A testing method for multilingual systems based on deep learning

Abstract

The present invention provides a testing method for a multilingual system based on deep learning, the method comprising: step 1, reading product language configuration information of the multilingual system, confirming the position, graphics, and text information of each control, interaction, and graphic of the tested multilingual system that needs to be displayed, and storing the preset information to be tested in a matching database; step 2, setting a test execution module, and performing intelligent test execution through the test execution module, wherein the intelligent test execution includes: performing automatic translation correctness detection, control functionality detection, and product business interaction detection in the language of the tested product; step 3, comparing the test result with the expected value, outputting differentiated function points whose actual value is different from the expected value, and forming a list of use cases of the difference; step 4, summarizing the test results and pushing the test report; the present invention can perform automated testing and reduce labor input costs.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-language system testing, and in particular to a multi-language system testing method based on deep learning. Background Art

[0002] With the development of society, many companies now have more overseas products abroad, supporting English and Arabic, but these two languages:

[0003] (1) Due to cost reasons, the product under test needs to support bilingual switching. In the Arabic language environment, the performance of various parts of the system is quite different from that in the non-Arabic environment, such as text typing direction, control position and display position. In daily testing and testing after version iteration, a lot of manual inspection costs are required.

[0004] (2) Arabic language checking is more difficult than English, and requires the testers to be more familiar with the language. In the case of a large number of product interface layouts, it is necessary to check the accuracy of the Arabic translation of the entire site, which is very difficult and risky.

[0005] (3) When the testing team considers product automation solutions, if UI automation is introduced for two languages ​​with completely different features, there may be high maintenance costs. Summary of the invention

[0006] In order to overcome the above problems, the purpose of the present invention is to provide a testing method for a multilingual system based on deep learning, which has high testing accuracy and low cost.

[0007] The present invention is implemented by the following scheme: a testing method for a multilingual system based on deep learning, the testing method comprising the following steps:

[0008] Step 1: Read the product language configuration information of the multi-language system, confirm the position, graphics, and text information of each control, interaction, and graphic of the multi-language system to be tested, and store the preset information to be tested in the matching database as the expected value of subsequent tests;

[0009] Step 2: Set up a test execution module, and perform intelligent test execution through the test execution module. The test execution module performs intelligent control acquisition and script triggering based on AI deep learning. The intelligent test execution includes: automatic translation correctness detection, control functionality detection, and product business interaction detection of the tested product in the language;

[0010] Step 3: Compare the test results with the expected values, output the differentiated functional points where the actual values ​​are different from the expected values, and form a list of use cases with differences;

[0011] Step 4: Summarize the test results and push the test report.

[0012] Furthermore, the step 1 further specifically includes: step 1.1, when obtaining the preset information of the multi-language system under test, determining whether the system under test has set the language, if yes, obtaining the set language and entering step 1.2; if no, obtaining the default language of the client and entering step 1.2;

[0013] Step 1.2: If the language obtained is English or Arabic, sort out the test cases to be executed according to the language and product business scenario;

[0014] Step 1.3, read the product language configuration information of the multi-language system, confirm the position and graphic information of each control, interaction and graphic of the multi-language system to be tested, and store the preset information to be tested in the matching database.

[0015] Furthermore, if the multilingual system under test is in Arabic, the preset value of each control graphic is: text input box display, displaying the corresponding input box or check box + form field name, and the cursor is on the left side of the text; and the text in each graphic of the multilingual system is in Arabic.

[0016] Furthermore, the test execution module can read and intelligently identify controls, and can intelligently identify and obtain the function menu controls, pictures, and button language information at all levels of the system under test. Based on the matching values ​​of the pre-trained AI deep learning library, it can identify the controls that appear on the interface of the system under test, and then combine the image recognition capabilities to identify the corresponding text information of the interface and the text information of the controls, and identify the control position information and control shape to prepare for the subsequent corresponding control triggering.

[0017] Further, the step 2 further specifically includes: according to the identified control list and text information, a comparison is made with the preset information to be tested, and a corresponding test comparison can be performed, wherein the corresponding test comparison is to perform an automatic translation correctness test, a control functionality test, and a product business interaction test on the tested product in the language; the automatic translation correctness test is: after obtaining the language in which the multilingual system is located at this time, it is determined whether the identified text information belongs to the language and the automatic translation correctness check can be performed on the tested product in the language; the control functionality test is: by analyzing the control list, the home page interface of the tested product can be identified, and it can be known from the preset information that the home page needs to display the necessary controls and menu information of the login button, the carousel, the application icon, the search input box, the resource card, and the page turning control, so as to complete the control function completeness check; the product business interaction test is: through the analysis of the current control, the current home page interface of the tested product can be obtained, and the product business scenario is automatically targeted, and the automated use cases to be tested are interactively listed according to the business scenarios that the home page needs to have, and the triggering of each control of the corresponding interface and the test of the scenario business process are completed according to the automated use cases.

[0018] Furthermore, the step 4 is further specified as follows: summarizing and analyzing the test results, outputting a multi-dimensional automated product test report, and pushing the test results and the number of test rounds, as well as a comparison form between the product test preset values ​​and actual values ​​to the tester.

[0019] The beneficial effects of the present invention are as follows: the present invention uses AI graphic recognition technology to intelligently identify the information of various controls, graphic resources, etc. in the multilingual form of the tested product, obtains the information of different controls of the tested product in a multilingual environment, combines the intelligent positioning technology of controls to obtain the position of controls, and completes the full-dimensional automated testing and comparison of the tested product in the specified language through information analysis and language configuration item reading. And through the comparison of automated translation technology, the accuracy check of internationalized languages ​​is completed, which greatly reduces the cost of labor input and reduces the threshold for the ability requirements of internationalized products for testers. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic flow chart of the method of the present invention.

[0021] Figure 2 It is a method flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The present invention will be further described below in conjunction with the accompanying drawings.

[0023] See also Figure 1 As shown, a testing method for a multi-language system based on deep learning of the present invention comprises the following steps:

[0024] Step 1, reading the product language configuration information of the multi-language system, confirming the position, graphics, and text information of each control, interaction, and graphic of the multi-language system to be tested, and storing the preset information to be tested in the matching database for storage as the expected value of subsequent tests; the step 1 further specifically includes: step 1.1, when obtaining the preset information to be tested of the multi-language system to be tested, judging whether the system to be tested has been set in a language, if yes, obtaining the set language and entering step 1.2; if no, obtaining the default language of the client and entering step 1.2;

[0025] Step 1.2: If the language obtained is English or Arabic, sort out the test cases to be executed according to the language and product business scenario;

[0026] Step 1.3, read the product language configuration information of the multi-language system, confirm the position and graphic information of each control, interaction and graphic of the multi-language system to be tested, and store the preset information to be tested in the matching database.

[0027] Step 2: Set up a test execution module, and perform intelligent test execution through the test execution module. The test execution module performs intelligent control acquisition and script triggering based on AI deep learning. The intelligent test execution includes: automatic translation correctness detection, control functionality detection, and product business interaction detection of the tested product in the language;

[0028] Step 3: Compare the test results with the expected values, output the differentiated functional points where the actual values ​​are different from the expected values, and form a list of use cases with differences;

[0029] Step 4: Summarize the test results and push the test report.

[0030] The present invention will be further described below in conjunction with a specific embodiment:

[0031] The multi-language system under test is a dual-client product for WEB and Android that supports dual-language switching. When the tested product is configured to point to a specific language, such as Arabic, after starting the tested product, even if the local system default language is not Arabic, the layout, text, pattern, and interaction in the product will display Arabic information. If the current product configuration does not point to a specific language, the language information in the product will be displayed in the current system default language. In the Arabic language environment, controls, text, and images are quite different from non-Arabic, for example:

[0032] 1. In the Arabic environment, the text in the input box is typed from right to left, while in the English environment, the text is displayed from left to right;

[0033] 2. In the Arabic language environment, the columns of the tree-shaped TAB bar are displayed on the right, while in the English language environment, the columns are displayed on the left;

[0034] 3. In Arabic, the “>” button means returning to the previous page, while in English, it means going to the next page;

[0035] Therefore, the completely different interactions and system displays bring more regression test branches to product testing. Therefore, this patent reads the product's automatic configuration, first stores the preset information of the product to be tested in the matching database, and uses it as the expected value of subsequent tests, and compares it with the actual results of subsequent automatic execution to automatically complete a series of product tests. The test execution module is mainly based on AI deep learning to obtain intelligent controls and trigger scripts to complete the automated testing of bilingual systems.

[0036] See also Figure 2 As shown, a test method of a multi-language system based on deep learning of the present invention comprises: testing a preset configuration reading operation and an intelligent test execution operation;

[0037] (1) Preparation before testing:

[0038] Mainly, before the test starts, by reading the product language configuration information, confirm the location, graphics, text and other information of each control, interaction and graphic of the tested system to be displayed, which is used as a matching reference for subsequent test execution. If the tested system is in Arabic, the preset value of each control graphic should be: text input box display, display the corresponding input box or check box + form field name, and the cursor is on the left side of the text; and the text in the graphics of various parts of the system is in Arabic, such as carousel pictures, resource thumbnails, etc. And according to the current configuration information, automatically match and output the test cases and expected values ​​to be executed.

[0039] (2) Intelligent identification and test execution module:

[0040] After obtaining the preset product information, the test system begins to perform intelligent testing, including control reading and intelligent identification, matching, and analysis, as well as targeted automated use case execution testing after analyzing control or image resource information:

[0041] ①Trigger control identification and matching: Start intelligent identification and acquisition of the function menu controls, pictures, and button language information at all levels of the system under test. According to the matching values ​​of the pre-trained AI deep learning library, identify the controls that appear on the interface of the system under test, and then combine the image recognition capabilities to identify the corresponding text information and control text information of the interface. Then, for these controls, automatically identify the control location information and control form to prepare for the subsequent corresponding control triggering. For example, recognize that the current interface has a login button, a carousel, an application icon, a search input box, a resource card, a page turning control, etc.

[0042] According to the current language situation of the product under test and the hierarchical relationship of the product under test, a list of test expectations is automatically generated, such as the position of graphics and text, and the next interface that will jump to after clicking the icon control. For example, the home page appears first, the home page includes a search input box, and entering keywords to trigger a search will jump to the search list interface, etc., which are the expected values ​​of the test cases.

[0043] ② Test execution and test result comparison: Based on the identified control list and text information, a comparison is made with the preset value in step (1) to perform a corresponding test comparison, wherein the corresponding test comparison is to perform an automatic translation correctness check, a control functionality check, and a product business interaction check on the tested product in the language. The automatic translation correctness check is to obtain the language of the multilingual system at this time, and then determine whether the identified text information belongs to the language, so as to perform an automatic translation correctness check on the tested product in the language. The control functionality check is to analyze the control list to identify the home page interface of the tested product, and to know from the preset information that the home page needs to display the necessary controls and menu information such as the login button, the carousel, the application icon, the search input box, the resource card, and the page turning control, so as to complete the control function completeness check. The product business interaction check is to analyze the current control to obtain the home page interface of the tested product, and automatically list the automated use cases to be tested interactively according to the business scenarios required for the home page according to the product business scenarios, and complete the test of the triggering of each control and the scenario business process of the corresponding interface according to the automated use cases. The specific test results are compared with the expected values. If the actual values ​​of the output differentiated functional points are different from the expected values, it means that this part of the test case has failed.

[0044] Among them, the identified control list and text information are compared with the preset values: A: If the current system language is English, the control arrangement position and text direction will be arranged from left to right. If the current system language is Arabic, the system controls, text, and interactions will be arranged from right to left. From the preset value in step 1, we can know that the current system is currently in Arabic. At this time, the icons in the product should be displayed on the left side of the text, such as If the current system language is English, the expected value of the test is that the icon is displayed on the left side of the text: Log out; In addition to the different interface positions, the interface interactions are also different. For example, when the system language is English, clicking the right icon means moving forward or entering the next environment →> moving forward. If the system language is Arabic, the expected value is the opposite. At this time, clicking left means returning: ←< moving forward. By reading the preset value, the expected value of the test is automatically updated, and the actual value read and the actual value of the page interaction are compared with the expected value, the correctness of the process can be verified.

[0045] ③ Test result collection and push: After completing the automated testing of the above-mentioned modules, the test results will be summarized and analyzed, and a multi-dimensional automated product test report will be output. The test results and the number of test rounds, as well as the comparison form between the product test preset value and the actual value will be pushed to the tester.

[0046] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made according to the scope of the patent application of the present invention should fall within the scope of the present invention.

Claims

1. A testing method for a multilingual system based on deep learning, characterized in that: The testing method comprises the following steps: Step 1: Read the product language configuration information of the multi-language system to confirm the position, graphics, and text information of each control, interaction, and graphic of the multi-language system to be tested; the position, graphics, and text information of each control, interaction, and graphic of the multi-language system to be tested are preset information to be tested; store the preset information to be tested in the matching database for storage as the expected value of subsequent tests; Step 2: Set up a test execution module, and perform intelligent test execution through the test execution module. The test execution module performs intelligent control acquisition and script triggering based on AI deep learning. The intelligent test execution includes: automatic translation correctness detection, control functionality detection, and product business interaction detection of the tested product in the language; Step 3: Compare the test results with the expected values, output the differentiated functional points where the actual values ​​are different from the expected values, and form a list of use cases with differences; Step 4: Summarize the test results and push the test report; The test execution module can read and intelligently identify controls, and can intelligently identify and obtain the function menu controls, pictures, and button language information of the system under test at all levels. According to the matching values ​​of the pre-trained AI deep learning library, it can identify the controls that appear on the interface of the system under test, and then combine the image recognition capability to identify the text information corresponding to the interface and the text information of the controls, and identify the control position information and control form to prepare for the subsequent corresponding control triggering; The step 2 further specifically includes: according to the identified control list and text information, a comparison is made with the preset information to be tested, and a corresponding test comparison can be performed, wherein the corresponding test comparison is to perform an automatic translation correctness test, a control functionality test, and a product business interaction test on the tested product in the language; the automatic translation correctness test is: after obtaining the language of the multilingual system at this time, it is determined whether the identified text information belongs to the language, and the automatic translation correctness check can be performed on the tested product in the language; the control functionality test is: by analyzing the control list, the home page interface of the tested product can be identified, and it can be known from the preset information that the home page needs to display the necessary controls and menu information of the login button, the carousel, the application icon, the search input box, the resource card, and the page turning control, so as to complete the control function completeness check; the product business interaction test is: through the analysis of the current control, the home page interface of the tested product can be obtained, and the automatic use cases to be tested are automatically listed interactively according to the business scenarios that the home page needs to have for the product business scenarios, and the triggering of each control of the corresponding interface and the test of the scenario business process are completed according to the automatic use cases.

2. The method for testing a multi-language system based on deep learning according to claim 1, characterized in that: The step 1 further specifically includes: step 1.1, when obtaining the preset information of the multi-language system under test, determining whether the system under test has been set to a language, if yes, obtaining the set language and proceeding to step 1.2; if no, obtaining the default language of the client and proceeding to step 1.2; Step 1.2: If the language obtained is English or Arabic, sort out the test cases to be executed according to the language and product business scenario; Step 1.3, read the product language configuration information of the multi-language system, confirm the position and graphic information of each control, interaction and graphic of the multi-language system to be tested, and store the preset information to be tested in the matching database.

3. The method for testing a multi-language system based on deep learning according to claim 2, characterized in that: The multi-language system under test is in Arabic, and the preset values ​​of each control graphic are: text input box display, displaying the corresponding input box or check box + form field name, and the cursor is on the left side of the text; and the text in each graphic of the multi-language system is in Arabic.

4. The method for testing a multi-language system based on deep learning according to claim 1, characterized in that: The step 4 is further specifically as follows: summarizing and analyzing the test results, outputting a multi-dimensional automated product test report, and pushing the test results and the number of test rounds, as well as a comparison form between the product test preset values ​​and actual values ​​to the tester.

Images