CN110825423B

CN110825423B - APP continuous improvement method based on user online comment emotion and preference analysis

Info

Publication number: CN110825423B
Application number: CN201911049834.0A
Authority: CN
Inventors: 陈世展; 肖建茂; 冯志勇; 薛霄
Original assignee: Tianjin University
Current assignee: Tianjin University
Priority date: 2019-10-31
Filing date: 2019-10-31
Publication date: 2023-09-29
Anticipated expiration: 2039-10-31
Also published as: CN110825423A

Abstract

The invention relates to an APP continuous improvement method based on user online comment emotion and preference analysis, which is characterized by comprising the following steps: the improvement method comprises the following steps: 1) Comment data preprocessing; 2) Comment emotion acquisition; 3) Obtaining preference characteristics; 4) Preference feature scores; 5) Analyzing emotion and preference characteristic evolution of the time sequence; 6) Emotion-preference feature association mapping analysis; 7) APP evolution and maintenance recommendation. The invention has scientific and reasonable design, can timely and accurately detect the emotion and preference characteristics of the user from the user comments, efficiently help the developer to analyze and decide the follow-up related matters about the evolution maintenance and follow-up version update of the APP, effectively improve the development efficiency and optimize the experience of the user using the APP.

Description

APP continuous improvement method based on user online comment emotion and preference analysis

Technical Field

The invention belongs to the field of software engineering, relates to software evolution and maintenance, and particularly relates to an APP continuous improvement method based on user online comment emotion and preference analysis.

Background

Along with the popularization of mobile intelligent terminals and the rapid development of mobile internet, application programs (hereinafter referred to as APP) become an indispensable part of life of users, and good user experience is a key to continuously maintaining competitiveness of APP. After the APP is used, the user can perform feedback on the APP, and because the APP developer and the user do not know each other, the feedback suggestion of the user group has great significance for the developer to perform software maintenance and update. Therefore, acquiring valuable APP usage feedback, enabling a user group to participate in the design and maintenance of software, is an important guarantee that software developers gain better benefits in the competitive software market.

Mobile application distributed platforms, such as google market, apple store, etc., allow users to submit feedback (such as scoring or text commentary) to downloaded APPs by scoring or text commentary, which feedback information explicitly or implicitly expresses the user's potential emotions and preferences for APPs, such as: express their satisfaction with a certain function, encounter holes or request new features of the function, etc. The emotion tendency reflects the identity of the user to the APP, and the preference expresses the intention of the user to the APP. The preference of the user can be reflected through the user comment characteristics, the emotion tendency and preference characteristics of the user on the APP can be timely and accurately mastered, and a developer can be helped to know the trend of how the emotion and preference characteristics of the user develop along with time in real time, and how the trends are related to the evolution and maintenance of the similar application programs in the future.

The online user comments of the APP software become important resources acquired by the software through feedback by virtue of the advantages of wide coverage users, rich content, strong timeliness and the like. The comments of the user directly reflect the visual feeling of the user using the APP, and in recent years, the comment mining of the APP is mainly focused on the feature mining and classification of the comment of the user, the analysis of the features of the user interested in the APP, the filtering of useless comment information and the like.

In fact, in the process of implementing the present invention, the inventor finds that the existing APP user comment mining technology has at least the following disadvantages:

1) The existing APP comment mining is only single comment feature considered or comment emotion analysis, and does not consider comment emotion and fine-granularity user preference features to conduct correlation analysis;

2) The existing APP comment mining only analyzes in a single dimension of comment content, and does not consider the change of emotion and preference of users in a time dimension. Factors that produce these changes may be due to an upgrade stack of software internal environment (e.g., new bug occurrences) development techniques or APP external market environment (e.g., competitors for the same APP occurrences) changes.

Aiming at the problems, the patent provides a method for analyzing user emotion from user comments in time based on time dimension, mining fine-grained user preference characteristics and mapping the user emotion and the preference characteristics in real time, which can timely and accurately recommend the user emotion and the preference characteristics for developers and how the preference characteristics are related to evolution maintenance of APP.

By searching for the published patent documents, a published patent document similar to the present patent application is not found.

Disclosure of Invention

The invention aims to overcome the defects of the prior art, provides an APP continuous improvement method based on user online comment emotion and preference analysis, and provides a user demand real-time acquisition tool based on user group feedback for a developer so as to realize continuous improvement of the APP and promote user experience. Based on the tool, a developer can detect the user emotion and preference feature scores in different time periods in real time, and can locate comment sentences of the user directly based on the mining of the user preference features, so that the developer can understand the actual intention expressed by the user very conveniently, and finally, the evolution trend of the user emotion and preference in the time dimension can be presented in a visual mode.

The invention solves the technical problems by the following technical proposal:

an APP continuous improvement method based on user online comment emotion and preference analysis is characterized by comprising the following steps: the improvement method comprises the following steps:

1) Comment data preprocessing: preprocessing comment data submitted by a user through a mobile terminal by adopting an NLP technology, wherein the preprocessing mainly comprises multi-language filtering, part-of-speech extraction, removal of deactivated words, morphological reduction and the like;

2) Comment emotion acquisition: carrying out user comment emotion analysis by using a SentiStrength tool, wherein the SentiStrength divides comment data into sentence levels and distributes corresponding positive or negative values, and the numerical range is-5;

3) Preference feature acquisition: extracting fine granularity features in user comment data by using a collocation word searching algorithm in NLTK;

4) Preference feature score): calculating a corresponding score for the fine-grained preference feature based on the emotion scores and the fine-grained preference feature mining of steps 2) and 3);

5) Emotion and preference feature evolution analysis of time series

a) User emotion analysis of time series: based on the step 2), the emotion score of each comment sentence can be obtained, a period of time is divided into a plurality of time slices based on a time sequence according to the needs of developers, each time slice is 2-3 days, the average score of the user comment emotion score in each time slice of a target APP is obtained, and for the same APP, the average score of the user comment emotion scores of all APPs in the same APP category is obtained, so that the evolution trend of the user emotion preference in a period of time is obtained;

b) Preference feature evolution analysis of time series: based on the steps 3) and 4), the preference characteristics and the characteristic scores of the user with fine granularity can be obtained, and a period of time is divided into a plurality of time slices according to the needs of developers, and each time slice is 2-3 days, so that a user preference characteristic score matrix is constructed; and for each APP, calculating a certain preference feature total score for each time segment (assumed to be Ti); for the APP in the same class, calculating the total score of all APP in the class in the Ti time segment, which shows the preference feature, so as to obtain the evolution trend of the user preference feature score in a period of time;

6) Emotion-preference feature association mapping analysis: based on the time dimension, establishing mapping relations between the emotion tendencies and preference characteristics mined in different time periods, continuously detecting the mapping relations between the emotion and preference of the user in each time period, and timely finding out important demands of the user;

7) APP evolution and maintenance recommendation: based on the step 6), the user preference characteristic change associated with the user emotion change under different time slices can be seen, and the emotion of the user emotion in the wave crest and the wave trough is recorded; and under other time slices, recording the frequency of the occurrence of the feature, and recommending the preference feature at the moment to a developer.

Moreover, the calculation method of the fine granularity preference feature corresponding score in the step 4) is as follows:

a) If the preference feature appears in the comment, its emotional score is equal to the positive or negative score of the sentence in which it is located;

b) If both positive and negative scores are present in the review, the score with the largest absolute value is selected as the score for the eigenvalue.

The invention has the advantages and beneficial effects that:

aiming at online comment data of a user and combining with user comment time, the invention provides an APP continuous improvement method based on online comment emotion and preference analysis of the user. The method can timely and accurately detect the emotion and preference characteristics of the user from the user comments, can efficiently help a developer to analyze and decide the follow-up related matters about evolution maintenance and follow-up version update of the APP, effectively improves development efficiency, and optimizes user experience of the APP.

Drawings

FIG. 1 is a block diagram of an improved method of the present invention;

FIG. 2 is a graph of user emotion change trend under a single APP of the invention;

FIG. 3 is a graph of the emotion change trend of a user under the same APP class;

FIG. 4 is a graph of the evolution trend of the user fine granularity preference feature of the present invention;

FIG. 5 is a schematic diagram of a user emotion-preference correlation map of the present invention;

FIG. 6 is a graph of emotional tendency versus preference characteristics in time according to the invention.

Detailed Description

The invention is further illustrated by the following examples, which are intended to be illustrative only and not limiting in any way.

As shown in fig. 1, an APP continuous improvement method based on user online comment emotion and preference analysis is innovative in that: the improvement method comprises the following steps:

1) And (3) data acquisition: based on user comment data of an application market, performing data crawling in real time, wherein the data comprises user comments, scores and the like, and storing the data into a mysql database;

2) Comment data preprocessing: preprocessing comment data submitted by a user through a mobile terminal by adopting an NLP technology, wherein the preprocessing mainly comprises multi-language filtering, part-of-speech extraction, removal of deactivated words, morphological reduction and the like;

4) Preference feature score): based on the emotion scores and the fine-granularity preference feature mining of the steps 2) and 3), calculating corresponding scores for the fine-granularity preference features, wherein the corresponding calculation method comprises the following steps:

b) If both positive and negative scores exist in the review, selecting the score with the largest absolute value as the score of the characteristic value;

5) Emotion and preference feature evolution analysis of time series

a) User emotion analysis of time series: based on the step 2), the emotion score of each comment sentence can be obtained, a period of time is divided into a plurality of time slices based on a time sequence according to the needs of a developer, each time slice is 2-3 days, and average score of user comment emotion scores in each time slice of a target APP is obtained, as shown in fig. 2; for the same class of APP, average scores of user comment emotion scores of all APP in the class are obtained, and a user emotion preference evolution trend in a period of time is obtained, as shown in FIG. 3;

b) Preference feature evolution analysis of time series: based on the steps 3) and 4), the preference characteristics and the characteristic scores of the user with fine granularity can be obtained, and a period of time is divided into a plurality of time slices according to the needs of developers, and each time slice is 2-3 days, so that a user preference characteristic score matrix is constructed; and for each APP, calculating a certain preference feature total score for each time segment (assumed to be Ti), as shown in fig. 4; for the APP in the same category, calculating the total score of all APP in the category under the Ti time segment, and obtaining the evolution trend of the score of the preference feature of the user in a period of time, wherein a developer can check the preference expressed by the user comment in each period of time and the score of the preference feature in real time, and the higher the score is, the more important the preference feature is, and the more important the developer should pay attention to;

6) Emotion-preference feature association mapping analysis: when the emotion of the user has peaks or troughs, the influence is caused by the change of user preference characteristics related to the APP, the mapping relation between the emotion tendency and the preference characteristics is established in different time periods based on the time dimension, the mapping relation between the emotion and the preference of the user in each time period is continuously detected, and the important requirement of the user is timely found out, as shown in fig. 5;

7) APP evolution and maintenance recommendation: based on the step 6), the user preference characteristic change related to the user emotion change under different time slices can be seen, the emotion of the user emotion at the wave crest and the wave trough is recorded, and when the emotion trend of the user under a certain time slice of the user is at the wave crest or the wave trough, a developer is focused on the emotion trend, which is an important value characteristic expressed by the user; and under other time slices, the frequency of the occurrence of the characteristics is recorded, and when the emotion trend is stable, the evaluation rate of the occurrence of the preference characteristics and the score thereof are focused, so that the characteristic is also a value characteristic commonly expressed by users, and the characteristic has great significance to the developer.

Meanwhile, in order to achieve more accurate recommendation, for each preference feature, a developer can directly position comment sentences where the feature is located based on the feature, so that reliability of the mined preference feature can be doubly guaranteed, recommendation recommended to the developer is guaranteed to be more timely and accurate, and a recommendation schematic diagram of the developer recommendation is shown in fig. 6.

Although the embodiments of the present invention and the accompanying drawings have been disclosed for illustrative purposes, those skilled in the art will appreciate that: various substitutions, changes and modifications are possible without departing from the spirit and scope of the invention and the appended claims, and therefore the scope of the invention is not limited to the embodiments and the disclosure of the drawings.

Claims

1. An APP continuous improvement method based on user online comment emotion and preference analysis is characterized by comprising the following steps: the improvement method comprises the following steps:

1) Comment data preprocessing: preprocessing comment data submitted by a user through a mobile terminal by adopting an NLP technology, wherein the preprocessing mainly comprises multi-language filtering, part-of-speech extraction, removal of deactivated words and morphological reduction;

4) Preference feature score: calculating a corresponding score for the fine-grained preference feature based on the emotion scores and the fine-grained preference feature mining of steps 2) and 3);

5) Emotion and preference feature evolution analysis of time series

a) User emotion analysis of time series: based on the step 2), the emotion score of each comment sentence can be obtained, a plurality of time slices are divided into a period of time according to the needs of developers based on a time sequence, each time slice is 2-3 days, the average score of the user comment emotion score in each time slice of a target APP is obtained, and for the same APP, the average score of the user comment emotion scores of all APPs in the same APP category is obtained, so that the evolution trend of the user emotion preference in a period of time is obtained;

b) Preference feature evolution analysis of time series: based on the steps 3) and 4), the preference characteristics and the characteristic scores of the user with fine granularity can be obtained, a plurality of time slices are divided for a period of time according to the needs of developers, each time slice is 2-3 days, and a user preference characteristic score matrix is constructed; and for each APP, calculate each time segment T _i A certain preference feature total score; for APP in the same class, then calculate all APP in the class at T _i The total score of the preference feature under the time segment is obtained, and the score evolution trend of the preference feature of the user in a period of time is obtained;

7) APP evolution and maintenance recommendation: based on the step 6), the user preference characteristic change related to the user emotion change under different time slices can be seen, the emotion of the user emotion at the wave crest and the wave trough is recorded, and when the emotion trend of the user under a certain time slice of the user is at the wave crest or the wave trough, a developer is focused on the emotion trend, which is an important value characteristic expressed by the user; under other time slices, the frequency of the feature occurrence is recorded, and when the emotion trend is stable, the frequency of the feature occurrence and the score thereof are focused; in order to achieve more accurate suggestion recommendation, for each preference feature, a developer can directly locate comment sentences where the feature is located based on the feature, and reliability of the mined preference feature can be doubly guaranteed.

2. The APP sustained improvement method based on user online comment emotion and preference analysis of claim 1, wherein: the calculation method of the fine granularity preference feature corresponding score in the step 4) comprises the following steps: