CN-121999475-A - Automatic schedule acquisition method for mobile operating system

CN121999475ACN 121999475 ACN121999475 ACN 121999475ACN-121999475-A

Abstract

The invention relates to an automatic schedule acquisition method for a mobile operating system, which comprises the steps of S100, S200, S300, S400, S500, wherein when a target application is detected to be started and is in a foreground operation, a real-time schedule acquisition flow is started, S200, the original text of a current interface of the target application is acquired, S300, the original text is purified to obtain a purified effective text, S400, the purified effective text is input into a potential schedule information buffer module to form a potential schedule text set, S500, the potential schedule text set is transmitted to a schedule extraction model, and S600, finally structured schedule information is output. The invention mainly aims to provide an automatic schedule acquisition method for a mobile operating system, which can finish automatic identification and extraction of cross-application schedule information without manual triggering of a user, and simultaneously ensures the stability and precision of text extraction.

Inventors

CHEN YUNFEI
GONG MINGZHI
LI CHAOYUE
WU HONGZHOU
MENG QINGBIN
TANG RUI

Assignees

麒麟软件有限公司

Dates

Publication Date: 20260508
Application Date: 20260409

Claims (9)

1. The automatic schedule obtaining method for the mobile operating system is applied to the electronic equipment carrying the mobile operating system and is characterized by comprising the following steps of: S100, a service monitoring module monitors the foreground activation state of a target application in real time, and when the target application is detected to be started and is in foreground operation, a real-time schedule acquisition process is started; The method comprises the steps of S200, a screen text acquisition module, a unified marking and recording module, wherein the screen text acquisition module acquires screen text of a current interface of a target application by preferentially acquiring the screen text of the current interface of the target application through barrier-free service, synchronously judges whether the target application has barrier-free service text acquisition limitation through a barrier-free service strategy monitoring and buffering module, and if the limitation exists or the text density acquired by the barrier-free service is lower than a preset threshold value, the screen text acquisition module switches to screen capturing and combines Optical Character Recognition (OCR) to acquire the screen text of the current interface of the target application; S300, inputting the original text obtained in the step S200 into a non-text content removing module, and removing non-text interference text in an interface to obtain a purified effective text; S400, inputting the purified effective text into a potential schedule information cache module, wherein the potential schedule information cache module stores the text according to a business scene corresponding to a target application by adopting a differential storage strategy matched with the business scene to form a potential schedule text set; s500, a text-to-schedule task trigger monitors the operation behaviors of a user and the running state of a target application in real time, and when a preset trigger condition matched with a current service scene is monitored, a potential schedule text set is transmitted to a schedule extraction model; s600, the schedule extraction model calls a standardized service template corresponding to the current service scene in the service template cache module, extracts schedule core information in a template matching mode, and calls a large language model LLM to extract the bottom of the schedule information if the template matching does not acquire effective schedule information, and finally outputs structured schedule information.
2. The automatic schedule obtaining method for mobile operating system according to claim 1, wherein in step S200, the method for determining whether there is a barrier-free service text obtaining restriction in the target application by the barrier-free service policy monitoring and caching module comprises the steps of: s210, periodically and respectively acquiring a component text extracted by barrier-free service and an aggregate text obtained by OCR (optical character recognition) of a screen snapshot for the same interface of a target application; s220, carrying out structural aggregation on OCR recognition results, namely calculating pixel row heights of each row of recognition texts, judging texts with row height differences within a preset allowable range as similar row texts, and merging the texts into a structural aggregation text from top to bottom and from left to right according to the coordinate sequence of the texts in a screen; And S230, comparing the text similarity of the component text acquired by the barrier-free service at the same interface position with the OCR aggregation text, judging that the barrier-free service text acquisition limit exists in the target application if the comparison result is lower than a preset similarity threshold, and updating the judgment result to the barrier-free service strategy monitoring and caching module.
3. The automatic schedule acquiring method for mobile operation system according to claim 2, wherein the preset similarity threshold is based on the measured general scene accuracy of the OCR model, and 10 percentage points are adjusted down as the decision threshold.
4. The automatic schedule obtaining method for mobile operation system according to claim 1, wherein in step S200, when screen text is obtained by combining screen capturing with OCR recognition, continuously captured screen frames are first filtered by a screen frame selecting and filtering module, specifically: And calculating the frame similarity of the continuous screen frames through a structural similarity SSIM algorithm or a mean hash aHash algorithm, judging the screen to be in a stable screen state when the continuous frame similarity reaches a preset similarity threshold value and the duration of the stable state meets a preset minimum duration threshold value, selecting a target frame in the stable state, executing subsequent OCR recognition, and recognizing that the obtained screen text is included in the original text category of the step S200.
5. The automatic schedule acquiring method for mobile operating system according to claim 1, wherein in step S300, the text culling rule of the non-text content removing module comprises: For the application scene of instant messaging, redundant time stamp texts in the dialogue interface, summary time texts on the top layer of the merging forwarding chat record, time identification texts attached to the file name of the message and interference texts identified from the head portrait area of the user are additionally removed.
6. The automatic schedule obtaining method for mobile operating system according to claim 1, wherein in step S400, the differential storage policy of the potential schedule information cache module is specifically: for instant messaging service scenes, continuously storing the dialogue text of the current chat window to form a time-ordered potential schedule text set, and for ticket purchase and order service scenes, covering the history cache text by using the effective text of the latest page, and only reserving the latest single potential schedule text.
7. The automatic schedule obtaining method for mobile operating system according to claim 1, wherein in step S500, the preset trigger condition of the text-to-schedule task trigger corresponds to the service scene one by one, specifically: Aiming at the instant messaging service scene, the preset trigger condition is that the user executes the operations of screen quenching, returning to the system desktop or exiting the current chat window, and aiming at the ticket buying and order service scene, the preset trigger condition is that the user completes the order payment operation and the order state is updated to be successful in payment.
8. The automatic schedule obtaining method for mobile operation system according to claim 1, wherein the business template buffer module pre-stores standardized business templates of common business scenes of travel, conference, ticket and hotel, when the matching success rate of the templates corresponding to the business scenes is lower than a preset threshold, the scene semantic understanding and formatting generating capacity of the large language model LLM is relied on to automatically complete the updating and optimizing of the corresponding business templates, and the updated templates are written into the business template buffer module.
9. The automatic schedule obtaining method for mobile operation system according to claim 1, wherein in step S600, after effective schedule information is successfully obtained through template matching, the text and schedule structuring result extracted this time are input into a large language model LLM synchronously, the time validity and keyword matching degree of the schedule information are checked, the final structuring schedule information is output after the checking is passed, and for atypical scenes without matching service templates, the large language model LLM is directly called to carry out semantic understanding of full text and schedule information deep extraction.

Description

Automatic schedule acquisition method for mobile operating system Technical Field The invention relates to the field of intelligent interaction of mobile operating systems, in particular to an automatic schedule acquisition method for a mobile operating system. Background Today, the mobile internet is rapidly developed, and users exchange information and manage schedules through various application programs (such as instant messaging software, emails, calendars and the like). In recent years, many operating system manufacturers have implemented functions that allow users to quickly add schedules by manual triggers (e.g., voice input, text input, screen shots, etc.), which provides convenience to the users. However, how to automatically acquire schedule information is still an urgent technical problem to be solved. The page reading method proposed by the Chinese patent CN119311982A realizes the extraction of the text in the screen through a series of gradual technical means. The method comprises the steps of firstly attempting to acquire a script of a current page so as to directly extract text, if the script fails to acquire, the system relies on barrier-free service to extract the text, and if the script fails to acquire the script, adopting an Optical Character Recognition (OCR) technology. The series of steps show the feasibility of acquiring the text from the user screen, and show that accurate text information can be successfully extracted through the implementation of reasonable technical means on screen content extraction. However, in instant messaging application or ticket buying software, due to the unique dynamic interactivity and information transmission mode, automatic acquisition of schedule information is still in an exploration stage. Instant messaging software typically includes real-time conversations and multi-level content structures, and text extraction is required to address more complex challenges than static web pages. Therefore, although the technical path of web page text extraction has been verified, it is still necessary to further research and develop the application of the web page text extraction to the extraction of schedule information of applications such as instant messaging, so as to overcome the potential problems. Chinese patent CN120338735A proposes a method for obtaining schedule information, which determines a service scenario corresponding to target text information based on comprehensive analysis of keywords, service entry information, application information, and intention description information of the target text information. Specifically, the patent designs various processing methods for target text information derived from pictures. For example, when a user drags a picture into a calendar application, the system can identify a specific business scene by identifying business entry information, and for a picture acquired by screen capturing, scene determination is performed based on program information of a foreground application. Meanwhile, the common target text information can be input into a pre-trained intention classification model, and the corresponding service scene is further analyzed by using the acquired intention description information, so that intelligent information processing is realized. Then, the patent carries out edge recognition through the target recognition model, comprehensively filters interference information irrelevant to schedules, and ensures that a user can obtain accurate schedule content. The method not only carries out deep analysis on different types of information such as instant messaging chat screenshot, notification card, order screenshot and the like, but also effectively extracts the real intention of the user. However, although the patent provides an effective schedule obtaining method, the process is designed mainly for manual triggering by a user, and various problems such as poor performance, high user experience delay, frequent local resource and remote API resource call and the like may occur when the process is forcedly applied to an automatic process. In summary, the prior art has the following core drawbacks: Text extraction effects are limited by the type of application-obtaining text data from dynamic applications presents a more complex challenge than static web pages. Different applications often employ unique design strategies and interactions, resulting in a variety of text content structures and presentations. In this case, it is often difficult to ensure the accuracy of text extraction by means of only a single unobstructed service or real-time OCR recognition technology. In addition, different types of applications do not specially set an effective text rejection strategy for the schedule acquisition task during design, so that the difficulty in the extraction process is further increased, and invalid information is mixed into or interferes with the extraction of effective data. Therefore, research and dev