CN111241815A - Text increment method and device and terminal equipment - Google Patents

Abstract

The present application is applicable to the technical field of natural language processing, and provides a text increment method, including: acquiring text to be incremented; performing feature extraction on the text to be incremented, and obtaining a feature matrix corresponding to the text to be incremented; determining the text topic of the text to be incremented; inputting the feature matrix into a variational autoencoder corresponding to the text topic to obtain the incremental text of the text to be incremented. The present application improves the correlation between the incremental text and the text to be incremented, thereby greatly improving the accuracy of the generated text.

Description

Text increment method, device and terminal device

technical field

The present application belongs to the technical field of natural language processing, and in particular, relates to a text increment method, an apparatus, a terminal device, and a computer-readable storage medium.

Background technique

Currently, in many artificial intelligence fields such as question answering systems, machine translation, etc., there is a need to generate other text data based on original text data. For example, in a human-machine question answering system, when a user asks a robot, the robot's answer must be related to the user's question, that is, the answer text data generated by the robot is required to be associated with the text data asked by the user.

However, the challenge faced by traditional text generation models is that the generated text is too random. Therefore, it is urgent to provide a new text increment scheme.

SUMMARY OF THE INVENTION

Embodiments of the present application provide a text increment method, apparatus, terminal device, and computer-readable storage medium, provide a new text increment scheme, and improve the correlation between incremented text and to-be- incremented text.

In a first aspect, an embodiment of the present application provides a text increment method, including:

Get the text to be incremented;

Perform feature extraction on the text to be incremented to obtain a feature matrix corresponding to the text to be incremented;

determining the text topic of the text to be incremented;

Inputting the feature matrix into a variational autoencoder corresponding to the text topic to obtain the incremental text of the text to be incremented.

In a second aspect, an embodiment of the present application provides a text increment device, including:

Get module, used to get the text to be incremented;

an extraction module, configured to perform feature extraction on the text to be incremented to obtain a feature matrix corresponding to the text to be incremented;

a determining module, used for determining the text topic of the text to be incremented;

an increment module, configured to input the feature matrix into a variational autoencoder corresponding to the text topic to obtain an increment text of the text to be incremented.

In a third aspect, an embodiment of the present application provides a terminal device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor executes the computer program When implementing the text increment method described in the first aspect.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the text increment according to the first aspect is implemented method.

In a fifth aspect, an embodiment of the present application provides a computer program product, which, when the computer program product runs on a terminal device, enables the terminal device to execute the text increment method described in the first aspect.

In the embodiment of the present application, by first extracting the feature matrix of the text to be incremented, the text subject of the text to be incremented is determined, and then the increment text is generated in combination with the VAE corresponding to the text subject. On the one hand, the VAE corresponding to the text topic is used to generate incremental text, and different topics are set to a different VAE; on the other hand, since the distribution calculated by the VAE depends on the input variables, all sampling of this distribution will be generated and input. Similar or related outputs can help to achieve certainty when generating text. Therefore, through the dual effects of these two aspects, the complete randomness when generating text is avoided, and the correlation between the incremental text and the text to be incremented is improved. This can greatly improve the quality of text generation.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present application more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only for the present application. In some embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without any creative effort.

1 is a schematic structural diagram of a mobile phone to which a text increment method provided by an embodiment of the present application is applicable;

2 is a schematic flowchart of a text increment method provided by an embodiment of the present application;

3 is a schematic flowchart of step 202 in the text increment method provided by an embodiment of the present application;

4 is a schematic structural diagram of a VAE in a text increment method provided by an embodiment of the present application;

5 is a schematic structural diagram of a text increment device provided by an embodiment of the present application;

FIG. 6 is a schematic structural diagram of a terminal device to which the text increment method provided by an embodiment of the present application is applicable.

Detailed ways

In the following description, for the purpose of illustration rather than limitation, specific details such as a specific system structure and technology are set forth in order to provide a thorough understanding of the embodiments of the present application.

In order to enable those skilled in the art to better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only The embodiments are part of the present application, but not all of the embodiments. Based on the embodiments in the present application, for those of ordinary skill in the art, all other embodiments obtained under the premise of no creative labor should fall within the protection scope of the present application. It should be noted that the embodiments in the present application and the features of the embodiments may be combined with each other in the case of no conflict.

In the following description, for the purpose of illustration rather than limitation, specific details such as a specific system structure and technology are set forth in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be practiced in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

It is to be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described feature, integer, step, operation, element and/or component, but does not exclude one or more other The presence or addition of features, integers, steps, operations, elements, components and/or sets thereof.

It will also be understood that, as used in this specification and the appended claims, the term "and/or" refers to and including any and all possible combinations of one or more of the associated listed items.

As used in the specification of this application and the appended claims, the term "if" may be contextually interpreted as "when" or "once" or "in response to determining" or "in response to detecting ". Similarly, the phrases "if it is determined" or "if the [described condition or event] is detected" may be interpreted, depending on the context, to mean "once it is determined" or "in response to the determination" or "once the [described condition or event] is detected. ]" or "in response to detection of the [described condition or event]".

In addition, in the description of the specification of the present application and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the description, and should not be construed as indicating or implying relative importance.

References in this specification to "one embodiment" or "some embodiments" and the like mean that a particular feature, structure or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in other embodiments," etc. in various places in this specification are not necessarily All refer to the same embodiment, but mean "one or more but not all embodiments" unless specifically emphasized otherwise. The terms "including", "including", "having" and their variants mean "including but not limited to" unless specifically emphasized otherwise.

The text increment method provided by the embodiments of the present application can be applied to mobile phones, tablet computers, wearable devices, in-vehicle devices, augmented reality (AR)/virtual reality (VR) devices, notebook computers, and super mobile personal On terminal devices such as a computer (ultra-mobile personal computer, UMPC), a netbook, a personal digital assistant (PDA), or a server, the embodiments of the present application do not limit the specific type of the terminal device. The servers include but are not limited to independent servers, cloud servers, distributed servers, and server clusters.

For example, the terminal device may be a station (STAION, ST) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (Session Initiation Protocol, SIP) phone, a Wireless Local Loop (WLL) station, a PDA , handheld devices with wireless communication capabilities, computing devices or other processing devices connected to wireless modems, in-vehicle devices, connected car terminals, computers, laptop computers, handheld communication devices, handheld computing devices, satellite wireless devices, wireless Modem cards, set top boxes (STBs), customer premise equipment (CPEs) and/or other equipment for communicating over wireless systems and next generation communication systems, e.g. mobile in 5G networks A terminal or a mobile terminal in a public land mobile network (PublicLand Mobile Network, PLMN) network that evolves in the future, and the like.

As an example and not a limitation, when the terminal device is a wearable device, the wearable device may also be a general term for the intelligent design of daily wear and the development of wearable devices using wearable technology, such as glasses, gloves, Watches, clothing and shoes, etc. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothing or accessories. Wearable device is not only a hardware device, but also realizes powerful functions through software support, data interaction, and cloud interaction. In a broad sense, wearable smart devices include full-featured, large-scale, complete or partial functions without relying on smart phones, such as smart watches or smart glasses, and only focus on a certain type of application function, which needs to be used in conjunction with other devices such as smart phones. , such as various types of smart bracelets and smart jewelry that monitor physical signs.

Take the terminal device as a mobile phone as an example. FIG. 1 is a block diagram showing a partial structure of a mobile phone provided by an embodiment of the present application. Referring to FIG. 1 , the mobile phone includes: a radio frequency (RF) circuit 110 , a memory 120 , an input unit 130 , a display unit 140 , a sensor 150 , an audio circuit 160 , a wireless fidelity (WiFi) module 170 , and a processor 180 , and components such as the power supply 190 . Those skilled in the art can understand that the structure of the mobile phone shown in FIG. 1 does not constitute a limitation on the mobile phone, and may include more or less components than the one shown, or combine some components, or arrange different components.

The following describes the various components of the mobile phone in detail with reference to Figure 1:

The RF circuit 110 can be used for receiving and sending signals during sending and receiving of information or during a call. In particular, after receiving the downlink information of the base station, it is processed by the processor 180; in addition, the designed uplink data is sent to the base station. Typically, the RF circuit includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a Low Noise Amplifier (LNA), a duplexer, and the like. In addition, the RF circuitry 110 may also communicate with networks and other devices via wireless communication. The above-mentioned wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile communication (GSM), General Packet Radio Service (General Packet Radio Service, GPRS), Code Division Multiple Access (Code Division Multiple Access) , CDMA), Wideband Code Division Multiple Access (Wideband Code Division Multiple Access, WCDMA), Long Term Evolution (Long Term Evolution, LTE), email, and Short Messaging Service (Short Messaging Service, SMS) and the like.

The memory 120 can be used to store software programs and modules, and the processor 180 executes various functional applications and data processing of the mobile phone by running the software programs and modules stored in the memory 120 . The memory 120 may mainly include a stored program area and a stored data area, wherein the stored program area may store an operating system, an application program required for at least one function (such as a sound playback function, an image playback function, etc.), a boot loader (Boot Loader) etc.; the storage data area can store data (such as audio data, phone book, etc.) created according to the use of the mobile phone, and the like. Additionally, memory 120 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. It can be understood that, in this embodiment of the present application, a program for text increment is stored in the memory 120 .

The input unit 130 may be used to receive input numerical or character information, and generate key signal input related to user settings and function control of the mobile phone 100 . Specifically, the input unit 130 may include a touch panel 131 and other input devices 132 . The touch panel 131, also referred to as a touch screen, can collect the user's touch operations on or near it (such as the user's finger, stylus, etc., any suitable objects or accessories on or near the touch panel 131 ). operation), and drive the corresponding connection device according to the preset program. Optionally, the touch panel 131 may include two parts, a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch orientation, detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into contact coordinates, and then sends it to the touch controller. To the processor 180, and can receive the commands sent by the processor 180 and execute them. In addition, the touch panel 131 can be implemented in various types such as resistive, capacitive, infrared, and surface acoustic waves. Besides the touch panel 131 , the input unit 130 may also include other input devices 132 . Specifically, other input devices 132 may include, but are not limited to, one or more of physical keyboards, function keys (such as volume control keys, switch keys, etc.), trackballs, mice, joysticks, and the like.

The display unit 140 may be used to display information input by the user or information provided to the user and various menus of the mobile phone. The display unit 140 may include a display panel 141, and optionally, the display panel 141 may be configured in the form of a liquid crystal display (Liquid Crystal Display, LCD), an organic light-emitting diode (Organic Light-Emitting Diode, OLED), or the like. Further, the touch panel 131 may cover the display panel 141, and when the touch panel 131 detects a touch operation on or near it, it transmits it to the processor 180 to determine the type of the touch event, and then the processor 180 determines the type of the touch event according to the touch event. Type provides corresponding visual output on display panel 141 . Although in FIG. 1, the touch panel 131 and the display panel 141 are used as two independent components to realize the input and input functions of the mobile phone, in some embodiments, the touch panel 131 and the display panel 141 can be integrated to form a Realize the input and output functions of the mobile phone.

The cell phone 100 may also include at least one sensor 150, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 141 according to the brightness of the ambient light, and the proximity sensor may turn off the display panel 141 and/or when the mobile phone is moved to the ear. or backlight. As a kind of motion sensor, the accelerometer sensor can detect the magnitude of acceleration in all directions (usually three axes), and can detect the magnitude and direction of gravity when it is stationary. games, magnetometer attitude calibration), vibration recognition related functions (such as pedometer, tapping), etc.; as for other sensors such as gyroscope, barometer, hygrometer, thermometer, infrared sensor, etc. Repeat.

The audio circuit 160, the speaker 161, and the microphone 162 can provide an audio interface between the user and the mobile phone. The audio circuit 160 can transmit the received audio data converted electrical signal to the speaker 161, and the speaker 161 converts it into a sound signal for output; on the other hand, the microphone 162 converts the collected sound signal into an electrical signal, which is converted by the audio circuit 160 After receiving, it is converted into audio data, and then the audio data is output to the processor 180 for processing, and then sent to, for example, another mobile phone through the RF circuit 110, or the audio data is output to the memory 120 for further processing.

WiFi is a short-distance wireless transmission technology, and the mobile phone can help users to send and receive emails, browse web pages and access streaming media through the WiFi module 170, which provides users with wireless broadband Internet access. Although FIG. 1 shows the WiFi module 170, it can be understood that it is not a necessary component of the mobile phone 100, and can be completely omitted as required within the scope of not changing the essence of the invention.

The processor 180 is the control center of the mobile phone, using various interfaces and lines to connect various parts of the entire mobile phone, by running or executing the software programs and/or modules stored in the memory 120, and calling the data stored in the memory 120. Various functions of the mobile phone and processing data, so as to monitor the mobile phone as a whole. Optionally, the processor 180 may include one or more processing units; preferably, the processor 180 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, etc. , the modem processor mainly deals with wireless communication. It can be understood that, the above-mentioned modulation and demodulation processor may not be integrated into the processor 180 . It can be understood that, in this embodiment of the present application, the memory 120 stores a program of text increment, and the processor 180 can be configured to call and execute the program of text increment stored in the memory 120, so as to realize the program of this embodiment of the present application. Text increment method.

The mobile phone 100 also includes a power source 190 (such as a battery) for supplying power to various components. Preferably, the power source can be logically connected to the processor 180 through a power management system, so as to manage charging, discharging, and power consumption management functions through the power management system.

Although not shown, the cell phone 100 may also include a camera. Optionally, the position of the camera on the mobile phone 100 may be front-mounted, rear-mounted, or built-in (can extend out of the body when in use), which is not limited in this embodiment of the present application.

Optionally, the mobile phone 100 may include a single camera, a dual camera, or a triple camera, etc., which is not limited in this embodiment of the present application. Cameras include, but are not limited to, wide-angle cameras, telephoto cameras, or depth cameras.

For example, the mobile phone 100 may include three cameras, wherein one is a main camera, one is a wide-angle camera, and one is a telephoto camera.

Optionally, when the mobile phone 100 includes multiple cameras, the multiple cameras may be all front-facing, or all rear-facing, or all built-in, or at least partially front-facing, or at least partially rear-facing, or at least partially built-in, etc. This is not limited in the application examples.

In addition, although not shown, the mobile phone 100 may also include a Bluetooth module, etc., which will not be described herein again.

FIG. 2 shows an implementation flowchart of a text increment method provided by an embodiment of the present application. The text increment method is applied to a terminal device. As an example and not a limitation, the text increment method can be applied to the mobile phone 100 having the above-mentioned hardware structure. The text increment method includes steps S201 to S204, and the specific implementation principles of each step are as follows.

S201, acquiring the text to be incremented.

In the embodiment of the present application, the text to be incremented is an object for text increment, such as sentence text and the like.

The text to be incremented can be the text that the user instantly enters through the input unit of the terminal device; it can also be the voice data that the user instantly collects through the audio collection unit of the terminal device; A picture of text; it can also be a picture including text instantly scanned by the user through the scanning device of the terminal device; it can also be the text originally stored in the terminal device; it can even be the terminal device through a wired or wireless network from other terminal devices The text obtained from the location, etc.

It should be noted that, for pictures including text, it is necessary to extract the text in the picture as the text to be incremented by enabling the picture recognition function of the terminal device; for voice data, it is necessary to activate the audio-to-text function of the terminal device to identify the text in the voice data. Text as the text to be incremented.

In a non-limiting usage scenario of the present application, after the user collects a piece of voice data input by the user's voice through the audio collection unit of the terminal device, the audio-to-text function is enabled to obtain the text input by the user. To perform text increment, the user can enable the text increment function of the terminal device by clicking the specific physical button or virtual button of the terminal device. In this mode, the terminal device will automatically follow steps S202 to S202 to the text input by the user in this mode. The process of S204 is processed to obtain incremental text. It should be noted here that the order in which the user enters the text and clicks the button can be interchanged, that is, the button can be clicked first, then the text input by the user can be obtained, and finally the text input by the user is automatically processed according to the process from step S202 to step S204. .

In another non-limiting usage scenario of the present application, when the user wants to increment the text already stored in the terminal device, the text increment function of the terminal device can be enabled by clicking a specific physical button or virtual button, If the text to be incremented is selected, the terminal device will automatically process the selected text to be incremented according to the process from step S202 to step S204 to obtain the incremented text. It should be noted here that the order of clicking the button and the selected text can be interchanged, that is, the text can be selected first, and then the text increment function of the terminal device can be enabled.

S202: Perform feature extraction on the text to be incremented to obtain a feature matrix corresponding to the text to be incremented.

Step S202 is the step of performing feature extraction on the text to be incremented, obtaining a feature matrix corresponding to the text to be incremented, and implementing a low-dimensional matrix to represent the text.

In some embodiments of the present application, feature extraction may be performed on the text to be incremented by using a word vector model to obtain a feature matrix corresponding to the text to be incremented. That is, the text to be incremented is converted into a feature matrix through the word vector model.

Word vector models include but are not limited to models such as word2vec (word to vector), ELMo, and BERT (Bidirectional Encoder Representation from Transformers). In the embodiment of the present application, through step S202, the word vector model is used to convert the abstract existing text in the real world into a vector or a matrix that can be operated by mathematical formulas. The processing of data into machine-processable data enables this application to be implemented.

It should be noted that, before using the word vector model, the training of the word vector model needs to be completed, and the word vector is generated by pre-training. In addition, in the training process of the word vector model, in order to obtain more accurate feature extraction results, the punctuation marks in the text to be incremented can be retained, and feature extraction is performed on the complete text to be incremented.

Exemplarily, in the training process of the BERT model, in order to be able to perform unsupervised pre-training on a huge dataset, 15% of the words in the training sentences used for training are randomly selected as words to be masked during the training process. Such a masking design is to allow the BERT model to fill in the blanks in the masked positions and achieve unsupervised training.

As a non-limiting example of the present application, step S202 includes: performing feature extraction on the text to be incremented by using a preset BERT model to obtain a feature matrix corresponding to the text to be incremented.

Exemplarily, the text to be incremented is converted into an N×768-dimensional feature matrix by a preset BERT model, and the preset BERT model includes 24 coding layers, that is to say, the BERT Large model is used, The number of Transformer blocks in this model is 24; wherein, the text to be incremented includes N characters, and N is a positive integer.

Each row in the feature matrix corresponds to a foreign character (or a Chinese character) of the text to be incremented. Obviously, the feature matrix can reflect the semantic features of the text to be incremented.

For example, the text to be incremented is "I love my motherland!", the word count of Chinese characters including punctuation marks is 7, and the word count of Chinese characters excluding punctuation marks is 6.

In another example, a BERT model including 12 coding layers, that is, a BERT Base model, is adopted, and the number of Transformer blocks in this model is 12.

As another non-limiting example of the present application, as shown in FIG. 3 , step S202 includes steps S2021 to S2023.

S2021, acquiring the keywords of the text to be incremented.

Among them, for the text to be incremented, firstly perform word segmentation and part-of-speech tagging, and then remove stop words according to a preset stop word dictionary, and remove non-prepositions, orientation words and modal particles according to the part of speech of the word after word segmentation. Feature words, get the keyword set of the text to be incremented.

Step S2021 filters some noisy data by acquiring the keywords of the text to be incremented, so as to ensure the accuracy of the results, and also appropriately reduce the amount of data, improve the processing efficiency, reduce the occupation of system resources, and reduce the cost of computing power.

S2022: Obtain a feature vector corresponding to each of the keywords.

The terminal device pre-stores the corresponding relationship between the keywords and the feature vectors, and obtains the feature vector corresponding to each keyword by searching for the corresponding relationship between the keywords and the feature vectors.

It should be noted that, before step S2022, the corresponding relationship between the keyword and the feature vector is established in advance, and the method for establishing the corresponding relationship is as follows:

First, the corpus crawled from various channels is organized into a document collection through web crawling technology.

Then, use the open source word segmentation tool to perform word segmentation and part-of-speech tagging on each document, and then remove stop words according to the preset stop word dictionary, and remove prepositions, orientation words and modal particles according to the part of speech of the word after segmentation. Non-feature words, get a keyword set.

Finally, the open source word vector training tool Word2Vec is used to train the above keyword sets to obtain feature vectors corresponding to different keywords, and the correspondence between keywords and feature vectors is stored in the word vector database. Exemplarily, each feature vector has the same dimension, and an N-dimensional (N is a positive integer) word vector is used, and the value of each word vector is between 0 and 1, or -1 and 1.

Through the above method, the corresponding relationship between keywords and feature vectors is established. By looking up the corresponding relationship, the feature vector corresponding to the keyword can be obtained, so that each keyword can be converted into a feature vector.

S2023, combine the feature vectors corresponding to all the keywords to generate a feature matrix.

Among them, combining the feature vectors corresponding to all the keywords is to splicing the feature vectors of all the keywords to generate a feature matrix.

For example, when the feature vector is 1×N-dimensional and the preset number is M (M is a positive integer), the feature matrix obtained by combining M 1×N-dimensional feature vectors can be M×N-dimensional or 1× (M+N) dimension.

In some embodiments of the present application, a deep learning network model is used to perform feature extraction on the text to be incremented, and a feature matrix corresponding to the text to be incremented is obtained.

The deep learning network model is used to extract the features of the text to be incremented. When the text to be incremented is input into the deep learning network model, the deep learning network model outputs the feature matrix corresponding to the text to be incremented. The deep learning network model may be a deep learning network model based on machine learning technology, including but not limited to a deep convolutional neural network model, a deep residual convolutional neural network model (Res Net), and the like. Among them, deep convolutional neural network models include but are not limited to AlexNet, VGG-Net, and DenseNet.

It is understandable that, before utilizing the deep learning network model, the training of the deep learning network model needs to be completed. In the process of training a deep learning network model, the loss function used may be one of a 0-1 loss function, an absolute value loss function, a logarithmic loss function, an exponential loss function and a hinge loss function, or at least a combination of the two.

It should be noted that the process of training the model, including the process of training the deep learning network model and the training word vector model, can be implemented on a terminal device, or can be implemented on other terminal devices that are communicatively connected to the terminal device. When the terminal device stores the trained model, or other terminal devices push the trained model to the terminal device, the terminal device can perform feature extraction on the acquired text to be incremented. It should be noted that the text to be incremented obtained by the terminal device during the text increment process can also be used to increase the data in the sample database of the training model, and perform further model optimization on the terminal device or other terminal devices. The terminal device then stores the further optimized model in the terminal device to replace the previous model. In this way, the model is optimized, the data breadth of the model is improved, and the applicable scope of the solution of the present application is improved.

S203: Determine the text topic of the text to be incremented.

In step 203, the text subject of the text to be incremented is determined, so that in the subsequent step S204, the feature matrix is incremented by a variational autoencoder (VAE) corresponding to the text subject.

In some embodiments of the present application, a document topic generation model (Latent Dirichlet Allocation, LDA) is used to identify the text topic of the text to be incremented. LDA is an unsupervised machine learning technique that can be used to identify hidden topic information in large-scale document collections or corpus.

It can be understood that, this is only an exemplary description, and should not be construed as a specific limitation to the present application. All ways of achieving the textual subject matter of the text to be incremented are applicable to this application.

It should be noted that, although step S202 and step S203 are divided into front and back in the description, and the labels are also divided into size, but the difference between the front and rear in the description and the size of the label do not mean that the sequence relationship between the steps is specifically limited. In this embodiment of the present application, step S202 may be performed before step S203, may be performed after step S203, or may be performed simultaneously with step S203, and the present application does not specifically limit the timing relationship between steps S202 and S203.

S204: Input the feature matrix into a variational autoencoder VAE corresponding to the text topic to obtain the incremental text of the text to be incremented.

In the embodiment of the present application, the terminal device pre-stores multiple VAEs, and each VAE corresponds to a text topic. After determining the text subject of the text to be incremented in step 203, a VAE corresponding to the text subject of the text to be incremented is determined from a plurality of pre-stored VAEs, so that the text to be incremented is incremented based on the determined VAE.

Input the feature matrix into the VAE corresponding to the text topic to obtain the incremental text of the text to be incremented. The text increment is performed based on the VAE corresponding to the text topic of the text to be incremented, which greatly improves the correlation between the incremented text and the to-be- incremented text, and greatly improves the accuracy of text generation.

As shown in Figure 4, VAE consists of two parts, including encoder and decoder. The encoder of VAE does not directly output the code, but considers that all codes conform to a normal distribution. The mean and variance calculation module of the encoder calculates the mean and variance of the normal distribution, and a normal distribution can be determined based on the mean and variance. , sample from the determined normal distribution to obtain a sample code, and then input this sample code into the generator of the decoder to generate incremental text data. That is to say, in the embodiment of the present application, it can be considered that each text to be incremented corresponds to a code in the normal distribution, and the normal distribution is estimated first through the existing training data, and then only the normal distribution needs to be estimated from the normal distribution. By sampling from the distribution, new encodings can be obtained to generate incremental text data.

As a non-limiting example of the present application, a relatively simple Recurrent Neural Network (RNN) is used as the encoder and decoder. The encoder receives the feature matrix as input, and outputs the variance and mean. The decoder determines the normal distribution based on the variance and mean, and performs sampling in the normal distribution to obtain the sampling code. The sampled encoding vector sampled from the normal distribution is input at each time step of the RNN of the decoder, so that the output of each time step is connected to a fully connected layer and softmax function to generate each word at that position The probability of occurrence, we choose the word with the highest probability as the word that appears at this time step. It should be noted that if the length of the generated text does not actually have that many time steps, a preset character will be generated for the part exceeding the length to represent padding.

Exemplarily, for the N×768-dimensional feature matrix output by the BERT Large model mentioned above, the encoder returns a vector with a dimension of 1×256 after receiving this feature matrix; this vector is then connected to two fully connected layers respectively. , the two fully connected layers output two vectors of size 1×256 respectively, and these two outputs are the mean and variance. Determine a normal distribution based on the mean and variance, sample the normal distribution, obtain the sample code, add a variance to the sample code and input it into the decoder, and the decoder will generate incremental text character by character. It should be noted that, the embodiment of the present application can generate incremental text because variance is added to the sampling code, so the exact same incremental text will not be generated.

In the above example, on the one hand, the high-dimensional text vector output by the BERT model contains extremely rich information, which is very suitable for the encoder of VAE to process it into semantic encoding. On the other hand, the distribution calculated by VAE depends on For the input variable, all sampling of this distribution will generate an output similar or related to the input, which itself can help achieve determinism when generating text, thus avoiding the randomness when generating text through the dual role of the BERT model and VAE combination, The correlation between the incremental text and the text to be incremented is greatly improved, so that the quality of text generation can be greatly improved.

It is understandable that the training of the VAE needs to be completed before utilizing the VAE for text increment.

In a non-limiting example of the present application, after obtaining a large-scale corpus for training the model, first classify the text topics on the corpus in the corpus, and then train a VAE for each category of corpus, so as to obtain multiple VAEs corresponding to different text topics.

In another non-limiting example of the present application, after obtaining a large-scale corpus for training the model, a basic VAE is first trained based on the corpus in the corpus; then, on the basis of text topic classification on the corpus in the corpus , for each category of corpus, retrain based on the basic VAE to obtain a VAE, thereby obtaining multiple VAEs corresponding to different text topics.

It can be understood that, in the above two non-limiting examples, in order to improve the accuracy of the VAE text incremental results, for each text topic, there is a corresponding large-scale corpus in the corpus.

In this embodiment of the present application, the feature matrix of the text to be incremented is first extracted, the text subject of the text to be incremented is determined, and then the incremented text is generated in combination with the VAE corresponding to the text subject. On the one hand, the incremental text is generated using the VAE corresponding to the text topic, and a different VAE is set for different topics; on the other hand, since the distribution calculated by the VAE depends on the input variables, all sampling of this distribution will generate the same value as the input Similar or related outputs can help to achieve certainty when generating text. Therefore, through the dual effects of these two aspects, the complete randomness when generating text is avoided, and the correlation between the incremental text and the text to be incremented is improved. This can greatly improve the quality of text generation.

It should be understood that the size of the sequence numbers of the steps in the above embodiments does not mean the sequence of execution, and the execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

Corresponding to the text increment method described in the above embodiment, FIG. 5 shows a structural block diagram of the text increment apparatus provided by the embodiment of the present application. part.

Referring to Figure 5, the device includes:

an acquisition module 51, for acquiring the text to be incremented;

an extraction module 52, configured to perform feature extraction on the text to be incremented, and obtain a feature matrix corresponding to the text to be incremented;

A determination module 53, configured to determine the text topic of the text to be incremented;

The increment module 54 is configured to input the feature matrix into a variational autoencoder corresponding to the text topic to obtain the increment text of the text to be incremented.

Wherein, the extraction module 52 is specifically used for:

The text to be incremented is converted into a feature matrix through a preset word vector model.

Wherein, the extraction module 52 is specifically used for:

Feature extraction is performed on the text to be incremented by using a preset BERT model, and a feature matrix corresponding to the text to be incremented is obtained.

Wherein, the extraction module 52 is specifically used for:

The text to be incremented is converted into an N×768-dimensional feature matrix by a preset BERT model, and the preset BERT model includes 24 coding layers; wherein the text to be incremented includes N characters, N is a positive integer.

Wherein, the extraction module 52 is specifically used for:

Obtain the keywords of the text to be incremented;

obtaining a feature vector corresponding to each of the keywords;

The feature vectors corresponding to all the keywords are combined to generate a feature matrix.

Wherein, the increment module 54 is specifically used for:

Inputting the feature matrix into an encoder of a variational autoencoder corresponding to the text topic, to obtain the mean and variance of the feature matrix;

Determine a normal distribution according to the mean value and the variance, and perform sampling from the normal distribution to obtain a sampling code;

The sample code is input into the decoder of the variational autoencoder to generate the incremental text of the text to be incremented.

It should be noted that the information exchange, execution process and other contents between the above modules/units are based on the same concept as the method embodiments of the present application. For specific functions and technical effects, please refer to the method embodiments section. It is not repeated here.

Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. Module completion, that is, dividing the internal structure of the device into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment may be integrated in one processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit, and the above-mentioned integrated units may adopt hardware. It can also be realized in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing from each other, and are not used to limit the protection scope of the present application. For the specific working processes of the units and modules in the above-mentioned system, reference may be made to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

FIG. 6 is a schematic structural diagram of a terminal device provided by an embodiment of the present application. As shown in FIG. 6 , the terminal device 6 in this embodiment includes: at least one processor 60 (only one processor is shown in FIG. 6 ), a memory 61 , and a memory 61 that is stored in the memory 61 and can be processed in the at least one processor A computer program 62 running on the processor 60, the processor 60 implements the steps in each of the above method embodiments when the computer program 62 is executed.

Embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the steps in the foregoing method embodiments can be implemented.

The embodiments of the present application provide a computer program product, when the computer program product runs on a mobile terminal, the steps in the foregoing method embodiments can be implemented when the mobile terminal executes the computer program product.

The integrated unit, if implemented in the form of a software functional unit and sold or used as an independent product, may be stored in a computer-readable storage medium. Based on this understanding, the present application realizes all or part of the processes in the methods of the above embodiments, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When executed by a processor, the steps of each of the above method embodiments can be implemented. Wherein, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file or some intermediate form, and the like. The computer-readable medium may include at least: any entity or device capable of carrying computer program codes to the photographing device/terminal device, recording medium, computer memory, read-only memory (ROM), random access memory (Random Access Memory, RAM), electrical carrier signals, telecommunication signals, and software distribution media. For example, U disk, mobile hard disk, disk or CD, etc. In some jurisdictions, under legislation and patent practice, computer readable media may not be electrical carrier signals and telecommunications signals.

In the foregoing embodiments, the description of each embodiment has its own emphasis. For parts that are not described or described in detail in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application.

In the embodiments provided in this application, it should be understood that the disclosed apparatus/network device and method may be implemented in other manners. For example, the apparatus/network device embodiments described above are only illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units. Or components may be combined or may be integrated into another system, or some features may be omitted, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the above-mentioned embodiments, those of ordinary skill in the art should understand that: it can still be used for the above-mentioned implementations. The technical solutions described in the examples are modified, or some technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions in the embodiments of the application, and should be included in the within the scope of protection of this application.

Claims (10)

1. a text increment method, is characterized in that, comprises: Get the text to be incremented; Perform feature extraction on the text to be incremented to obtain a feature matrix corresponding to the text to be incremented; determining the text topic of the text to be incremented; Inputting the feature matrix into a variational autoencoder corresponding to the text topic to obtain the incremental text of the text to be incremented. 2. The text increment method according to claim 1, characterized in that, performing feature extraction on the text to be incremented to obtain a feature matrix corresponding to the text to be incremented, comprising: The text to be incremented is converted into a feature matrix through a preset word vector model. 3. The text increment method as claimed in claim 2, wherein, converting the text to be incremented into a feature matrix by a preset word vector model, comprising: Feature extraction is performed on the text to be incremented by using a preset BERT model, and a feature matrix corresponding to the text to be incremented is obtained. 4. The text increment method according to claim 3, wherein the feature extraction is performed on the text to be incremented by a preset BERT model, and a feature matrix corresponding to the text to be incremented is obtained, comprising: : The text to be incremented is converted into an N×768-dimensional feature matrix by a preset BERT model, and the preset BERT model includes 24 coding layers; wherein the text to be incremented includes N characters, N is a positive integer. 5. The text increment method according to claim 2, wherein the text to be incremented is converted into a feature matrix by a preset word vector model, comprising: Obtain the keywords of the text to be incremented; obtaining a feature vector corresponding to each of the keywords; The feature vectors corresponding to all the keywords are combined to generate a feature matrix. 6. The text increment method according to claim 1, wherein the feature matrix is input into a variational autoencoder corresponding to the text topic to obtain the increment text of the text to be incremented ,include: Inputting the feature matrix into an encoder of a variational autoencoder corresponding to the text topic, to obtain the mean and variance of the feature matrix; Determine a normal distribution according to the mean value and the variance, and perform sampling from the normal distribution to obtain a sampling code; The sample code is input into the decoder of the variational autoencoder to generate the incremental text of the text to be incremented. 7. A text increment device, characterized in that, comprising: Get module, used to get the text to be incremented; an extraction module, configured to perform feature extraction on the text to be incremented to obtain a feature matrix corresponding to the text to be incremented; a determining module, used for determining the text topic of the text to be incremented; an increment module, configured to input the feature matrix into a variational autoencoder corresponding to the text topic to obtain an increment text of the text to be incremented. 8. The text increment device according to claim 7, wherein the extraction module is specifically used for: Feature extraction is performed on the text to be incremented by using a preset BERT model, and a feature matrix corresponding to the text to be incremented is obtained. 9. A terminal device, comprising a memory, a processor and a computer program stored in the memory and running on the processor, wherein the processor implements the computer program as claimed in the claims when executing the computer program The text increment method described in any one of 1 to 6. 10. A computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, the text extension according to any one of claims 1 to 6 is implemented. quantitative method.

Images