changing of presentation of documents
- G06F40/117—Tagging; Marking up; Designating a block; Setting of attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
- G06F40/186—Templates
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Machine Translation (AREA)
Abstract
The invention provides an interpretable emotion tracing method, which aims at an emotion analysis problem, firstly provides an emotion tracing concept, and the emotion tracing refers to finding a complete reasoning link and supporting a text on the reasoning link in the emotion analysis process; the rule template is utilized, so that the emotion analysis process is more interpretable; the text support is found for the extracted rule by using the tracing evidence model, and the invention considers that only the rule with the text support can be used as the correct rule and added into the result, so that the result is more accurate; the trace evidence model is verified to be helpful for emotion tracing, so that emotion analysis is facilitated, and the method has good practicability.
Description
Technical Field
The invention relates to the field of natural language of computer technology, in particular to an interpretable emotion tracing method.
Background
Trace reasoning is studied in both formal logic and natural language processing. Causal reasoning refers to the fact that given a theory C and possibly implications Q, this implication cannot be demonstrated in theory C. There is a new fact f that adding f to C gives an indication Q, and in NLP f can be masked from C by masking.
Emotion analysis has recently been attracting attention from many researchers due to the rapid development of social media, and along with the rapid development of social media, information such as comments, forum discussions, and social platform utterances appear in large quantities, so that the emotion of expressive persons on things, events or subjects and other entities is mined for the information, and the understanding of the views of the crowd and the choices made by them is facilitated. From the beginning of year 2000, emotion analysis has become one of the most active fields of research for natural language processing.
Emotion analysis is to classify information emotionally, and some tasks are classified into two types: either positive or negative, there is a task to add neutral classification on this basis. Early emotion analysis methods included supervised methods based on conventional machine learning methods such as support vector machines, maximum entropy, naive bayes, etc., and unsupervised methods such as emotion dictionaries, etc. Later, with the development of machine learning, machine learning methods were also used in the task of emotion analysis, for example Kalchbrenner et al proposed a dynamic CNN model (DCNN) that uses a dynamic K-Max pooling algorithm as a nonlinear sub-sampling function, and feature maps were generated by a network that was able to capture the relationships between words. With the development of emotion analysis, thet et al have presented the task of aspect emotion analysis, where emotion and target information are considered simultaneously, and the target is usually an entity or aspect, i.e., a sentence and a target aspect are given, and aspect-level emotion classification aims at deducing the emotion polarity of a sentence for a given target aspect.
The existing method is basically an unsupervised emotion dictionary method or a supervised deep learning network method, but the existing model has the problem of weak interpretation and also has the problem of weak mobility, and for texts in another field, a great deal of labeling and training can be needed again by using the supervised method.
Disclosure of Invention
The invention provides an interpretable emotion tracing method, which solves the problem of low interpretability of an emotion analysis model in the past and obtains more convincing results through emotion tracing.
In order to achieve the technical effects, the technical scheme of the invention is as follows:
an interpretable emotion tracing method comprises the following steps:
S1: collecting data and preprocessing the collected data;
S2: training a rule extraction model;
S3: training a trace evidence model;
s4: forming emotion tracing results;
s5: and inputting unlabeled text to obtain emotion tracing results.
Further, the specific process of the step S1 is:
s11: crawling comment information from a social platform or a shopping platform;
S12: for the crawled information, cleaning and removing the parts which do not relate to evaluation, and reserving information expressing the emotion or opinion of the commodity or entity by the user;
S13: manually marking the information obtained by crawling, wherein the marked content comprises: the sentence related aspect category and the corresponding emotion polarity, the reasoning process for obtaining the result, the rule template obtained by templating the reasoning process, and marking which clauses belong to the evidence needing to be searched in the text and which belong to the external knowledge in the rule template.
Further, the specific process of step S2 is as follows:
s21: dividing the data set into a training set and a testing set, wherein the training set is 70% and the testing set is 30%;
s22: the rule extraction model is a sequence labeling question-answering model, wherein a linear layer of sequence labeling is added on the basis of a pre-training language model Roberta model;
S23: the model training can cover the part belonging to the template variable in each marked reasoning rule clause, convert the part into question sentences, take the marked variable result as the answer of the question, input the question sentences and the text, and train the model to enable the highest evaluation sequence to be the correct answer. All related clauses of all training sets are used as one complete training;
s24: in the training process, the model trained each time is recorded, and the model with the highest test set accuracy and single f1 value is obtained through test of the test set.
Further, the specific process of step S3 is as follows:
S31: constructing a data set of the step: the same category in the sentence is replaced by the same type of replacement by the variables involved in the rule template to obtain a new inference clause. The clause generated in this way is marked as false, and the originally marked inference chain clause is marked as true; training set accounts for 70% and test set accounts for 30%;
S32: the trace evidence model is characterized in that sentences in the text are spliced into corresponding encoded clauses, the middle of the clauses is divided by using Roberta special separators, after the encoding, vectors of < s > are added together in a Roberta pre-training language model, and the added vectors are input into a linear layer and softmax for normalization to obtain a judgment whether corresponding variables conform to the clause description condition;
s33: in the training process, once training of all samples is completed, storing a model, and finally using the model with the highest accuracy of the test set.
Further, the specific process of step S4 is as follows:
s41: the score of each clause obtained in the step S3 is used for obtaining the score of the whole emotion tracing link;
s42: comparing the scores of the links, and taking the link with the highest score in each aspect as a final emotion tracing result;
S43: after the emotion tracing result is obtained, the emotion analysis result of the segment of speech on the aspect can be obtained through the conclusion of the emotion tracing reasoning chain.
Further, the specific process of step S5 is:
S51: for each rule template, splitting clauses, covering variables and constructing a problem text;
S52: inputting the questions and the texts obtained in the previous step into a rule extraction model to obtain answers to the questions, replacing variables, constructing a next-step question text, and if no answer exists, considering that emotion tracing of the text cannot be performed by using the template;
s53: repeating S51 and S52 until all variables of all templates have been confirmed, i.e., confirmed as specific contents or confirmed as not found;
S54: inputting clauses related to the evidence to be searched from the text and sentences with corresponding variables mentioned in the text into a traceability evidence model to obtain scores of the clauses, thereby obtaining scores of results;
s54: and obtaining the score of the complete emotion tracing link from the score obtained in the previous step, thereby obtaining a final result.
Further, in the step S1, the crawler tool is used to crawl comment information from social websites and shopping websites, only comments expressing the emotion of the user are reserved, the data are manually marked with aspect emotion polarities, the polarities are classified into positive, negative and neutral, and rules are added into the data for use, and meanwhile, the marked rules also need to be used as input by extracting rule templates; in the step S2, the rule template is split and converted into question sentences, and a question-answering model is trained to obtain information of corresponding variables in the template; training a traceability evidence model aiming at the result obtained in the step S2, and scoring all sentences referring to related variables in the text by proving sentence information rule clause scoring in the step S2 through model calculation so as to score the combined rules; and (3) setting a threshold value for the score in the step S4 to obtain the score of each reasoning link, and determining a final emotion reasoning link to obtain a final result.
Further, in step S5, the variables in the rule template are covered by matching the unlabeled text with the rule template, designed into question sentences, the result of the variables is obtained by rule extraction model, the result of the next variable is continuously obtained in the template, and the result found by all the variables in the template sentences is known to obtain a plurality of emotion tracing reasoning links. And (3) obtaining the score of each link clause through the trace evidence model according to the result obtained in the last step, using the highest score in the same aspect as an emotion trace result, and using the link result as a conclusion of emotion analysis.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
Aiming at the emotion analysis problem, the invention provides an emotion tracing concept for the first time, wherein emotion tracing means that a complete reasoning link is found in the emotion analysis process and the reasoning link has text support; the rule template is utilized, so that the emotion analysis process is more interpretable; the text support is found for the extracted rule by using the tracing evidence model, and the invention considers that only the rule with the text support can be used as the correct rule and added into the result, so that the result is more accurate; the trace evidence model is verified to be helpful for emotion tracing, so that emotion analysis is facilitated, and the method has good practicability.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a flow chart of the application phase of the method of the present invention.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
for the purpose of better illustrating the embodiments, certain elements of the drawings may be omitted, enlarged or reduced and do not represent the actual product dimensions;
It will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
As shown in fig. 1-2, an interpretable emotion tracing method includes the following steps:
Step 1, preprocessing such as data crawling, cleaning, labeling and the like: the method comprises the steps of crawling comment information from social websites, shopping websites and the like by using a crawler tool, cleaning some meaningless texts, only keeping comments expressing the emotion of a user, manually marking the emotion polarity of the data, classifying the polarity into positive, negative and neutral, and adding using rules into the positive, negative and neutral. Meanwhile, the labeling also needs to take the rule template extracted by the labeled rule as input:
s11: crawling comment information from a social platform or a shopping platform;
S12: for the crawled information, cleaning and removing the parts which do not relate to evaluation, and reserving information expressing the emotion or opinion of the commodity or entity by the user;
S13: manually marking the information obtained by crawling, wherein the marked content comprises: the sentence related aspect category and the corresponding emotion polarity, the reasoning process for obtaining the result, the rule template obtained by templating the reasoning process, and marking which clauses belong to the evidence needing to be searched in the text and which belong to the external knowledge in the rule template.
Step 2: training an evidence extraction model: through the paragraphs obtained in the step 2, a evidence extraction model is used for extracting relevant evidence, the evidence is expressed in a triplet mode, and the relation is defined in the process of constructing a data set:
S21: dividing the data set into a training set and a testing set, wherein the training set accounts for 70% and the testing set accounts for 30%;
s22: the rule extraction model is a sequence labeling question-answering model, wherein a linear layer of sequence labeling is added on the basis of a pre-training language model Roberta model;
S23: the model training can cover the part belonging to the template variable in each marked reasoning rule clause, convert the part into question sentences, take the marked variable result as the answer of the question, input the question sentences and the text, and train the model to enable the highest evaluation sequence to be the correct answer. All related clauses of all training sets are used as one complete training;
S24: in the training process, recording a model trained each time, and testing through a test set to obtain a model with highest test set accuracy and single f1 value;
S25: in the using process, the value of the variable is obtained through the answer obtained from the last question and then is carried into the next clause to form the details of the next clause. For example, for rule templates: if A is positive, B is A, B belongs to C, then C is positive, question of the first clause: what is positive? After obtaining answer a, let A equal the answer, replace to the second clause, the question becomes: what is a? And so on until each variable of the entire rule template finds a specific answer.
Step 3: and (3) forming emotion tracing results: setting a threshold value for the score in the step 4 to obtain the score of each reasoning link, and determining a final emotion reasoning link to obtain a final result:
S31: constructing a data set of the step: the same category in the sentence is replaced by the same type of replacement by the variables involved in the rule template to obtain a new inference clause. The clause generated in this way is marked as false and the originally marked inference chain clause is marked as true. Training set accounts for 70% and test set accounts for 30%;
S32: the trace evidence model is characterized in that sentences in the text are spliced into corresponding encoded clauses, the middle of the clauses is divided by using Roberta special separators, after the encoding, vectors of < s > are added together in a Roberta pre-training language model, and the added vectors are input into a linear layer and softmax for normalization to obtain a judgment whether corresponding variables conform to the clause description condition;
S33: in the training process, the random gradient descent mode is used for updating parameters, once training of all samples is completed, a model is stored, and finally the model with the highest accuracy of the test set is used.
Step 4: and (3) forming emotion tracing results: setting a threshold value for the score in the step 4 to obtain the score of each reasoning link, and determining a final emotion reasoning link to obtain a final result:
s41: the score of each clause obtained in the step3 is used for obtaining the score of the whole emotion tracing link;
s42: comparing the scores of the links, and taking the link with the highest score in each aspect as a final emotion tracing result;
S43: after the emotion tracing result is obtained, the emotion analysis result of the segment of speech on the aspect can be obtained through the conclusion of the emotion tracing reasoning chain.
Step 5: inputting unlabeled text to obtain emotion tracing results: the method comprises the steps of matching unlabeled texts with a rule template, covering variables in the rule template, designing a question sentence, obtaining a variable result through a rule extraction model, inputting the result into the template, continuously obtaining a next variable result, knowing that all variables in the template sentence have found results, and obtaining a plurality of emotion tracing reasoning links. And (3) obtaining the score of each link clause through the trace evidence model according to the result obtained in the last step, using the highest score in the same aspect as the emotion trace result, and using the link result as the conclusion of emotion analysis:
S51: for each rule template, splitting clauses, covering variables and constructing a problem text;
S52: inputting the questions and the texts obtained in the previous step into a rule extraction model to obtain answers to the questions, replacing variables, constructing a next-step question text, and if no answer exists, considering that emotion tracing of the text cannot be performed by using the template;
s53: repeating S51 and S52 until all variables of all templates have been confirmed, i.e., confirmed as specific contents or confirmed as not found;
S54: inputting clauses related to the evidence to be searched from the text and sentences with corresponding variables mentioned in the text into a traceability evidence model to obtain scores of the clauses, thereby obtaining scores of results;
s54: and obtaining the score of the complete emotion tracing link from the score obtained in the previous step, thereby obtaining a final result.
The same or similar reference numerals correspond to the same or similar components;
The positional relationship depicted in the drawings is for illustrative purposes only and is not to be construed as limiting the present patent;
It is to be understood that the above examples of the present invention are provided by way of illustration only and not by way of limitation of the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (2)
1. An interpretable emotion tracing method is characterized by comprising the following steps of:
S1: collecting data and preprocessing the collected data;
S2: training a rule extraction model;
S3: training a trace evidence model;
s4: forming emotion tracing results;
S5: inputting unlabeled text to obtain emotion tracing results;
The specific process of the step S1 is as follows:
s11: crawling comment information from a social platform or a shopping platform;
S12: for the crawled information, cleaning and removing the parts which do not relate to evaluation, and reserving information expressing the emotion or opinion of the commodity or entity by the user;
S13: manually marking the information obtained by crawling, wherein the marked content comprises: the method comprises the steps of obtaining the aspect category related to sentences and the corresponding emotion polarity, obtaining the reasoning process of the result, templating the reasoning process to obtain a rule template, and marking which clauses belong to evidences needing to be searched in texts and which belong to external knowledge in the rule template;
the specific process of the step S2 is as follows:
S21: dividing the data set into a training set and a testing set, wherein the training set is 70% and the testing set is 30%;
s22: the rule extraction model is a sequence labeling question-answering model, wherein a linear layer of sequence labeling is added on the basis of a pre-training language model Roberta model;
S23: the model training can cover the part belonging to the template variable in each marked reasoning rule clause, convert the part into question sentences, take the marked variable result as the answer of the question, input the question sentences and the text, train the model to make the highest evaluation sequence be the correct answer, and take all the related clauses of all training sets as one-time complete training;
S24: in the training process, recording a model trained each time, and testing through a test set to obtain a model with highest test set accuracy and single f1 value;
the specific process of the step S3 is as follows:
S31: constructing a data set of the step: the method comprises the steps of substituting the variables related in a rule template in a similar substitution mode to obtain a new reasoning clause, marking the clause generated in the mode as false, and marking the originally marked reasoning chain clause as true; training set accounts for 70% and test set accounts for 30%;
S32: the trace evidence model is characterized in that sentences in the text are spliced into corresponding encoded clauses, the middle of the clauses is divided by using Roberta special separators, after the encoding, vectors of < s > are added together in a Roberta pre-training language model, and the added vectors are input into a linear layer and softmax for normalization to obtain a judgment whether corresponding variables conform to the clause description condition;
s33: in the training process, once training of all samples is completed, storing a model, and finally using the model with the highest accuracy of the test set;
the specific process of the step S4 is as follows:
s41: the score of each clause obtained in the step S3 is used for obtaining the score of the whole emotion tracing link;
s42: comparing the scores of the links, and taking the link with the highest score in each aspect as a final emotion tracing result;
s43: after the emotion tracing result is obtained, an emotion analysis result is obtained through the conclusion of the emotion tracing reasoning chain;
the specific process of the step S5 is as follows:
S51: for each rule template, splitting clauses, covering variables and constructing a problem text;
S52: inputting the questions and the texts obtained in the previous step into a rule extraction model to obtain answers to the questions, replacing variables, constructing a next-step question text, and if no answer exists, considering that emotion tracing of the text cannot be performed by using the template;
s53: repeating S51 and S52 until all variables of all templates have been confirmed, i.e., confirmed as specific contents or confirmed as not found;
S54: inputting clauses related to the evidence to be searched from the text and sentences with corresponding variables mentioned in the text into a traceability evidence model to obtain scores of the clauses, thereby obtaining scores of results;
s54: and obtaining the score of the complete emotion tracing link from the score obtained in the previous step, thereby obtaining a final result.
2. The interpretable emotion tracing method of claim 1, wherein in step S13, the emotion polarity is manually marked, and the polarities are classified into positive, negative and neutral.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110743688.2A CN113590744B (en) | 2021-06-30 | 2021-06-30 | Expandable emotion tracing method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110743688.2A CN113590744B (en) | 2021-06-30 | 2021-06-30 | Expandable emotion tracing method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113590744A CN113590744A (en) | 2021-11-02 |
CN113590744B true CN113590744B (en) | 2024-05-24 |
Family
ID=78245568
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110743688.2A Active CN113590744B (en) | 2021-06-30 | 2021-06-30 | Expandable emotion tracing method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113590744B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109101597A (en) * | 2018-07-31 | 2018-12-28 | 中电传媒股份有限公司 | A kind of electric power news data acquisition system |
CN110083700A (en) * | 2019-03-19 | 2019-08-02 | 北京中兴通网络科技股份有限公司 | A kind of enterprise's public sentiment sensibility classification method and system based on convolutional neural networks |
CN110427549A (en) * | 2019-06-20 | 2019-11-08 | 北京清博大数据科技有限公司 | A kind of network public opinion Source Tracing method, apparatus, terminal and storage medium |
CN110612525A (en) * | 2017-05-10 | 2019-12-24 | 甲骨文国际公司 | Enabling thesaurus analysis by using an alternating utterance tree |
CN110968699A (en) * | 2019-11-01 | 2020-04-07 | 数地科技(北京)有限公司 | Logic map construction and early warning method and device based on event recommendation |
CN112000804A (en) * | 2020-08-18 | 2020-11-27 | 安徽理工大学 | Microblog hot topic user group emotion tendentiousness analysis method |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170011029A1 (en) * | 2013-05-09 | 2017-01-12 | Moodwire, Inc. | Hybrid human machine learning system and method |
US20140365208A1 (en) * | 2013-06-05 | 2014-12-11 | Microsoft Corporation | Classification of affective states in social media |
-
2021
- 2021-06-30 CN CN202110743688.2A patent/CN113590744B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110612525A (en) * | 2017-05-10 | 2019-12-24 | 甲骨文国际公司 | Enabling thesaurus analysis by using an alternating utterance tree |
CN109101597A (en) * | 2018-07-31 | 2018-12-28 | 中电传媒股份有限公司 | A kind of electric power news data acquisition system |
CN110083700A (en) * | 2019-03-19 | 2019-08-02 | 北京中兴通网络科技股份有限公司 | A kind of enterprise's public sentiment sensibility classification method and system based on convolutional neural networks |
CN110427549A (en) * | 2019-06-20 | 2019-11-08 | 北京清博大数据科技有限公司 | A kind of network public opinion Source Tracing method, apparatus, terminal and storage medium |
CN110968699A (en) * | 2019-11-01 | 2020-04-07 | 数地科技(北京)有限公司 | Logic map construction and early warning method and device based on event recommendation |
CN112000804A (en) * | 2020-08-18 | 2020-11-27 | 安徽理工大学 | Microblog hot topic user group emotion tendentiousness analysis method |
Non-Patent Citations (1)
Title |
---|
基于情感词典与语义规则的微博情感分析;陈国兰;情报探索;20160229(第2期);正文第1-6页 * |
Also Published As
Publication number | Publication date |
---|---|
CN113590744A (en) | 2021-11-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107066446B (en) | Logic rule embedded cyclic neural network text emotion analysis method | |
Mukhtar et al. | Urdu sentiment analysis using supervised machine learning approach | |
CN106886580B (en) | Image emotion polarity analysis method based on deep learning | |
Tang et al. | Multi-label patent categorization with non-local attention-based graph convolutional network | |
CN109726745B (en) | Target-based emotion classification method integrating description knowledge | |
Wahid et al. | Cricket sentiment analysis from Bangla text using recurrent neural network with long short term memory model | |
CN110929034A (en) | Commodity comment fine-grained emotion classification method based on improved LSTM | |
CN107679110A (en) | The method and device of knowledge mapping is improved with reference to text classification and picture attribute extraction | |
CN113742733B (en) | Method and device for extracting trigger words of reading and understanding vulnerability event and identifying vulnerability type | |
Kandhro et al. | Sentiment analysis of students’ comment using long-short term model | |
CN116860978B (en) | Primary school Chinese personalized learning system based on knowledge graph and large model | |
Shao et al. | An optimized mining algorithm for analyzing students’ learning degree based on dynamic data | |
CN110851593A (en) | Complex value word vector construction method based on position and semantics | |
CN114417851A (en) | Emotion analysis method based on keyword weighted information | |
Saddam et al. | Sentiment analysis of flood disaster management in Jakarta on Twitter using support vector machines | |
Jeon et al. | Dropout prediction over weeks in MOOCs via interpretable multi-layer representation learning | |
CN109858550B (en) | Machine identification method for potential process failure mode | |
Karrabi et al. | Sentiment analysis of informal persian texts using embedding informal words and attention-based lstm network | |
Albesta et al. | The impact of sentiment analysis from user on Facebook to enhanced the service quality. | |
Sinapoy et al. | Comparison of lstm and indobert method in identifying hoax on twitter | |
CN117216221A (en) | Intelligent question-answering system based on knowledge graph and construction method | |
CN117592563A (en) | Power large model training and adjusting method with field knowledge enhancement | |
CN113590744B (en) | Expandable emotion tracing method | |
Aliyanto et al. | Supervised probabilistic latent semantic analysis (sPLSA) for estimating technology readiness level | |
CN115879463A (en) | Course element recognition model training and recognition method based on text mining |
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 |