CN109948665B - Human activity type classification method and system based on long-time and short-time memory neural network - Google Patents
Human activity type classification method and system based on long-time and short-time memory neural network Download PDFInfo
- Publication number
- CN109948665B CN109948665B CN201910152363.XA CN201910152363A CN109948665B CN 109948665 B CN109948665 B CN 109948665B CN 201910152363 A CN201910152363 A CN 201910152363A CN 109948665 B CN109948665 B CN 109948665B
- Authority
- CN
- China
- Prior art keywords
- activity type
- data
- human activity
- matrix
- comment
- 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
Images
Abstract
The invention discloses a human activity type classification method and system based on a long-time memory neural network. Aiming at the problem that the prior commenting website only considers the specific contents and things of commenting but ignores the activity types of the crowds behind different comments, the invention provides a method for combining a word2vec model and TF-IDF (Trans-IdF) to infer and identify the current activity types and states of people who send different comments on the commenting website. Constructing a comment data set by using a comment website, wherein the comment data set comprises a comment content and a label of an activity type, and a Word vector representation of the comment content is obtained by using a Word2Vec model; secondly, calculating keywords of each activity classification by using TF-IDF, and constructing a keyword dictionary; secondly, establishing a weight matrix by counting whether the words in each comment appear in the dictionary or not; and finally, splicing the word vectors of the comment contents with the weight matrix, and inputting the word vectors into a long-time memory neural network for training so as to obtain a human activity classification module.
Description
Technical Field
The invention relates to the field of data analysis, in particular to a human activity type classification method and system based on a long-time memory neural network.
Background
With the continuous development of computer networks, people can obtain information from the internet and directly or indirectly publish a large amount of comment information on various network platforms. The comment information expresses various contents and reflects the current activity state of the person who makes the comment on the back. Such as eating, traveling, shopping, or performing recreational activities, etc.
The content expressed by the comments is analyzed, so that the current activity type of the person who sends the comments can be known, and meanwhile, the current activity place of the person who sends the comments can be probably inferred. For example, if a webster posts a comment on a food on a comment website, it can be basically inferred that the webster is eating, and it can be roughly determined that the current activity place of the webster is a restaurant. Meanwhile, the change of the basic activity type and the change of the activity track of a person can be estimated by analyzing the comments of the person in a period of time. Abundant space structure information and user behavior rule information are hidden behind a large amount of user position and behavior track data. By deeply mining and utilizing the information, not only the daily behavior rules of the individual users and the common behavior characteristics of the group users can be found, but also the social relationship information can be mastered. The method has important significance for intelligent transportation, advertisement recommendation and the like.
Disclosure of Invention
By analyzing the comment data, the invention provides a human activity type classification method based on a long-time memory neural network by combining word2vec and TF-IDF methods so as to know the current activity type of the person who makes the comment.
According to one aspect of the present invention, a human activity type classification method based on a long-time and short-time memory neural network for solving the technical problem of the present invention includes the following steps:
s1, collecting different types of comment data from the comment website, and constructing an activity label for each piece of comment data to form a data set; each activity label corresponds to a human activity type;
s2, preprocessing the data set to remove noise in the data;
s3, extracting keywords of each human body activity type from the data set by using a TF-IDF method, and constructing a keyword dictionary containing each human body activity type;
s4, obtaining vector representation of the comment data by using a vector generation model;
s5, counting the number of keywords contained in different human activity types in each piece of comment data to construct a weight matrix;
s6, correspondingly splicing the weight matrix and the vector representation of each piece of comment data, and forming a spliced matrix after splicing each piece of comment data;
s7, taking each spliced matrix and the corresponding activity label as training data, inputting all the training data together to form a training set to a long-time memory neural network for training to obtain a human activity type classification model, and classifying the comment data to be processed by the obtained human activity type to obtain the human activity type.
Further, in the human activity type classification method based on the long-and-short memory neural network of the present invention, in step S2, the data preprocessing further includes: and dividing the connected words and punctuation marks by using a blank space.
Further, in the human activity type classification method based on the long-and-short memory neural network of the present invention, in step S4: the vector is expressed as a 1 x N vector, N is a positive integer greater than 1 and greater than the number M of human activity types;
in step S5: if k1, k2, … and kM keywords are respectively in a first human activity type, a second human activity type, … and an M human activity type of a keyword dictionary in one piece of comment data, the weight matrix is [ k1, k2, …, kP, 0,0, … and 0 ]; wherein the number of 0's after kM is N-M to expand the weight matrix to a vector of 1 × N;
in step S6: selecting the first N words of each piece of comment data, splicing the vector representation of each word as a row to form an N x N matrix, and then using the weight matrix as the last row to form an (N +1) x N matrix as a final spliced matrix; or, the first N words of each piece of comment data are selected, the vector representation of each word is spliced as a column to form a matrix with the size of N x N, and then the weight matrix is used as the last column to form a matrix with the size of N x N +1 as the final spliced matrix.
Further, in the human activity type classification method based on the long-term memory neural network of the present invention, in step S4, the vector generation model is a Word2Vec model.
Further, in the human activity type classification method based on the long-and-short memory neural network, the commenting website specifically refers to a Yelp commenting website.
According to another aspect of the present invention, to solve the technical problem, the adopted human activity type classification system based on a long-time and short-time memory neural network comprises the following modules:
the system comprises a data set forming module, a data processing module and a data processing module, wherein the data set forming module is used for collecting different types of comment data from a comment website and constructing an active tag for each piece of comment data to form a data set; each activity label corresponds to a human activity type;
the preprocessing module is used for preprocessing the data of the data set and removing noise in the data;
the dictionary building module is used for extracting the key words of each human body activity type from the data set by using a TF-IDF method and building a key word dictionary containing each human body activity type;
the vector representation module is used for obtaining vector representation of the comment data by utilizing the vector generation model;
the weight matrix construction module is used for counting the number of keywords respectively contained in different human activity types in each piece of comment data so as to construct a weight matrix;
the splicing module is used for correspondingly splicing the weight matrix and the vector representation of each piece of comment data respectively, and each piece of comment data is spliced to form a spliced matrix;
and the model training module is used for taking each spliced matrix and the corresponding activity label as training data, all the training data form a training set together and are input into the long-time memory-based neural network for training to obtain a human activity type classification model, and the obtained human activity type is used for classifying the comment data to be processed to obtain the human activity type.
Further, in the long-and-short-term memory neural network-based human activity type classification system of the present invention, in the preprocessing module, the data preprocessing further includes: and dividing the connected words and punctuation marks by using a blank space.
Further, in the long-and-short-term memory neural network-based human activity type classification system of the present invention, the vector representation module comprises: the vector is expressed as a 1 x N vector, N is a positive integer greater than 1 and greater than the number M of human activity types;
in the weight matrix construction module: if k1, k2, … and kM keywords are respectively in a first human activity type, a second human activity type, … and an M human activity type of a keyword dictionary in one piece of comment data, the weight matrix is [ k1, k2, …, kP, 0,0, … and 0 ]; wherein the number of 0's after kM is N-M to expand the weight matrix to a vector of 1 × N;
in the splicing module: selecting the first N words of each piece of comment data, splicing the vector representation of each word as a row to form an N x N matrix, and then using the weight matrix as the last row to form an (N +1) x N matrix as a final spliced matrix; or, the first N words of each piece of comment data are selected, the vector representation of each word is spliced as a column to form a matrix with the size of N x N, and then the weight matrix is used as the last column to form a matrix with the size of N x N +1 as the final spliced matrix.
Further, in the human activity type classification system based on the long-and-short memory neural network, in the vector representation module, the vector generation model refers to a Word2Vec model.
Further, in the human activity type classification system based on the long-and-short memory neural network, the review website specifically refers to a Yelp review website.
Aiming at the problem that the prior commenting website only considers the specific contents and things of commenting but ignores the activity types of the crowds behind different comments, the invention provides a method for combining a word2vec model and TF-IDF (Trans-IdF) to infer and identify the current activity types and states of people who send different comments on the commenting website. Constructing a comment data set by using a comment website, wherein the comment data set comprises a comment content and a label of an activity type, and a Word vector representation of the comment content is obtained by using a Word2Vec model; secondly, calculating keywords of each activity classification by using TF-IDF, and constructing a keyword dictionary; secondly, establishing a weight matrix by counting whether the words in each comment appear in the dictionary or not; and finally, splicing the word vectors of the comment contents with the weight matrix, and inputting the word vectors into a long-time memory neural network for training so as to obtain a human activity classification module.
Drawings
The invention will be further described with reference to the accompanying drawings and examples, in which:
FIG. 1 is a flowchart of an embodiment of a human activity type classification method based on a long-term and short-term memory neural network according to the present invention;
FIG. 2 is a diagram of the activity classification implementation steps of FIG. 1
Fig. 3 is a schematic diagram of an embodiment of the human activity type classification system based on a long-and-short memory neural network according to the present invention.
Detailed Description
For a more clear understanding of the technical features, objects and effects of the present invention, embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
Referring to fig. 1-2, a human activity type classification method based on a long-and-short-term memory neural network includes the following steps:
s1, collecting different types of comment data from the comment website, and constructing an activity label for each piece of comment data to form a data set; each activity tag corresponds to a type of human activity. In the present embodiment, the review website is selected as the Yelp review website. Types of human activities are for example: four, beauty, art & understanding, travel, cropping, services, active life, health, automotive, night life, pets, duration, churches, massmedia, etc. 14 human activity types, and 7000 activity tags in total, form a data set.
S2, preprocessing the data set, removing noise in the data, and dividing the connected words and punctuation marks by spaces;
s3, extracting the key words of each human activity type from the data set by using a TF-IDF method to obtain 14 types of different key words, and constructing a key word dictionary containing each human activity type.
And S4, obtaining the vector representation of the comment data by using a vector generation model-Word 2Vec model.
S5, counting the number of keywords contained in each comment data in 14 human activity types respectively to construct a weight matrix; for example, 10 words in a piece of data appear in a first class of the dictionary, 8 words appear in a second class of the dictionary, and so on, the weight matrix may be represented as [10, 8. A vector of size 1 x 14, each value in the vector being the number of words in the review data that occur in each class of the lexicon. The vector of size 1 x 14 is then expanded to a vector of 1 x 100. The remaining portion is entirely filled with 0, i.e., the weight matrix is changed from [10, 8.. eta.. n ] (size 1 × 14) to [10, 8.. eta.. n, 0,0,0,0,0.. eta. ] (size 1 × 100).
S6, correspondingly splicing the weight matrix and the vector representation of each piece of comment data, and forming a spliced matrix after splicing each piece of comment data: selecting the first 100 words of each piece of comment data, using the vector representation of each word as a row for splicing to form a matrix with the size of 100 x 100, using the weight matrix as the last row, and forming a matrix with the size of 101 x 100 as a final spliced matrix; or, the first 100 words of each piece of comment data are selected, the vector representation of each word is spliced as a column to form a matrix with the size of 100 × 100, the weight matrix is used as the last column, and a matrix with the size of 100 × 101 is formed as the final spliced matrix.
S7, taking each spliced matrix and the corresponding activity label as training data, inputting all the training data together to form a training set to a long-time memory neural network for training to obtain a human activity type classification model, and classifying the comment data to be processed by the obtained human activity type to obtain the human activity type.
Referring to fig. 3, according to another aspect of the present invention, in order to solve the technical problem, the human activity type classification system based on a long-time memory neural network includes the following modules: a dataset forming module 31, a preprocessing module 32, a dictionary building module 33, a vector representation module 34, a weight matrix building module 35, a stitching module 36, and a model training module 37.
The data set forming module 31 is used for collecting different types of comment data from a comment website-Yelp comment website, and constructing an activity label for each piece of comment data to form a data set; each activity label corresponds to a human activity type; the preprocessing module 32 is used for preprocessing the data of the data set, removing noise in the data, and segmenting words and punctuation marks which are connected together by using spaces; the dictionary building module 33 is configured to extract keywords of each human activity type from the data set by using a TF-IDF method, and build a keyword dictionary containing each human activity type; the vector representation module 34 is used for obtaining the vector representation of the comment data by utilizing a vector generation model-Word 2Vec model; the weight matrix construction module 35 is configured to count the number of keywords respectively contained in different human activity types in each piece of comment data to construct a weight matrix; the splicing module 36 is configured to splice the weight matrix and the vector representation of each piece of comment data correspondingly, and form a spliced matrix after each piece of comment data is spliced; the model training module 37 is configured to use each spliced matrix and the corresponding activity label as a training data, and all the training data together form a training set and input the training set to the long-time-based memory neural network for training to obtain a human activity type classification model, where the obtained human activity type is used to classify the comment data to be processed to obtain a human activity type.
The specific implementation manner of the splicing part is as follows:
in the vector representation module: the vector is expressed as a 1 x N vector, N is a positive integer greater than 1 and greater than the number M of human activity types;
in the weight matrix construction module: if k1, k2, … and kM keywords are respectively in a first human activity type, a second human activity type, … and an M human activity type of a keyword dictionary in one piece of comment data, the weight matrix is [ k1, k2, …, kP, 0,0, … and 0 ]; wherein the number of 0's after kM is N-M to expand the weight matrix to a vector of 1 × N;
in the splicing module: selecting the first N words of each piece of comment data, splicing the vector representation of each word as a row to form an N x N matrix, and then using the weight matrix as the last row to form an (N +1) x N matrix as a final spliced matrix; or, the first N words of each piece of comment data are selected, the vector representation of each word is spliced as a column to form a matrix with the size of N x N, and then the weight matrix is used as the last column to form a matrix with the size of N x N +1 as the final spliced matrix.
The innovative aspects of the present invention will be elucidated below.
Innovation in data. In past work, people only pay attention to specific content expressed by comment data, and rarely link specific activity types and places of comment backers. The method collects various comment data from the Yelp comment website, and manually marks corresponding activity type labels on the data. A first set of complete comment data and a data set corresponding to human activities are formed, and the method can be used for a series of work such as human activity type recognition, human activity track analysis and the like.
Innovation in character expression. Generally, words are not mathematically processable, which makes them difficult for computers to process. The invention uses different vector generation models to convert characters into vector form, so that the characters can be subjected to mathematical operation. And commenting data and key Word information of two different levels, wherein corresponding vector representation is obtained by respectively using a Word2Vec model and a TF-IDF model, so that the characters become computable, and original semantic information of the characters is kept.
Innovation in information fusion. For the classification problem, previous methods rely only on current textual information. This may cause the problem of incomplete information, resulting in a low accuracy of classification. The invention not only utilizes the information of the comment text, but also utilizes the key word information which is extracted by the TF-IDF method and represents each different category, and the information of the comment text and the key word information is coded and fused, so that the expression of the text information is more complete, and the classification accuracy is improved.
While the present invention has been described with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, which are illustrative and not restrictive, and it will be apparent to those skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (8)
1. A human activity type classification method based on a long-time and short-time memory neural network is characterized by comprising the following steps:
s1, collecting different types of comment data from the comment website, and constructing an activity label for each piece of comment data to form a data set; each activity label corresponds to a human activity type;
s2, preprocessing the data set to remove noise in the data;
s3, extracting keywords of each human body activity type from the data set by using a TF-IDF method, and constructing a keyword dictionary containing each human body activity type;
s4, obtaining vector representation of the comment data by using a vector generation model;
s5, counting the number of keywords contained in different human activity types in each piece of comment data to construct a weight matrix;
s6, correspondingly splicing the weight matrix and the vector representation of each piece of comment data, and forming a spliced matrix after splicing each piece of comment data;
s7, taking each spliced matrix and the corresponding activity label as training data, inputting all the training data together to form a training set to a long-time memory-based neural network for training to obtain a human activity type classification model, and classifying the comment data to be processed by the obtained human activity type to obtain the human activity type;
in step S4: the vector is expressed as a 1 x N vector, N is a positive integer greater than 1 and greater than the number M of human activity types;
in step S5: if k1, k2, … and kM keywords are respectively in a first human activity type, a second human activity type, … and an Mth human activity type of the keyword dictionary in one piece of comment data, the weight matrix is [ k1, k2, …, kM, 0,0, … and 0 ]; wherein the number of 0's after kM is N-M to expand the weight matrix to a vector of 1 × N;
in step S6: selecting the first N words of each piece of comment data, splicing the vector representation of each word as a row to form an N x N matrix, and then using the weight matrix as the last row to form an (N +1) x N matrix as a final spliced matrix; or, the first N words of each piece of comment data are selected, the vector representation of each word is spliced as a column to form a matrix with the size of N x N, and then the weight matrix is used as the last column to form a matrix with the size of N x N +1 as the final spliced matrix.
2. The human activity type classification method based on long-and-short memory neural network as claimed in claim 1, wherein in step S2, the data preprocessing further comprises: and dividing the connected words and punctuation marks by using a blank space.
3. The human activity type classification method based on the long-and-short term memory neural network as claimed in claim 1, wherein in step S4, the vector generation model is Word2Vec model.
4. The long-and-short memory neural network-based human activity type classification method as claimed in claim 1, wherein the review website is specifically a Yelp review website.
5. A human activity type classification system based on a long-time memory neural network is characterized by comprising the following modules:
the system comprises a data set forming module, a data processing module and a data processing module, wherein the data set forming module is used for collecting different types of comment data from a comment website and constructing an active tag for each piece of comment data to form a data set; each activity label corresponds to a human activity type;
the preprocessing module is used for preprocessing the data of the data set and removing noise in the data;
the dictionary building module is used for extracting the key words of each human body activity type from the data set by using a TF-IDF method and building a key word dictionary containing each human body activity type;
the vector representation module is used for obtaining vector representation of the comment data by utilizing the vector generation model;
the weight matrix construction module is used for counting the number of keywords respectively contained in different human activity types in each piece of comment data so as to construct a weight matrix;
the splicing module is used for correspondingly splicing the weight matrix and the vector representation of each piece of comment data respectively, and each piece of comment data is spliced to form a spliced matrix;
the model training module is used for taking each spliced matrix and the corresponding activity label as training data, all the training data form a training set together and are input into the long-time memory-based neural network for training to obtain a human activity type classification model, and the obtained human activity type is used for classifying the comment data to be processed to obtain the human activity type;
in the vector representation module: the vector is expressed as a 1 x N vector, N is a positive integer greater than 1 and greater than the number M of human activity types;
in the weight matrix construction module: if k1, k2, … and kM keywords are respectively in a first human activity type, a second human activity type, … and an Mth human activity type of the keyword dictionary in one piece of comment data, the weight matrix is [ k1, k2, …, kM, 0,0, … and 0 ]; wherein the number of 0's after kM is N-M to expand the weight matrix to a vector of 1 × N;
in the splicing module: selecting the first N words of each piece of comment data, splicing the vector representation of each word as a row to form an N x N matrix, and then using the weight matrix as the last row to form an (N +1) x N matrix as a final spliced matrix; or, the first N words of each piece of comment data are selected, the vector representation of each word is spliced as a column to form a matrix with the size of N x N, and then the weight matrix is used as the last column to form a matrix with the size of N x N +1 as the final spliced matrix.
6. The long-and-short-term memory neural network-based human activity type classification system as claimed in claim 5, wherein in the preprocessing module, the data preprocessing further comprises: and dividing the connected words and punctuation marks by using a blank space.
7. The long-and-short-term memory neural network-based human activity type classification system as claimed in claim 5, wherein in the vector representation module, the vector generation model is a Word2Vec model.
8. The long-and-short memory neural network-based human activity type classification system as claimed in claim 5, wherein the review website is specifically a Yelp review website.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910152363.XA CN109948665B (en) | 2019-02-28 | 2019-02-28 | Human activity type classification method and system based on long-time and short-time memory neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910152363.XA CN109948665B (en) | 2019-02-28 | 2019-02-28 | Human activity type classification method and system based on long-time and short-time memory neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109948665A CN109948665A (en) | 2019-06-28 |
| CN109948665B true CN109948665B (en) | 2020-11-27 |
Family
ID=67007038
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910152363.XA Active CN109948665B (en) | 2019-02-28 | 2019-02-28 | Human activity type classification method and system based on long-time and short-time memory neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109948665B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111191099B (en) * | 2019-12-30 | 2023-04-07 | 中国地质大学（武汉） | User activity type identification method based on social media |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682220A (en) * | 2017-01-04 | 2017-05-17 | 华南理工大学 | Online traditional Chinese medicine text named entity identifying method based on deep learning |
| CN108364028A (en) * | 2018-03-06 | 2018-08-03 | 中国科学院信息工程研究所 | A kind of internet site automatic classification method based on deep learning |
| CN108573047A (en) * | 2018-04-18 | 2018-09-25 | 广东工业大学 | A kind of training method and device of Module of Automatic Chinese Documents Classification |
| CN108763477A (en) * | 2018-05-29 | 2018-11-06 | 厦门快商通信息技术有限公司 | A kind of short text classification method and system |
| CN109033433A (en) * | 2018-08-13 | 2018-12-18 | 中国地质大学（武汉） | A kind of comment data sensibility classification method and system based on convolutional neural networks |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10460201B2 (en) * | 2015-12-31 | 2019-10-29 | Microsoft Technology Licensing, Llc | Structure and training for image classification |
| CN105740349B (en) * | 2016-01-25 | 2019-03-08 | 重庆邮电大学 | A kind of sensibility classification method of combination Doc2vec and convolutional neural networks |
| CN106547885B (en) * | 2016-10-27 | 2020-04-10 | 桂林电子科技大学 | Text classification system and method |
| CN109213861B (en) * | 2018-08-01 | 2022-03-29 | 上海电力学院 | Traveling evaluation emotion classification method combining At _ GRU neural network and emotion dictionary |
-
2019
- 2019-02-28 CN CN201910152363.XA patent/CN109948665B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682220A (en) * | 2017-01-04 | 2017-05-17 | 华南理工大学 | Online traditional Chinese medicine text named entity identifying method based on deep learning |
| CN108364028A (en) * | 2018-03-06 | 2018-08-03 | 中国科学院信息工程研究所 | A kind of internet site automatic classification method based on deep learning |
| CN108573047A (en) * | 2018-04-18 | 2018-09-25 | 广东工业大学 | A kind of training method and device of Module of Automatic Chinese Documents Classification |
| CN108763477A (en) * | 2018-05-29 | 2018-11-06 | 厦门快商通信息技术有限公司 | A kind of short text classification method and system |
| CN109033433A (en) * | 2018-08-13 | 2018-12-18 | 中国地质大学（武汉） | A kind of comment data sensibility classification method and system based on convolutional neural networks |
Non-Patent Citations (3)
| Title |
|---|
| "Sentiment Analysis of Online Reviews Using Bag-of-Words and LSTM Approaches";James Barry;《ASIC 2017.Computer Science》;20171231;第1-12页 * |
| "基于word2vec和LSTM的饮食健康文本分类研究";赵明 等;《农业机械学报》;20171031;第48卷(第10期);第202-208页 * |
| "基于word2vec的一种文档向量表示";唐明 等;《计算机科学》;20160630;第43卷(第6期);第214-217页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109948665A (en) | 2019-06-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11687728B2 (en) | Text sentiment analysis method based on multi-level graph pooling | |
| CN110633373B (en) | Automobile public opinion analysis method based on knowledge graph and deep learning | |
| CN111784455A (en) | Article recommendation method and recommendation equipment | |
| CN105574067A (en) | Item recommendation device and item recommendation method | |
| CN111767725B (en) | Data processing method and device based on emotion polarity analysis model | |
| US20200210442A1 (en) | Identifying and extracting addresses within content | |
| CN110096681B (en) | Contract term analysis method, apparatus, device and readable storage medium | |
| CN111783394A (en) | Training method of event extraction model, event extraction method, system and equipment | |
| CN111309936A (en) | Method for constructing portrait of movie user | |
| CN109584006B (en) | Cross-platform commodity matching method based on deep matching model | |
| CN104376010A (en) | User recommendation method and user recommendation device | |
| CN113254652B (en) | Social media posting authenticity detection method based on hypergraph attention network | |
| CN111666496A (en) | Group recommendation method based on comment text | |
| CN111191099A (en) | User activity type identification method based on social media | |
| CN115017303A (en) | Method, computing device and medium for enterprise risk assessment based on news text | |
| Stock et al. | Detecting geospatial location descriptions in natural language text | |
| Upreti et al. | Online content match-making in B2B markets: Application of neural content modeling | |
| CN116737922A (en) | Tourist online comment fine granularity emotion analysis method and system | |
| KR102185733B1 (en) | Server and method for automatically generating profile | |
| CN114722810A (en) | Real estate customer portrait method and system based on information extraction and multi-attribute decision | |
| CN109947894B (en) | Text label extraction system | |
| CN109948665B (en) | Human activity type classification method and system based on long-time and short-time memory neural network | |
| CN113343712A (en) | Social text emotional tendency analysis method and system based on heterogeneous graph | |
| KR20190066156A (en) | Bigdata and machine learning based smart travel chatting robot system | |
| CN113741759B (en) | Comment information display method and device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |