CN111914540A - Statement identification method and device, storage medium and processor - Google Patents

Abstract

The invention discloses a sentence identification method and device, a storage medium and a processor. Wherein, the method includes: obtaining a statement to be processed; inputting the statement to be processed into a target encoder, and processing to obtain a semantic representation of the statement to be processed, wherein the target encoder is an encoder shared by a grammatical error detection task and a grammatical error correction task ; Input the semantic representation of the statement to be processed into the processing model to obtain a processing result, wherein the processing result includes one of the following: a grammatical error detection result and a grammatical error error correction result. The invention solves the technical problem of poor performance of the system processing tasks due to the difficulty of directly mixing and using the labeling data of different tasks.

Description

Statement authentication method and device, storage medium and processor

technical field

The present invention relates to the technical field of information processing, and in particular, to a sentence identification method and device, a storage medium and a processor.

Background technique

In the task of grammatical error detection and error correction, in the past, there were certain differences in the labeled data due to different task divisions and specific requirements. In the grammatical error detection scenario, similar to the spelling check in Word, it only needs to find and identify errors in the user's input, and give prompts. Taking this as a requirement, some error locations and types in the data are marked, for example, the sentence "I like to eat flat fruit" is marked as the fourth word "flat fruit" error. In another grammatical error correction scenario, the user's input needs to be corrected for errors, and based on this, the grammatically correct form of another batch of data is marked, such as the sentence "I like to eat flat fruit" and its correct form" I like to eat apples". These two kinds of labeling data are not the same in themselves and cannot be directly mixed, so they are called heterogeneous labeling data. However, the detection and correction of grammatical errors for these two data are related to each other, and it is difficult to directly mix and use the labeled data of different tasks, resulting in poor performance of the system processing tasks.

For the above problems, no effective solution has been proposed yet.

SUMMARY OF THE INVENTION

Embodiments of the present invention provide a sentence identification method and device, a storage medium, and a processor to at least solve the technical problem of poor performance of system processing tasks due to difficulty in directly mixing and using different annotation data for different tasks.

According to an aspect of the embodiments of the present invention, there is provided a statement identification method, comprising: acquiring a statement to be processed; inputting the statement to be processed into a target encoder, and processing to obtain a semantic representation of the statement to be processed, wherein the The target encoder is an encoder shared by the grammatical error detection task and the grammatical error correction task; the semantic representation of the statement to be processed is input into the processing model to obtain a processing result, wherein the processing result includes one of the following: Error detection results, grammatical error correction results.

Further, the method further includes: before inputting the to-be-processed sentence into a target encoder and processing to obtain a semantic representation of the to-be-processed sentence, acquiring a plurality of marked sentences, wherein the plurality of The annotated sentences include: first-type sentences that have been marked with grammatical errors and second-type sentences that have been corrected for grammatical errors; inputting the first-type sentences to the target encoder, and processing to obtain the first type of sentences The semantic representation of the class sentence, the second class sentence is input into the target encoder, the semantic representation of the second class sentence is obtained by processing, and the semantic representation of the first class sentence is input to the syntax error detection sub-model , obtain the grammatical error detection result of the sentence of the first type, input the semantic representation of the sentence of the second type to the target decoder, and obtain the grammatical error correction result of the sentence of the second type; The grammatical error detection result of the sentence is compared with the marked grammatical error to obtain a first comparison result, and the grammatical error correction result of the second type of sentence is compared with the grammatical error that has been corrected to obtain a second comparison result. Comparison result; according to the first comparison result and the second comparison result, adjust the parameters of the target encoder.

Further, the method further includes: adjusting parameters of the syntax error detection sub-model according to the first comparison result; adjusting parameters of the target decoder according to the second comparison result.

Further, according to the first comparison result and the second comparison result, adjusting the parameters of the target encoder includes: calculating a first loss value according to the first comparison result and the cross entropy loss function; The second comparison result and the cross-entropy loss function are used to calculate a second loss value; the first loss value is calculated by a back-propagation algorithm to calculate the first parameter to be adjusted, and the second loss value is passed through the The back-propagation algorithm calculates the second parameter to be adjusted; according to the first parameter and the second parameter, the parameter of the target encoder is adjusted.

Further, according to the first parameter and the second parameter, adjusting the parameters of the target encoder includes: determining a first weight value of the syntax error detection task and a second weight of the syntax error correction task value; adjust the parameters of the target encoder according to the first parameter, the first weight value, the second parameter and the second weight value.

Further, the target encoder is a multi-layer Bi-LSTM encoder, and the target decoder is a multi-layer Bi-LSTM decoder.

Further, if the statement to be processed is a statement to be subjected to grammatical error detection, the processing model is a grammatical error detection sub-model, and if the statement to be processed is a statement to be subjected to grammatical error correction, the processing The model is a multi-layer Bi-LSTM decoder.

According to another aspect of the embodiments of the present invention, a sentence identification device is further provided, including: a first acquiring unit, configured to acquire a to-be-processed sentence; and a first processing unit, configured to input the to-be-processed sentence into a target code is processed to obtain the semantic representation of the statement to be processed, wherein, the target encoder is an encoder shared by a syntax error detection task and a syntax error error correction task; a second processing unit is used to convert the syntax of the statement to be processed The semantic representation is input to the processing model to obtain a processing result, wherein the processing result includes one of the following: a grammatical error detection result and a grammatical error correction result.

Further, the device further comprises: a second obtaining unit, configured to obtain a plurality of marked sentences before inputting the to-be-processed sentence into the target encoder and processing to obtain the semantic representation of the to-be-processed sentence, Wherein, the multiple marked sentences include: a first type of sentences marked with grammatical errors and a second type of sentences that have been corrected for grammatical errors; a third processing unit, configured to convert the first type of sentences Input to the target encoder, process to obtain the semantic representation of the first type of sentence, input the second type of sentence into the target encoder, process to obtain the semantic representation of the second type of sentence, and thirdly obtain a unit, configured to input the semantic representation of the first type of sentence into the grammatical error detection sub-model, obtain the grammatical error detection result of the first type of sentence, and input the semantic representation of the second type of sentence to the target decoder to obtain the grammatical error correction result of the second type of statement; the comparison unit is used to compare the grammatical error detection result of the first type of statement with the marked grammatical error, obtain the first comparison result, and set the The grammatical error correction result of the second type of sentence is compared with the grammatical error that has been corrected to obtain a second comparison result; a first adjustment unit is used for comparing the first comparison result and the second comparison result Compare the results and adjust the parameters of the target encoder.

Further, the apparatus further includes: a second adjustment unit, configured to adjust the parameters of the syntax error detection sub-model according to the first comparison result; and a third adjustment unit, configured to adjust the parameters of the syntax error detection sub-model according to the second comparison result Adjust the parameters of the target decoder.

Further, the first adjustment unit includes: a first calculation module, configured to calculate a first loss value according to the first comparison result and the cross entropy loss function, and according to the second comparison result and the cross entropy The loss function calculates the second loss value; the second calculation module is used to calculate the first parameter to be adjusted by the back propagation algorithm, and the second loss value is calculated by the back propagation algorithm a second parameter to be adjusted; an adjustment module configured to adjust the parameter of the target encoder according to the first parameter and the second parameter.

Further, the adjustment module includes: a determination sub-module for determining a first weight value of the grammatical error detection task and a second weight value of the grammatical error correction task; The first parameter, the first weight value, the second parameter, and the second weight value adjust the parameters of the target encoder.

According to another aspect of the embodiments of the present invention, a storage medium is further provided, wherein the storage medium includes a stored program, wherein when the program is run, the device where the storage medium is located is controlled to execute any one of the above The sentence identification method described in item.

According to another aspect of the embodiments of the present invention, a processor is further provided, wherein the processor is configured to run a program, wherein, when the program runs, any one of the statement identification methods described above is executed.

In the embodiment of the present invention, the to-be-processed sentence is acquired; the to-be-processed sentence is input into a target encoder, and the semantic representation of the to-be-processed sentence is obtained by processing, wherein the target encoder is shared by the grammatical error detection task and the grammatical error error correction task. The encoder; inputs the semantic representation of the sentence to be processed into the processing model, and obtains a processing result, wherein the processing result includes one of the following: a grammatical error detection result and a grammatical error correction result, which achieves the goal of passing the grammatical error detection task and grammatical error correction. The encoder shared by wrong tasks enables heterogeneous data to obtain information from the other party, thereby achieving the technical effect of improving system performance, and solving the problem of the performance of system processing tasks caused by the difficulty of directly mixing different labeled data for different tasks. Bad technical issues.

Description of drawings

The accompanying drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the attached image:

1 is a block diagram of a hardware structure of a computer terminal provided according to an embodiment of the present invention;

2 is a flowchart of a statement identification method provided according to an embodiment of the present invention;

3 is a schematic diagram of a sentence identification method provided according to an embodiment of the present invention;

4 is a schematic diagram of a sentence identification device provided according to an embodiment of the present invention; and

FIG. 5 is a structural block diagram of an optional computer terminal according to an embodiment of the present invention.

Detailed ways

In order to make those skilled in the art better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only Embodiments are part of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

It should be noted that the terms "first", "second" and the like in the description and claims of the present invention and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific sequence or sequence. It is to be understood that the data so used may be interchanged under appropriate circumstances such that the embodiments of the invention described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "having" and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those expressly listed Rather, those steps or units may include other steps or units not expressly listed or inherent to these processes, methods, products or devices.

First of all, some nouns or terms that appear in the process of describing the embodiments of the present application are suitable for the following explanations:

Heterogeneous labeled data: refers to the data that is labeled for the same or related tasks to a certain extent, but the content of the data itself is different. Taking grammatical error correction as an example, a piece of data annotates some sentences and their corresponding grammatically correct forms, and a piece of data annotates sentences and their grammatical error locations. These two pieces of data are heterogeneous labeling data for grammatical error correction. .

Grammatical error detection: refers to the detection of grammatical errors in sentences.

Grammatical error correction: refers to the correction of grammatical errors in sentences.

Example 1

According to an embodiment of the present invention, an embodiment of a method for identifying a sentence is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although A logical order is shown in the flowcharts, but in some cases steps shown or described may be performed in an order different from that herein.

The method embodiment provided in Embodiment 1 of the present application may be executed in a mobile terminal, a computer terminal, or a similar computing device. FIG. 1 shows a block diagram of the hardware structure of a computer terminal (or mobile device) for implementing the sentence identification method. As shown in FIG. 1, the computer terminal 10 (or the mobile device 10) may include one or more processors 102 (represented by 102a, 102b, . A processor MCU or a processing device such as a programmable logic device FPGA), a memory 104 for storing data, and a transmission device for communication functions. In addition, may also include: display, input/output interface (I/O interface), universal serial bus (USB) port (may be included as one of the ports of the I/O interface), network interface, power supply and/or camera. Those of ordinary skill in the art can understand that the structure shown in FIG. 1 is only a schematic diagram, which does not limit the structure of the above electronic device. For example, the computer terminal 10 may also include more or fewer components than shown in FIG. 1 , or have a different configuration than that shown in FIG. 1 .

It should be noted that the one or more processors 102 and/or other data processing circuits described above may generally be referred to herein as "data processing circuits." The data processing circuit may be embodied in whole or in part as software, hardware, firmware or any other combination. Furthermore, the data processing circuitry may be a single stand-alone processing module, or incorporated in whole or in part into any of the other elements in the computer terminal 10 (or mobile device). As referred to in the embodiments of the present application, the data processing circuit acts as a kind of processor control (eg, selection of a variable resistance termination path connected to an interface).

The memory 104 can be used to store software programs and modules of application software, such as a program instruction/data storage device corresponding to the statement identification method in the embodiment of the present invention, the processor 102 executes the software programs and modules stored in the memory 104 by running Various functional applications and data processing, that is, the sentence identification method for realizing the above-mentioned application program. Memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, memory 104 may further include memory located remotely from processor 102, which may be connected to computer terminal 10 through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

Transmission means 106 are used to receive or transmit data via a network. A specific example of the above-mentioned network may include a wireless network provided by a communication provider of the computer terminal 10 . In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (Radio Frequency, RF) module, which is used for wirelessly communicating with the Internet.

The display may be, for example, a touch screen type liquid crystal display (LCD) that enables a user to interact with the user interface of the computer terminal 10 (or mobile device).

Under the above operating environment, the present application provides a sentence identification method as shown in FIG. 2 . FIG. 2 is a flowchart of a sentence identification method according to Embodiment 1 of the present invention.

Step S201, acquiring the to-be-processed statement.

For example, the pending statement is: I like to eat flat fruit.

Step S202, inputting the sentence to be processed into the target encoder, and processing to obtain a semantic representation of the sentence to be processed, wherein the target encoder is an encoder shared by the task of grammatical error detection and the task of grammatical error correction.

The above-mentioned sentences to be processed may be sentences of grammatical error detection or error correction data. After a shared target decoder (eg, a multi-layer Bi-LSTM encoder), the semantic representation of each word in the sentence is obtained. The multi-layer Bi-LSTM encoder contains: 1) The words of the sentence are mapped to vector representation through the word vector matrix, and the word vector sequence of the sentence is obtained; 2) The word vector sequence is passed through the multi-layer Bi-LSTM network to obtain the semantic representation of each word .

That is, input the sentence to be processed into the target encoder, and obtain a matrix whose semantic representation is N*D dimension, where N and D are natural numbers. For example, the to-be-processed sentence "I like to eat flat fruit" includes four words: I, like, eat, flat fruit, which is a 4*100-dimensional matrix after its semantic representation.

Optionally, in the sentence identification method provided by the embodiment of the present invention, the method further includes: before inputting the to-be-processed sentence into the target encoder and processing to obtain a semantic representation of the to-be-processed sentence, acquiring a plurality of marked sentences , wherein the multiple marked sentences include: the first type of sentences that have been marked with grammatical errors and the second type of sentences that have been corrected for grammatical errors; the first type of sentences are input into the target encoder, and the first Semantic representation of class sentences, input the second class of sentences to the target encoder, process to obtain the semantic representation of the second class of sentences, input the semantic representation of the first class of sentences to the syntax error detection sub-model, and obtain the grammar of the first class of sentences Error detection results, input the semantic representation of the second type of sentences to the target decoder, and obtain the grammatical error correction results of the second type of sentences; compare the grammatical error detection results of the first type of sentences with the marked grammatical errors, Obtain the first comparison result, compare the grammatical error correction result of the second type of sentence with the grammatical error that has been corrected, and obtain the second comparison result; according to the first comparison result and the second comparison result, adjust Parameters for the target encoder.

The above-mentioned multiple marked sentences include the first type of sentences marked with grammatical errors and the second type of sentences that have been corrected for grammatical errors. For example, the multiple marked sentences include: Sentence one" Today's weather is really Qin Lang", the grammatical error is marked as the fourth word "Qin Lang", and the grammatical error correction is "Today's weather is really sunny"; the second sentence "The call is short due to poor signal", the grammatical error is marked as the fifth The word "medium and short" is wrong, and the grammatical error is corrected as "the call is interrupted due to poor signal". Input the sentence 1 "today's weather is really Qin Lang", which has been marked as the fourth word "Qin Lang" by a grammatical error, into the target encoder, and obtain the semantic representation 1 of sentence 1, and correct the grammatical error as "today's weather". The sentence 1 of "It's really sunny" and "Today's weather is really cool" is input to the target encoder, and the semantic representation 2 of the sentence 1 is obtained after processing, and the semantic representation 1 of the sentence 1 is input to the grammatical error detection sub-model, and the grammatical error detection result of the sentence 1 is obtained. , input the semantic representation 2 of sentence 1 to the target decoder, and obtain the grammatical error correction result of sentence 1; compare the grammatical error detection result of sentence 1 with the marked grammatical errors to obtain the first comparison result, The grammatical error correction result of sentence one is compared with the grammatical error that has been corrected to obtain a second comparison result; according to the first comparison result and the second comparison result, the parameters of the target encoder are adjusted.

According to the above scheme, the parameters of the target encoder are adjusted according to the first comparison result of statement 1 and the second comparison result of statement 1, and the target encoder is adjusted according to the first comparison result of statement 2 and the second comparison result of statement 2. The parameters of the encoder, ... According to the first comparison result of the sentence N and the second comparison result of the sentence 1, the parameters of the target encoder are adjusted. The cross-entropy loss function is calculated according to the first comparison result and the second comparison result, and the parameters of the target encoder are updated by reverse gradient propagation.

As shown in Figure 3, this method extracts the semantic representation of sentences by using a shared multi-layer Bi-LSTM encoder on the data of grammatical error detection and grammatical error error correction, and jointly trains the encoder with heterogeneous data. The encoder is fully learned and can extract error detection and error correction information at the same time. Different data independent syntax error detection sub-models and multi-layer Bi-LSTM decoders ensure the difference of different data outputs. At the same time, the two grammatical error detection tasks and grammatical error correction tasks of this scheme have no cascading relationship, no error accumulation will occur, and different data can be jointly trained, so that heterogeneous data can obtain information from each other, and ultimately improve the system performance.

Step S203: Input the semantic representation of the statement to be processed into the processing model, and obtain a processing result, wherein the processing result includes one of the following: a grammatical error detection result and a grammatical error error correction result.

In the statement identification method provided by the embodiment of the present invention, if the statement to be processed is a statement to be subjected to grammatical error detection, the processing model is a grammatical error detection sub-model; The processing model is a multi-layer Bi-LSTM decoder.

If a 4*100-dimensional matrix after the semantic representation of the to-be-processed sentence "I like to eat flat fruit" is input into the grammatical error detection sub-model, the grammatical error detection result of the to-be-processed sentence "I like to eat flat fruit" is marked as No. 4 words "flat fruit" wrong. If the semantic representation of the to-be-processed sentence "I like to eat flat fruit" is a 4*100-dimensional matrix input to the multi-layer Bi-LSTM decoder, the grammatical error correction result of the to-be-processed sentence "I like to eat flat fruit" is " I like to eat apples".

To sum up, the sentence identification method provided by the embodiment of the present invention obtains the to-be-processed sentence; inputs the to-be-processed sentence into the target encoder, and processes to obtain the semantic representation of the to-be-processed sentence, wherein the target encoder is the task of grammatical error detection The encoder shared with the grammatical error correction task; the semantic representation of the sentence to be processed is input into the processing model, and the processing result is obtained, wherein the processing result includes one of the following: the grammatical error detection result, the grammatical error error correction result, and the achievement of passing The shared encoder between the grammatical error detection task and the grammatical error error correction task enables heterogeneous data to obtain information from the other party, thereby achieving the technical effect of improving system performance, and solving the problem of difficulty in directing direct data due to different labeling data for different tasks. Mixed use of technical issues that lead to poor performance of the system processing tasks.

Optionally, in the sentence identification method provided by the embodiment of the present invention, the method further includes: adjusting the parameters of the syntax error detection sub-model according to the first comparison result; and adjusting the parameters of the target decoder according to the second comparison result.

In the above scheme, the parameters of the syntax error detection sub-model are adjusted according to the first comparison result; the parameters of the target decoder are adjusted according to the second comparison result, and the parameters of the syntax error detection sub-model and the target decoder are continuously updated to ensure that The accuracy of the syntax error detection sub-model and the output of the target decoder.

Optionally, in the sentence identification method provided by the embodiment of the present invention, adjusting the parameters of the target encoder according to the first comparison result and the second comparison result includes: calculating the first comparison result according to the first comparison result and the cross entropy loss function. a loss value; calculate the second loss value according to the second comparison result and the cross-entropy loss function; calculate the first parameter to be adjusted by the back-propagation algorithm, and calculate the second loss value by the back-propagation algorithm The second parameter to be adjusted; adjust the parameters of the target encoder according to the first parameter and the second parameter.

In the above solution, the first parameter to be adjusted and the second parameter to be adjusted include the parameter to be adjusted and the value to be updated. The parameters of the encoder are adjusted, which plays the role of jointly training the target encoder.

Optionally, in the sentence identification method provided in the embodiment of the present invention, adjusting the parameters of the target encoder according to the first parameter and the second parameter includes: determining the first weight value of the grammatical error detection task and the grammatical error correction task. The second weight value; according to the first parameter, the first weight value, the second parameter and the second weight value, the parameters of the target encoder are adjusted.

For example, the first weight value of the syntax error detection task is 0.5, the second weight value of the syntax error correction task is 1.0, and the final adjustment target is obtained by combining the first parameter to be adjusted*0.5+the second parameter to be adjusted*1.0 Encoder parameters. By presetting the weight values of different tasks, the training of the trained target encoder is more targeted, and the accuracy of the output result of the target encoder at the training place is guaranteed.

It should be noted that, for the sake of simple description, the foregoing method embodiments are all expressed as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence. As in accordance with the present invention, certain steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

From the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is better implementation. Based on this understanding, the technical solutions of the present invention can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products are stored in a storage medium (such as ROM/RAM, magnetic disk, CD-ROM), including several instructions to make a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

Example 2

According to an embodiment of the present invention, an apparatus for implementing the above sentence identification method is also provided. As shown in FIG. 4 , the apparatus includes:

a first obtaining unit 401, configured to obtain a statement to be processed;

The first processing unit 402 is used for inputting the statement to be processed into a target encoder, and processing to obtain the semantic representation of the statement to be processed, wherein the target encoder is shared by the grammatical error detection task and the grammatical error correction task the encoder;

The second processing unit 403 is configured to input the semantic representation of the to-be-processed statement into a processing model to obtain a processing result, wherein the processing result includes one of the following: a grammatical error detection result and a grammatical error correction result.

To sum up, in the sentence identification method provided by the embodiment of the present invention, the to-be-processed sentence is obtained through the first obtaining unit 401; the first processing unit 402 inputs the to-be-processed sentence into the target encoder, and processes the to-be-processed sentence to obtain the to-be-processed sentence The semantic representation of , wherein, the target encoder is an encoder shared by the grammatical error detection task and the grammatical error correction task; the second processing unit 403 inputs the semantic representation of the statement to be processed into the processing model, and obtains the processing result, Wherein, the processing result includes one of the following: a grammatical error detection result and a grammatical error error correction result, achieving an encoder shared by a grammatical error detection task and a grammatical error error correction task, so that heterogeneous data can be obtained from each other. Therefore, the technical effect of improving the performance of the system is realized, and the technical problem of poor performance of the system processing tasks due to the difficulty of direct mixed use of the labeled data of different tasks is solved.

Optionally, in the sentence identification device provided in the embodiment of the present invention, the device further includes: a second obtaining unit, configured to input the to-be-processed sentence into a target encoder, and process to obtain the semantics of the to-be-processed sentence Before the expression, obtain a plurality of marked sentences, wherein the plurality of marked sentences include: the first type of sentences marked with grammatical errors and the second type of sentences with grammatical errors corrected; the third The processing unit is configured to input the sentence of the first type into the target encoder, process to obtain the semantic representation of the sentence of the first type, input the sentence of the second type to the target encoder, and obtain the the semantic representation of the sentence of the second type, and a third acquisition unit, configured to input the semantic representation of the sentence of the first type into the grammatical error detection sub-model, obtain the grammatical error detection result of the sentence of the first type, The semantic representation of the sentence of the second type is input to the target decoder to obtain the grammatical error correction result of the sentence of the second type; the comparison unit is used to compare the grammatical error detection result of the sentence of the first type with the marked grammatical error Carry out comparison, obtain the first comparison result, and compare the grammatical error correction result of the second type of statement with the grammatical error that has been corrected to obtain the second comparison result; the first adjustment unit is used for according to The first comparison result and the second comparison result adjust the parameters of the target encoder.

Optionally, in the sentence identification device provided by the embodiment of the present invention, the device further includes: a second adjustment unit, configured to adjust the parameters of the grammatical error detection sub-model according to the first comparison result; a third An adjustment unit, configured to adjust the parameters of the target decoder according to the second comparison result.

Optionally, in the sentence identification device provided in the embodiment of the present invention, the first adjustment unit includes: a first calculation module, configured to calculate a first loss value according to the first comparison result and a cross-entropy loss function, Calculate the second loss value according to the second comparison result and the cross-entropy loss function; the second calculation module is configured to calculate the first parameter to be adjusted by using the back-propagation algorithm to calculate the first loss value, The second loss value is calculated by the back-propagation algorithm to calculate the second parameter to be adjusted; the adjustment module is configured to adjust the parameter of the target encoder according to the first parameter and the second parameter.

Optionally, in the sentence identification device provided by the embodiment of the present invention, the adjustment module includes: a determination sub-module, configured to determine the first weight value of the grammatical error detection task and the first weight of the grammatical error error correction task. Two weight values; an adjustment sub-module, configured to adjust the parameters of the target encoder according to the first parameter, the first weight value, the second parameter and the second weight value.

Optionally, in the sentence identification device provided by the embodiment of the present invention, the target encoder is a multi-layer Bi-LSTM encoder, and the target decoder is a multi-layer Bi-LSTM decoder.

Optionally, in the statement identification device provided by the embodiment of the present invention, if the to-be-processed statement is a statement to be subjected to grammatical error detection, the processing model is a grammatical error detection sub-model, and if the to-be-processed statement is For sentences to be corrected for syntax errors, the processing model is a multi-layer Bi-LSTM decoder.

It should be noted here that the above-mentioned first acquisition unit 401, first processing unit 402 and second processing unit 403 correspond to step S201, steps S202 to S203 in Embodiment 1, and the two modules correspond to the corresponding steps. The implementation examples and application scenarios are the same, but are not limited to the content disclosed in the first embodiment. It should be noted that, as a part of the apparatus, the above-mentioned modules may run in the computer terminal 10 provided in the first embodiment.

Example 3

Embodiments of the present invention may provide a computer terminal, and the computer terminal may be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the above-mentioned computer terminal may also be replaced by a terminal device such as a mobile terminal.

Optionally, in this embodiment, the above-mentioned computer terminal may be located in at least one network device among multiple network devices of a computer network.

In this embodiment, the above-mentioned computer terminal can execute the program code of the following steps in the statement identification method of the application program: obtaining the statement to be processed; inputting the statement to be processed into the target encoder, and processing to obtain the semantic representation of the statement to be processed, wherein, The target encoder is an encoder shared by the grammatical error detection task and the grammatical error error correction task; the semantic representation of the sentence to be processed is input into the processing model, and the processing result is obtained, wherein the processing result includes one of the following: grammatical error detection result, grammar Error correction result.

The above-mentioned computer terminal can also execute the program code of the following steps in the statement identification method of the application program: the method further includes: before inputting the statement to be processed into the target encoder, and processing to obtain the semantic representation of the statement to be processed, acquiring a plurality of Annotated sentences, wherein the multiple annotated sentences include: a first-type sentence that has been marked with grammatical errors and a second-type sentence that has been corrected for grammatical errors; input the first-type sentences to the target encoder, and process Obtain the semantic representation of the first type of sentence, input the second type of sentence into the target encoder, process to obtain the semantic representation of the second type of sentence, input the semantic representation of the first type of sentence into the syntax error detection sub-model, and obtain the first type of sentence. The grammatical error detection result of the sentence, input the semantic representation of the second type sentence to the target decoder, and obtain the grammatical error correction result of the second type sentence; compare the grammatical error detection result of the first type sentence with the marked grammatical error Compare, obtain the first comparison result, compare the grammatical error correction result of the second type of sentence with the grammatical error that has been corrected, and obtain the second comparison result; according to the first comparison result and the second comparison As a result, the parameters of the target encoder are adjusted.

The above-mentioned computer terminal can also execute the program code of the following steps in the statement identification method of the application program: the method further comprises: adjusting the parameters of the syntax error detection sub-model according to the first comparison result; adjusting the parameters of the target decoder according to the second comparison result; parameter.

The above-mentioned computer terminal can also execute the program code of the following steps in the statement identification method of the application program: according to the first comparison result and the second comparison result, adjusting the parameters of the target encoder includes: according to the first comparison result and the cross entropy loss The function calculates the first loss value; calculates the second loss value according to the second comparison result and the cross-entropy loss function; calculates the first parameter to be adjusted through the back propagation algorithm, and calculates the second loss value through the reverse propagation algorithm. The propagation algorithm calculates the second parameter to be adjusted; according to the first parameter and the second parameter, the parameter of the target encoder is adjusted.

The above-mentioned computer terminal can also execute the program code of the following steps in the statement identification method of the application program: according to the first parameter and the second parameter, adjusting the parameters of the target encoder includes: determining the first weight value of the grammatical error detection task and the grammatical error correction. The second weight value of the wrong task is adjusted; the parameters of the target encoder are adjusted according to the first parameter, the first weight value, the second parameter and the second weight value.

The above-mentioned computer terminal can also execute the program code of the following steps in the sentence identification method of the application program: the target encoder is a multi-layer Bi-LSTM encoder, and the target decoder is a multi-layer Bi-LSTM decoder.

The above-mentioned computer terminal can also execute the program code of the following steps in the statement identification method of the application program: if the statement to be processed is a statement to be subjected to grammatical error detection, the processing model is a grammatical error detection submodel, and if the statement to be processed is a grammar to be performed For sentences with error correction, the processing model is a multi-layer Bi-LSTM decoder.

Optionally, FIG. 5 is a structural block diagram of a computer terminal according to an embodiment of the present invention. As shown in FIG. 5 , the computer terminal A may include: one or more (only one is shown in the figure) processors and memories.

The memory can be used to store software programs and modules, such as program instructions/modules corresponding to the statement identification method and device in the embodiment of the present invention, and the processor executes various functional applications by running the software programs and modules stored in the memory. And data processing, that is, to realize the above sentence identification method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory located remotely from the processor, and these remote memories may be connected to Terminal A through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

The processor can call the information and application program stored in the memory through the transmission device to perform the following steps: acquiring the statement to be processed; inputting the statement to be processed into the target encoder, and processing to obtain the semantic representation of the statement to be processed, wherein the target encoder It is an encoder shared by the grammatical error detection task and the grammatical error error correction task; the semantic representation of the sentence to be processed is input into the processing model, and the processing result is obtained, wherein the processing result includes one of the following: grammatical error detection result, grammatical error error correction result.

Optionally, the above-mentioned processor may also execute the program code of the following steps: the method further includes: before inputting the statement to be processed into the target encoder and processing to obtain the semantic representation of the statement to be processed, acquiring a plurality of marked statements , wherein the multiple marked sentences include: the first type of sentences that have been marked with grammatical errors and the second type of sentences that have been corrected for grammatical errors; the first type of sentences are input into the target encoder, and the first Semantic representation of class sentences, input the second class of sentences to the target encoder, process to obtain the semantic representation of the second class of sentences, input the semantic representation of the first class of sentences to the syntax error detection sub-model, and obtain the grammar of the first class of sentences Error detection results, input the semantic representation of the second type of sentences to the target decoder, and obtain the grammatical error correction results of the second type of sentences; compare the grammatical error detection results of the first type of sentences with the marked grammatical errors, Obtain the first comparison result, compare the grammatical error correction result of the second type of sentence with the grammatical error that has been corrected, and obtain the second comparison result; according to the first comparison result and the second comparison result, adjust Parameters for the target encoder.

Optionally, the above processor may further execute the program code of the following steps: the method further includes: adjusting the parameters of the syntax error detection sub-model according to the first comparison result; and adjusting the parameters of the target decoder according to the second comparison result.

Optionally, the above-mentioned processor may also execute the program code of the following steps: according to the first comparison result and the second comparison result, adjusting the parameters of the target encoder includes: calculating the first comparison result and the cross entropy loss function according to the first comparison result. a loss value; calculate the second loss value according to the second comparison result and the cross-entropy loss function; calculate the first parameter to be adjusted by the back-propagation algorithm, and calculate the second loss value by the back-propagation algorithm The second parameter to be adjusted; adjust the parameters of the target encoder according to the first parameter and the second parameter.

Optionally, the above-mentioned processor can also execute the program code of the following steps: according to the first parameter and the second parameter, adjusting the parameters of the target encoder includes: determining the first weight value of the grammatical error detection task and the grammatical error correction task. The second weight value; according to the first parameter, the first weight value, the second parameter and the second weight value, the parameters of the target encoder are adjusted.

Optionally, the above processor may also execute the program code of the following steps: the target encoder is a multi-layer Bi-LSTM encoder, and the target decoder is a multi-layer Bi-LSTM decoder.

Optionally, the above processor may also execute the program code of the following steps: if the statement to be processed is a statement to be subjected to grammatical error detection, the processing model is a grammatical error detection sub-model, and if the statement to be processed is a grammatical error correction to be performed sentence, the processing model is a multi-layer Bi-LSTM decoder.

The embodiment of the present invention provides a statement identification solution, by acquiring the statement to be processed; inputting the statement to be processed into a target encoder, and processing to obtain the semantic representation of the statement to be processed, wherein the target encoder is a grammatical error detection task The encoder shared with the grammatical error correction task; the semantic representation of the sentence to be processed is input into the processing model, and the processing result is obtained, wherein the processing result includes one of the following: the grammatical error detection result, the grammatical error error correction result, and the achievement of passing The shared encoder between the grammatical error detection task and the grammatical error error correction task enables heterogeneous data to obtain information from the other party, thereby achieving the technical effect of improving system performance, and solving the problem of difficulty in directing direct data due to different labeling data for different tasks. Mixed use of technical issues that lead to poor performance of the system processing tasks.

Those of ordinary skill in the art can understand that the structure shown in FIG. 5 is only a schematic diagram, and the computer terminal can also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, an applause computer, and a mobile internet device (Mobile Internet Devices, MID). ), PAD and other terminal equipment. FIG. 5 does not limit the structure of the above electronic device. For example, the computer terminal 10 may also include more or fewer components than those shown in FIG. 5 (eg, network interfaces, display devices, etc.), or have a different configuration than that shown in FIG. 5 .

Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium can Including: flash disk, read-only memory (Read-Only Memory, ROM), random access device (RandomAccess Memory, RAM), magnetic disk or optical disk, etc.

Example 4

Embodiments of the present invention also provide a storage medium. Optionally, in this embodiment, the above-mentioned storage medium may be used to store the program code executed by the statement identification method provided in the above-mentioned first embodiment.

Optionally, in this embodiment, the above-mentioned storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps: acquiring the statement to be processed; inputting the statement to be processed into the target encoder, and processing to obtain the semantic representation of the statement to be processed, wherein , the target encoder is an encoder shared by the grammatical error detection task and the grammatical error error correction task; the semantic representation of the sentence to be processed is input into the processing model, and the processing result is obtained, wherein the processing result includes one of the following: grammatical error detection result, Syntax error correction result.

Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps: the method further includes: before inputting the to-be-processed statement into the target encoder and processing to obtain the semantic representation of the to-be-processed statement , to obtain a plurality of marked sentences, wherein the plurality of marked sentences include: the first type of sentences that have been marked with grammatical errors and the second type of sentences that have been corrected for grammatical errors; input the first type of sentences To the target encoder, processing to obtain the semantic representation of the first type of sentence, inputting the second type of sentence to the target encoder, processing to obtain the semantic representation of the second type of sentence, and inputting the semantic representation of the first type of sentence to the syntax error detector. model, obtain the grammatical error detection results of the first type of sentences, input the semantic representation of the second type of sentences to the target decoder, and obtain the grammatical error correction results of the second type of sentences; compare the grammatical error detection results of the first type of sentences with The grammatical errors that have been marked are compared to obtain the first comparison result, and the grammatical error correction results of the second type of sentences are compared with the grammatical errors that have been corrected to obtain the second comparison result; according to the first comparison The result is compared with the second result, and the parameters of the target encoder are adjusted.

Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps: the method further includes: adjusting the parameters of the syntax error detection sub-model according to the first comparison result; according to the second comparison The result adjusts the parameters of the target decoder.

Optionally, in this embodiment, the storage medium is configured to store program codes for performing the following steps: according to the first comparison result and the second comparison result, adjusting the parameters of the target encoder includes: according to the first comparison result Calculate the first loss value according to the result and the cross-entropy loss function; calculate the second loss value according to the second comparison result and the cross-entropy loss function; calculate the first loss value through the back-propagation algorithm to calculate the first parameter to be adjusted, The second loss value is calculated by the back-propagation algorithm to calculate the second parameter to be adjusted; according to the first parameter and the second parameter, the parameter of the target encoder is adjusted.

Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps: according to the first parameter and the second parameter, adjusting the parameters of the target encoder includes: determining the first parameter of the syntax error detection task The weight value and the second weight value of the syntax error correction task; according to the first parameter, the first weight value, the second parameter and the second weight value, the parameters of the target encoder are adjusted.

Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps: the target encoder is a multi-layer Bi-LSTM encoder, and the target decoder is a multi-layer Bi-LSTM decoder.

Optionally, in this embodiment, the storage medium is configured to store program codes for executing the following steps: if the statement to be processed is a statement to be subjected to syntax error detection, the processing model is a syntax error detection sub-model; The processing sentence is a sentence to be corrected for syntax errors, and the processing model is a multi-layer Bi-LSTM decoder.

The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages or disadvantages of the embodiments.

In the above-mentioned embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are only illustrative, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units or components may be combined or Integration into another system, or some features can be ignored, 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 units or modules, and may be in electrical 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.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

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 technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes .

The above are only the preferred embodiments of the present invention. It should be pointed out that for those skilled in the art, without departing from the principles of the present invention, several improvements and modifications can be made. It should be regarded as the protection scope of the present invention.

Claims (14)

1. a statement identification method, is characterized in that, comprises: Get the pending statement; Inputting the statement to be processed into a target encoder, and processing to obtain a semantic representation of the statement to be processed, wherein the target encoder is an encoder shared by a grammatical error detection task and a grammatical error correction task; The semantic representation of the to-be-processed statement is input to the processing model to obtain a processing result, wherein the processing result includes one of the following: a grammatical error detection result and a grammatical error error correction result. 2. The method according to claim 1, wherein the method further comprises: Before the to-be-processed sentence is input into the target encoder, and before the semantic representation of the to-be-processed sentence is obtained by processing, a plurality of marked sentences are obtained, wherein the plurality of marked sentences include: The first category of grammatical errors and the second category of grammatical errors that have been corrected; Inputting the sentence of the first type into the target encoder, processing to obtain the semantic representation of the sentence of the first type, inputting the sentence of the second type into the target encoder, and processing to obtain the sentence of the second type the semantic representation of , Inputting the semantic representation of the first type of sentence into the grammatical error detection sub-model, obtaining the grammatical error detection result of the first type of sentence, and inputting the semantic representation of the second type of sentence into the target decoder to obtain the The grammatical error correction result of the second type of sentence; Compare the grammatical error detection result of the first type of statement with the marked grammatical error to obtain the first comparison result, and compare the grammatical error error correction result of the second type of statement with the grammatical error that has been corrected. Compare to obtain a second alignment result; The parameters of the target encoder are adjusted according to the first comparison result and the second comparison result. 3. The method according to claim 2, wherein the method further comprises: Adjust the parameters of the syntax error detection sub-model according to the first comparison result; The parameters of the target decoder are adjusted according to the second comparison result. 4. The method according to claim 2, wherein, according to the first comparison result and the second comparison result, adjusting the parameters of the target encoder comprises: Calculate a first loss value according to the first comparison result and the cross-entropy loss function; Calculate a second loss value according to the second alignment result and the cross-entropy loss function; Calculate the first parameter to be adjusted by using the back-propagation algorithm for the first loss value, and calculate the second parameter to be adjusted by using the back-propagation algorithm for the second loss value; The parameters of the target encoder are adjusted according to the first parameter and the second parameter. 5. The method according to claim 4, wherein, according to the first parameter and the second parameter, adjusting the parameters of the target encoder comprises: determining a first weight value of the grammatical error detection task and a second weight value of the grammatical error correction task; The parameters of the target encoder are adjusted according to the first parameter, the first weight value, the second parameter and the second weight value. 6. The method according to claim 2, wherein the target encoder is a multi-layer Bi-LSTM encoder, and the target decoder is a multi-layer Bi-LSTM decoder. 7. The method according to claim 6, wherein, if the statement to be processed is a statement to be processed for grammatical error detection, then the processing model is a grammatical error detection submodel, and if the statement to be processed is a statement to be processed For sentences with syntax error correction, the processing model is a multi-layer Bi-LSTM decoder. 8. A sentence identification device, comprising: a first obtaining unit, used to obtain the to-be-processed statement; The first processing unit is configured to input the statement to be processed into a target encoder, and process to obtain the semantic representation of the statement to be processed, wherein the target encoder is shared by the task of grammatical error detection and grammatical error correction. Encoder; The second processing unit is configured to input the semantic representation of the to-be-processed statement into a processing model to obtain a processing result, wherein the processing result includes one of the following: a grammatical error detection result and a grammatical error correction result. 9. The apparatus of claim 8, wherein the apparatus further comprises: The second acquiring unit is configured to acquire a plurality of annotated sentences before inputting the to-be-processed sentence into a target encoder and obtaining a semantic representation of the to-be-processed sentence, wherein the plurality of annotated sentences The sentences include: the first type of sentences that have been marked with grammatical errors and the second type of sentences that have been corrected for grammatical errors; The third processing unit is configured to input the sentence of the first type into the target encoder, process to obtain the semantic representation of the sentence of the first type, input the sentence of the second type to the target encoder, process to obtain the semantic representation of the second class of sentences, a third acquiring unit, configured to input the semantic representation of the first type of sentence into the grammatical error detection sub-model, obtain the grammatical error detection result of the first type of statement, and input the semantic representation of the second type of statement to the grammatical error detection sub-model The target decoder obtains the grammatical error correction result of the second type of sentence; The comparison unit is used to compare the grammatical error detection result of the first type of statement with the marked grammatical error, obtain the first comparison result, and compare the grammatical error correction result of the second type of statement with the grammatical error that has been marked. Error correction syntax errors are compared to obtain a second comparison result; A first adjustment unit, configured to adjust parameters of the target encoder according to the first comparison result and the second comparison result. 10. The apparatus of claim 9, wherein the apparatus further comprises: a second adjustment unit, configured to adjust parameters of the syntax error detection sub-model according to the first comparison result; A third adjustment unit, configured to adjust the parameters of the target decoder according to the second comparison result. 11. The device according to claim 9, wherein the first adjustment unit comprises: a first calculation module, configured to calculate a first loss value according to the first comparison result and the cross-entropy loss function, and calculate a second loss value according to the second comparison result and the cross-entropy loss function; a second calculation module, configured to calculate the first parameter to be adjusted by using the back-propagation algorithm for the first loss value, and calculate the second parameter to be adjusted by using the back-propagation algorithm for the second loss value; An adjustment module, configured to adjust the parameters of the target encoder according to the first parameter and the second parameter. 12. The apparatus according to claim 11, wherein the adjustment module comprises: A determination submodule for determining the first weight value of the grammatical error detection task and the second weight value of the grammatical error correction task; An adjustment submodule, configured to adjust the parameters of the target encoder according to the first parameter, the first weight value, the second parameter and the second weight value. 13. A storage medium, characterized in that the storage medium comprises a stored program, wherein when the program is run, a device where the storage medium is located is controlled to execute the statement identification according to any one of claims 1 to 7 method. 14. A processor, wherein the processor is configured to run a program, wherein the statement identification method according to any one of claims 1 to 7 is executed when the program is run.

Images