CN116257618A - Multi-source intelligent travel recommendation method based on fine granularity emotion analysis - Google Patents
Multi-source intelligent travel recommendation method based on fine granularity emotion analysis Download PDFInfo
- Publication number
- CN116257618A CN116257618A CN202211100576.6A CN202211100576A CN116257618A CN 116257618 A CN116257618 A CN 116257618A CN 202211100576 A CN202211100576 A CN 202211100576A CN 116257618 A CN116257618 A CN 116257618A
- Authority
- CN
- China
- Prior art keywords
- user
- emotion
- sentence
- preference
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000008451 emotion Effects 0.000 title claims abstract description 88
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000004458 analytical method Methods 0.000 title claims abstract description 36
- 238000012549 training Methods 0.000 claims abstract description 46
- 239000013598 vector Substances 0.000 claims description 44
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 9
- 230000003993 interaction Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000013136 deep learning model Methods 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 230000014509 gene expression Effects 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 230000008485 antagonism Effects 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000011156 evaluation Methods 0.000 abstract description 4
- 230000000694 effects Effects 0.000 abstract description 2
- 238000005065 mining Methods 0.000 abstract 1
- 238000012545 processing Methods 0.000 description 4
- 238000007405 data analysis Methods 0.000 description 2
- 238000007418 data mining Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 238000010200 validation analysis 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
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
- G06F16/337—Profile generation, learning or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/14—Travel agencies
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses a multisource intelligent travel recommendation method based on fine granularity emotion analysis, which comprises the following steps of: marking category labels and emotion labels for each sentence on a sentence level for the travel field data set; establishing a fine granularity emotion analysis model of the BERT combined countermeasure training network, and deeply mining fine granularity evaluation information implicit in user comments; and constructing a user portrait and a scenic spot portrait based on the fine-grained emotion analysis model, and recommending scenic spots according to the preference of the user. According to the invention, the countermeasure training method is added, so that a key effect is played on the improvement of the precision of the model, so that the model learns countermeasure samples while training, and emotion information in a user comment text is more accurately mined; the fine-granularity emotion attributes are added into the recommendation system as additional attribute values, so that accuracy of travel recommendation can be effectively improved, and travel experience of a user is optimized.
Description
Technical Field
The invention belongs to the technical fields of text data mining, data analysis and intelligent recommendation, and particularly relates to a multisource intelligent travel recommendation method based on fine granularity emotion analysis by combining dynamic travel evaluation text data with travel tracks.
Background
More and more people share own travel experiences on social media or leave own comments on travel service websites, and the comment information provides a larger reference value for later users. Users tend to review experiences and opinions of other guests on different network platforms while planning their travels. However, there is a problem that social media comment data is rapidly growing in size, variety, diversity, etc., and massive text data is generated every day, and it is difficult for a user to acquire effective information from the massive text data for assisting travel planning. At the same time, different categories may be mentioned in the comments, such as locations, features and services of scenic spots, variability of different users, ways in which comments are written by reviewers, etc., which makes it difficult to retrieve useful information from these data. The manual processing of huge amounts of user generated data is impractical, and related scholars propose some data mining tools and algorithms, and the natural language processing technology is utilized to extract information from large-scale unstructured text data, so that the method is used for extracting attention objects and emotion attributes of users from the large data, and is applied to the fields of data analysis, travel recommendation and the like. However, the conventional recommendation method based on emotion analysis generally uses a text analysis method based on statistics, such as topic modeling, and methods such as TF-IDF and the like often summarize travel experience of users according to shallow text features such as word frequency and the like, and all implicit semantic information in comment texts is ignored.
Disclosure of Invention
In order to solve the technical problems in the prior art, the invention provides a multisource intelligent travel recommendation method based on fine granularity emotion analysis by combining dynamic travel evaluation text data with travel tracks, a deep learning method based on BERT combined countermeasure training, fine granularity evaluation information implicit in user comments is deeply mined, user portraits and scenic spot portraits are constructed based on a fine granularity emotion analysis model, and scenic spot recommendation is performed according to user preferences.
In order to achieve the technical purpose, the invention adopts the following technical scheme:
the invention comprises the following steps:
step 1: collecting text information of unstructured data related to travel in a social media website; preprocessing the collected text information, removing repeated and invalid data, and obtaining a self-constructed travel field data set; invalid data is data in which the content is empty and the content contains only expressions, symbols, or pictures. Labeling the data of the travel field data set, and marking a category label for each sentence on the sentence level; then, each sentence in the travel field dataset is labeled with an emotion label, which is classified into "very dissatisfied", "intermediate attitude", "satisfied" and "very satisfied". And marking a category label and an emotion label on part of data in the travel field data set, and dividing the data marked with the category label and the emotion label into a training set, a verification set and a test set.
Step 2: and constructing a fine granularity emotion analysis model of the BERT combined countermeasure training network, and training, verifying and testing the fine granularity emotion analysis model. The process of constructing the fine granularity emotion analysis model of the BERT combined antagonism training network is specifically as follows:
(1) the BERT-based deep learning model processes comment data of a user, and for a comment R= { R containing n sentences 1 ，R 2 ，R i ，…，R n Each sentence R i Adds special marks [ CLS ] to the head and tail of sentence]And [ SEP ]]The word embedder through the deep learning model converts into sentence vectors of m×768 dimensions, where M is the word number average of sentences in the comment data.
(2) Generating word embedding vectors with the dimension of M multiplied by 768 for each sentence vector on the basis of the sentence vectors generated in the step (1) so as to specify the position of each element in the sentence vector; and adding the generated word embedding vector and the sentence vector to obtain a combined vector.
(3) Adding an countermeasure disturbance delta x into each combination vector x by adopting an countermeasure training method of a rapid gradient method, and defending against attack to generate an countermeasure sample x+delta x; updating the training set with (x+Δx, y), wherein y represents a class label and an emotion label vector of a sentence corresponding to x; then, the parameter θ of the countermeasure training network E is updated using a gradient descent method.
(4) The class to which the combined vector belongs and the emotion score are output using a Softmax function.
The specific formula of the rapid gradient method is as follows:
wherein D represents a training set, θ represents an E parameter of the countermeasure training network, L (x, y; θ) is a loss function of the combination vector x, Δx is an opposing disturbance satisfying the condition Δx ε; epsilon is a constant for constraint, taking a value between 0.5 and 1; the value of Deltax is Deltaj. L2 norm of x.
Step 3: the method comprises the steps that unlabeled user comment data in the travel field data set are used as input of a fine-granularity emotion analysis model, and the fine-granularity emotion analysis model outputs the category and emotion score of each sentence in the user comment; counting the category and emotion score of each sentence in the user comment to obtain an emotion score matrix R s ＝{r 1 ，r 2 ，r i ，…，r n }, wherein vector r i ＝{c i ,s i }，c i Class s for the ith sentence i Is the emotion score of the ith sentence.
Step 4: taking a weighted average value of the emotion score matrixes of the users obtained in the step 3 as scenic spot images aiming at each scenic spot; taking a weighted average value of the emotion score matrix of each user for each scenic spot, which is obtained in the step 3, as a user image, wherein the emotion score matrix weight of the comment is higher as the number of sentences contained in the comment is larger. Recommending scenic spots conforming to the preference for the user by using a collaborative filtering algorithm based on scenic spot portraits and user portraits for the user with access records; and for the user without access records, constructing a preference matrix according to each preference category and preference value selected by the user in the user interaction interface, calculating the cosine similarity of the scenery spot portrait of each scenery spot and the user preference matrix, and recommending the scenery spot with the maximum calculated cosine similarity to the user.
Preferably, Δx is expressed as:
in the method, in the process of the invention,
representing the gradient of the loss function L (x, y; θ) with respect to the combined vector x, the sign function sign being used for the gradient
Performing standardization;
Δx must satisfy constraint conditions Δx ε, and finally, by applying the constraint conditions Δx ε to
Normalization and multiplication with a constant epsilon yields deltax:
preferably, preference selection is performed in a "preference input" panel of the user interaction interface, and the user selects any number of preferences according to his own preferences and selects preference values. After selecting the preference, clicking a 'confirm' button, namely calculating the cosine similarity between the user preference matrix and the scenery spot portrait of each scenery spot according to the preference category and the preference value input by the user, and presenting the scenery spot with the maximum calculated cosine similarity on a user interaction interface for recommendation to the user.
The invention has the beneficial effects that:
the invention applies the improved BERT deep learning network, firstly, the comments are classified into different categories, secondly, the comments are subjected to emotion analysis, and finally, the comments are recommended to suitable tourist attractions based on the preference input by the user. For emotion analysis tasks, the robustness and generalization ability of the model under small sample training are improved by using an countermeasure training method. The addition of the countermeasure training method plays a key role in improving the precision of the model, so that the model learns countermeasure samples while training, and emotion information in user comment texts is further accurately mined. The convergence rate of the model is slightly slower after the countermeasure training method is added, but the accuracy is obviously higher than that of the original model. The fine-granularity emotion attributes are added into the recommendation system as additional attribute values, so that accuracy of travel recommendation can be effectively improved, and travel experience of a user is optimized.
Drawings
FIG. 1 is a flow chart of the present invention.
FIG. 2 is a flow chart of a fine granularity emotion analysis model for building a BERT binding challenge training network in accordance with the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
As shown in fig. 1, a multi-source intelligent travel recommendation method based on fine granularity emotion analysis comprises the following steps:
step 1: collecting text information of unstructured data (data inconvenient to be represented by a two-dimensional logic table of a database) related to travel in a social media website, wherein the unstructured data related to the travel are basic information of scenic spots, basic information of users and comment data of the users; the basic information of the scenic spot comprises a scenic spot name, a scenic spot ID and a geographic position, and the basic information of the user comprises a user ID, a scenic spot set accessed by the user at present, time accessed by the user at present and a history of access by the user; preprocessing the collected text information, removing repeated and invalid data, and obtaining a self-constructed travel field data set; invalid data is data in which the content is empty and the content contains only expressions, symbols, or pictures. Labeling the data of the travel field data set, labeling each sentence on the sentence level, classifying the content of the sentence description into 34 categories, wherein the 34 categories are listed in table 1, and labeling the sentences belonging to a plurality of categories simultaneously to improve the classification accuracy. Then, each sentence in the travel field dataset is labeled with an emotion label, which is classified into "very dissatisfied", "intermediate attitude", "satisfied" and "very satisfied". The five-level emotion tag has the advantage of being able to more accurately mine the emotion of a user than a five-level emotion tag that only contains "positive" and "negative" two-level emotion attributes. Five-level emotion tags are respectively represented by 1 to 5, and the larger the score is, the more positive the emotion attribute of the sentence is, and the more negative the emotion attribute is represented by the opposite. The method comprises the steps of marking 10% of data in the tourist area data set with category labels and emotion labels, dividing the data marked with the category labels and emotion labels into a training set, a verification set and a test set, and using the training set, the verification set and the test set for training, verifying and testing a fine-grained emotion analysis model, wherein the rest of data without the category labels and emotion labels in the tourist area data set are used for generating scenic spot portraits and user portraits; the ratio of training set, validation set to test set may be set to 7:2:1.
TABLE 1
Step 2: building a fine granularity emotion analysis model of a BERT combined countermeasure training (AT, adversarial training) network, training, verifying and testing the fine granularity emotion analysis model, wherein the batch size is set to be 32 during training; wherein the countermeasure training is used to improve the model accuracy under small samples.
The process of constructing the fine-grained emotion analysis model of the BERT-combined countermeasure training (AT, adversarial training) network is specifically as follows:
(1) processing comment data of a user based on a deep learning model of BERT (Bidirectional Encoder Representation from Transformers), wherein for a comment R= { R containing n sentences 1 ，R 2 ，R i ，…，R n Each sentence R i Adds special marks [ CLS ] to the head and tail of sentence]And [ SEP ]]Word embedder through deep learning model converts into M×768-dimensional sentence vector for processing comment data into vector information convenient for computer processing, wherein M is word number average value of sentences in comment data。
(2) Generating word embedding vectors with the dimension of M multiplied by 768 for each sentence vector on the basis of the sentence vectors generated in the step (1) so as to specify the position of each element in the sentence vector; and adding the generated word embedding vector and the sentence vector to obtain a combined vector.
(3) Adding a countermeasure disturbance delta x into each combination vector x by adopting a countermeasure training method of a fast gradient method (FGM, fast Gradient Method), and performing defense countermeasure attack to generate a countermeasure sample x+delta x; updating the training set with (x+Δx, y), wherein y represents a class label and an emotion label vector of a sentence corresponding to x; then, the parameter θ of the countermeasure training network E is updated using a gradient descent method.
(4) The class to which the combined vector belongs and the emotion score are output using a Softmax function.
The specific formula of the rapid gradient method is as follows:
wherein D represents a training set, θ represents an E parameter of the countermeasure training network, L (x, y; θ) is a loss function of the combination vector x, Δx is an opposing disturbance satisfying the condition Δx ε; epsilon is a constant for constraint, taking a value between 0.5 and 1; the L2 norm of Δx is the L Δx; max (max) ||Δx||≤ε L (x+Deltax, y; θ) acts to obtain a disturbance Deltax that maximizes the loss and satisfies the condition, thereby optimizing the parameters θ of the countermeasure training network E under the disturbance of the disturbance Deltax so that the loss of the countermeasure training network E is minimized.
Step 3: the method comprises the steps that unlabeled user comment data in the travel field data set are used as input of a fine-granularity emotion analysis model, and the fine-granularity emotion analysis model outputs the category and emotion score of each sentence in the user comment; counting the category and emotion score of each sentence in the user comment to obtain an emotion score matrix R s ＝{r 1 ，r 2 ，r i ，…，r n }, wherein vector r i ＝{c i ,s i }，c i Class s for the ith sentence i For the ith sentenceEmotion score of son is 0.ltoreq.c i ≤33,1≤s i And is less than or equal to 5. In order to verify the accuracy of the calculated emotion score, comment data with scoring data are input into a fine granularity emotion analysis model, the average value of emotion scores of all sentences output by the comment data is compared with the score of the comment data under the condition that the category is not considered, and the result shows that the emotion score calculated by the fine granularity emotion analysis model is very close to the score given by a user, so that the fine granularity emotion analysis model is reasonable and practical for the emotion score calculation method.
Step 4: taking a weighted average value of the emotion score matrixes of the users obtained in the step 3 as scenic spot images aiming at each scenic spot; taking a weighted average value of the emotion score matrix of each user for each scenic spot, which is obtained in the step 3, as a user image, wherein the emotion score matrix weight of the comment is higher as the number of sentences contained in the comment is larger. Recommending scenic spots conforming to the preference of the user by using a collaborative filtering algorithm based on scenic spot portraits and user portraits for the user with access records; and for the user without access records, constructing a preference matrix according to each preference category and preference value selected by the user in the user interaction interface, calculating the cosine similarity of the scenery spot portrait of each scenery spot and the user preference matrix, and recommending the scenery spot with the maximum calculated cosine similarity to the user.
As a preferred embodiment, Δx is expressed as:
in the method, in the process of the invention,
representing the gradient of the loss function L (x, y; θ) with respect to the combined vector x, the sign function sign being used for the gradient
Performing standardization; Δx must satisfy the constraint Δx ε to achieve nearly the same effect as the salient features of the original combined vector.
Final Deltax pass pair
Normalized and multiplied by a constant epsilon, namely:
as a preferred embodiment, preference selection is made within a "preference input" panel of the user interactive interface, and the user can select any number of preferences according to his own preferences and select preference values. After selecting the preference, clicking a 'confirm' button, and calculating the cosine similarity between the user preference matrix and the scenery spot portrait of each scenery spot according to the preference category and the preference value input by the user, and presenting the scenery spot with the maximum calculated cosine similarity on a user interaction interface for recommendation to the user.
Claims (3)
1. A multisource intelligent travel recommendation method based on fine granularity emotion analysis is characterized in that: the method comprises the following steps:
step 1: collecting text information of unstructured data related to travel in a social media website; preprocessing the collected text information, removing repeated and invalid data, and obtaining a self-constructed travel field data set; the invalid data is data in which the content is empty and the content only contains expressions, symbols or pictures; labeling the data of the travel field data set, and marking a category label for each sentence on the sentence level; then, labeling each sentence in the travel field data set with emotion tags, wherein the emotion tags are classified into 'dissatisfaction', 'intermediate attitude', 'satisfaction' and 'satisfaction'; the method comprises the steps of marking part of data in the travel field data set with category labels and emotion labels, and dividing the data marked with the category labels and the emotion labels into a training set, a verification set and a test set;
step 2: building a fine granularity emotion analysis model of the BERT combined countermeasure training network, and training, verifying and testing the fine granularity emotion analysis model; the process of constructing the fine granularity emotion analysis model of the BERT combined antagonism training network is specifically as follows:
(1) the BERT-based deep learning model processes comment data of a user, and for a comment R= { R containing n sentences 1 ，R 2 ，R i ，…，R n Each sentence R i Adds special marks [ CLS ] to the head and tail of sentence]And [ SEP ]]The word embedder through the deep learning model converts the word embedder into M×768-dimensional sentence vectors, wherein M is the word number average value of sentences in comment data;
(2) generating word embedding vectors with the dimension of M multiplied by 768 for each sentence vector on the basis of the sentence vectors generated in the step (1) so as to specify the position of each element in the sentence vector; adding the generated word embedding vector and the sentence vector to obtain a combined vector;
(3) adding an countermeasure disturbance delta x into each combination vector x by adopting an countermeasure training method of a rapid gradient method, and defending against attack to generate an countermeasure sample x+delta x; updating the training set with (x+Δx, y), wherein y represents a class label and an emotion label vector of a sentence corresponding to x; then, updating the parameter θ of the countermeasure training network E using a gradient descent method;
(4) outputting the category and emotion score to which the combination vector belongs by using a Softmax function;
the specific formula of the rapid gradient method is as follows:
wherein D represents a training set, θ represents an E parameter of the countermeasure training network, L (x, y; θ) is a loss function of the combination vector x, Δx is an opposing disturbance satisfying the condition Δx ε; epsilon is a constant for constraint, taking a value between 0.5 and 1; the L2 norm of Δx is the L Δx;
step 3: the method comprises the steps that unlabeled user comment data in the travel field data set are used as input of a fine-granularity emotion analysis model, and the fine-granularity emotion analysis model outputs the category and emotion score of each sentence in the user comment; counting category and each sentence in user commentsEmotion score, obtain emotion score matrix R s ＝{r 1 ，r 2 ，r i ，…，r n }, wherein vector r i ＝{c i ,s i }，c i Class s for the ith sentence i An emotion score for the ith sentence;
step 4: taking a weighted average value of the emotion score matrixes of the users obtained in the step 3 as scenic spot images aiming at each scenic spot; taking a weighted average value of the emotion score matrix of each user for each scenic spot, which is obtained in the step 3, as a user image, wherein the emotion score matrix weight of the comment is higher as the number of sentences contained in the comment is larger; recommending scenic spots conforming to the preference for the user by using a collaborative filtering algorithm based on scenic spot portraits and user portraits for the user with access records; and for the user without access records, constructing a preference matrix according to each preference category and preference value selected by the user in the user interaction interface, calculating the cosine similarity of the scenery spot portrait of each scenery spot and the user preference matrix, and recommending the scenery spot with the maximum calculated cosine similarity to the user.
2. The multi-source intelligent travel recommendation method based on fine granularity emotion analysis of claim 1, wherein the method comprises the following steps of: Δx is expressed as:
in the method, in the process of the invention,
representing the gradient of the loss function L (x, y; θ) with respect to the combined vector x, the sign function sign being used for the gradient +.>
Performing standardization;
Δx must satisfy constraint conditions Δx ε, and finally, by applying the constraint conditions Δx ε to
Normalization and multiplication with a constant epsilon yields deltax:
3. the multi-source intelligent travel recommendation method based on fine-grained emotion analysis according to claim 1 or 2, wherein the method comprises the following steps of: selecting preferences in a preference input panel of the user interaction interface, and selecting any number of preferences and preference values by a user according to own preferences; after selecting the preference, clicking a 'confirm' button, namely calculating the cosine similarity between the user preference matrix and the scenery spot portrait of each scenery spot according to the preference category and the preference value input by the user, and presenting the scenery spot with the maximum calculated cosine similarity on a user interaction interface for recommendation to the user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211100576.6A CN116257618A (en) | 2022-09-09 | 2022-09-09 | Multi-source intelligent travel recommendation method based on fine granularity emotion analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211100576.6A CN116257618A (en) | 2022-09-09 | 2022-09-09 | Multi-source intelligent travel recommendation method based on fine granularity emotion analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116257618A true CN116257618A (en) | 2023-06-13 |
Family
ID=86683230
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211100576.6A Pending CN116257618A (en) | 2022-09-09 | 2022-09-09 | Multi-source intelligent travel recommendation method based on fine granularity emotion analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116257618A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117077901A (en) * | 2023-10-17 | 2023-11-17 | 北京铭洋商务服务有限公司 | Travel data processing method and system |
-
2022
- 2022-09-09 CN CN202211100576.6A patent/CN116257618A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117077901A (en) * | 2023-10-17 | 2023-11-17 | 北京铭洋商务服务有限公司 | Travel data processing method and system |
| CN117077901B (en) * | 2023-10-17 | 2024-01-05 | 北京铭洋商务服务有限公司 | Travel data processing method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111753060B (en) | Information retrieval method, apparatus, device and computer readable storage medium | |
| CN108021616B (en) | Community question-answer expert recommendation method based on recurrent neural network | |
| CN111966917B (en) | Event detection and summarization method based on pre-training language model | |
| CN105279495B (en) | A kind of video presentation method summarized based on deep learning and text | |
| CN110765260A (en) | Information recommendation method based on convolutional neural network and joint attention mechanism | |
| CN111797898B (en) | Online comment automatic reply method based on deep semantic matching | |
| CN102955848A (en) | Semantic-based three-dimensional model retrieval system and method | |
| CN112989802B (en) | Bullet screen keyword extraction method, bullet screen keyword extraction device, bullet screen keyword extraction equipment and bullet screen keyword extraction medium | |
| CN110825850B (en) | Natural language theme classification method and device | |
| CN110750648A (en) | Text emotion classification method based on deep learning and feature fusion | |
| CN113610680A (en) | AI-based interactive reading material personalized recommendation method and system | |
| CN111737427A (en) | Mu lesson forum post recommendation method integrating forum interaction behavior and user reading preference | |
| CN113486645A (en) | Text similarity detection method based on deep learning | |
| CN115659947A (en) | Multi-item selection answering method and system based on machine reading understanding and text summarization | |
| CN113934835B (en) | Retrieval type reply dialogue method and system combining keywords and semantic understanding representation | |
| CN111666374A (en) | Method for integrating additional knowledge information into deep language model | |
| CN116257618A (en) | Multi-source intelligent travel recommendation method based on fine granularity emotion analysis | |
| CN112966518B (en) | High-quality answer identification method for large-scale online learning platform | |
| CN115481313A (en) | News recommendation method based on text semantic mining | |
| CN111274354B (en) | Referee document structuring method and referee document structuring device | |
| CN112950414A (en) | Legal text representation method based on decoupling legal elements | |
| CN115964560B (en) | Information recommendation method and equipment based on multi-mode pre-training model | |
| Nguyen et al. | Analyzing customer experience in hotel services using topic modeling | |
| Hsu et al. | Xiao-Shih: the educational intelligent question answering bot on Chinese-based MOOCs | |
| CN115359486A (en) | Method and system for determining custom information in document image |
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 |