Natural language processing system and method, and computer-readable storage medium
Through a natural language processing system, analyzing the failure history of semiconductor equipment is generated and an abnormal model classification table is solved, which solves the problem of engineers wasting time due to relying on subjective experience when repairing semiconductor equipment, and achieves the efficiency of quickly finding repair solutions.
Patent Information
- Application Number
- CN202110279214.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2020-06-07
- Filing Date
- 2021-03-16
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2041-03-16
AI Technical Summary
Semiconductor equipment often needs repair due to failures, and engineers waste time finding repair methods because they rely on subjective experience.
Design a natural language processing system to generate an exception model classification table by storing fault history and using potential Dirichlet allocation model analysis to help engineers quickly find repair solutions.
Through natural language processing of semiconductor equipment failure history, exception model classification tables are generated to help engineers quickly identify fault topics and corresponding solutions, improving repair efficiency.
Smart Images

Figure CN113761187B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a natural language processing system and method and a computer-readable storage medium. Background Art
[0002] The electrical conductivity of semiconductor materials is between that of conductors and insulators. Generally speaking, semiconductor equipment can be used to manufacture semiconductor devices.
[0003] However, semiconductor devices often fail due to various problems, and engineers often waste a lot of time repairing semiconductor devices due to their unreliable subjective experience. Summary of the invention
[0004] The purpose of the present invention is to provide a natural language processing system and method and a computer-readable storage medium, which can help engineers quickly find a way to repair the semiconductor device when the semiconductor device fails again.
[0005] In one embodiment of the present invention, the natural language processing system proposed by the present invention includes a storage device and a processor. The storage device is used to store multiple records of the fault history of the semiconductor device, wherein the multiple records of the fault history of the semiconductor device include natural language. The processor is electrically connected to the storage device, and the processor is used to perform natural language processing on the multiple records of the fault history of the semiconductor device to generate an abnormal model classification table.
[0006] In one embodiment of the present invention, a storage device stores a dictionary file, and a processor searches for multiple valid words in multiple records of the fault history of a semiconductor device according to the dictionary file to generate a record word matrix, which represents the word frequency of multiple valid words in each of the multiple records of the fault history of the semiconductor device.
[0007] In one embodiment of the present invention, the storage device stores a latent Dirichlet allocation model, and the processor analyzes the record word matrix based on the latent Dirichlet allocation model to generate a topic word matrix and a record topic matrix.
[0008] In one embodiment of the present invention, the subject word matrix represents the respective ratio of each valid word in each different subject to all multiple valid words, and the record subject matrix represents the respective correlation between each different subject and each record in multiple records of the failure history of semiconductor equipment.
[0009] In one embodiment of the present invention, the processor generates an abnormal model classification table based on a subject word matrix and a record subject matrix, wherein the abnormal model classification table represents the respective prevalence rates of each different subject and at least one related word of each different subject, wherein the at least one related word is selected from a group consisting of multiple valid words.
[0010] In one embodiment of the present invention, the natural language processing system further comprises a communication device. The communication device is electrically connected to the processor and is used to send the abnormal model classification table to the computer device so that the display screen of the computer device displays the abnormal model classification table.
[0011] In one embodiment of the present invention, a communication device receives a plurality of records of failure histories of semiconductor devices.
[0012] In one embodiment of the present invention, the natural language processing method proposed by the present invention includes: preloading multiple records of the failure history of semiconductor equipment, wherein the multiple records of the failure history of semiconductor equipment include natural language; performing natural language processing on the multiple records of the failure history of semiconductor equipment to generate an abnormal model classification table.
[0013] In one embodiment of the present invention, the step of performing natural language processing includes: searching for multiple valid words in multiple records of the fault history of the semiconductor device according to a dictionary file to generate a record word matrix, which represents the word frequency of multiple valid words in each of the multiple records of the fault history of the semiconductor device.
[0014] In one embodiment of the present invention, the step of performing natural language processing further includes: analyzing the record word matrix based on a latent Dirichlet allocation model to generate a topic word matrix and a record topic matrix.
[0015] In one embodiment of the present invention, the subject word matrix represents the respective ratio of each valid word in each different subject to all multiple valid words, and the record subject matrix represents the respective correlation between each different subject and each record in multiple records of the failure history of semiconductor equipment.
[0016] In one embodiment of the present invention, the step of performing natural language processing also includes: generating an abnormal model classification table based on a subject word matrix and a record subject matrix, wherein the abnormal model classification table represents the respective prevalence rates of each different subject and at least one related word of each different subject, and the at least one related word is selected from a group consisting of multiple valid words.
[0017] In one embodiment of the present invention, the natural language processing method further includes: sending the abnormal model classification table to a computer device, so that a display screen of the computer device displays the abnormal model classification table.
[0018] In one embodiment of the present invention, the natural language processing method further includes: receiving a plurality of records of failure history of semiconductor equipment.
[0019] In one embodiment of the present invention, the computer-readable storage medium proposed by the present invention stores multiple instructions to enable a computer to execute a natural language processing method, which includes: preloading multiple records of the failure history of semiconductor equipment, wherein the multiple records of the failure history of semiconductor equipment include natural language; performing natural language processing on the multiple records of the failure history of semiconductor equipment to generate an abnormal model classification table.
[0020] In one embodiment of the present invention, the step of performing natural language processing includes: searching for multiple valid words in multiple records of the fault history of the semiconductor device according to a dictionary file to generate a record word matrix, which represents the word frequency of multiple valid words in each of the multiple records of the fault history of the semiconductor device.
[0021] In one embodiment of the present invention, the step of performing natural language processing further includes: analyzing the record word matrix based on a latent Dirichlet allocation model to generate a topic word matrix and a record topic matrix.
[0022] In one embodiment of the present invention, the subject word matrix represents the respective ratio of each valid word in each different subject to all multiple valid words, and the record subject matrix represents the respective correlation between each different subject and each record in multiple records of the failure history of semiconductor equipment.
[0023] In one embodiment of the present invention, the step of performing natural language processing also includes: generating an abnormal model classification table based on a subject word matrix and a record subject matrix, wherein the abnormal model classification table represents the respective prevalence rates of each different subject and at least one related word of each different subject, and the at least one related word is selected from a group consisting of multiple valid words.
[0024] In one embodiment of the present invention, the natural language processing method further includes: sending the abnormal model classification table to a computer device, so that a display screen of the computer device displays the abnormal model classification table.
[0025] In summary, the technical solution of the present invention has obvious advantages and beneficial effects compared with the prior art. The natural language processing system and method and the computer-readable storage medium classify the topics of various problems of semiconductor devices and provide corresponding solutions. Therefore, the abnormal model classification table can help engineers quickly find ways to repair semiconductor devices when the semiconductor devices fail again.
[0026] The above description will be described in detail below with reference to implementation examples, and a further explanation will be provided for the technical solution of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] In order to make the above and other objects, features, advantages and embodiments of the present invention more clearly understood, the accompanying drawings are described as follows:
[0028] Figure 1 is a block diagram of a natural language processing system according to an embodiment of the present invention; and
[0029] Figure 2 is a flow chart of a natural language processing method according to an embodiment of the present invention.
[0030] Description of main reference numerals:
[0031] 100 - natural language processing system, 110 - storage device, 120 - processor, 130 - communication device, 170 - computer device, 172 - display screen, 190 - semiconductor device, 200 - natural language processing method, S201, S203, S205, S207 - steps. DETAILED DESCRIPTION
[0032] In order to make the description of the present invention more detailed and complete, reference can be made to the attached drawings and various embodiments described below, in which the same numbers represent the same or similar elements. On the other hand, well-known elements and steps are not described in the embodiments to avoid unnecessary limitations on the present invention.
[0033] In the embodiments and the scope of the patent application, the description involving “connection” may generally refer to one element being indirectly coupled to another element through other elements, or one element being directly connected to another element without going through other elements.
[0034] In the embodiments and the scope of the patent application, the description involving "connection" may generally refer to an element indirectly communicating with another element via other elements through wired and / or wireless communication, or an element being physically connected to another element without going through other elements.
[0035] In the embodiments and patent claims, unless the context specifically limits the articles, "a", "an" and "the" may refer to one or more.
[0036] The terms "about", "approximately" or "roughly" used herein are used to modify any quantity that may vary slightly, but such slight variations do not change its essence. In the embodiments, if there is no special description, the error range of the value modified by "about", "approximately" or "roughly" is generally allowed within 20%, preferably within 10%, and more preferably within 5%.
[0037] Figure 1 FIG. 1 is a block diagram of a natural language processing system 100 according to an embodiment of the present invention. Figure 1As shown, the natural language processing system 100 includes a storage device 110, a processor 120, and a communication device 130. For example, the storage device 110 may be a hardware storage device (such as a hard disk, a solid state drive), the processor 120 may be a central processing unit, and the communication device 130 may be a wired or wireless communication circuit.
[0038] In terms of architecture, the communication device 130 is electrically connected to the processor 120, and the processor 120 is electrically connected to the storage device 110. The communication device 130 can establish communication with the computer device 170 and / or the semiconductor device 190. The computer device 170 has a display screen 172. For example, the display screen 172 can be a liquid crystal display or other displays.
[0039] In practice, the semiconductor device 190 can be used to manufacture semiconductor element devices. Every time the semiconductor device 190 fails, the engineer needs to repair the semiconductor device 190. The engineer can record the failure history of the semiconductor device 190 in natural language through the computer device 170, the semiconductor device 190, or other devices. Therefore, the failure history of the semiconductor device includes natural language.
[0040] For example, the failure history of a semiconductor device is recorded as follows.
[0041]
[0042] When in use, the communication device 130 receives multiple records of the fault history of the semiconductor device from the computer device 170, the semiconductor device 190 or other devices, and the storage device 110 preloads multiple records of the fault history of the semiconductor device 190. The processor 120 is used to perform natural language processing on the multiple records of the fault history of the semiconductor device 190 to generate an abnormal model classification table. Then, the communication device 130 is used to send the abnormal model classification table to the computer device 170, so that the display screen 172 of the computer device 170 displays the abnormal model classification table. The abnormal model classification table classifies the topics of various problems of the semiconductor device 190 and provides corresponding solutions. Therefore, the abnormal model classification table can help engineers quickly find ways to repair the semiconductor device 190 when the semiconductor device 190 fails again.
[0043] In one embodiment of the present invention, the storage device 110 stores a dictionary file, and the processor 120 searches for multiple valid words in multiple records of the fault history of the semiconductor device 190 according to the dictionary file to generate a record word matrix, which represents the word frequency of multiple valid words in each of the multiple records of the fault history of the semiconductor device 190. For example, the dictionary file defines invalid words, so the remaining words in the multiple records are valid words.
[0044] In one embodiment of the present invention, the storage device 110 stores a latent Dirichlet allocation (LDA) model, and the processor 120 analyzes the record word matrix based on the latent Dirichlet allocation model to generate a topic word matrix and a record topic matrix.
[0045] In one embodiment of the present invention, the subject word matrix represents the respective ratio of each effective word in each different subject to all multiple effective words. For example, a partial subject word matrix is shown below, where t_1, t_2, and t_3 represent different subjects, and asi_pca, psg_psg, and pressure_graph are different effective words.
[0046]
[0047] In one embodiment of the present invention, the record topic matrix represents the respective correlation between each different topic and each record in the plurality of records of the failure history of the semiconductor device 190. For example, a portion of the record topic matrix is shown below, where t_1, t_2, and t_3 represent different topics, and 3, 6, and 33 are the numbers of the corresponding records.
[0048]
[0049] In one embodiment of the present invention, the processor 120 generates an abnormal model classification table based on the subject word matrix and the record subject matrix, and the abnormal model classification table represents the respective prevalence of each different subject and at least one related word of each different subject, and the at least one related word is selected from a group consisting of multiple valid words, and the related word can be, for example, the most or more words in the valid words. For example, a portion of the subject word matrix is shown below, where the hot terms are the related words of each subject.
[0050]
[0051] To further explain the natural language processing method performed by the natural language processing system 100, please also refer to Figure 1-2 , Figure 2 FIG. 2 is a flow chart of a natural language processing method 200 according to an embodiment of the present invention. Figure 2 As shown, the natural language processing method 200 includes steps S201, S203, S205, and S207 (it should be understood that the order of the steps mentioned in this embodiment, except for those whose order is specifically stated, can be adjusted according to actual needs, and can even be executed simultaneously or partially simultaneously).
[0052] The natural language processing method 200 may be in the form of a computer program product on a computer-readable storage medium, the computer-readable recording medium having a plurality of computer-readable instructions contained in the medium. Suitable recording media may include any of the following: non-volatile memory, such as read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electronically erasable programmable read-only memory (EEPROM); volatile memory, such as static random access memory (SRAM), dynamic random access memory (DRAM), double data rate synchronous dynamic random access memory (DDR-RAM); optical storage device, such as read-only compact disc (CD-ROM), read-only digital versatile video disc (DVD-ROM); magnetic storage device, such as hard disk drive, floppy disk drive.
[0053] In practice, every time a semiconductor device 190 fails, an engineer needs to repair the semiconductor device 190. The engineer can record the failure history of the semiconductor device 190 in natural language through the computer device 170, the semiconductor device 190, or other devices. In step S201, multiple records of the failure history of the semiconductor device are received from the computer device 170, the semiconductor device 190, or other devices, and the multiple records of the failure history of the semiconductor device include natural language.
[0054] Next, in step S203, multiple records of the fault history of the semiconductor device 190 are preloaded. In step S205, natural language processing is performed on the multiple records of the fault history of the semiconductor device 190 to generate an abnormal model classification table. In step S207, the abnormal model classification table is sent to the computer device 170, so that the display screen 172 of the computer device 170 displays the abnormal model classification table. The abnormal model classification table classifies the topics of various problems of the semiconductor device 190 and provides corresponding solutions. Therefore, the abnormal model classification table can help engineers quickly find a way to repair the semiconductor device 190 when the semiconductor device 190 fails again.
[0055] In one embodiment of the present invention, in step S205, multiple valid words are searched in multiple records of the fault history of the semiconductor device 190 according to the dictionary file to generate a record word matrix, which represents the word frequency of multiple valid words in each of the multiple records of the fault history of the semiconductor device 190.
[0056] In one embodiment of the present invention, in step S205, the record word matrix is analyzed based on the latent Dirichlet allocation model to generate a topic word matrix and a record topic matrix.
[0057] In one embodiment of the present invention, the subject word matrix represents the respective ratio of each valid word in each different subject to all multiple valid words, and the record subject matrix represents the respective correlation between each different subject and each record in the multiple records of the failure history of the semiconductor device 190.
[0058] In one embodiment of the present invention, in step S205, an abnormal model classification table is generated based on the subject word matrix and the record subject matrix, and the abnormal model classification table represents the respective prevalence rates of each different subject and at least one related word of each different subject, and the at least one related word is selected from a group consisting of multiple valid words.
[0059] In summary, the natural language processing system and method and computer-readable storage medium of the present invention have obvious advantages and beneficial effects compared with the prior art. The natural language processing system 100 and the natural language processing method 200 classify the topics of various problems of the semiconductor device 190 and provide corresponding solutions. Therefore, the abnormal model classification table can help engineers quickly find a way to repair the semiconductor device 190 when the semiconductor device 190 fails again.
[0060] Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the claims.
Claims
1. A natural language processing system, characterized in that: include: A storage device for storing a plurality of records of a failure history of a semiconductor device, wherein the plurality of records of the failure history of the semiconductor device include a natural language; as well as a processor, electrically connected to the storage device, and used to perform natural language processing on the multiple records of the fault history of the semiconductor device to generate an abnormal model classification table, the storage device stores a dictionary file, the processor searches for multiple valid words in the multiple records of the fault history of the semiconductor device according to the dictionary file to generate a record word matrix, the record word matrix represents the word frequency of the multiple valid words in each of the multiple records of the fault history of the semiconductor device, the storage device stores a latent Dirichlet allocation model, and the processor analyzes the record word matrix based on the latent Dirichlet allocation model The processor generates a subject word matrix and a record subject matrix, wherein the subject word matrix represents the respective ratio of each valid word in each different subject to all the multiple valid words, and the record subject matrix represents the respective correlation between each different subject and each record in the multiple records of the fault history of the semiconductor device. The processor generates the abnormal model classification table based on the subject word matrix and the record subject matrix, and the abnormal model classification table represents the respective prevalence rate of each different subject and at least one related word of each different subject, and the at least one related word is selected from the group consisting of the multiple valid words.
2. The natural language processing system according to claim 1, characterized in that Also includes: The communication device is electrically connected to the processor and is used to send the abnormal model classification table to the computer device so that the display screen of the computer device displays the abnormal model classification table.
3. The natural language processing system according to claim 2, characterized in that: The communication device receives the plurality of records of the failure history of the semiconductor device.
4. A natural language processing method, characterized in that: include: preloading a plurality of records of a failure history of a semiconductor device, wherein the plurality of records of the failure history of the semiconductor device include natural language; as well as Performing natural language processing on the plurality of records of the fault history of the semiconductor device to generate an abnormal model classification table, wherein the step of performing the natural language processing comprises: searching for a plurality of valid words in the plurality of records of the failure history of the semiconductor device according to a dictionary file to generate a record word matrix, the record word matrix indicating word frequencies of the plurality of valid words in each of the plurality of records of the failure history of the semiconductor device; and The record word matrix is analyzed based on a latent Dirichlet allocation model to generate a subject word matrix and a record topic matrix, wherein the subject word matrix represents the respective ratio of each valid word in each different topic to all the multiple valid words, and the record topic matrix represents the respective correlation between each different topic and each of the multiple records of the fault history of the semiconductor device, and the natural language processing step further includes: Based on the subject word matrix and the record subject matrix, the abnormal model classification table is generated, and the abnormal model classification table represents the respective prevalence rates of each different subject and at least one related word of each different subject, and the at least one related word is selected from the group consisting of the multiple valid words.
5. The natural language processing method according to claim 4, characterized in that: Also includes: The abnormal model classification table is sent to a computer device so that a display screen of the computer device displays the abnormal model classification table.
6. The natural language processing method according to claim 4, characterized in that: Also includes: The plurality of records of the failure history of the semiconductor device are received.
7. A computer-readable storage medium storing a plurality of instructions for causing a computer to execute a natural language processing method, characterized in that: The natural language processing method comprises: preloading a plurality of records of failure histories of semiconductor devices, wherein the plurality of records of the failure histories of the semiconductor devices include natural language; and Performing natural language processing on the plurality of records of the fault history of the semiconductor device to generate an abnormal model classification table, wherein the step of performing the natural language processing comprises: searching for a plurality of valid words in the plurality of records of the failure history of the semiconductor device according to a dictionary file to generate a record word matrix, the record word matrix indicating word frequencies of the plurality of valid words in each of the plurality of records of the failure history of the semiconductor device; and The record word matrix is analyzed based on a latent Dirichlet allocation model to generate a subject word matrix and a record topic matrix, wherein the subject word matrix represents the respective ratio of each valid word in each different topic to all the multiple valid words, and the record topic matrix represents the respective correlation between each different topic and each of the multiple records of the fault history of the semiconductor device, and the natural language processing step further includes: Based on the subject word matrix and the record subject matrix, the abnormal model classification table is generated, and the abnormal model classification table represents the respective prevalence rates of each different subject and at least one related word of each different subject, and the at least one related word is selected from the group consisting of the multiple valid words.
8. The computer-readable storage medium of claim 7, wherein: The natural language processing method further comprises: The abnormal model classification table is sent to a computer device so that a display screen of the computer device displays the abnormal model classification table.
