CN112149680A

CN112149680A - Wrong word detection and identification method and device, electronic equipment and storage medium

Info

Publication number: CN112149680A
Application number: CN202011037561.0A
Authority: CN
Inventors: 李全恩; 李雪冬
Original assignee: Wuhan Yuexuebang Network Technology Co ltd
Current assignee: Wuhan Yuexuebang Network Technology Co ltd
Priority date: 2020-09-28
Filing date: 2020-09-28
Publication date: 2020-12-29
Anticipated expiration: 2040-09-28
Also published as: CN112149680B

Abstract

The present disclosure provides a method, an apparatus, an electronic device and a storage medium for detecting and identifying a wrong word, wherein the method for detecting and identifying the wrong word comprises the following steps: acquiring composition picture information of a user; detecting and identifying the composition picture information to obtain text content corresponding to the composition picture information; and carrying out wrong word detection on the text content according to a pre-trained semantic model, and generating corresponding prompt information according to a detection result. The wrongly written characters in the composition can be detected by acquiring the handwritten composition picture of the user, and corresponding prompt information is generated according to the detection result to remind the user, so that a teacher or parents can intuitively and quickly find the wrongly written characters when reading in batch, the trouble of review caused by the wrongly written characters is avoided, and the review efficiency is improved.

Description

Wrong word detection and identification method and device, electronic equipment and storage medium

Technical Field

The present disclosure relates to the field of artificial intelligence, and in particular, to a method and an apparatus for detecting and identifying a wrong word, an electronic device, and a storage medium.

Background

At present, in the manuscript checking stage, examination and reading are carried out word by sentence by examination and approval personnel aiming at articles handwritten by students, students or researchers. However, because the drafter inevitably has carelessness and miswritten situations in the process of writing, written articles are inevitably wrongly written and wrongly written, and wrongly written characters and incorrectly expressed places are caused, which brings about great troubles to examination and approval personnel.

Disclosure of Invention

The embodiment of the disclosure at least provides a wrong word detection and identification method, a wrong word detection and identification device, electronic equipment and a storage medium.

In a first aspect, an embodiment of the present disclosure provides a method for detecting and identifying a wrong word, including:

acquiring composition picture information of a user;

detecting and identifying the composition picture information to acquire text content corresponding to the composition picture information;

and carrying out wrong word detection on the text content according to a pre-trained semantic model, and generating corresponding prompt information according to a detection result.

In the embodiment of the disclosure, by acquiring the composition picture information of the user, the composition picture information is detected and identified to acquire the text content corresponding to the composition picture information, and the text content is subjected to wrong word detection according to the pre-trained semantic model, the wrong words in the composition can be automatically detected, and corresponding prompt information is generated according to the detection result to remind the user, so that a teacher or parents can intuitively and quickly find wrong words when reading in batches, troubles caused by wrong words are avoided, and the review efficiency is improved.

According to the first aspect, in a possible implementation manner, the detecting and identifying the composition picture information to obtain text content corresponding to the composition picture information includes:

segmenting the composition picture information to obtain a plurality of single-character image blocks, and determining the position of each single-character image block in the composition picture information;

detecting and identifying each single-character image block to obtain an identification candidate character corresponding to each single-character image block and the probability of each identification candidate character;

determining the recognition candidate character with the highest probability and larger than a specific threshold in the recognition candidate characters corresponding to each single character image block as a single character text corresponding to the single character image block;

and combining a plurality of single-word texts according to the position of each single-word text to obtain the text content.

In the embodiment of the disclosure, a plurality of single-character image blocks can be obtained by segmenting the composition picture information, a single-character text corresponding to each single-character image block is obtained by identifying each single-character image block, and then the plurality of single-character texts are combined according to the position of each single-character text, so that the extraction and conversion from the picture to the text content are realized.

In a possible implementation manner, after determining, as the single-word text corresponding to the single-word image block, the recognition candidate word having the highest probability and being greater than a specific threshold from among the recognition candidate words corresponding to each single-word image block, the method for detecting and recognizing a wrong word further includes:

and based on the recognized single word text, recognizing the single word image blocks without a certain recognition candidate character according to a pre-trained semantic model to obtain the corresponding single word text, wherein the probability of the single word image blocks is greater than the specific threshold.

In the embodiment of the disclosure, when the single character text at a certain position cannot be recognized, the single character text at the position is further confirmed through the semantic model, so that the accuracy of recognizing the composition picture is further improved.

In a possible implementation manner, the identifying, based on recognized word texts, a word image block, for which a probability that a recognition candidate word does not exist is greater than a specific threshold, according to a pre-trained semantic model, so as to obtain a corresponding word text, includes:

based on the recognized single word text, recognizing the single word image block without a recognition candidate word according to the semantic model, wherein the probability of the single word image block is greater than the specific threshold value, so as to obtain the semantic candidate word corresponding to the single word image block and the probability of each semantic candidate word;

and adjusting the probability of the recognition candidate words according to the semantic candidate words and the probability thereof, and confirming the recognition candidate word with the highest adjusted probability as the single word text corresponding to the single word image block.

In the embodiment of the disclosure, the probability of identifying the candidate word is adjusted through the semantic candidate word and the probability thereof, and the identified candidate word with the highest adjusted probability is determined as the single word text corresponding to the single word image block, so that high-precision text extraction is realized.

segmenting the composition picture information to obtain a plurality of single-character image blocks, and determining the coordinates of each single-character image block;

combining a plurality of single-character image blocks in the same row according to the coordinates of each single-character image block to obtain a recombined single-row image strip;

segmenting the composition picture information to obtain a plurality of original single-line image strips, and determining the coordinates of each original single-line image strip;

detecting and identifying each original single-line image strip to obtain an identification candidate word corresponding to each position in each original single-line image strip and the probability of each identification candidate word;

if the recognition candidate character with the maximum probability at the corresponding position in the recombined single-line image strip is the same as the recognition candidate character with the maximum probability at the corresponding position in the original single-line image strip, determining the recognition candidate character with the maximum probability as the single character text corresponding to the position;

and combining a plurality of single-word texts according to the coordinates of each single-word text to obtain the text content.

In the embodiment of the disclosure, the single-word detection and recognition and the single-line detection and recognition are combined, and under the condition that the maximum probability recognition candidate words detected by the two are the same, the recognition candidate words with the maximum probability are determined as the words written by the drafter, so that the recognition accuracy is further ensured.

In a possible implementation manner, before the combining the multiple single-word texts according to the coordinates of each single-word text to obtain the text content, the method for detecting and identifying the wrong word further includes:

and if the recognition candidate words with the maximum probability of the corresponding positions in the recombined single-line image strip are different from those in the original single-line image strip, performing semantic recognition on the positions according to a pre-trained semantic model based on the recognized single character texts to obtain the corresponding single character texts. In the embodiment of the disclosure, the single-word detection and identification and the single-line detection and identification are combined, and when the two are different, semantic detection is required to further judge, so that the identification precision of the composition picture is further improved. According to the first aspect, in a possible implementation manner, the performing wrong word detection on the text content according to a pre-trained semantic model, and generating corresponding prompt information according to a detection result includes:

performing semantic detection on the single word text on each position in sequence according to a predetermined sequence on the text content according to the semantic model to obtain semantic candidate words of each position and the probability of each semantic candidate word;

judging whether the current single character text at each position belongs to a candidate character to be selected at the position; the candidate word to be selected is a semantic candidate word with the probability at the position larger than a reference threshold;

if the current single character text at a certain position does not belong to the candidate character to be selected at the position, determining the single character text at the position as a wrong character, and simultaneously displaying the wrong character and the correct character; wherein, the correct word is the semantic candidate word with the highest probability at the position.

In the embodiment of the disclosure, semantic detection is performed on the single word texts in the recognized texts one by one, so that semantic candidate words and probabilities thereof at each position can be obtained, and whether the current single word text is a wrong word can be determined by comparing the current text at each position with the semantic candidate words, so that the method is simple and rapid, and the efficiency of wrong word detection is improved.

According to the first aspect, in a possible implementation manner, after determining whether the current single word text at each position belongs to a candidate word to be selected at the position, the method for detecting and identifying a wrong word further includes:

and if the current single character text at each position belongs to the candidate character to be selected at the position, determining that no wrong character exists in the text content, and generating prompt information.

In the embodiment of the disclosure, when it is determined that no wrong word exists in the text content, the prompt message is sent to remind the user that no wrong word exists in the current composition text, so that the composition text is visual and vivid, and the user experience is improved.

According to the first aspect, in a possible implementation manner, the performing incorrect word detection on the text content according to the pre-trained semantic model includes:

splitting the text content by taking a sentence as a unit;

and performing semantic detection on each split sentence according to the semantic model so as to identify whether wrong characters exist in the current sentence.

In a second aspect, an embodiment of the present disclosure further provides an apparatus for detecting and identifying a wrong word, including:

the image acquisition module is used for acquiring composition picture information of a user;

the detection and identification module is used for detecting and identifying the composition picture information so as to obtain text content corresponding to the composition picture information;

and the error correction prompt module is used for carrying out wrong word detection on the text content according to the pre-trained semantic model and generating corresponding prompt information according to the detection result.

According to the second aspect, in a possible implementation manner, the detection and identification module is specifically configured to:

According to the second aspect, in a possible implementation manner, the detection and identification module is further specifically configured to:

and if the recognition candidate words with the maximum probability of the corresponding positions in the recombined single-line image strip are different from those in the original single-line image strip, performing semantic recognition on the positions according to a pre-trained semantic model based on the recognized single character texts to obtain the corresponding single character texts. According to the second aspect, in a possible implementation manner, the error correction prompting module is specifically configured to:

According to the second aspect, in a possible implementation manner, the error correction prompting module is further specifically configured to:

According to the second aspect, in a possible implementation manner, the error correction prompting module is specifically configured to:

splitting the text content by taking a sentence as a unit;

In a third aspect, an embodiment of the present disclosure further provides an electronic device, including: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, the processor and the memory communicate via the bus when the electronic device is running, and the machine-readable instructions, when executed by the processor, perform the steps of the method for identifying and detecting a faulty word according to the first aspect or any one of the possible embodiments of the first aspect.

In a fourth aspect, this disclosed embodiment further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method for detecting and identifying a faulty word described in the above first aspect or any one of the possible implementation manners of the first aspect are performed.

For the description of the effects of the above-mentioned device, electronic device and computer-readable storage medium, reference is made to the above description of the method for detecting and identifying wrong words, and no further description is given here.

In order to make the aforementioned objects, features and advantages of the present disclosure more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly described below, and the drawings herein incorporated in and forming a part of the specification illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the technical solutions of the present disclosure. It is appreciated that the following drawings depict only certain embodiments of the disclosure and are therefore not to be considered limiting of its scope, for those skilled in the art will be able to derive additional related drawings therefrom without the benefit of the inventive faculty.

Fig. 1 is a flowchart illustrating a method for detecting and identifying a wrong word according to an embodiment of the present disclosure;

fig. 2 shows a flowchart of a method for detecting an identification picture and obtaining text content according to an embodiment of the present disclosure.

FIG. 3 is a flow chart illustrating another method for detecting a recognition picture and obtaining text content according to an embodiment of the disclosure;

FIG. 4 is a flow chart illustrating a method for detecting incorrect words and generating prompt messages according to an embodiment of the present disclosure;

FIG. 5 is a schematic diagram illustrating an apparatus for detecting and identifying a wrong word according to an embodiment of the present disclosure;

fig. 6 shows a schematic diagram of an electronic device provided by an embodiment of the present disclosure.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present disclosure more clear, the technical solutions of the embodiments of the present disclosure will be described clearly and completely with reference to the drawings in the embodiments of the present disclosure, and it is obvious that the described embodiments are only a part of the embodiments of the present disclosure, not all of the embodiments. The components of the embodiments of the present disclosure, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present disclosure, presented in the figures, is not intended to limit the scope of the claimed disclosure, but is merely representative of selected embodiments of the disclosure. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the disclosure without making creative efforts, shall fall within the protection scope of the disclosure.

According to research, examination and reading are carried out by examination and approval personnel on articles handwritten by students, students or researchers word by word and sentence by sentence in the manuscript checking stage. However, because the drafter inevitably has carelessness and miswritten situations in the process of writing, written articles are inevitably wrongly written and wrongly written, and wrongly written characters and incorrectly expressed places are caused, which brings about great troubles to examination and approval personnel. For example, in the teaching of Chinese, composition teaching is to develop the writing ability of students. However, in the process of writing a composition, miswritten characters often occur due to miswriting or carelessness, which brings a certain trouble to teachers or parents when checking or reading the composition.

Based on the research, the method for detecting and identifying the wrong words comprises the steps of obtaining the composition picture information of the user, detecting and identifying the composition picture information to obtain the text content corresponding to the composition picture information, detecting the wrong words of the text content according to a pre-trained semantic model, automatically detecting the wrong words in the composition, further reducing the trouble of an approver in the review process, saving the time for the approver to find the wrong words, and improving the review efficiency.

The above-mentioned drawbacks are the results of the inventor after practical and careful study, and therefore, the discovery process of the above-mentioned problems and the solutions proposed by the present disclosure to the above-mentioned problems should be the contribution of the inventor in the process of the present disclosure.

It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.

To facilitate understanding of the present embodiment, first, a method for detecting and identifying a wrong word disclosed in the embodiments of the present disclosure is described in detail, where an execution subject of the method for detecting and identifying a wrong word provided in the embodiments of the present disclosure is generally a computer device with certain computing capability, and the computer device includes, for example: a terminal device including, but not limited to, other portable devices such as a mobile phone with a touch sensitive surface (e.g., a touch screen display and/or a touch pad), a laptop computer, a family computer, or a tablet computer, or a server or other processing device. It should also be understood that in some embodiments the terminal device is not a portable communication device, but is a desktop computer having a touch-sensitive surface.

In some possible implementations, the method for detecting and identifying a wrong word can be implemented by a processor calling computer readable instructions stored in a memory.

The method for detecting and identifying wrong words provided by the embodiment of the present disclosure is described below by taking an execution subject as a terminal device as an example.

Referring to fig. 1, a flowchart of a method for detecting and identifying a wrong word provided by the embodiment of the present disclosure is shown, where the method includes steps S101 to S103:

s101, composition picture information of the user is obtained.

Specifically, the terminal device may photograph the composition of the user through the camera to obtain the composition picture information, or may obtain the composition picture information in a manner of scanning the composition of the user. When the composition length is long, for example, when the same composition is distributed on two pages of paper, different pages can be photographed respectively to acquire different composition picture information respectively, and the terminal device can mark according to the sequence of acquiring the pictures to facilitate subsequent judgment processing.

It should be noted that the composition mentioned in the embodiments of the present application is not limited to the composition itself, and other texts (such as diary, letter, novel, story, pen-on, paper, technical document or operation manual) recorded in the form of text paragraphs may be referred to as the composition.

In some possible embodiments, after the composition is photographed or scanned, the obtained picture may be further enhanced to improve the recognition accuracy.

In some possible embodiments, the composition language may be chinese, arabic, alphabetic, english, japanese, korean, and the like, and of course, the composition language may also be a mixture of various languages, and the details thereof are not repeated.

S102, detecting and identifying the composition picture information to obtain text content corresponding to the composition picture information.

Specifically, the text of each word may be obtained by detecting and recognizing a single word in the composition picture information, and then the obtained texts of each word are combined in sequence to obtain the text content corresponding to the composition picture information.

S103, carrying out wrong word detection on the text content according to a pre-trained semantic model, and generating corresponding prompt information according to a detection result.

The semantic model refers to a network model which is obtained through training and can understand potential features of a language. For example, models with different functions, such as a model capable of finding out time, place name, and character name in a text, a model capable of matching an ancient poem, a model capable of matching a grammar detection, a model capable of matching a chiropractic manipulation, and a model capable of matching an english fixed match, may be trained for different tasks. The semantic model in the embodiment of the application is a semantic model capable of predicting the current word based on the context, and is obtained by improvement on the basis of a bert bidirectional semantic model.

In the embodiment of the disclosure, by acquiring the composition picture information of the user, detecting and identifying the composition picture information to acquire the text content corresponding to the composition picture information, and performing wrong character detection on the text content according to the pre-trained semantic model, the wrong characters in the composition can be automatically detected, and corresponding prompt information is generated according to the detection result to remind the user, so that an approver (or a reviewer, such as a teacher, a parent, an editor and the like) can intuitively and quickly find the wrong characters during reviewing, the time for the approver to find the wrong characters is saved, the reviewing efficiency is improved, and further the trouble of the approver in the reviewing process is reduced.

The above-mentioned S102 to S103 will be described in detail with reference to specific embodiments.

As to the above S102, when detecting and identifying the composition picture information to obtain the text content corresponding to the composition picture information, as shown in fig. 2, the following S1021 to S1026 may be included:

and S1021, segmenting the composition picture information to obtain a plurality of single-character image blocks, and determining the position of each single-character image block in the composition picture information.

For example, a coordinate system may be established according to the composition picture information, then the coordinates of each line of text in the composition picture information are determined, then each line of text is segmented from the composition picture information according to the coordinates, and then each line of text is segmented to obtain a plurality of single-character image blocks. Because each single-character image block is provided with coordinate information, the position of each single-character image block in the composition picture information can be determined according to the coordinate information.

And S1022, detecting and identifying each single-character image block to obtain an identification candidate word corresponding to each single-character image block and the probability of each identification candidate word.

For example, a plurality of recognition candidate words and a probability of each recognition candidate word possible at each word position may be recognized according to a trained recognition network model. For example, the segmented image blocks of a plurality of single characters are "i", "love", "mouth", and "country", respectively, and the probabilities of the recognition candidate words at each position and each recognition candidate word after being recognized by the recognition network model are shown in table 1 below:

table 1: recognition candidate word for each position and probability table thereof

And S1023, determining the recognition candidate words with the highest probability and larger than a specific threshold in the recognition candidate words corresponding to each single-word image block as the single-word texts corresponding to the single-word image blocks.

Specifically, at a certain position, if the probability corresponding to the recognition candidate word with the highest probability is greater than a specific threshold, it is determined that the recognition candidate word with the highest probability is the single-word text corresponding to the single-word image block, that is, the recognition candidate word with the highest probability is the word written by the contributor at the certain position.

It is understood that the specific threshold may be set according to the identification network model specifically applied at present, for example, in the embodiment of the present application, the specific threshold may be 0.9. Of course, in other embodiments, the specific threshold may be other values (e.g., 0.8), which is not limited herein. However, in order to ensure the text recognition accuracy, the specific threshold should be set as high as possible, for example, about 0.9. Therefore, as shown in table 1 above, the single character text corresponding to position 1 is me, the single character text corresponding to position 2 cannot be determined, the single character text corresponding to position 3 is kou, and the single character text corresponding to position 4 is country.

Since the sum of the probabilities of all the recognition candidate words is 1, when the specific threshold is set to 0.5 or more, there is only one recognition candidate word having the highest probability and being greater than the specific threshold, and there are no two or more recognition candidate words.

And S1024, based on the recognized single character texts, recognizing the single character image blocks of which the probability of all recognized candidate characters is smaller than the specific threshold value according to a pre-trained semantic model to obtain the corresponding single character texts.

Since the probability of each recognition candidate word at the position 2 is not greater than a specific threshold, the single word text at the position 2 cannot be confirmed according to the recognition network model, and therefore, further recognition is needed to confirm the single word text at the position 2.

Specifically, the position 2 is detected according to the semantic model, a plurality of semantic candidate words and the probability of each semantic candidate word at the position 2 can be obtained, the probability of identifying the candidate words is adjusted according to the semantic candidate words and the probability thereof, and the identification candidate word with the highest probability after adjustment is confirmed as the single-word text at the position 2.

For example, the continuous, longest, recognized text can be obtained, and the single word text at position 2 can be identified according to the semantic model. For example, according to the previous word "me", and the following two word mouths and countries, through semantic retrieval and analysis, it is determined that semantic candidate words "love", "hate", and "cherish" may be included between "me" and "country", and assuming that the probability of "love" is 0.63, the probability of "hate" is 0.34, and the probability of "cherish" is 0.02, and since "love" also appears in the recognition candidate words, the probability of recognizing "love" in the candidate words can be adjusted according to the semantic candidate words and the probabilities thereof, for example, the adjusted probability of recognizing "love" in the candidate words reaches 0.8 and is the highest, and then the "love" word is determined as the single word text of the position 2. The method for adjusting the probability of identifying the candidate word according to the semantic candidate word and the probability thereof is not limited herein, and for example, a normalization processing method may be adopted.

It is understood that, in some embodiments, if the single-word texts corresponding to all the single-word image blocks can be identified through the step S1023, the step S1024 may be omitted.

S1025, combining the single-word texts according to the position of each single-word text to obtain the text content.

Specifically, since each single character text carries coordinate information, a plurality of single character texts can be combined according to the position of each single character text to obtain text content consistent with the content in the composition picture information.

In the embodiment of the application, a plurality of single-character image blocks can be obtained by segmenting composition picture information, a single-character text corresponding to each single-character image block is obtained by identifying each single-character image block, and when the single-character text at a certain position cannot be identified, the single-character text at the position is further confirmed through a semantic model, so that the accuracy of composition picture identification is improved.

In order to improve the recognition accuracy of the handwritten composition picture, in another embodiment, regarding step S102, when detecting and recognizing the composition picture information to obtain the text content corresponding to the composition picture information, as shown in fig. 3, the method may include the following steps S102a to S102 i:

s102a, segmenting the composition picture information to obtain a plurality of single-character image blocks, and determining coordinates of each single-character image block.

The step is similar to the step S1021, and is not described herein again.

S102b, performing detection and identification on each single-word image block to obtain an identification candidate word corresponding to each single-word image block and a probability of each identification candidate word.

The steps are similar to the step S1022 described above, and are not described herein again.

And S102c, combining a plurality of single-character image blocks in the same row according to the coordinates of each single-character image block to obtain a recombined single-row image strip.

In the present embodiment, after each single-character image block is detected and recognized, it is reconstructed in units of lines.

S102d, performing segmentation on the composition picture information to obtain a plurality of original single-line image strips, and determining coordinates of each original single-line image strip.

The step is similar to the step S1021, and is not described herein again.

S102e, performing detection and identification on each original single-line image strip to obtain an identification candidate word corresponding to each position in each original single-line image strip and a probability of each identification candidate word.

Similar to the single-character recognition, in the present embodiment, the detection and recognition of the character at each position in each original single-line image strip are also performed in units of lines.

S102f, judging whether the recognition candidate words with the maximum probability of the corresponding positions in the recombined single-line image strip and the original single-line image strip are the same or not; if yes, go to step S102 g; if not, go to step S102 h.

It can be understood that, because the recognition detection method adopted by each position in the recombined single-line image strip is different from the recognition detection method adopted by the corresponding position in the original single-line image strip, if the recognition candidate words with the maximum probability at the same position are the same, the recognition candidate word with the maximum probability is the word written by the contributor; if the reconstructed single-line image block does not have the recognition candidate character at the X position, but the original single-line image strip does have the corresponding recognition candidate character at the X position, it indicates that there is a missing detection condition in the single character recognition process, and then step S102h may be executed to further detect and recognize the position, thereby further ensuring the recognition accuracy.

S102g, determining the recognition candidate word with the highest probability as the single character text corresponding to the position.

S102h, based on the recognized single character text, performing semantic recognition on the position according to a pre-trained semantic model to obtain the corresponding single character text.

The step is similar to the step S1024, and is not described herein again.

S102i, combining a plurality of single-word texts according to the coordinates of each single-word text to obtain the text content.

The step is similar to the step S1025, and is not described herein again.

In the embodiment of the disclosure, because the single-word detection and recognition and the single-line detection and recognition are combined, and the recognition candidate word with the maximum probability is determined as the character written by the manuscript writer under the condition that the results of the recognition candidate words with the maximum probability detected by the two are the same, and if the two recognition candidate words are different, the semantic detection is required to further judge, so that the recognition precision of the composition picture is improved. For the above S103, when performing wrong word detection on the text content according to the pre-trained semantic model and generating corresponding prompt information according to the detection result, as shown in fig. 4, steps S1031 to S1034 may be included:

and S1031, performing semantic detection on the single-word text on each position in sequence according to a predetermined sequence on the text content according to the semantic model to obtain semantic candidate words of each position and the probability of each semantic candidate word.

Specifically, semantic detection is performed on the text content identified in step S102 one by one according to the semantic model to obtain semantic candidate words at each position and a probability of each semantic candidate word. For example, when semantic detection is performed on a single word w at a position i, the current word w is shielded and is regarded as unknown, and then semantic candidate words corresponding to the position i and the probability of each semantic candidate word are obtained according to the context semantics of the words before and after the unknown i.

S1032, judging whether the current single character text at each position belongs to candidate characters to be selected at the position; the candidate word to be selected is a semantic candidate word with the probability at the position larger than a reference threshold; if yes, go to step S1033; if not, go to step S1034.

The semantic candidate words with the probability larger than the reference threshold are determined as candidate words to be selected, if the current single word text is in the candidate words to be selected, the word written by the contributor is a correct word, and if the current single word text is not in the candidate words to be selected, the word written by the contributor is a wrongly written word. It is to be understood that the reference threshold may be set according to a specific semantic model, and is not limited herein.

S1033, determining that no wrong word exists in the text content, and generating prompt information.

When it is determined that there is no wrong word in the text content, a simple prompt is given to the user. Wherein, the prompt message may include at least one of voice prompt message, text prompt message, animation prompt message and flashing light, so when prompting the user through the terminal device, the method may include: the user is prompted in at least one of a voice form, a text form, an animation form, and a flashing form.

S1034, determining the single character text at the position as a wrong character, and displaying the wrong character and the correct character at the same time; wherein, the correct word is the semantic candidate word with the highest probability at the position.

With reference to step S1022, the word text at position 3 is "mouth", and the semantic detection on position 3 yields that the probability of the semantic candidate word "middle" is 0.54, the probability of the "ancestor" is 0.43, the probability of "beauty" is 0.02, etc., and the "mouth" does not belong to the candidate word "middle" at the position, so that the word text "mouth" at position 3 is determined to be wrong, and the wrong word "mouth" and the corresponding correct word "middle" are displayed at the same time. Therefore, wrong words can be prompted, and a correct result can be indicated, so that a user can understand the composition well.

Furthermore, the semantic recognition can be performed on sentences by taking the sentences as units, and further, wrong words can be recognized. The sentences may be divided in units of punctuation marks, phrases or short sentences, which is not limited herein. Therefore, in some embodiments, the performing incorrect word detection on the text content according to the pre-trained semantic model may further include: splitting the text content by taking a sentence as a unit; and performing semantic detection on each split sentence according to the semantic model so as to identify whether wrong characters exist in the current sentence.

For example, "i is a small contributor" should be written, but since the contributor wrongly writes, "i is a wood contributor", it is easy to determine that wrongly written characters are wood; for example, writing 'I love mom' into 'I love good', easily determining that good is wrongly written words; for example, the 'grass on the departure circle' is written as 'grass on the departure circle', and because of the fixed collocation, the wrongly written characters can be easily determined.

It will be understood by those skilled in the art that in the method of the present invention, the order of writing the steps does not imply a strict order of execution and any limitations on the implementation, and the specific order of execution of the steps should be determined by their function and possible inherent logic.

Based on the same inventive concept, the embodiment of the present disclosure further provides a device for detecting and identifying a wrong word corresponding to the method for detecting and identifying a wrong word, and since the principle of solving the problem of the device in the embodiment of the present disclosure is similar to that of the method for detecting and identifying a wrong word in the embodiment of the present disclosure, the implementation of the device can refer to the implementation of the method, and repeated details are not repeated.

Referring to fig. 5, a schematic diagram of an architecture of an apparatus for detecting and identifying a wrong word provided in an embodiment of the present disclosure is shown, where the apparatus includes: an image acquisition module 501, a detection and identification module 502 and an error correction prompt module 503; wherein the content of the first and second substances,

According to the text content detection method and device, the composition picture information of the user is obtained, the composition picture information is detected and identified to obtain the text content corresponding to the composition picture information, the text content is subjected to wrong word detection according to the pre-trained semantic model, the wrong words in the composition can be automatically detected, corresponding prompt information is generated according to the detection result, the user is reminded, a teacher or parents can find the wrong words intuitively and quickly when the wrong words are read and approved, the time for finding the wrong words by an approver is saved, the review efficiency is improved, and the trouble of the approver in the review process is further reduced.

In a possible implementation manner, the detection and identification module is specifically configured to:

In a possible implementation manner, the detection and identification module is further specifically configured to:

and if the recognition candidate words with the maximum probability of the corresponding positions in the recombined single-line image strip are different from those in the original single-line image strip, performing semantic recognition on the positions according to a pre-trained semantic model based on the recognized single character texts to obtain the corresponding single character texts.

In a possible implementation manner, the error correction prompting module is specifically configured to:

In a possible implementation manner, the error correction prompting module is further specifically configured to:

splitting the text content by taking a sentence as a unit;

The description of the processing flow of each module in the device and the interaction flow between the modules may refer to the related description in the above method embodiments, and will not be described in detail here.

Based on the same technical concept, the embodiment of the disclosure also provides an electronic device. Referring to fig. 6, a schematic structural diagram of an electronic device 700 provided in the embodiment of the present disclosure includes a processor 701, a memory 702, and a bus 703. The memory 702 is used for storing execution instructions and includes a memory 7021 and an external memory 7022; the memory 7021 is also referred to as an internal memory and temporarily stores operation data in the processor 701 and data exchanged with an external memory 7022 such as a hard disk, and the processor 701 exchanges data with the external memory 7022 via the memory 7021.

In this embodiment, the memory 702 is specifically configured to store application program codes for executing the scheme of the present application, and is controlled by the processor 701 to execute. That is, when the electronic device 700 is operating, communication between the processor 701 and the memory 702 is via the bus 703, which enables the processor 701 to execute application program code stored in the memory 702.

The Memory 702 may be, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Read-Only Memory (EPROM), an electrically Erasable Read-Only Memory (EEPROM), and the like. The memory 702 is configured to store a program, and the processor 703 executes the program after receiving an execution instruction, and a method executed by the electronic device 200 defined by a flow disclosed in any embodiment of the invention described later may be applied to the processor 703 or implemented by the processor 703.

The processor 701 may be an integrated circuit chip having signal processing capabilities. The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

It is to be understood that the illustrated structure of the embodiment of the present application does not specifically limit the electronic device 700. In other embodiments of the present application, the electronic device 700 may include more or fewer components than shown, or combine certain components, or split certain components, or a different arrangement of components. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

The embodiments of the present disclosure also provide a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the computer program performs the steps of the method for detecting and identifying a wrong word described in the above method embodiments. The storage medium may be a volatile or non-volatile computer-readable storage medium.

The computer program product of the method for detecting and identifying a wrong word provided by the embodiment of the present disclosure includes a computer readable storage medium storing a program code, where instructions included in the program code may be used to execute the steps of the method for detecting and identifying a wrong word described in the above method embodiment, which may be referred to the above method embodiment specifically, and are not described herein again.

The embodiments of the present disclosure also provide a computer program, which when executed by a processor implements any one of the methods of the foregoing embodiments. The computer program product may be embodied in hardware, software or a combination thereof. In an alternative embodiment, the computer program product is embodied in a computer storage medium, and in another alternative embodiment, the computer program product is embodied in a Software product, such as a Software Development Kit (SDK), or the like.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the system and the apparatus described above may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system, apparatus, and method may be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one logical division, and there may be other divisions when actually implemented, and for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or units through some communication interfaces, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present disclosure may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such understanding, the technical solution of the present disclosure may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing an electronic device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present disclosure. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

Finally, it should be noted that: the above-mentioned embodiments are merely specific embodiments of the present disclosure, which are used for illustrating the technical solutions of the present disclosure and not for limiting the same, and the scope of the present disclosure is not limited thereto, and although the present disclosure is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive of the technical solutions described in the foregoing embodiments or equivalent technical features thereof within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present disclosure, and should be construed as being included therein. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.

Claims

1. A method for detecting and identifying wrong words is characterized by comprising the following steps:

acquiring composition picture information of a user;

2. The method according to claim 1, wherein the detecting and identifying the composition picture information to obtain the text content corresponding to the composition picture information comprises:

3. The method according to claim 2, wherein after determining the recognition candidate word with the highest probability and larger than a specific threshold from among the recognition candidate words corresponding to each single-word image block as the single-word text corresponding to the single-word image block, the method for detecting and recognizing wrong words further comprises:

and based on the recognized single word text, recognizing the single word image blocks of which the probability of recognizing the candidate words is smaller than the specific threshold according to a pre-trained semantic model to obtain the corresponding single word text.

4. The method of claim 3, wherein the identifying, based on the recognized word texts, the word image blocks with all candidate words recognition probabilities smaller than the specific threshold according to the trained semantic model to obtain the corresponding word texts comprises:

based on the recognized single word text, recognizing the single word image blocks of which the probabilities of all recognized candidate words are smaller than the specific threshold according to the semantic model to obtain the corresponding semantic candidate words and the probability of each semantic candidate word;

5. The method according to claim 1, wherein the detecting and identifying the composition picture information to obtain the text content corresponding to the composition picture information comprises:

6. The method according to claim 5, wherein before combining a plurality of single-word texts according to the coordinates of each single-word text to obtain the text content, the method for detecting and identifying wrong words further comprises:

7. The method according to any one of claims 1 to 6, wherein the detecting the wrong word of the text content according to the pre-trained semantic model and generating the corresponding prompt information according to the detection result comprises:

8. The method as claimed in claim 7, wherein after determining whether the current word text at each position belongs to the candidate word to be selected at the position, the method for detecting and identifying wrong words further comprises:

9. The method according to any one of claims 1-6, wherein the detecting the text content for miswords according to the pre-trained semantic model comprises:

splitting the text content by taking a sentence as a unit;

10. An apparatus for detecting and recognizing a wrong word, comprising:

11. An electronic device, comprising: a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory communicating via the bus when the electronic device is operating, the machine-readable instructions when executed by the processor performing the steps of the method for detecting and identifying erroneous words according to any one of claims 1 to 9.

12. A computer-readable storage medium, having stored thereon a computer program for executing the steps of the method according to any one of claims 1 to 9 when executed by a processor.