Sensitive information identification method and device, equipment, medium and product

Abstract

The invention discloses a sensitive information identification method, apparatus and device, a medium and a product. The method comprises the steps of obtaining a to-be-judged voice stream in a current conference scene and a currently configured preset sensitive field library; performing voice recognition on the to-be-judged voice stream to obtain a voice recognition result containing context text information; performing key field retrieval on the voice recognition result based on a preset sensitive field library, and obtaining a retrieval result in combination with the context text information; under the condition that the retrieval result contains the sensitive field, sensitive field reminding information is generated, and participants in the current conference scene are reminded. According to the method, the to-be-judged voice stream generated in real time in the conference process is subjected to voice recognition, complex semantic phenomena such as polysemy are captured, whether a voice recognition result contains a sensitive field or not is determined, and reminding is carried out when the sensitive field is contained. According to the invention, real-time identification of the sensitive information in the conference process and timely reminding when the sensitive information is involved are realized, so that the occurrence of secret leakage in the conference process is reduced.

Description

Technical Field

[0001] The present invention relates to the field of speech processing technology, and in particular to a sensitive information identification method, device, equipment, medium and product. Background Art

[0002] With the development of communication technology and the popularization of smart terminals, the functions of terminals are becoming more and more powerful. Even if people are not in the same place, they can still use terminals to hold remote meetings anytime and anywhere, which greatly improves the efficiency of communication between people.

[0003] However, in existing terminal remote conferences, conference information is generally exchanged through group voice data, especially in large multinational companies, where information leaks frequently occur during meetings. Summary of the invention

[0004] The present invention provides a sensitive information identification method, device, equipment, medium and product, which automatically judge and remind sensitive information of voice stream during a meeting by referring to context information, so as to reduce the leakage of information during the meeting and avoid the misjudgment of sensitive information.

[0005] According to a first aspect of the present invention, there is provided a sensitive information identification method, comprising:

[0006] Acquire the voice stream to be judged in the current conference scene and the currently configured preset sensitive field library; the preset sensitive field library matches the conference content of the current conference scene;

[0007] Performing speech recognition on the speech stream to be judged to obtain a speech recognition result containing context text information;

[0008] Performing a key field search on the speech recognition result based on the preset sensitive field library, and obtaining a search result in combination with contextual text information in the speech recognition result;

[0009] In the case where the search result contains sensitive fields, sensitive field reminder information is generated and the participants in the current conference scenario are reminded.

[0010] According to a second aspect of the present invention, there is provided a sensitive information identification device, comprising:

[0011] An information acquisition module, used to acquire the voice stream to be judged in the current conference scene and the currently configured preset sensitive field library; the preset sensitive field library matches the conference content of the current conference scene;

[0012] A first determination module is used to perform speech recognition on the speech stream to be determined to obtain a speech recognition result containing context text information;

[0013] A second determination module is used to perform a key field search on the speech recognition result based on the preset sensitive field library, and obtain a search result in combination with the context text information in the speech recognition result;

[0014] The information reminder module is used to generate sensitive field reminder information and remind the participants in the current meeting scenario when the search results contain sensitive fields.

[0015] According to a third aspect of the present invention, there is provided an electronic device, the electronic device comprising:

[0016] at least one processor; and

[0017] a memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the sensitive information identification method described in any embodiment of the present invention.

[0019] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a processor to implement the sensitive information identification method described in any embodiment of the present invention when executed.

[0020] According to the fifth aspect of the present invention, an embodiment of the present invention further provides a computer program product, the computer program product includes a computer program, and when the computer program is executed by a processor, it implements the sensitive information identification method of any embodiment of the present invention.

[0021] The technical solution of the embodiment of the present invention is to obtain the voice stream to be judged in the current conference scene and the preset sensitive field library currently configured; the preset sensitive field library is matched with the conference content of the current conference scene; the voice stream to be judged is subjected to speech recognition to obtain a speech recognition result containing contextual text information; the speech recognition result is subjected to key field retrieval based on the preset sensitive field library, and the retrieval result is obtained in combination with the contextual text information in the speech recognition result; when the retrieval result contains sensitive fields, sensitive field reminder information is generated and the participants in the current conference scene are reminded. By performing speech recognition on the voice stream to be judged generated in real time during the conference, complex semantic phenomena such as polysemy of a word are captured, and it is determined whether the voice recognition result carrying contextual information contains sensitive fields, so as to remind the participants when sensitive fields are contained. Real-time recognition of sensitive information during the conference and timely reminders when sensitive information is involved are realized to reduce the occurrence of leaks during the conference.

[0022] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 is a flow chart of a sensitive information identification method provided according to Embodiment 1 of the present invention;

[0025] Figure 2 is a flow chart of a sensitive information identification method provided according to Embodiment 2 of the present invention;

[0026] Figure 3 is a schematic diagram of the structure of a sensitive information identification device provided according to Embodiment 3 of the present invention;

[0027] Figure 4 It is a schematic diagram of the structure of an electronic device implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme 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 only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] Embodiment 1

[0031] Figure 1 A flowchart of a sensitive information identification method is provided for the first embodiment of the present invention. This embodiment is applicable to the sensitive information judgment of the voice stream during the conference. The method can be executed by a sensitive information identification device. The sensitive information identification device can be implemented in the form of hardware and / or software. The sensitive information identification device can be configured in an electronic device. Figure 1 As shown, the method includes:

[0032] S110: Acquire the voice stream to be judged in the current conference scene and the currently configured preset sensitive field library.

[0033] In this embodiment, the current meeting scene can be understood as all scenes involved in the current meeting. For example, if participants participate in the meeting at different locations, the current meeting scene includes the meeting scenes of all participants at different locations. The voice stream to be determined can be understood as the voice stream obtained in real time during the meeting. The preset sensitive field library can be understood as being used to pre-record sensitive fields in this meeting, and the preset sensitive field library matches the meeting content of the current meeting scene.

[0034] It is known that before the meeting, relevant information about the meeting can be collected in advance to determine the sensitive fields that need to be paid attention to in this meeting from the information. For example, sensitive fields can be core confidential information such as the company's "customer name", "customer project name" and "project quotation". After the sensitive fields are collected, the processor can establish a small database for management, and the processor can dynamically adjust the content in the database. If the project has been mass-produced, the keyword information of the project can be removed from the database. If there is a new project, new field information can be added as a new preset sensitive field library.

[0035] Specifically, before a meeting is held, the processor can search for a preset sensitive field library corresponding to the meeting through a database or storage medium, and the processor can obtain the voice stream to be judged in real time during the meeting.

[0036] S120 , performing speech recognition on the speech stream to be judged, and obtaining a speech recognition result including context text information.

[0037] In this embodiment, the contextual text information can be understood as the preceding and following information corresponding to each frame of audio in the speech stream to be judged, which helps to understand the deep meaning and context of the text, so as to understand the text more accurately and comprehensively. The speech recognition result can be understood as the recognition result converted into text form.

[0038] Specifically, the processor can split each frame of audio in the voice stream to be judged to obtain audio frames, and extract audio features from each audio frame. For example, the audio feature extraction model can be used to filter out background audio that is not related to the meeting content, and the audio features can be used to convert and analyze text to obtain a speech recognition result containing contextual text information.

[0039] 130. Perform key field retrieval on the speech recognition result based on the preset sensitive field library, and obtain the retrieval result in combination with the contextual text information in the speech recognition result.

[0040] In this embodiment, the search result can be understood as a result used to reflect whether the speech recognition result contains a sensitive field.

[0041] Specifically, the processor can perform a key field search on the speech recognition results based on all sensitive fields contained in the preset sensitive field library. Since the speech recognition results contain contextual text information, the processor can refer to the contextual text information to determine whether the speech recognition results contain fields that match the content in the preset sensitive field library. If so, the sensitive fields are included as the search results; if not, the non-sensitive fields are not included as the search results.

[0042] S140: When the search results contain sensitive fields, sensitive field reminder information is generated and the participants in the current meeting scenario are reminded.

[0043] In this embodiment, the sensitive field reminder information can be understood as information used to remind participants of the sensitive fields involved in the current meeting. Participants can be understood as all persons participating in the current meeting or managers related to the meeting.

[0044] Specifically, when the determined search results contain sensitive fields, the processor can determine the sensitive field reminder information and give a reminder according to a preset reminder method, for example, by displaying the sensitive field reminder information (for example, the currently mentioned sensitive field) in a pop-up window on the terminal of the participant or by using sound and light reminders.

[0045] The technical solution of the embodiment of the present invention is to obtain the voice stream to be judged in the current conference scene and the preset sensitive field library currently configured; the preset sensitive field library is matched with the conference content of the current conference scene; the voice stream to be judged is subjected to speech recognition to obtain a speech recognition result containing contextual text information; the speech recognition result is subjected to key field retrieval based on the preset sensitive field library, and the retrieval result is obtained in combination with the contextual text information in the speech recognition result; when the retrieval result contains sensitive fields, sensitive field reminder information is generated and the participants in the current conference scene are reminded. By performing speech recognition on the voice stream to be judged generated in real time during the conference, complex semantic phenomena such as polysemy of a word are captured, and it is determined whether the voice recognition result carrying contextual information contains sensitive fields, so as to remind the participants when sensitive fields are contained. Real-time recognition of sensitive information during the conference and timely reminders when sensitive information is involved are realized to reduce the occurrence of leaks during the conference.

[0046] Embodiment 2

[0047] Figure 2 This is a flow chart of a sensitive information identification method provided in Embodiment 2 of the present invention. This embodiment is a further refinement of the above embodiment. Figure 2 As shown, the method includes:

[0048] S201. Obtain the voice stream to be judged in the current conference scene and the currently configured preset sensitive field library.

[0049] S202: Input the speech stream to be judged into a pre-trained audio feature extraction model, extract audio features from the speech stream to be judged, and obtain audio feature results.

[0050] In this embodiment, the audio feature extraction model can be understood as a model for extracting audio features from a speech stream, such as a DNN acoustic model in a deep neural network model, etc. The audio feature result can be understood as a feature result of audio related to the conference content.

[0051] Specifically, the processor can input the voice stream to be judged into a pre-trained audio feature extraction model, and use the audio feature extraction model to filter the voice stream to be judged, thereby extracting only the audio feature results related to the conference content.

[0052] Further, based on the above embodiment, the steps of inputting the speech stream to be judged into the pre-trained audio feature extraction model, extracting the audio features of the speech stream to be judged, and obtaining the audio feature results can be refined as follows:

[0053] The speech stream to be judged is input into the audio feature extraction model; the speech stream to be judged is framed by the audio feature extraction model to obtain the framed audio frames; the audio frames are windowed and smoothed to obtain the initial audio frames; the initial audio frames are processed by Mel-spectrogram cepstral coefficients to obtain the audio feature results.

[0054] In this embodiment, the audio frame can be understood as the audio data in frames constituting the voice stream to be determined. The initial audio frame can be understood as the audio frame after windowing. Mel-Frequency Cepstral Coefficients (MFCC) can be understood as an audio feature extraction method based on human hearing, which is used to capture the spectral characteristics of audio signals.

[0055] Specifically, the processor can input the speech stream to be judged into the pre-trained audio feature extraction model, and perform frame processing on the speech stream to be judged through the audio feature extraction model to obtain the framed audio frame, and perform windowing and smoothing processing on the framed audio frame to obtain the initial audio frame, so as to achieve smoothing of the audio frame and reduce the boundary effect in signal processing. The combination of framing and windowing not only retains the transient characteristics of the speech signal, but also provides a short-time stable signal block for further analysis. The audio feature result is obtained by performing fast Fourier transform, Mel filter, discrete cosine transform (inverse DFT, inverse discrete Fourier transform) and other Mel spectrum cepstrum coefficient processing on the initial audio frame.

[0056] S203: Convert the audio feature result into text to obtain a text result.

[0057] In this embodiment, the text result can be understood as a result in text form.

[0058] Specifically, the processor can convert the audio feature results into text form through a pronunciation dictionary and other methods to obtain text results. Among them, the pronunciation dictionary is based on acoustic box linguistics and describes the relationship between words and pronunciation letters. Taking Chinese occurrence words as an example, by specifying the relationship between the words and pinyin in the vocabulary, the speech recognition dictionary is composed of vowel primitives and syllable primitives, which can convert audio signals into pinyin symbols, decode the pinyin into Chinese characters, and obtain the corresponding Chinese vocabulary after processing.

[0059] S204: Input the text result into a pre-trained dynamic word embedding model, perform semantic recognition on the text result, and obtain a speech recognition result containing contextual text information.

[0060] In this embodiment, the dynamic word embedding model can be understood as a model that can dynamically generate word vectors according to the context to capture complex semantic phenomena such as polysemy.

[0061] Specifically, the processor can input the text results into a pre-trained dynamic word embedding model, perform context prediction on the text results through the dynamic word embedding model, and then output a speech recognition result carrying context information.

[0062] Further, based on the above embodiment, the dynamic word embedding model includes a forward long short-term memory network, a backward long short-term memory network and a splicing module, and the text result can be input into the pre-trained dynamic word embedding model, and the text result can be semantically recognized to obtain the speech recognition result containing contextual text information. The steps are refined as follows:

[0063] The text result is input into the forward long short-term memory network, and the audio feature result is semantically recognized based on the first order to obtain the forward prediction result; the text result is input into the backward long short-term memory network, and the audio feature result is semantically recognized based on the second order to obtain the backward prediction result; the first order is opposite to the second order; the forward prediction result and the backward prediction result are input into the splicing module for feature splicing to obtain the speech recognition result containing contextual text information.

[0064] In the present embodiment, the forward long short-term memory network (LSTM) can be understood as a model for predicting the front-side vocabulary of the current vocabulary. The first order can be understood as the processing order of the forward long short-term memory network, for example, it can be the previous vocabulary direction of the current vocabulary. The forward prediction result can be understood as the result obtained after predicting the front-side vocabulary of the current vocabulary. The backward long short-term memory network can be understood as a model for predicting the back-side vocabulary of the current vocabulary. The second order can be understood as the processing order of the backward long short-term memory network, for example, it can be the reverse of the next vocabulary of the current vocabulary. The backward prediction result can be understood as the result obtained after predicting the back-side vocabulary of the current vocabulary. The splicing module can be understood as a module for vector splicing. In order to ensure that the context information of the sentence can be fully understood, the first order is opposite to the second order.

[0065] Specifically, the processor can input the text result into a pre-trained dynamic word embedding model, which includes a forward LSTM and a backward LSTM. The forward LSTM will process the sentence in the first order, for example, the first order is to process the sentence from left to right, predict the word to the right of each word, and obtain the forward prediction result. The backward LSTM will process the sentence in the second order, for example, the second order is to process the sentence from right to left, predict the word to the left of each word, and obtain the backward prediction result, so that the context information can be added to the word embedding vector representation, so that the meaning of the target word can be comprehensively inferred whether it is a sensitive field, or whether there are other meanings, to avoid false masking. The forward LSTM includes multiple layers of RNN, and there is a weighted layer between every two layers of RNN. The backward LSTM also has the same structure. The vectors output by the forward LSTM and the backward LSTM are spliced ​​through a splicing module to obtain a word embedding vector representation, that is, a speech recognition result containing contextual text information is obtained, wherein the weights of the weighted layers of the forward LSTM and the backward LSTM may be different.

[0066] S205: Perform a key field search on the speech recognition result based on a preset sensitive field library to obtain a preliminary search result.

[0067] In this embodiment, the preliminary search result can be understood as the result after a simple search.

[0068] Specifically, the processor can perform a key field search on the speech recognition result based on the preset sensitive field library. For example, the key field search can be performed through binary search to determine whether the speech recognition result contains the content in the preset sensitive field library to obtain a preliminary search result.

[0069] S206, performing pronunciation fuzzy recognition on the speech recognition result and the preset sensitive field library through a similarity algorithm to obtain a fuzzy recognition result.

[0070] In this embodiment, the similarity algorithm can be understood as an algorithm for determining the degree of similarity between two fields.

[0071] It is known that participants from different regions may have accent problems, which may lead to inaccurate pronunciation of a word in a word. Simply searching through the vocabulary in the speech recognition results may result in inaccurate recognition. Therefore, fuzzy recognition of pronunciation can be performed to achieve more accurate recognition.

[0072] Specifically, the processor can use a similarity algorithm to perform fuzzy pronunciation recognition on the pinyin of each field of the speech recognition result and the pinyin of the preset sensitive fields contained in the preset sensitive field library, determine the similarity between the two fields, and then identify the fields whose similarity reaches a threshold as similar, and those that do not reach the threshold as dissimilar, thereby obtaining a fuzzy recognition result.

[0073] Further, based on the above embodiment, the steps of performing pronunciation fuzzy recognition on the speech recognition result and the preset sensitive field library through a similarity algorithm to obtain the fuzzy recognition result can be refined as follows:

[0074] For words contained in the speech recognition result, determine the edit distance between the first pinyin string of the word and the second pinyin string of each preset sensitive field in the preset sensitive field library; if the edit distance is less than the preset distance, the similarity with the sensitive field is regarded as the fuzzy recognition result; otherwise, the dissimilarity with the sensitive field is regarded as the fuzzy recognition result.

[0075] In this embodiment, the first pinyin string can be understood as the string corresponding to the pinyin of the word in the speech recognition result. The second pinyin string can be understood as the string corresponding to the pinyin of the preset sensitive field. The edit distance can be understood as a method for characterizing the difference between two strings. The smaller the edit distance, the more similar the two strings are.

[0076] Specifically, for words included in the speech recognition results, the processor can determine the edit distance between the first pinyin string of the word and the second pinyin string of each preset sensitive field in the preset sensitive field library. The edit distance is defined as the minimum number of editing operations required to convert one string into another string, including editing operations such as inserting, deleting, and replacing characters. If the edit distance is less than the preset distance, it can be determined that the word is a sensitive word, and the processor can use the similarity with the sensitive field as the fuzzy recognition result; otherwise, the processor can use the dissimilarity with the sensitive field as the fuzzy recognition result.

[0077] S207. Based on the preliminary search results and the fuzzy recognition results, determine whether the speech recognition results contain sensitive fields. If so, the sensitive fields are used as search results.

[0078] Specifically, the processor can combine the preliminary search results and the fuzzy recognition results to jointly determine whether the speech recognition results contain sensitive fields and the recognition results corresponding to the sensitive fields. If so, the processor can use the identified sensitive fields as the search results.

[0079] S208. If not, the search result will be a search result that does not include sensitive fields.

[0080] Specifically, if no sensitive field is identified in the preliminary search results and the fuzzy recognition results, the processor may take the search results that do not include the sensitive field as the search results.

[0081] S209: When the search results contain sensitive fields, sensitive field reminder information is generated and the participants in the current meeting scenario are reminded.

[0082] The technical solution of the embodiment of the present invention performs frame processing and windowing and smoothing processing on the speech stream to be judged through the audio feature extraction model to obtain the initial audio frame, which reduces the boundary effect in the audio signal processing and improves the accuracy of subsequent analysis. The initial audio frame is memorized by the Mel spectrum cepstral coefficient to better capture the spectral characteristics in the audio signal and obtain the audio feature result. The audio feature result is converted into text to obtain a text result in text form, and the text result is contextually embedded and analyzed by the pre-trained dynamic word embedding model to obtain a speech recognition result containing contextual text information. By considering the context information, a dynamically changing word vector is generated for each word, so as to better capture complex semantic phenomena such as polysemy. Based on the preset sensitive field library, the speech recognition result is searched for key fields to obtain a preliminary search result. The edit distance between the first pinyin string of the word in the speech recognition result of the reference context information and the second pinyin string of the preset sensitive field is combined with the preset distance to determine the fuzzy recognition result, so as to achieve accurate recognition of the sensitive field in the case of inaccurate pronunciation. The fuzzy recognition result is combined with the preliminary search result to comprehensively determine whether the sensitive field is mentioned in the speech stream to be judged, and then generate sensitive field reminder information and remind when the sensitive field exists. It realizes real-time identification of sensitive information during meetings and timely reminders when sensitive information is involved, so as to reduce the occurrence of leaks during meetings and avoid misjudgment of sensitive information.

[0083] Embodiment 3

[0084] Figure 3 This is a schematic diagram of the structure of a sensitive information identification device provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes: an information acquisition module 31, a first determination module 32, a second determination module 33 and an information reminder module 34.

[0085] The information acquisition module 31 is used to acquire the voice stream to be judged in the current conference scene and the currently configured preset sensitive field library; the preset sensitive field library matches the conference content of the current conference scene;

[0086] A first determination module 32 is used to perform speech recognition on the speech stream to be determined to obtain a speech recognition result containing context text information;

[0087] A second determination module 33 is used to perform a key field search on the speech recognition result based on the preset sensitive field library, and obtain a search result in combination with the context text information in the speech recognition result;

[0088] The information reminder module 34 is used to generate sensitive field reminder information and remind the participants in the current conference scenario when the search result contains sensitive fields.

[0089] The technical solution of the embodiment of the present invention is to obtain the voice stream to be judged in the current conference scene and the preset sensitive field library currently configured; the preset sensitive field library is matched with the conference content of the current conference scene; the voice stream to be judged is subjected to speech recognition to obtain a speech recognition result containing contextual text information; the speech recognition result is subjected to key field retrieval based on the preset sensitive field library, and the retrieval result is obtained in combination with the contextual text information in the speech recognition result; when the retrieval result contains sensitive fields, sensitive field reminder information is generated and the participants in the current conference scene are reminded. By performing speech recognition on the voice stream to be judged generated in real time during the conference, complex semantic phenomena such as polysemy of a word are captured, and it is determined whether the voice recognition result carrying contextual information contains sensitive fields, so as to remind the participants when sensitive fields are contained. Real-time recognition of sensitive information during the conference and timely reminders when sensitive information is involved are realized to reduce the occurrence of leaks during the conference.

[0090] Further, the first determining module 32 includes:

[0091] A first determination unit is used to input the speech stream to be determined into a pre-trained audio feature extraction model, extract audio features from the speech stream to be determined, and obtain an audio feature result;

[0092] A second determining unit, configured to convert the audio feature result into text to obtain a text result;

[0093] The third determination unit inputs the text result into a pre-trained dynamic word embedding model, performs semantic recognition on the text result, and obtains a speech recognition result containing contextual text information.

[0094] Wherein, the first determining unit is specifically used for:

[0095] Inputting the speech stream to be judged into the audio feature extraction model;

[0096] Performing frame processing on the speech stream to be judged by using the audio feature extraction model to obtain framed audio frames;

[0097] Performing windowing and smoothing processing on the audio frame to obtain an initial audio frame;

[0098] The initial audio frame is processed by Mel-spectrogram cepstral coefficients to obtain an audio feature result.

[0099] The dynamic word embedding model includes a forward long short-term memory network, a backward long short-term memory network and a splicing module, and the third determination unit is specifically used for:

[0100] Inputting the text result into the forward long short-term memory network, performing semantic recognition on the audio feature result based on the first order, and obtaining a forward prediction result;

[0101] Inputting the text result into the backward long short-term memory network, performing semantic recognition on the audio feature result based on a second order, and obtaining a backward prediction result; the first order is opposite to the second order;

[0102] The forward prediction result and the backward prediction result are input into the splicing module for feature splicing to obtain a speech recognition result containing contextual text information.

[0103] Furthermore, the second determining module 33 includes:

[0104] A fourth determining unit, configured to perform a key field search on the speech recognition result based on the preset sensitive field library to obtain a preliminary search result;

[0105] A fifth determining unit, configured to perform pronunciation fuzzy recognition on the speech recognition result and the preset sensitive field library by using a similarity algorithm to obtain a fuzzy recognition result;

[0106] a sixth determination unit, configured to determine whether the speech recognition result includes a sensitive field based on the preliminary search result and the fuzzy recognition result, and if so, to use the sensitive field as the search result;

[0107] The seventh determining unit is used to, if not, include no sensitive fields as the search result.

[0108] The fifth determining unit is specifically configured to:

[0109] For a word included in the speech recognition result, determining the edit distance between a first pinyin string of the word and a second pinyin string of each preset sensitive field in the preset sensitive field library;

[0110] If the edit distance is less than the preset distance, the similarity with the sensitive field is regarded as a fuzzy recognition result;

[0111] Otherwise, the dissimilarities with the sensitive fields are regarded as fuzzy recognition results.

[0112] The sensitive information identification device provided in the embodiment of the present invention can execute the sensitive information identification method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0113] Embodiment 4

[0114] Figure 4A schematic diagram of the structure of an electronic device 40 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), smart speakers and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0115] like Figure 4 As shown, the electronic device 40 includes at least one processor 41, and a memory connected to the at least one processor 41, such as a read-only memory (ROM) 42, a random access memory (RAM) 43, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 41 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 42 or the computer program loaded from the storage unit 48 to the random access memory (RAM) 43. In the RAM 43, various programs and data required for the operation of the electronic device 40 can also be stored. The processor 41, the ROM 42, and the RAM 43 are connected to each other through a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0116] A number of components in the electronic device 40 are connected to the I / O interface 45, including: an input unit 46, such as a keyboard, a mouse, etc.; an output unit 47, such as various types of displays, speakers, etc.; a storage unit 48, such as a disk, an optical disk, etc.; and a communication unit 49, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 49 allows the electronic device 40 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0117] The processor 41 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 41 performs the various methods and processes described above, such as the sensitive information identification method.

[0118] In some embodiments, the sensitive information identification method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 48. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 40 via the ROM 42 and / or the communication unit 49. When the computer program is loaded into the RAM 43 and executed by the processor 41, one or more steps of the sensitive information identification method described above may be performed. Alternatively, in other embodiments, the processor 41 may be configured to perform the sensitive information identification method in any other appropriate manner (e.g., by means of firmware).

[0119] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0120] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0121] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0122] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0123] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0124] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0125] In one embodiment, the embodiment of the present invention also includes a computer program product, the computer program product includes a computer program, and the computer program implements the sensitive information identification method of any embodiment of the present invention when executed by a processor.

[0126] In the process of implementation, the computer program product can be written in one or more programming languages ​​or a combination thereof to perform the computer program code of the present invention, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0127] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0128] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A sensitive information identification method, characterized in that: include: Obtain the voice stream to be judged in the current conference scenario and the currently configured preset sensitive field library; The preset sensitive field library matches the conference content of the current conference scene; Performing speech recognition on the speech stream to be judged to obtain a speech recognition result containing context text information; Performing a key field search on the speech recognition result based on the preset sensitive field library, and obtaining a search result in combination with contextual text information in the speech recognition result; In the case where the search result contains sensitive fields, sensitive field reminder information is generated and the participants in the current conference scenario are reminded.

2. The method according to claim 1, characterized in that The performing speech recognition on the speech stream to be judged to obtain a speech recognition result containing context text information includes: Inputting the speech stream to be judged into a pre-trained audio feature extraction model, performing audio feature extraction on the speech stream to be judged, and obtaining an audio feature result; Convert the audio feature result into text to obtain a text result; The text result is input into a pre-trained dynamic word embedding model, semantic recognition is performed on the text result, and a speech recognition result containing contextual text information is obtained.

3. The method according to claim 2, characterized in that The step of inputting the to-be-judged voice stream into a pre-trained audio feature extraction model, extracting audio features from the to-be-judged voice stream, and obtaining an audio feature result includes: Inputting the speech stream to be judged into the audio feature extraction model; Performing frame processing on the speech stream to be judged by using the audio feature extraction model to obtain framed audio frames; Performing windowing and smoothing processing on the audio frame to obtain an initial audio frame; The initial audio frame is processed by Mel-spectrogram cepstral coefficients to obtain an audio feature result.

4. The method according to claim 2, characterized in that: The dynamic word embedding model includes a forward long short-term memory network, a backward long short-term memory network and a splicing module; the text result is input into the pre-trained dynamic word embedding model, and semantic recognition is performed on the text result to obtain a speech recognition result containing contextual text information, including: Inputting the text result into the forward long short-term memory network, performing semantic recognition on the audio feature result based on the first order, and obtaining a forward prediction result; Inputting the text result into the backward long short-term memory network, performing semantic recognition on the audio feature result based on a second order, and obtaining a backward prediction result; the first order is opposite to the second order; The forward prediction result and the backward prediction result are input into the splicing module for feature splicing to obtain a speech recognition result containing contextual text information.

5. The method according to claim 1, characterized in that The performing a key field search on the speech recognition result based on the preset sensitive field library and obtaining a search result in combination with contextual text information in the speech recognition result includes: Performing a key field search on the speech recognition result based on the preset sensitive field library to obtain a preliminary search result; Performing pronunciation fuzzy recognition on the speech recognition result and the preset sensitive field library by using a similarity algorithm to obtain a fuzzy recognition result; Based on the preliminary search result and the fuzzy recognition result, determine whether the speech recognition result contains sensitive fields, and if so, take the sensitive fields as the search result; If not, the search result will not include sensitive fields.

6. The method according to claim 5, characterized in that The method of performing pronunciation fuzzy recognition on the speech recognition result and the preset sensitive field library by using a similarity algorithm to obtain a fuzzy recognition result includes: For a word included in the speech recognition result, determining the edit distance between a first pinyin string of the word and a second pinyin string of each preset sensitive field in the preset sensitive field library; If the edit distance is less than the preset distance, the similarity with the sensitive field is regarded as a fuzzy recognition result; Otherwise, the dissimilarities with the sensitive fields are regarded as fuzzy recognition results.

7. A sensitive information identification device, characterized in that: include: An information acquisition module is used to obtain the voice stream to be judged in the current conference scene and the currently configured preset sensitive field library; The preset sensitive field library matches the conference content of the current conference scene; A first determination module is used to perform speech recognition on the speech stream to be determined to obtain a speech recognition result containing context text information; A second determination module is used to perform a key field search on the speech recognition result based on the preset sensitive field library, and obtain a search result in combination with the context text information in the speech recognition result; The information reminder module is used to generate sensitive field reminder information and remind the participants in the current meeting scenario when the search results contain sensitive fields.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the sensitive information identification method described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the sensitive information identification method described in any one of claims 1-6 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the sensitive information identification method according to any one of claims 1 to 6.

Images