Multi-party sentiment analysis system and method based on pre-training model

Through a multi-party sentiment analysis system based on the pre-trained model, the problem of low decision-making efficiency caused by the large number of hotel room reviews is solved, and the accurate extraction and analysis of emotional tendencies is achieved, intelligent evaluation conclusions are generated, and consumer decision-making efficiency is improved.

CN119067082APending Publication Date: 2024-12-03SUZHOU LANGJIETONG INTELLIGENT TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411195273.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

In the hotel room reservation system, there are too many comments in a single hotel room, which makes it difficult for customers to quickly analyze the proportion of positive and negative emotions, affecting decision-making efficiency.

Method used

Adopt a multi-party sentiment analysis system based on pre-trained models, including model creation module, text analysis module, classification storage module and comment analysis module. Through emotion recognition model training, emotion label generation, comment classification and intelligent evaluation conclusion generation, comments are extracted and analyzed. emotional tendencies in the generation of room recommendation levels, facility evaluation and emotional trends.

Benefits of technology

It realizes accurate extraction and analysis of emotional tendencies, improves consumers' decision-making efficiency in matching their own needs based on comments, and simplifies the analysis process of multiple comments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067082A_ABST
    Figure CN119067082A_ABST
Patent Text Reader

Abstract

The invention discloses a multi-party sentiment analysis system based on a pre-training model, and the system comprises a model building module which is used for training a sentiment recognition model according to a sentiment set database, and obtaining the pre-training model; the text analysis module is used for generating an emotion label and identifying the comments according to an emotion analysis result obtained after the comments are input into the pre-training model; the classified storage module is used for storing the comments in a classified manner according to the emotion labels of the comments; and the comment analysis module is used for generating an intelligent evaluation conclusion according to the classified and stored comments. The invention further discloses a multi-party sentiment analysis method based on the multi-party sentiment analysis system. Compared with the prior art, the problem that the decision-making efficiency is affected due to the fact that a large amount of existing comment information of guest rooms lacks sentiment analysis for reference is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of sentiment analysis systems, and in particular to a multi-party sentiment analysis system and method based on a pre-training model. Background Art

[0002] Hotels usually use a three-tiered room reservation system based on "browser-web server-backend database" to realize hotel room reservation. The specific reservation process is:

[0003] 1) Log in, check rooms, book rooms, check out, etc. in the browser;

[0004] 2) Transmit customer information and room information to the database for background management;

[0005] 3) The database information is then transmitted to the display page via the Web server to facilitate customer selection, thereby achieving the purpose of customer information entry, room reservation, and real-time feedback of room information, thereby facilitating customer check-in.

[0006] In order to help customers choose a room that meets their needs before booking, the display page will not only display fixed room type information, but also previous customers' evaluation information for reference.

[0007] Due to the excessive number of reviews for a single hotel room, it is difficult for customers to quickly analyze the positive and negative sentiments of all room reviews, as well as the ratio of positive to negative sentiments, when referring to the reviews. This makes it difficult to match their own needs based on previous reviews, affecting the decision-making efficiency of customers when making reservations. Summary of the invention

[0008] The purpose of the present invention is to provide a multi-party sentiment analysis system based on a pre-trained model to solve the problem that a large amount of existing guest room review information lacks sentiment analysis for reference, which affects decision-making efficiency.

[0009] In order to achieve the above objectives, on the one hand, the present application provides a multi-party sentiment analysis system based on a pre-trained model, comprising:

[0010] A model creation module is used to train the emotion recognition model according to the emotion set database to obtain a pre-trained model;

[0011] A text analysis module, which is used to generate sentiment tags and identify comments based on the sentiment analysis results obtained after the comments are input into the pre-trained model;

[0012] A classification storage module, which is used to classify and store comments according to their sentiment tags;

[0013] The comment analysis module is used to generate intelligent evaluation conclusions based on the classified stored comments.

[0014] As a further description of the above technical solution:

[0015] The classification storage module includes a label recognition unit, a keyword extraction unit and a storage unit. The label recognition unit is used to recognize emotional labels, the keyword extraction unit is used to extract the set hotel room interior facilities vocabulary included in the comment text, and the storage unit is used to store the classified comments.

[0016] As a further description of the above technical solution:

[0017] The review analysis module includes a recommendation level analysis unit, a facility level analysis unit and a sentiment trend analysis unit. The recommendation level analysis unit is used to generate a room recommendation level according to the proportion of positive comments, negative comments and comments with no sentiment tendency in all the comments on a single room. The facility level analysis unit is used to generate a room facility evaluation according to the proportion of positive comments, negative comments and comments with no sentiment tendency in the corresponding comments on a single facility in the hotel room. The sentiment trend analysis unit is used to generate a room sentiment trend according to the changing trend of positive comments, negative comments and comments with no sentiment tendency in the past period of time.

[0018] On the other hand, the present invention also provides a multi-party sentiment analysis method based on the multi-party sentiment analysis system, comprising the following steps:

[0019] S1. Input all reviews of a single hotel room into the pre-trained model, identify the reviews with sentiment tags based on the sentiment analysis results, and classify the reviews into positive reviews, negative reviews, and reviews with no sentiment tendency;

[0020] S2, the classification storage module uses sentiment tags to classify and store all the comments about the guest rooms;

[0021] S3, the review analysis module generates a room recommendation grade based on the proportion of positive reviews, negative reviews, and comments without sentiment tendency in all reviews of a single room;

[0022] S4, the review analysis module calculates the proportion of positive reviews, negative reviews, and reviews without sentiment tendency in all reviews corresponding to a single facility in the hotel guest room, and generates a guest room facility evaluation of a single facility;

[0023] S5. According to the time of comment release, the comment analysis module counts the proportion of positive comments, negative comments, and comments without sentiment tendency in the guest room comments counted in a natural month, analyzes the trend of the proportion of positive comments, negative comments, and comments without sentiment tendency in the guest room comments in the past year, and generates the guest room sentiment trend;

[0024] S6. Generate intelligent evaluation conclusions based on the room recommendation level, room facility evaluation, and room sentiment trend obtained from the review analysis.

[0025] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0026] 1. In the present invention, based on the pre-trained emotion recognition model, the emotional tendency in the comments is accurately extracted. The comment analysis module generates a room recommendation grade based on the classified stored comments, based on the proportion of positive comments, negative comments and comments without emotional tendency in all comments on a single guest room. The room facility evaluation of a single facility is generated by the proportion of positive comments, negative comments and comments without emotional tendency in all comments corresponding to a single facility in the hotel room. The proportion of positive comments, negative comments and comments without emotional tendency in the guest room reviews counted on a natural month basis is counted, and the changing trend of the proportion of positive comments, negative comments and comments without emotional tendency in the guest room reviews in the past year is analyzed to generate the guest room emotional trend. The analysis of the guest room based on the emotional tendency in all past comments is realized, and an intelligent evaluation conclusion is generated, which is convenient for consumers to analyze the matching degree between their own needs and the guest room, thereby improving the decision-making efficiency of consumers. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0028] Figure 1 The system architecture diagram of a multi-party sentiment analysis system based on a pre-trained model.

[0029] Figure 2 Flowchart of a multi-party sentiment analysis system based on a pre-trained model. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0031] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] Example 1

[0033] See also Figure 1-2 On the one hand, the present application provides a multi-party sentiment analysis system based on a pre-trained model, comprising:

[0034] A model creation module is used to train the emotion recognition model according to the emotion set database to obtain a pre-trained model;

[0035] A text analysis module, which is used to generate sentiment tags and identify comments based on the sentiment analysis results obtained after the comments are input into the pre-trained model;

[0036] A classification storage module, which is used to classify and store comments according to their sentiment tags;

[0037] The comment analysis module is used to generate intelligent evaluation conclusions based on the classified stored comments.

[0038] In order to pre-train the model of the emotion recognition model, the model creation module uses text segmentation to generate an emotion set database, selects more than 3,000 comment emotion samples and more than 1,500 emotion reading samples, and selects more than 1,000 emotion reading model samples and more than 1,500 pre-training model samples. The emotion intervention data set is used for simulation testing to verify the error rate and data information reading accuracy of the emotion recognition model, and to test whether the final emotion recognition model data set reasonably reads the emotion comments in the database during the reader reading process. Verify that each comment emotion analysis data set is consistent with the comment emotion analysis to form a multi-party data set.

[0039] The classification storage module includes a label recognition unit, a keyword extraction unit and a storage unit. The label recognition unit is used to recognize the emotional label, the keyword extraction unit is used to extract the set hotel room interior facilities vocabulary included in the comment text, and the storage unit is used to store the classified comments. The storage unit classifies and stores all comments according to the emotional label recognized by the label recognition unit and the room interior facilities text extracted by the keyword extraction unit. For example, the keyword extraction unit extracts all comments including a certain facility (such as air conditioning) inside the hotel room, and then stores all comments involving the hotel room interior facilities, which is convenient for analyzing the quality of the hotel room interior facilities.

[0040] The classification storage module also includes a time extraction unit, which is used to extract the release time of the comment information so as to classify and store the comments according to the release time of the comment information.

[0041] The review analysis module includes a recommendation level analysis unit, a facility level analysis unit and a sentiment trend analysis unit. The recommendation level analysis unit is used to generate a room recommendation level according to the proportion of positive comments, negative comments and comments with no sentiment tendency in all the comments on a single room. The facility level analysis unit is used to generate a room facility evaluation according to the proportion of positive comments, negative comments and comments with no sentiment tendency in the corresponding comments on a single facility in the hotel room. The sentiment trend analysis unit is used to generate a room sentiment trend according to the changing trend of positive comments, negative comments and comments with no sentiment tendency in the past period of time.

[0042] The recommendation level analysis unit is used to classify guest rooms, so that they can be sorted when displaying guest room information, which is convenient for users to make decisions. For example, if the proportion of positive comments in all reviews of a single guest room reaches 90% and the proportion of negative comments is less than 5%, it is the highest recommendation level, and if the proportion of positive comments in all reviews of a single guest room reaches 80% and the proportion of negative comments is less than 10%, it is the second recommendation level, and so on.

[0043] On the other hand, the present invention also provides a multi-party sentiment analysis method based on the multi-party sentiment analysis system, comprising the following steps:

[0044] S1. Input all reviews of a single hotel room into the pre-trained model, identify the reviews with sentiment tags based on the sentiment analysis results, and classify the reviews into positive reviews, negative reviews, and reviews with no sentiment tendency;

[0045] S2, the classification storage module uses sentiment tags to classify and store all the comments about the guest rooms;

[0046] S3, the review analysis module generates a room recommendation grade based on the proportion of positive reviews, negative reviews, and comments without sentiment tendency in all reviews of a single room;

[0047] S4, the review analysis module calculates the proportion of positive reviews, negative reviews, and reviews without sentiment tendency in all reviews corresponding to a single facility in the hotel guest room, and generates a guest room facility evaluation of a single facility;

[0048] S5. According to the time of comment release, the comment analysis module counts the proportion of positive comments, negative comments, and comments without sentiment tendency in the guest room comments counted in a natural month, analyzes the trend of the proportion of positive comments, negative comments, and comments without sentiment tendency in the guest room comments in the past year, and generates the guest room sentiment trend;

[0049] S6. Generate intelligent evaluation conclusions based on the room recommendation level, room facility evaluation, and room sentiment trend obtained from the review analysis.

[0050] Working principle: Based on the pre-trained sentiment recognition model, the sentiment tendency in the comments can be accurately extracted. The comment analysis module generates a room recommendation grade based on the classified stored comments, and calculates the proportion of positive comments, negative comments, and comments without sentiment tendency in all comments on a single room. It also generates a room facility evaluation for a single facility based on the proportion of positive comments, negative comments, and comments without sentiment tendency in all comments corresponding to a single facility in the hotel room. It also calculates the proportion of positive comments, negative comments, and comments without sentiment tendency in the room reviews counted on a natural month basis, analyzes the changing trend of the proportion of positive comments, negative comments, and comments without sentiment tendency in the room reviews in the past year, generates the room sentiment trend, realizes the analysis of the room based on the sentiment tendency in all past comments, and generates intelligent evaluation conclusions, which is convenient for consumers to analyze the degree of match between their own needs and the room, and improves the decision-making efficiency of consumers.

[0051] Example 2

[0052] The rest is the same as Example 1, except that: the review analysis module includes an improvement analysis unit, which is used to analyze the interior facilities of the guest room that need to be replaced based on the proportion of negative reviews of the interior facilities of a single guest room in all negative reviews of the guest room and the change trend.

[0053] If the proportion of negative comments about a single room's internal facilities in all negative comments about the room exceeds 30%, it means that the room's internal facilities are an important cause of negative comments about the room, and need to be focused on and adjusted. On the other hand, the proportion of negative comments about a single room's internal facilities in all negative comments about the room is counted in a natural month as a single cycle. By comparing the changes in the proportion of negative comments about a single room's internal facilities in the past year, the impact of the facility on the generation of negative comments can be judged. If the proportion gradually increases, it means that the room's internal facilities need to be replaced.

[0054] Example 3

[0055] The rest is the same as Example 1, except that: the review analysis module also includes an advantage recommendation unit, which is used to generate a room recommendation tag when the number of positive reviews of a single facility inside a hotel room and the proportion of positive reviews in the corresponding reviews of a single facility inside a hotel room both reach a threshold.

[0056] Different consumers attach different importance to the internal facilities of different guest rooms. In order to attract consumers and help them quickly find a room with high satisfaction of internal facilities, when there are many positive comments on a single facility in all guest room reviews and the proportion of positive comments on all reviews about the facility is high, it means that the facility is the advantage of the guest room. It is marked with a room recommendation tag. One room recommendation tag corresponds to one facility advantage. The specific number of room recommendation tags can be flexibly set according to demand.

[0057] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A multi-party sentiment analysis system based on a pre-trained model, characterized in that: include: A model creation module is used to train the emotion recognition model according to the emotion set database to obtain a pre-trained model; A text analysis module, which is used to generate sentiment tags and identify comments based on the sentiment analysis results obtained after the comments are input into the pre-trained model; A classification storage module, which is used to classify and store comments according to their sentiment tags; The comment analysis module is used to generate intelligent evaluation conclusions based on the classified stored comments.

2. A multi-party sentiment analysis system based on a pre-training model according to claim 1, characterized in that: The classification storage module includes a label recognition unit, a keyword extraction unit and a storage unit. The label recognition unit is used to recognize emotional labels, the keyword extraction unit is used to extract the set hotel room interior facilities vocabulary included in the comment text, and the storage unit is used to store the classified comments.

3. A multi-party sentiment analysis system based on a pre-training model according to claim 2, characterized in that: The classification storage module also includes a time extraction unit, and the time extraction unit is used to extract the release time of the comment information.

4. The multi-party sentiment analysis system based on a pre-training model according to claim 1, characterized in that: The review analysis module includes a recommendation level analysis unit, a facility level analysis unit and a sentiment trend analysis unit. The recommendation level analysis unit is used to generate a guest room recommendation level according to the proportion of positive comments, negative comments and comments with no sentiment tendency in all the comments on a single guest room. The facility level analysis unit is used to generate a guest room facility evaluation according to the proportion of positive comments, negative comments and comments with no sentiment tendency in the corresponding comments on a single facility in the hotel guest room. The sentiment trend analysis unit is used to generate a guest room sentiment trend according to the changing trend of positive comments, negative comments and comments with no sentiment tendency in the guest room over a period of time in the past.

5. A multi-party sentiment analysis system based on a pre-training model according to claim 4, characterized in that: The review analysis module includes an improvement analysis unit, which is used to analyze the proportion of negative reviews of internal facilities of a single guest room in all negative reviews of the guest room and the change trend.

6. A multi-party sentiment analysis system based on a pre-training model according to claim 4, characterized in that: The review analysis module also includes an advantage recommendation unit, which is used to generate a room recommendation tag when the number of positive reviews of a single facility inside a hotel room and the proportion of positive reviews in the corresponding reviews of the single facility inside the hotel room both reach a threshold.

7. A multi-party sentiment analysis method based on the multi-party sentiment analysis system according to claims 1-6, characterized in that: The following steps are involved: S1. Input all reviews of a single hotel room into the pre-trained model, identify the reviews with sentiment tags based on the sentiment analysis results, and classify the reviews into positive reviews, negative reviews, and reviews with no sentiment tendency; S2, the classification storage module uses sentiment tags to classify and store all the comments about the guest rooms; S3, the review analysis module generates a room recommendation grade based on the proportion of positive reviews, negative reviews, and comments without sentiment tendency in all reviews of a single room; S4, the review analysis module calculates the proportion of positive reviews, negative reviews, and reviews without sentiment tendency in all reviews corresponding to a single facility in the hotel guest room, and generates a guest room facility evaluation of a single facility; S5. According to the time of comment release, the comment analysis module counts the proportion of positive comments, negative comments, and comments without sentiment tendency in the guest room comments counted in a natural month, analyzes the trend of the proportion of positive comments, negative comments, and comments without sentiment tendency in the guest room comments in the past year, and generates the guest room sentiment trend; S6. Generate intelligent evaluation conclusions based on the room recommendation level, room facility evaluation, and room sentiment trend obtained from the review analysis.