Recording text mark management method and device and storage medium

Abstract

The invention discloses a recording text mark management method and device and a storage medium, and belongs to the technical field of recording text mark management. The method comprises the following steps: constructing a first feature according to a sound recording file, and constructing a second feature according to a sound recording text corresponding to the sound recording file; constructing a fused audio feature according to the first feature and the second feature; determining a classification label based on a classification result corresponding to the fused audio feature; and associating the classification label with the identifier of the recording text, and establishing an index for the identifier and the storage position of the recording text. According to the method, the first audio feature is extracted from the recording file, and the second audio feature is extracted after the recording file is converted into the text, so that the label of the recording text is determined from multiple angles, the label and the text are subjected to composite dimension storage, the recording text is conveniently managed, the searching time of the voice text is shortened, and the efficiency is improved. And the searching efficiency of the sound recording file is improved.

Description

Technical Field

[0001] The present application relates to the technical field of text storage management, and in particular to a method, device and storage medium for recording text tag management. Background Art

[0002] At present, based on the management requirements for recordings of customer service and customer communications, it is necessary to save the recording files of calls with customers and review the service processes of the agents. In related technologies, when the text of the recording files is stored, the storage location of the recording text is divided only according to the agent number and recording date corresponding to the recording text, and the content of the recording text is not identified. Therefore, when the text search speed of the quality inspector is limited, when there are a large number of recording texts with the same agent number on the same day, it takes a lot of browsing time to search for the recording text, resulting in a long time to search for the recording text and low search efficiency.

[0003] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0004] The main purpose of this application is to provide a method, device and storage medium for recording text tag management, aiming to solve the technical problem in the related art that recording files are migrated to other devices for storage, the recordings are not classified, and the search efficiency is low.

[0005] To achieve the above-mentioned purpose, the present application provides a method for managing recorded text marking, which includes the following steps: constructing a first feature based on a recorded file, and constructing a second feature based on the recorded text corresponding to the recorded file; constructing a fused audio feature based on the first feature and the second feature; determining a classification label based on a classification result corresponding to the fused audio feature; and associating the classification label with the recorded file and saving it to enable reading of the recorded file based on the classification label.

[0006] In one embodiment, the step of constructing a fused audio feature according to the first feature and the second feature includes: Standardizing the first feature to determine a first standard feature, and standardizing the second feature to determine a second standard feature; The first standard feature and the second standard feature are fused to determine the fused audio feature.

[0007] In one embodiment, the step of fusing the first standard feature and the second standard feature to determine the fused audio feature includes: Determining a weight allocation strategy based on the correlation between the first standard feature and the second standard feature; Based on the weight allocation strategy, the first standard feature and the second standard feature are weightedly summed to determine the fused audio feature.

[0008] In one embodiment, the step of determining a classification label based on the classification result corresponding to the fused audio feature includes: Inputting the fused audio features into a preset classification model to determine the classification result; Based on the classification result, at least one classification label is output.

[0009] In one embodiment, the step of constructing the first feature according to the recording file, and constructing the second feature according to the recording text corresponding to the recording file, includes: Extracting the first feature of the recording file based on the audio attributes of the recording file, the audio attributes including: spectral characteristics, frequency, audio duration and audio level; The recorded text is recognized based on natural language processing technology to extract the second feature.

[0010] In one embodiment, before the step of constructing a fused audio feature according to the first feature and the second feature, the method further includes: Construct the third feature based on customer complaint information and call conditions; The step of constructing a fused audio feature according to the first feature and the second feature comprises: The fused audio feature is constructed according to the first feature, the second feature and the third feature.

[0011] In one embodiment, the step of associating and saving the classification tag with the recording file to read the recording file based on the classification tag includes: Based on the text features of the audio recording text, generating the identifier of the audio recording text based on a preset algorithm, the text features including the text name and creation time; The identifier is associated with the classification label and the index is established with the storage location.

[0012] In one embodiment, after the step of associating the identifier with the classification label and establishing the index with the storage location, the method further includes: In response to the query operation of the classification tag, displaying the storage location of the audio recording text corresponding to the classification tag; In response to a click operation on the storage location, the text content of the recorded text is displayed.

[0013] In addition, to achieve the above-mentioned purpose, the present application also provides a recording text marking management device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the recording text marking management method as described above.

[0014] In addition, to achieve the above-mentioned purpose, the present application also provides a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the recording text marking management method as described above are implemented.

[0015] The present application provides a method for managing recording text tags. The present application analyzes the audio features of the recording file and the semantic features of the recording text, defines tags from multiple angles, and establishes indexes for the identifiers of tags and texts, so as to facilitate quality inspectors to locate the pre-searched text according to the multi-dimensional tags, effectively realize the recording-to-text storage and set corresponding tags to match the recording text, thus facilitating the management of the recording text, reducing the search time for the voice text, and improving the search efficiency of the recording files. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0018] Figure 1 A flowchart of the first embodiment of the recording text tagging management method of the present application is provided; Figure 2 A flowchart of the second embodiment of the recording text tag management method of the present application is provided; Figure 3 A flowchart of the third embodiment of the recording text tag management method of the present application is provided; Figure 4 A flowchart of the fourth embodiment of the recording text tag management method of the present application is provided; Figure 5 A flowchart of Embodiment 5 of the method for managing recording text markup of this application is provided; Figure 6 A flowchart of Embodiment 6 of the method for managing recording text markup of this application is provided; Figure 7 A flowchart of a recording text query process provided in Example 6 of the recording text tag management method of the present application; Figure 8 Schematic diagram of the device structure of the hardware operating environment involved in the recording text marking management method in the embodiment of the present application.

[0019] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0020] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0021] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0022] The recording text number needs to take into account the audio features of the recording file and the semantic features of the recording text, where the audio features include but are not limited to time domain features and frequency domain features, and the semantic features for the recording text capture the semantic information in the data. In the recording text management system, the files that need to be saved are often saved according to the naming or text attributes. For example, the user creates a folder based on the file name and text generation time, and puts the text in it.

[0023] However, when there are too many seats serving on the same day, storing one recording file in a folder greatly reduces storage efficiency, and it is difficult to effectively manage file storage based on the text generation time. Therefore, when the quality inspector has limited text search speed and there are many recording texts with the same seat number on the same day, it takes a lot of browsing time to search for the recording text, resulting in a long time to search for the recording text and low search efficiency.

[0024] To solve the above problems, the present application proposes a method for managing audio text tags, which obtains audio files and their corresponding audio texts, extracts audio features and semantic features, generates classification labels based on these features, and then associates the classification labels with the identifiers of the audio texts, and establishes an index of the identifiers and the storage locations of the audio texts, so as to facilitate subsequent classification management and rapid retrieval. In the above scheme, by converting the audio files into audio texts, classifying the audio texts with labels, and storing the audio texts and label names in the form of composite dimensions, the problem of low efficiency in searching the audio files after storage can be solved.

[0025] In order to better understand the above technical solution, exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0026] It should be noted that the execution subject of this embodiment can be a recording text mark management device, or a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, etc., and this embodiment does not specifically limit this. The following takes the recording text mark management device as an example to illustrate this embodiment and the following embodiments.

[0027] Based on this, the first embodiment of the present application provides a method for managing recording text marks, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the recording text marking management method of this application.

[0028] In this embodiment, the recording text mark management method includes steps S100 to S400: Step S100, constructing a first feature according to the recording file, and constructing a second feature according to the recording text corresponding to the recording file.

[0029] First, use a recording device such as a mobile phone or voice recorder to record the audio, then save the recording file and upload it to the recording text tag management system. For the first feature extraction of the recording file, first convert the recording file into a format suitable for processing through format conversion, then use an audio processing tool to sample the audio file, obtain a digital representation of the audio data, and then extract the corresponding features.

[0030] As an optional solution to extract the features of the recording file, the recording file is pre-processed by sampling, quantization, etc. to ensure the quality of the audio signal. De-noising is performed to reduce the impact of background noise on feature extraction, and then feature extraction is performed on the pre-processed recording file.

[0031] Specifically, the fundamental frequency characteristics are determined based on the frequency of the main frequency components in the audio signal to characterize the speaker's pitch; the sound intensity characteristics are determined based on the amplitude or strength of the audio signal to reflect the loudness of the sound; and the rhythm characteristics are determined based on the speed and regularity of the sound changes in the audio signal.

[0032] Exemplarily, the recording text tagging management system is labeling and classifying the recording files of the agent service process. First, the audio processing tool is used to detect the recording files. It is found that the frequency of the audio signal of the agent speaking in a certain period of time is relatively high, and the amplitude of the audio signal can be relatively large, and the sound in the audio signal changes rapidly. Based on the first feature of the recording file, it can be preliminarily determined that the agent service process is in a relatively exciting situation.

[0033] For the second feature extraction of the audio text, the audio text is first preprocessed, including word segmentation, stop word removal, stem extraction, etc., and then feature extraction is performed on the preprocessed audio text.

[0034] As an optional solution for feature extraction of preprocessed audio text, the semantic relationship between words is captured through word embedding. Each word in the audio text is represented as a vector of fixed length through the CBOW (Continuous Bag of Words) model and the Skip-gram model. These vectors have specific directions and distances in space, which can reflect the semantic similarity and correlation between words.

[0035] It should be noted that the CBOW model predicts the target vocabulary through the context, while Skip-gram predicts the context based on the target vocabulary.

[0036] Exemplarily, after extracting the first feature of the recording file, it is preliminarily determined based on the first feature that the agent is in a relatively agitated state during the service process, and the recording file is converted into a recording text. After preprocessing the recording text, the corresponding semantic relationship between words in the recording text is captured through the word embedding method. First, the semantic relationship is predicted by combining the context with the CBOW model. Combined with the fact that the tone of the first feature is at a high level, it can be preliminarily inferred that the agent is in a state of agitation during the service process, and may even be dissatisfied. From the first and second features, it can be concluded that the customer service in the recording text is in an agitated state.

[0037] Step S200: construct a fused audio feature based on the first feature and the second feature.

[0038] First, the extracted first feature and the second feature are preliminarily standardized, and the fused audio feature is constructed by fusing the first feature extracted from the recording file and the second feature extracted from the recording text.

[0039] As an alternative to normalizing the first feature and the second feature, the first feature and the second feature are converted to the same vector length or dimension by aligning the first feature and the second feature in time.

[0040] For example, when converting audio to text using speech recognition technology, the timestamp of each word is marked, and the marked text can be matched during the second feature extraction process. At the same time, when performing the second feature extraction on the text, the same vector length is maintained.

[0041] As an optional solution for constructing fused audio features, the first feature extracted from the recording file and the second feature extracted from the recording text are mapped to a common feature space through machine learning technology, and then the features are fused in this space.

[0042] Exemplarily, a neural network model is constructed, which has two input branches. The two input branches of the model are used as feature input ports of the recording file and the recording text respectively, and the first feature and the second feature are received. The distance between them is mapped to the same feature space by minimizing the contrast loss function. In the common feature space, the first feature and the second feature are fused by simple concatenation, weighted summation or attention mechanism.

[0043] Optionally, the fused feature vector is trained through a speech recognition model (Recurrent Neural Network, RNN) to optimize model parameters, and the fused audio features are adjusted and optimized according to the model performance.

[0044] Step S300: determining a classification label based on the classification result corresponding to the fused audio feature.

[0045] Through a preset classification model, the fused audio fusion features are used as a data set. According to the characteristics of the data, including the type and structure, scale and distribution of the data, a suitable classification model is selected. It can be a traditional machine learning model, including logistic regression, support and decision trees, etc., or a deep learning model, including neural networks, convolutional neural networks and recurrent neural networks, etc. The fused data features are classified through the classification model, and the model outputs the probability of one or more classification labels, and the classification label with the highest probability in each input sample is output as the classification result.

[0046] As an alternative to classifying the fused data features through a classification model, first create a data loader, pass the fused audio features as the data input of the model, perform forward propagation through the model to output a probability distribution of one or more classification labels, and then use the softmax function to extract the probability of the classification label from the output of the model, and select the classification label with the highest probability as the prediction result for each input sample.

[0047] Optionally, the calculated classification label with the highest probability is used as the prediction result to determine the output classification label.

[0048] Step S400, associating the classification label with the identifier of the audio recording text, and indexing the identifier with the storage location of the audio recording text.

[0049] After the recorded text is stored, the identifier corresponding to the generated text is associated with the text label, and the identifier is used as a foreign key to establish an index relationship with the storage location of the recorded text.

[0050] It should be noted that an identifier is a unique code or serial number used to identify and distinguish each audio recording in a database or file system. An index usually refers to a pointer to a data storage location in a database so that the data can be quickly accessed.

[0051] As an alternative to associating the identifier corresponding to the text with the text label and using the identifier as a foreign key to index the storage location of the recording text, design a relational database and create a master table to store the recording text and its metadata. Design the table structure, including fields such as identifier, label, storage_location, etc. Associate the classification label with the identifier of the recording text in the database. Use the identifier as a foreign key to establish an index relationship with the storage location.

[0052] For example, after the received customer service recording is tagged and classified, the tag is associated with the identifier Test01 of the recording text. At the same time, the identifier is associated with the location where the recording text is stored, such as https: / / localstorge.com / recordings / Test01, and the quality inspector can quickly retrieve the corresponding recording text and file storage location through the tag.

[0053] This embodiment provides a method for managing recording text tags, by extracting features from recording files, extracting second features from recording texts, and then fusing the extracted first features and second features. The fused audio features can better express the overall audio features of the recording files, and more accurately match the labels corresponding to the recording files. When quality inspectors detect recording files, they can accurately find the corresponding recording files based on the labels, greatly improving the detection rate.

[0054] Based on the first embodiment, the second embodiment of the present application provides a method for managing recording text marks, referring to Figure 2 , Figure 2 This is a flow chart of Embodiment 2 of the recording text tag management method of this application, step S200 includes steps S210 to S220: Step S210, standardizing the first feature to determine a first standard feature, and standardizing the second feature to determine a second standard feature.

[0055] Step S220: Fusing the first standard feature and the second standard feature to determine the fused audio feature.

[0056] The audio features extracted from the recording file are preprocessed, and then the mean value is subtracted from each eigenvalue to make the center of the eigenvalue 0, and each eigenvalue is divided by its standard deviation to make the distribution of the eigenvalue have unit variance. For the recording text, the L2 normalization method is used for standardization. After standardization, the first standard feature and the second standard feature are obtained, and the first standard feature and the second standard feature are fused to determine the fused audio feature.

[0057] For example, L2 normalization is used to process two word vectors whose original lengths are sqrt(12+32+52)=sqrt(77). After L2 normalization, these vectors are converted to (1 / sqrt(14), 2 / sqrt(14), 3 / sqrt(14)) and (4 / sqrt(77), 5 / sqrt(77), 6 / sqrt(77)), scaling the length to 1.

[0058] Optionally, step S220 includes steps S221-S222: Step S221, determining a weight allocation strategy based on the correlation between the first standard feature and the second standard feature.

[0059] Step S222: Based on the weight allocation strategy, weighted sum of the first standard feature and the second standard feature is performed to determine the fused audio feature.

[0060] First, it is necessary to quantify the similarity between the first standard feature and the second standard feature, evaluate the importance of the first standard feature and the second standard feature, and assign weights to the features.

[0061] As an alternative to quantifying the first standard feature and the second standard feature, the similarity is quantified by calculating the cosine value of the angle between the two features.

[0062] As an optional scheme for assigning weights to features, for highly correlated features, if their importance to the target variable is similar, one of them is selected as the representative and assigned a high weight; for features with low correlation but each contributing to the target variable, a medium weight is assigned; for features that perform well in both correlation analysis and importance, a high weight is assigned.

[0063] In this embodiment, the original first feature is converted into the first standard feature through standardization processing, and the second feature is converted into the second standard feature, and then the first standard feature and the second standard feature are fused.

[0064] This embodiment provides a method for managing audio recording text tags, which converts the original first feature into a first standard feature through standardization processing, and also converts the second feature into a second standard feature. The comparability of these features in terms of values ​​is determined, and the advantages of audio and text information are combined to determine the fused audio feature.

[0065] Based on Example 1, Example 3 of the present application provides a method for managing recording text marks, referring to Figure 3 , Figure 3 This is a flow chart of Embodiment 3 of the recording text tag management method of this application. Before step S200, it also includes steps S180 to S190. Figure 3 The flowchart of steps S180 to S190 is as follows: Step S180, constructing a third feature based on customer complaint information and call status.

[0066] In the recording file tag management system, complaints are divided into different types according to the content of the complaints, the number of complaints made by customers within a certain period of time is counted, the duration of each call is recorded, and the number of calls made by customers within a certain period of time is counted. Thus, the third feature is constructed based on the complaint situation and the call situation.

[0067] As an optional solution to construct the third feature through complaint status and call status, the agent service status is determined based on the customer complaint content and call records, and the correlation and consistency between customer complaints and calls are identified.

[0068] Exemplarily, after receiving a customer's complaint record, the system prioritizes identifying and processing the complaint record, wherein the complaint may be multiple missed calls from customers, long time intervals between responses to customers, interruptions to customers, etc.

[0069] Specifically, after step S180 constructs the third feature, step S190 is further included: Step S190: construct the fused audio feature according to the first feature, the second feature and the third feature.

[0070] The fusion feature of the three is determined by combining the third feature with the first feature of the audio analysis and the second feature of the recorded text analysis, and at the same time, a fused audio feature is jointly constructed.

[0071] Optionally, the third feature is obtained by combining the customer complaint situation and the call situation, and is combined with the extracted first feature and second feature to jointly construct the third feature.

[0072] As an optional implementation method for jointly constructing fused audio features, the information provided by the customer during the complaint process is collected, including the specific content of the complaint, which can be a business complaint or an attitude complaint based on the customer's complaint to the agent, and the agent's call status is collected, which can be multiple hang-ups and multiple missed customer calls.

[0073] Furthermore, based on the content and category of the complaint, the emotional state during the complaint process is extracted through text analysis technology, and a fused emotional state feature is constructed by combining the first feature and the second feature.

[0074] Exemplarily, customer complaint records are collected, and based on the third feature described in the complaint record that the agent's response time interval is long, the tone and volume of the recording file are too low, and the content of the recording text is discontinuous, the fused audio feature is determined to indicate that the customer service is in a silent state.

[0075] This embodiment provides a recording text tag management method, which integrates the first feature, the second feature and the third feature. By comprehensively considering these three features, a more comprehensive fusion feature is determined, which reflects the audio attributes of the audio and the voice information of the text content, and also incorporates the agent's behavior pattern and problem-solving efficiency during the customer call process.

[0076] Based on the same inventive concept, the fourth embodiment of the present application provides a method for managing recording text marks, referring to Figure 4 , Figure 4 This is a flowchart of Embodiment 4 of the recording text tag management method of this application, step S300 includes steps S310~S320: Step S310: input the fused audio features into a preset classification model to determine the classification result.

[0077] Step S320: output at least one classification label based on the classification result.

[0078] The fused audio features are input into a preset classification model, the input data is classified by the preset classification model, the classification result is output, and at least one classification label is output according to the classification result.

[0079] It should be noted that the classification result is a judgment on the category to which the audio signal belongs, which can be a discrete category label or a probability distribution.

[0080] As an optional solution for generating a preset classification model, audio data containing different tones, volumes, and speaking speeds are collected, and the audio data is manually labeled to indicate its characteristics such as pitch, volume, and speaking speed. A vector machine is used to identify the pitch, volume, and speaking speed, and the labeled audio data set is used to train the model. The model will learn how to map the characteristics of the audio signal to labels such as pitch, volume, and speaking speed.

[0081] Furthermore, the fused audio features are classified using a preset classification model, and the model is able to identify abnormalities in the pitch, volume, and speech speed of the first feature.

[0082] Similarly, ASR technology is used to convert the audio file into text, and the audio text is semantically annotated, such as part-of-speech tagging, named entity recognition, and semantic role labeling. The text is then converted into a vector representation using the word vector model (Bidirectional Encoder Representations from Transformers, BERT). The text is then encoded using a deep learning model to capture its semantic features. Finally, a suitable classification model is selected to input the annotated data set for training. Optionally, the classification model includes vector machines, random forests, neural networks, etc.

[0083] Optionally, the recorded text of the agent service is annotated based on previous customer complaint information and call situations, and the annotated training set is input into the training model for training, so that the training model can be trained based on customer complaint information and call situations.

[0084] It should be noted that the preset classification model is a single model, and the training data is simultaneously input into the model with multiple dimensional data for training, so that the number of identifiable features is increased, and a highly integrated preset classification model is obtained.

[0085] This embodiment provides a method for recording file tagging management, which inputs the fused audio features into a preset classification model, performs complex calculations and analyses based on the various dimensional information in its feature vector, and thus determines the classification category to which the feature vector belongs. Based on the classification results, the system outputs at least one classification label, providing customer service personnel with important clues for quickly identifying customer problems and needs.

[0086] Based on the same inventive concept, the fifth embodiment of the present application provides a method for managing recording text marks, referring to Figure 5 , Figure 5 This is a flowchart of Embodiment 5 of the recording text tag management method of this application, step S100 includes steps S110~S120: Step S110: extracting the first feature of the recording file based on the audio attributes of the recording file, wherein the audio attributes include spectral characteristics, frequency, audio duration and audio level.

[0087] By analyzing the audio properties of the recording file, the spectral characteristics, frequency, audio duration and audio level corresponding to the recording file are determined.

[0088] It should be noted that the spectral characteristics are the energy distribution of audio signals at different frequencies, which are used to distinguish different types of audio signals, including speech, music, noise, etc.; frequency represents the number of times the signal vibrates per unit time, and the frequency information can be a high pitch with a high vibration frequency, and a low pitch with a low vibration frequency; the audio duration is the total duration of the recording file, which is useful for long sentence detection and timestamp marking of the recording file; the audio level refers to the strength and loudness of the audio signal, expressed in decibels (dB), which is used to quantify the ratio of the power and amplitude of the sound signal relative to a reference value.

[0089] As an optional solution for analyzing the audio properties of a recording file and determining the spectral characteristic frequency, audio duration and audio level corresponding to the recording file, the spectral characteristics are obtained by dividing the audio signal into several small segments through short-time Fourier transform, and Fourier transform is performed on each small segment to obtain the spectral information that changes with time.

[0090] Similarly, for frequency acquisition, the frequency with the largest amplitude in the Fourier transform result is used to obtain the main frequency, and the weighted average of the Fourier transform result is used to obtain the average frequency. It should be noted that the main frequency is the strongest frequency component in the audio signal, and the average frequency is the average of all frequency components in the audio signal.

[0091] The duration of the audio signal can be obtained by calculating the duration of the audio signal.

[0092] The audio level is obtained by calculating the peak amplitude of the audio signal, which can reflect the strength of the audio signal.

[0093] Step S120: Recognize the recorded text based on natural language processing technology and extract the second feature.

[0094] After extracting the audio features of the recording file, the recording file needs to be converted into recording text, and the text content of the recording text needs to be recognized and analyzed to extract the second feature.

[0095] As a solution for converting audio files into audio text, sound signals are first collected through a microphone or other audio equipment, and the collected audio signals are preprocessed, including noise reduction, echo removal, and voice signal enhancement. The preprocessed audio signals are input into a speech recognition model through speech-to-text technology (Automatic Speech Recognition, ASR), the speech features are identified, and then these features are converted into text form. Optionally, the recognized text is post-processed, such as correcting spelling errors and adjusting grammatical structures, to improve the readability and accuracy of the text.

[0096] As an optional solution for analyzing the text content of the recording and extracting the second feature, the word embedding method uses a neural network to map words to a high-dimensional vector space, so that similar words appear closer in the vector space, thereby capturing the semantic relationship between words.

[0097] Optionally, a pre-trained language model based on the Transfermer architecture can capture the contextual information of words and generate more accurate word embeddings.

[0098] As a second feature extraction method based on syntactic and semantic analysis, a syntactic parser is used to convert the text into a syntactic tree to extract the structural information of the sentence, such as subject, predicate, object, etc. On the basis of syntactic analysis, the semantic role of the sentence is further identified.

[0099] For example, for the text "Zhang San attended the meeting", the word segmentation and annotation are performed, "Zhang San / NR (proper noun)", "Participate / VV (verb)", "Late / AS (particle)", "Meeting / NN (noun)". Based on the syntactic analysis, semantic role annotation can be further performed to identify the semantic roles in the sentence, and the annotation "Participate" is a verb, with "Agent (the executor of the action)", the agent is "Zhang San", "Patient (the recipient of the action)", the patient is "meeting", and "Late" indicates the voice, indicating that the action has been completed.

[0100] As an optional solution to extract the second feature based on sentiment analysis and emotion recognition, sentiment dictionary or machine information model is used to analyze the sentiment tendency of text and extract the sentiment features of the text. Sentiment features can reflect the emotional color of the text, such as positive, negative and neutral.

[0101] This embodiment provides a method for managing recording file tagging, which extracts a first feature based on the audio attributes of the recording file; at the same time, natural language processing technology is used to identify the text content of the recording, and a second feature is extracted to understand the semantic information in the recording, which helps to comprehensively analyze the recording file and provide key data support for subsequent applications such as speech analysis, emotion recognition, and content retrieval.

[0102] Based on the same inventive concept, the sixth embodiment of the present application provides a method for managing recording text marks, referring to Figure 6 , Figure 6 This is a flowchart of Embodiment 6 of the recording text tag management method of this application, step S400 includes steps S410~S420: Step S410: generating the identifier of the recorded text based on a preset algorithm based on text features of the recorded text, wherein the text features include a text name and a creation time.

[0103] Step S420, associating the identifier with the classification label, and establishing the index with the storage location.

[0104] It should be noted that text features refer to the attributes of the text itself and have nothing to do with the text content.

[0105] As an optional implementation, text features are extracted from the audio recording, and a preset algorithm, such as a hash function, is used to generate a corresponding identifier through the extracted text features. The generated identifier is then associated with the classification label of the audio recording (such as "customer consultation", "technical support") in the database, and an index relationship is established between the identifier and the storage location of the audio recording (such as a file path or a database record ID).

[0106] For example, customer service calls are automatically transcribed into text and stored in a database. Each piece of text has a name (such as "Call_01") and a creation time (such as "2024-10-28 15:00"). The system uses these features to generate a hash value as an identifier (such as "h123456789") and associates this identifier with the classification label "Customer Consultation". At the same time, the system creates an index for this identifier in the database and associates it with the local database, which can be " / local_storage / calls / 2024 / 10 / Call_01.txt".

[0107] This embodiment provides a method for managing audio recording text marking, which generates a unique identifier for each audio recording text through a preset algorithm. The identifier is based on the text features of the audio recording text, and then associates the audio recording text with a corresponding classification label to reflect the content and attributes of the text. The identifier is indexed with the storage location to achieve rapid positioning and access to specific audio recording files in the audio recording file database, thereby improving the efficiency of data management.

[0108] Optionally, in this embodiment, after step S420, steps S430 to S440 are further included, referring to Figure 7 , Figure 7 The flowchart of steps S430 to S440 is as follows: Step S430, in response to the query operation of the classification tag, display the storage location of the recorded text corresponding to the classification tag.

[0109] The user submits a query operation through the interface or API, which includes the category label to be queried. After receiving the query operation submitted by the user, the audio text tag management system parses the category label and executes the query operation in the database. After matching the identifier corresponding to the category label, it locates the location identifier of the audio text. The system constructs this location identifier into a response message and returns it to the user through the interface display.

[0110] It should be noted that a query operation refers to a request made by a user or system to a database or data storage for the purpose of retrieving, searching or obtaining information matching a specific condition. In this embodiment, the query operation involves searching for relevant audio text records according to classification tags.

[0111] Step S440, in response to a click operation on the storage location, displaying the text content of the recorded text.

[0112] After obtaining the connection, button or clickable element of the storage location corresponding to the classification label input by the user, the system receives the user click event, parses the storage location information clicked, thereby locating the database record corresponding to the recorded text, and then renders the text content to the display interface for the user to view.

[0113] It should be noted that a click operation refers to an action instruction given by a user to an operating system using an external device, which may be closing a window or opening a window, etc.

[0114] As an optional implementation, after receiving the user's query operation, the corresponding recording text is matched according to the tag input by the user, and the storage location of the recording text is displayed on the client operated by the user. The user finds the corresponding recording text according to the displayed storage location, and the content of the recording text is displayed.

[0115] As an optional implementation, when a user submits a query request for a classification tag, the backend executes a database query and returns the storage location of the recorded text associated with the tag. After the user clicks on the storage location, the system retrieves the corresponding text content from the database and displays it on the interface.

[0116] For example, the quality inspector enters the classification tag of "bad review", finds multiple bad review recordings, and then enters "violation" to determine that the operation is not standardized and obtain a recording of the bad review seat. The web interface displays a list of storage locations for recordings related to this tag. After clicking on one of the locations, the system retrieves and displays the content of the recording from the database.

[0117] This embodiment provides a method for managing audio recording text tags, which enables users to query based on classification tags. After responding to this operation, the system displays the storage locations of all audio recording texts associated with the classification tags. Then, when the user clicks on a specific storage location, the system further displays the specific content of the corresponding audio recording text. This effectively improves the convenience and intuitiveness of data retrieval, while allowing users to quickly obtain and view audio recording text information, thereby promoting effective management of audio recording texts.

[0118] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the recording text marking management method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0119] The present application provides a recording text marking management device, which includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the recording text marking management method in the above-mentioned embodiment 1.

[0120] Reference below Figure 8 , which shows a schematic diagram of the structure of a recording text mark management device suitable for implementing the embodiment of the present application. The recording text mark management device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8 The recorded text tag management device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0121] like Figure 8As shown, the recording text mark management device may include a processing device 1001 (e.g., a core processor, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the recording text mark management device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the recording text mark management device to communicate with other devices wirelessly or by wire to exchange data. Although the recording text mark management device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0122] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0123] The recording text mark management device provided by the present application adopts the recording text mark management method in the above embodiment, which can solve the technical problem of recording text mark management. Compared with the prior art, the beneficial effects of the recording text mark management device provided by the present application are the same as the beneficial effects of the recording text mark management method provided by the above embodiment, and other technical features in the recording text mark management device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0124] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0125] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0126] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the recording text marking management method in the above-mentioned embodiment.

[0127] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM: Random Access Memory), a read-only memory (ROM: Read Only Memory), an erasable programmable read-only memory (EPROM: Erasable Programmable Read Only Memory or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM: CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency: Radio Frequency), etc., or any suitable combination of the above.

[0128] The computer-readable storage medium may be included in the recording text mark management device; or may exist independently without being assembled into the recording text mark management device.

[0129] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the recording text mark management device, the recording text mark management device: Constructing a first feature according to the recording file, and constructing a second feature according to the recording text corresponding to the recording file; Constructing a fused audio feature according to the first feature and the second feature; Determining a classification label based on the classification result corresponding to the fused audio feature; The classification tag is associated with an identifier of the audio recording text, and the identifier is indexed with a storage location of the audio recording text.

[0130] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may 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 may 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 may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0131] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0132] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0133] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned recording text mark management method, and can solve the technical problem of recording text mark management. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the recording text mark management method provided in the above-mentioned embodiment, and will not be described in detail here.

[0134] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for managing recording text markings, It is characterized in that The recording text mark management method comprises the following steps: Constructing a first feature according to the recording file, and constructing a second feature according to the recording text corresponding to the recording file; Constructing a fused audio feature according to the first feature and the second feature; Determining a classification label based on the classification result corresponding to the fused audio feature; The classification tag is associated with an identifier of the audio recording text, and the identifier is indexed with a storage location of the audio recording text.

2. The method for managing recording text marks according to claim 1, It is characterized in that The step of constructing a fused audio feature according to the first feature and the second feature includes: Standardizing the first feature to determine a first standard feature, and standardizing the second feature to determine a second standard feature; The first standard feature and the second standard feature are fused to determine the fused audio feature.

3. The method for managing recording text marks according to claim 2, It is characterized in that The step of fusing the first standard feature and the second standard feature to determine the fused audio feature comprises: Determining a weight allocation strategy based on the correlation between the first standard feature and the second standard feature; Based on the weight allocation strategy, the first standard feature and the second standard feature are weightedly summed to determine the fused audio feature.

4. The method for managing recording text marks according to claim 1, It is characterized in that The step of determining a classification label based on the classification result corresponding to the fused audio feature comprises: Inputting the fused audio features into a preset classification model to determine the classification result; Based on the classification result, at least one classification label is output.

5. The method for managing recording text marks according to claim 1, It is characterized in that The step of constructing the first feature according to the recording file, and constructing the second feature according to the recording text corresponding to the recording file, comprises: Extracting the first feature of the recording file based on the audio attributes of the recording file, the audio attributes including: spectral characteristics, frequency, audio duration and audio level; The recorded text is recognized based on natural language processing technology to extract the second feature.

6. The method for managing recording text marks according to claim 1, It is characterized in that Before the step of constructing a fused audio feature according to the first feature and the second feature, the method further includes: Construct the third feature based on customer complaint information and call conditions; The step of constructing a fused audio feature according to the first feature and the second feature comprises: The fused audio feature is constructed according to the first feature, the second feature and the third feature.

7. The method for managing recording text marks according to claim 1, It is characterized in that The step of associating and saving the classification tag with the recording file to read the recording file based on the classification tag includes: Based on the text features of the audio recording text, generating the identifier of the audio recording text based on a preset algorithm, the text features including the text name and creation time; The identifier is associated with the classification label and the index is established with the storage location.

8. The method for managing recording text marks according to claim 7, It is characterized in that After the step of associating the identifier with the classification label and establishing the index with the storage location, the method further includes: In response to the query operation of the classification tag, displaying the storage location of the audio recording text corresponding to the classification tag; In response to a click operation on the storage location, the text content of the recorded text is displayed.

9. A recording text tag management device, It is characterized in that The recorded text mark management device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the recorded text mark management method according to any one of claims 1 to 8.

10. A storage medium, It is characterized in that The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the recording text marking management method according to any one of claims 1 to 8 are implemented.

Images