Method and device for analyzing the interpretation of an object

The preset model trained through sentence splicing method, combined with syntactic analysis and candidate interpretation matching, solves the problem of low vocabulary translation efficiency in the existing technology, and achieves efficient and accurate vocabulary translation effect.

CN114841174BActive Publication Date: 2024-12-06GUANGZHOU SHIYUAN ELECTRONICS CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110142998.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-02-02
Publication Date
2024-12-06
Estimated Expiration
2041-02-02

AI Technical Summary

Technical Problem

Existing translation software and electronic dictionaries are inefficient and have low accuracy when translating vocabulary in sentences, and require training a large number of labels and data samples, resulting in large overhead for model training and storage.

Method used

The preset model is trained by sentence splicing, and through syntax analysis and candidate interpretation matching, the scope of candidate interpretation is narrowed, the translation efficiency and accuracy are improved, and the data volume requirements of model training are reduced.

Benefits of technology

It improves the efficiency and accuracy of vocabulary translation, reduces model training and storage requirements, simplifies the data sampling process, and realizes efficient translation of vocabulary in sentences.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114841174B_ABST
    Figure CN114841174B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and device for analyzing the interpretation of an object. The method comprises: obtaining a target sentence and words to be analyzed in the target sentence; performing syntactic analysis on the target sentence to determine the target part of speech of the words to be analyzed in the target sentence; determining the candidate interpretation corresponding to the words to be analyzed according to the target part of speech; processing the target sentence, the words to be analyzed and the candidate interpretations based on a preset model to obtain the target interpretation of the words to be analyzed in the target sentence, wherein the preset model is trained by sentence splicing. The present invention solves the technical problem of low translation efficiency when translating vocabulary in a sentence in the prior art.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of text translation, and in particular to a method and device for analyzing the interpretation of an object. Background Art

[0002] Currently, many translation software and electronic dictionaries have translation functions. However, most translation software translates the entire text input by the user based on the entire sentence or a single word. When an electronic dictionary translates a word, it outputs the entire meaning of the word to be translated. In the prior art, only a few translation software and electronic dictionaries can translate the meaning of a specified word in a sentence.

[0003] At present, a classification model is usually used to translate designated words in a sentence, that is, examples of words with different meanings are used, and the meanings of the words are used as labels to train the classification model. Therefore, for a new sentence and designated words, it is only necessary to input the sentence and the designated words into the corresponding classification model to obtain the meaning of the classification model words in the sentence.

[0004] However, in the prior art, if a classification model is trained, the labels of the classification model are all the interpretations of all the words, resulting in a very large number of interpretations. For example, if a word has at least 2 interpretations, then for k words, there are at least 2k labels. For 10,000 words, it is necessary to train a classification model with at least 20,000 labels. Moreover, the obtained classification model not only has a low translation accuracy when translating words, but also has low translation efficiency.

[0005] In addition, if a separate classification model is trained for each word, at least hundreds of examples must be collected for each meaning of each word, which affects the data sampling. In addition, the existing classification model training cannot guarantee the training effect of a model with only hundreds to thousands of training data. Moreover, since the vocabulary in daily life is large, if a separate classification model is trained for each word, the cost of model training and storage will increase.

[0006] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0007] The embodiment of the present invention provides a method and device for analyzing the interpretation of an object, so as to at least solve the technical problem of low translation efficiency when translating words in a sentence in the prior art.

[0008] According to one aspect of an embodiment of the present invention, a method for analyzing the interpretation of an object is provided, including: obtaining a target sentence and words to be analyzed in the target sentence; performing syntactic analysis on the target sentence to determine a target part of speech of the words to be analyzed in the target sentence; determining candidate interpretations corresponding to the words to be analyzed according to the target part of speech; processing the target sentence, the words to be analyzed, and the candidate interpretations based on a preset model to obtain a target interpretation of the words to be analyzed in the target sentence, wherein the preset model is trained by sentence concatenation.

[0009] Furthermore, the method for analyzing the interpretation of an object also includes: in a preset model, obtaining a candidate sentence corresponding to the candidate interpretation, wherein the candidate sentence at least includes the word to be analyzed; splicing the candidate sentence with the target sentence to obtain a splicing result; and determining the target interpretation of the word to be analyzed in the target sentence from the candidate interpretation according to the splicing result.

[0010] Furthermore, the method for analyzing the interpretation of an object also includes: setting a separator between the candidate sentence and the target sentence, splicing the target sentence and the candidate sentence in the order of classification label, target sentence, separator, and candidate sentence to obtain a first splicing result; obtaining a first vector corresponding to the word to be analyzed in the target sentence and a second vector corresponding to the word to be analyzed in the candidate sentence from the first splicing result; determining the value of the classification label corresponding to the candidate sentence based on the first vector and the second vector; determining whether the first interpretation of the word to be analyzed in the target sentence and the second interpretation of the word to be analyzed in the candidate sentence are the same based on the value of the classification label to obtain a splicing result.

[0011] Furthermore, the method for analyzing the interpretation of an object also includes: determining a scoring value corresponding to the second interpretation based on the first vector and the second vector, wherein the scoring value represents the accuracy of the second interpretation; when the scoring value is greater than a preset scoring value, determining the value of the classification label to be a first preset value; when the scoring value is less than or equal to the preset scoring value, determining the value of the classification label to be a second preset value.

[0012] Furthermore, the method for analyzing the object interpretation also includes: when the value of the classification label is a first preset value and the number of candidate interpretations is one, determining the candidate interpretation corresponding to the candidate sentence as the target interpretation.

[0013] Furthermore, the method for analyzing object interpretation also includes: when the value of the classification label is a first preset value and the number of candidate interpretations is multiple, obtaining the candidate score value corresponding to the first vector and the second vector under each candidate interpretation; and determining the candidate interpretation with the largest candidate score value as the target interpretation.

[0014] Furthermore, the method for analyzing object interpretation also includes: when the values ​​of the classification labels corresponding to all candidate sentences are the second preset values, obtaining the candidate score values ​​corresponding to the first vector and the second vector under each candidate interpretation; and determining the candidate interpretation with the largest candidate score value as the target interpretation.

[0015] Furthermore, the method for analyzing the interpretation of an object also includes: before processing the target sentence, the words to be analyzed and the candidate interpretations based on a preset model to obtain the target interpretation of the words to be analyzed in the target sentence, constructing a data sample, wherein the data sample includes at least a positive data sample and a negative data sample; and training the data sample to obtain the preset model.

[0016] Furthermore, the method for analyzing the interpretation of an object also includes: obtaining a first preset sentence; performing word segmentation processing on the first preset sentence and determining words to be replaced in the first preset sentence; obtaining synonyms corresponding to the words to be replaced; replacing the words to be replaced with synonyms according to preset grammatical rules to obtain a second preset sentence; and constructing a positive data sample based on the first preset sentence and the second preset sentence.

[0017] According to another aspect of an embodiment of the present invention, a method for analyzing the interpretation of an object is also provided, comprising: obtaining a sentence to be analyzed and a word to be analyzed in the sentence to be analyzed; concatenating the sentence to be analyzed with a candidate sentence to obtain a concatenated result, wherein the candidate sentence at least includes the word to be analyzed; determining a label corresponding to the concatenated result based on a first interpretation of the word to be analyzed in the sentence to be analyzed and a second interpretation of the word to be analyzed in the candidate sentence; and performing model training based on the sentence to be analyzed, the word to be analyzed and the labels corresponding to the concatenated result to obtain a preset model, so that the preset model outputs a target interpretation corresponding to the word to be analyzed.

[0018] According to another aspect of an embodiment of the present invention, there is also provided a device for analyzing the interpretation of an object, including: an acquisition module, used to acquire a target sentence and words to be analyzed in the target sentence; an analysis module, used to perform syntactic analysis on the target sentence and determine the target part of speech of the words to be analyzed in the target sentence; a determination module, used to determine the candidate interpretations corresponding to the words to be analyzed according to the target part of speech; and a processing module, used to process the target sentence, the words to be analyzed and the candidate interpretations based on a preset model to obtain the target interpretation of the words to be analyzed in the target sentence, wherein the preset model is trained by sentence splicing.

[0019] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned method for analyzing object interpretation when running.

[0020] According to another aspect of an embodiment of the present invention, a processor is provided, which is used to run a program, wherein the program is configured to execute the above-mentioned method for analyzing object interpretation when running.

[0021] In an embodiment of the present invention, a preset model is obtained by sentence concatenation so that the preset model predicts the target interpretation of the words to be analyzed in the target sentence. After obtaining the target sentence and the words to be analyzed in the target sentence, the target part of speech of the words to be analyzed in the target sentence is determined by performing syntactic analysis on the target sentence, and the candidate interpretations corresponding to the words to be analyzed are determined according to the target part of speech. Then, the target sentence, the words to be analyzed and the candidate interpretations are processed based on the preset model to obtain the target interpretation of the words to be analyzed in the target sentence.

[0022] In the above process, before translating the words to be analyzed, the word orientation of the words to be analyzed in the target sentence is first determined, which narrows the range of candidate interpretations corresponding to the words to be analyzed, improves the translation efficiency of the preset model for translating the words to be analyzed, and improves the accuracy of the translation. In addition, the preset model is trained based on sentence concatenation, so that only one model needs to be trained for all words, and there is no need to train models for all words separately. Moreover, training the preset model by sentence concatenation reduces the data volume requirements for model training and simplifies the data sampling process, thereby improving the training efficiency of the preset model.

[0023] It can be seen that the solution provided in the present application achieves the purpose of translating words in a sentence, thereby realizing the technical effect of improving the translation efficiency of words, and further solving the technical problem of low translation efficiency when translating words in a sentence in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this 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 drawings:

[0025] Figure 1 is a flow chart of a method for analyzing object interpretation according to an embodiment of the present invention;

[0026] Figure 2 is a flow chart of an optional method for analyzing object interpretation according to an embodiment of the present invention;

[0027] Figure 3 is a flow chart of an optional method for analyzing object interpretation according to an embodiment of the present invention;

[0028] Figure 4 is an optional sentence concatenation schematic diagram according to an embodiment of the present invention;

[0029] Figure 5 is a schematic diagram of training an optional preset model according to an embodiment of the present invention;

[0030] Figure 6 is a flow chart of a method for analyzing object interpretation according to an embodiment of the present invention;

[0031] Figure 7 2 is a schematic diagram of a device for analyzing the interpretation of an object according to an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] Example 1

[0035] According to an embodiment of the present invention, a method embodiment for analyzing the interpretation of an object 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 flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0036] In addition, it should be noted that the translation client can be used as the execution subject of the method improved by this embodiment, wherein the translation client can be but not limited to translation software, electronic dictionary, etc. The translation client can run in a terminal device, such as a computer, a mobile phone, a tablet, etc.

[0037] Figure 1is a flow chart of a method for analyzing object interpretation according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0038] Step S102, obtaining a target sentence and words to be analyzed in the target sentence.

[0039] In step S102, the word to be analyzed is the word to be translated, and the target sentence is the example sentence containing the word to be analyzed. For example, for the example sentence "Let me take your coat", "take" is the word to be analyzed, and the above example sentence is the target sentence.

[0040] Optional, Figure 2 A flowchart of an optional method for analyzing object interpretation is shown in Figure 2 In the example, the user inputs the target sentence "Let me take your coat" into the translation client, and determines that the word to be analyzed in the target sentence is "take". For example, the word to be analyzed in the target sentence can be determined by inputting the word to be analyzed into the translation client, or by specially marking the word to be analyzed in the target sentence. Then, the translation client can translate the target sentence and output the target interpretation of the word to be analyzed in the target sentence.

[0041] Step S104, performing syntactic analysis on the target sentence to determine the target part of speech of the word to be analyzed in the target sentence.

[0042] Optional, such as Figure 2 As shown, the translation client can use syntactic analysis tools such as NLTK / WordNet / jieba to perform syntactic analysis on the target sentence, and perform word segmentation and part-of-speech tagging on the target sentence, so as to obtain the target part-of-speech of the word to be analyzed in the target sentence.

[0043] It should be noted that analyzing the target part of speech of the word to be analyzed in the target sentence can narrow the scope of candidate interpretations and improve the efficiency of the preset model.

[0044] Step S106, determining candidate definitions corresponding to the word to be analyzed according to the target part of speech.

[0045] Optional, such as Figure 2As shown, the translation client obtains candidate interpretations according to the target part of speech corresponding to the word to be analyzed, for example, for the target sentence "Let me take your coat." and the word to be analyzed "take". First, the target sentence is segmented and POS tagged, and the part of speech of "take" in the target sentence is obtained as "V" verb. Therefore, the candidate interpretations are all interpretation entities of take as verbs, including: "v. take", "v. need", "v. take", "v. accept", "v. think", "v. spend", "v. ride", "v. execute", "v. understand", "v. take (course)", "v. eat (drink)".

[0046] Step S108, processing the target sentence, the word to be analyzed and the candidate interpretations based on a preset model to obtain a target interpretation of the word to be analyzed in the target sentence, wherein the preset model is trained by sentence concatenation.

[0047] Optional, such as Figure 2 As shown, after obtaining the candidate interpretations, the translation client inputs the target sentence, the words to be analyzed, and the candidate interpretations into the preset model in sequence, so that the preset model predicts the interpretation of the words to be analyzed in the target sentence, obtains the target interpretation, and outputs the target interpretation.

[0048] It should be noted that, in step S108, sentence concatenation refers to a method of concatenating the target sentence with the example sentences corresponding to the candidate interpretations, so as to determine the target interpretation of the word to be analyzed in the target sentence according to the candidate interpretation of the word to be analyzed in the example sentences. It is easy to notice that, by using sentence concatenation to predict the interpretation of the word to be analyzed in the target sentence, only one prediction model needs to be trained, and there is no need to train models for all words separately. Moreover, by using sentence concatenation to train the preset model, the requirement for the amount of data for model training is reduced, and the process of data sampling is simplified, thereby improving the training efficiency of the preset model.

[0049] Based on the scheme defined by the above steps S102 to S108, it can be known that in an embodiment of the present invention, a preset model is obtained by sentence concatenation so that the preset model predicts the target interpretation of the words to be analyzed in the target sentence. After obtaining the target sentence and the words to be analyzed in the target sentence, the target part of speech of the words to be analyzed in the target sentence is determined by performing syntactic analysis on the target sentence, and then the target sentence, the words to be analyzed and the target part of speech are processed based on the preset model to obtain the target interpretation of the words to be analyzed in the target sentence.

[0050] It is easy to notice that in the above process, before translating the words to be analyzed, the word orientation of the words to be analyzed in the target sentence is first determined, which narrows the range of candidate interpretations corresponding to the words to be analyzed, improves the translation efficiency of the preset model for translating the words to be analyzed, and improves the accuracy of translation. In addition, the preset model is trained based on sentence concatenation, so that only one model needs to be trained for all words, and there is no need to train models for all words separately. Moreover, the use of sentence concatenation to train the preset model reduces the data volume requirements for model training and simplifies the data sampling process, thereby improving the training efficiency of the preset model.

[0051] It can be seen that the solution provided in the present application achieves the purpose of translating words in a sentence, thereby realizing the technical effect of improving the translation efficiency of words, and further solving the technical problem of low translation efficiency when translating words in a sentence in the prior art.

[0052] In an optional embodiment, after performing syntactic analysis on the target sentence, determining the target part of speech of the word to be analyzed in the target sentence, and determining the candidate interpretation corresponding to the word to be analyzed based on the target part of speech, the translation client processes the target sentence, the word to be analyzed and the candidate interpretation based on a preset model to obtain the target interpretation of the word to be analyzed in the target sentence.

[0053] Specifically, in the preset model, a candidate sentence corresponding to the candidate interpretation is obtained, the candidate sentence is spliced ​​with the target sentence to obtain a splicing result, and the target interpretation of the word to be analyzed in the target sentence is determined from the candidate interpretation according to the splicing result. Among them, the candidate sentence includes at least the word to be analyzed. For example, the part of speech of "take" in the target sentence "Let me take your coat" is a verb, and the candidate interpretation "v. Expend" of "take" corresponds to the candidate sentence "I had heard an appeal could take years"; the candidate interpretation "v. Take" of "take" corresponds to the candidate sentence "I take a handkerchief from my pocket". "Let me take your coat" is spliced ​​with "I had heard an appeal could take years", and "Let me take your coat" is spliced ​​with "I take a handkerchief from my pocket" to obtain multiple splicing results, and finally the splicing results are analyzed to determine the target interpretation of "take" in the target sentence "Let me take your coat".

[0054] In an optional embodiment, if Figure 3 An optional method flow chart of analyzing the interpretation of an object is shown, comprising Figure 3 It can be seen that after the translation client obtains the target sentence and the words to be analyzed and generates candidate interpretations corresponding to the words to be analyzed, it extracts candidate sentences for each candidate interpretation, and concatenates the candidate sentences with the target sentence to obtain the concatenation result, and determines the target interpretation of the words to be analyzed in the target sentence based on the concatenation result.

[0055] Specifically, a separator is set between the candidate sentence and the target sentence, and the target sentence and the candidate sentence are spliced ​​in the order of classification label, target sentence, separator, and candidate sentence to obtain a first splicing result. Then, a first vector corresponding to the word to be analyzed in the target sentence and a second vector corresponding to the word to be analyzed in the candidate sentence are obtained from the first splicing result, and the value of the classification label corresponding to the candidate sentence is determined according to the first vector and the second vector. Finally, according to the value of the classification label, it is determined whether the first interpretation of the word to be analyzed in the target sentence is the same as the second interpretation of the word to be analyzed in the candidate sentence, and a splicing result is obtained.

[0056] Optional, such as Figure 4 In the sentence concatenation diagram shown in the figure, the target sentence (i.e. Figure 4 sentence1) is "Let me take your coat", and the candidate sentence (i.e. Figure 4 The target sentence and the candidate sentence are concatenated, and the first concatenation result can be expressed as: [CLS]+sentence1 after word segmentation+[SEP]+sentence2 after word segmentation+[SEP], where [CLS] represents the classification label, which contains sentence-level information, and the numerical value corresponding to the classification label indicates whether the meaning of the word to be analyzed in the target sentence is the same as that in the candidate sentence; the first vector of the word to be analyzed in the target sentence and the second vector of the word to be analyzed in the candidate sentence represent the meaning of the word to be analyzed in the target sentence and the candidate sentence respectively; [SEP] is used as a separator to separate two sentences / words.

[0057] After obtaining the target sentence and the candidate sentence, the translation client inputs the target sentence and the candidate sentence into the Bert model. In this process, the present application has transformed the training method of the Bert model. For ordinary text matching tasks, it is only necessary to input two sentences into the Bert model and directly train a model that predicts whether the two sentences match. In the present application, it is predicted whether the words to be analyzed have the same meaning in the two sentences (i.e., the target sentence and the candidate sentence). After inputting the target sentence and the candidate sentence into the Bert model and undergoing a round of training, the second splicing result can be obtained: [CLS] + first vector + [SEP] + second vector + [SEP] Continue training. Finally, the vector at the [CLS] position is taken to predict the model label, that is, the output classification target is 0 / 1. Among them, 0 indicates that the meaning of the words to be analyzed in the two sentences is inconsistent, and 1 indicates that the meaning of the words to be analyzed in the two sentences is consistent.

[0058] For example, input "[CLS]Let me take your coat[SEP]I take a handkerchief frommy pocket[SEP]" into the Bert model. After one round of training, take: the vector at the [CLS] position, the vector at the "take" position in the first sentence (i.e. the first vector), the vector at the [SEP] position, the vector at the "take" position in the second sentence (i.e. the second vector), and the vector at the [SEP] position at the end as the input for the second round of training. Continue training, and finally take the vector at the [CLS] position after the second round of training to fit the final label (here label = 1), and predict whether the meaning of "take" in the two sentences is consistent.

[0059] Further, after obtaining the first vector and the second vector, the value of the classification label is determined according to the first vector and the second vector. Specifically, the score value corresponding to the second interpretation is determined according to the first vector and the second vector, wherein, when the score value is greater than the preset score value, the value of the classification label is determined to be the first preset value; when the score value is less than or equal to the preset score value, the value of the classification label is determined to be the second preset value. Optionally, the score value represents the accuracy of the second interpretation, wherein the similarity can be used to determine the score value. For example, if the similarity between the first vector and the second vector is greater than the preset similarity, the value of the classification label is determined to be 1, indicating that the interpretation of the word to be analyzed in the candidate sentence is the same as the interpretation of the word to be analyzed in the target sentence; if the similarity between the first vector and the second vector is less than or equal to the preset similarity, the value of the classification label is determined to be 0, indicating that the interpretation of the word to be analyzed in the candidate sentence is different from the interpretation of the word to be analyzed in the target sentence.

[0060] Optionally, when the value of the classification label is the first preset value and the number of candidate interpretations is one, the candidate interpretation corresponding to the candidate sentence is determined as the target interpretation. That is, if only one candidate interpretation with a classification label value of 1 is obtained, then the candidate interpretation is the target interpretation of the word to be analyzed in the target sentence.

[0061] Optionally, when the value of the classification label is a first preset value and there are multiple candidate interpretations, the candidate score values ​​corresponding to the first vector and the second vector under each candidate interpretation are obtained, and the candidate interpretation with the largest candidate score value is determined as the target interpretation. For example, if there are multiple candidate interpretations with a classification label value of 1, the score values ​​of the candidate interpretations are sorted, and the candidate interpretation with the highest score value is taken as the target interpretation of the word to be analyzed in the target sentence.

[0062] Optionally, when the values ​​of the classification labels corresponding to all candidate sentences are the second preset values, the candidate score values ​​corresponding to the first vector and the second vector are obtained under each candidate interpretation, and the candidate interpretation with the largest candidate score value is determined as the target interpretation. For example, if the values ​​of all the classification labels are 0, they can still be sorted according to the score value, and the candidate interpretation with the highest score value is taken as the target interpretation. However, in this scenario, there is a great possibility that the interpretation of the word to be analyzed in the target sentence is a new interpretation that does not appear in the library. Therefore, in this way, the vocabulary interpretation in the library can also be updated.

[0063] It should be noted that, in practical applications, the preset score value may be adjusted to avoid the situation where the candidate interpretation is the same as the target interpretation when the values ​​of the classification labels are all 0.

[0064] In an optional embodiment, before processing the target sentence, the word to be analyzed and the candidate interpretations based on the preset model to obtain the target interpretation of the word to be analyzed in the target sentence, it is also necessary to construct a data sample and train the data sample to obtain the preset model. The data sample includes at least a positive data sample and a negative data sample.

[0065] Optional, such as Figure 5 The training diagram of the preset model shown in Figure 5 It can be seen that after constructing the data sample, positive samples and negative samples can be generated, and the number of positive samples and negative samples can be balanced through data enhancement, and training, testing, and validation data sets can be generated at the same time. Then, the above data sets are input into the model for training to obtain the preset model.

[0066] In an alternative embodiment, during the process of constructing positive data samples, first obtain a first preset sentence, perform word segmentation on the first preset sentence, and determine the words to be replaced in the first preset sentence. Then, obtain the synonymous words corresponding to the words to be replaced, and replace the words to be replaced with the synonymous words according to the preset grammar rules to obtain a second preset sentence. Finally, construct positive data samples based on the first preset sentence and the second preset sentence.

[0067] Optionally, during the process of constructing training, validation, and test data sets, split the existing context, entity reference (i.e., the word to be analyzed), target entity (i.e., the target interpretation), and candidate entity (i.e., the candidate interpretation) into multiple samples. Among them, the sample format can be expressed as: sentence1, sentence2, entity reference, label. To ensure the balance of positive and negative samples, data augmentation needs to be performed on the positive data samples.

[0068] It should be noted that for the training data of a general entity linking model, since usually multiple candidate interpretations can be found for a word to be analyzed, each piece of data can be split into one positive sample and multiple negative samples. To ensure the balance of the number of positive and negative samples during model training, each piece of data can be sampled according to the ratio of positive to negative samples of 1:2. When the candidate interpretation is the target interpretation, the sample is a positive sample, and the corresponding value of the classification label is 1; otherwise, it is a negative sample, and the corresponding value of the classification label is 0.

[0069] In this application, since the candidate interpretation is a Chinese interpretation and the target sentence containing the word to be analyzed is an English example sentence, considering that the semantic spaces of the Chinese and English vectors trained are different, in this application, the meaning of the word to be analyzed in the target sentence is used to replace the candidate Chinese interpretation, thus turning the translation problem into a text matching problem. For example:

[0070] For the input target sentence “Let me take your coat” and the word to be analyzed “take”, at this time, the target interpretation should be “v. take”. For the target interpretation “v. take”, select other sentences of “take” under this target interpretation. For example, “I take a handkerchief from my pocket” is used to represent the “v. take” interpretation of “take”. At this time, the meaning of “take” in these two sentences is the same. Therefore, the “label” (i.e., the classification label) is 1, and it is a positive sample. The specific format corresponding to the sample is: ["Let me take your coat", "I take a handkerchief from my pocket", "take", "1"].

[0071] Among other candidate interpretations, other examples of these interpretations are selected in turn. For example, the example sentence of the interpretation of "v. spend" is: "I had heard an appeal that could take years". At this time, the interpretations of "take" in these two sentences are inconsistent, so "label" is 0, and the specific format corresponding to the sample is: ["Let me take your coat.","I had heard an appeal could take years.","take","0"]. According to this method, a series of negative samples can be constructed.

[0072] In the process of constructing samples, since there is only one correct interpretation among the candidate interpretations and the rest are negative samples, negative samples are easy to construct. In order to balance the number of positive and negative samples, negative samples can be sampled to balance the number of positive and negative samples, and data enhancement can also be performed on positive samples.

[0073] Optionally, in the present application, synonyms are used to generate positive sample data. For example, in the example sentence "Let me take your coat.", while ensuring the existence of the word "take", 1-2 words are randomly replaced with synonyms, for example, "coat" is replaced with "umbrella", "me" is replaced with "him", etc.

[0074] It should be noted that when performing synonym replacement, grammatical rules need to be set. For example, the tense and singular and plural types of the replaced words and the original words must be consistent, so as to avoid the problem of inconsistent subject and predicate in the replaced sentences to a large extent. After completing the data format conversion, the data can be shuffled and split into training, verification, and test data in a ratio of 6:2:2.

[0075] In addition, it should be noted that most dictionaries currently return all definitions when translating words. This algorithm for predicting the corresponding definition of a word in a sentence can not only return the precise definition of the word in the sentence, but also enrich the example sentence information in the library and automatically label the example sentences with the meaning of the word. Because it is very simple to crawl vocabulary examples on the Internet, but the data containing the meaning of the word in the example sentences is very limited. Using the method provided in this application, you can automatically label the example sentences with the meaning of the word when you only have partially complete data, and you can also verify the accuracy of the crawled data (the example sentence vocabulary definition labels).

[0076] In addition, the solution provided by this application can also be used in the recommendation scenario of vocabulary learning. For example, for primary school students, they may only be required to master the most basic meaning of "take", which is "take, take", while the meanings of "take" such as "eat, drink, think, spend" are not learned by primary school students. In this scenario, the translation client can use the solution provided by this application to label the example sentences, and only return the example sentences with specific meanings when recommending sentences for learning. For high school students, when they have learned richer meanings, the solution provided by this application can recommend richer example sentences.

[0077] As can be seen from the above content, the scheme provided by the present application introduces part-of-speech tagging, first predicts the part-of-speech of the vocabulary, narrows the scope of the candidate interpretations, and improves the efficiency of the model. In addition, the interpretation of the vocabulary in the sentence is predicted by sentence splicing, and only one model needs to be trained. Since the final model is a 0 / 1 binary classification model, there is no need to extract hundreds of data for each vocabulary under each interpretation. It is only necessary to ensure that each vocabulary has data under each interpretation, and the training data under the 0 / 1 category is as balanced as possible, thereby reducing the requirements of the training model for the amount of data, and there is no need to perform complex data sampling. In the data sampling part, for the vocabulary interpretation with only one example sentence, data enhancement is performed by synonym replacement. In addition, the method of sentence splicing does not need to forcibly label the prediction results. When the meaning of the vocabulary in the example sentence is inconsistent with that in the library, the classification model will definitely give a result, but the result is likely to be wrong, and the robustness of the entity linking model is stronger. Finally, the present application can also complete the vocabulary interpretation that does not appear in the library and improve the vocabulary interpretation in the library.

[0078] Example 2

[0079] According to an embodiment of the present invention, a method embodiment for analyzing object interpretation is also provided, wherein: Figure 6 is a flow chart of a method for analyzing object interpretation according to an embodiment of the present invention. Figure 6 As shown, the method comprises the following steps:

[0080] Step S602: Obtain the sentence to be analyzed and the words to be analyzed in the sentence to be analyzed.

[0081] In step S602, the word to be analyzed is the word to be translated, and the sentence to be analyzed is the example sentence containing the word to be analyzed. For example, for the example sentence "Let me take your coat", "take" is the word to be analyzed, and the above example sentence is the sentence to be analyzed.

[0082] Optionally, the dictionary on the Internet may be crawled to obtain the sentences and words to be analyzed, or the sentences and words to be analyzed may be obtained through other open source free resources.

[0083] Step S604, concatenating the sentence to be analyzed with the candidate sentences to obtain a concatenation result, wherein the candidate sentences at least include the words to be analyzed.

[0084] It should be noted that, in step S604, sentence concatenation refers to a method of concatenating the sentence to be analyzed with the example sentences corresponding to the candidate interpretations, so as to determine the target interpretation of the word to be analyzed in the sentence to be analyzed according to the candidate interpretation of the word to be analyzed in the example sentences. It is easy to notice that, in order to predict the interpretation of the word to be analyzed in the sentence to be analyzed by sentence concatenation, only one prediction model needs to be trained, and there is no need to train models for all words separately. Moreover, training the preset model by sentence concatenation reduces the data volume requirement for model training, simplifies the data sampling process, and thus improves the training efficiency of the preset model.

[0085] Optionally, in step S604, a separator is set between the candidate sentence and the sentence to be analyzed, and the sentence to be analyzed and the candidate sentence are spliced ​​in the order of classification label, word to be analyzed, separator, and candidate sentence to obtain a first splicing result. Then, a first vector corresponding to the word to be analyzed in the sentence to be analyzed and a second vector corresponding to the word to be analyzed in the candidate sentence are obtained from the first splicing result, and the value of the classification label corresponding to the candidate sentence is determined based on the first vector and the second vector. Finally, it is determined based on the value of the classification label whether the first interpretation of the word to be analyzed in the sentence to be analyzed and the second interpretation of the word to be analyzed in the candidate sentence are the same to obtain a splicing result.

[0086] Step S606: Determine a label corresponding to the concatenation result according to the first interpretation of the word to be analyzed in the sentence to be analyzed and the second interpretation of the word to be analyzed in the candidate sentence.

[0087] Optionally, the score value corresponding to the second interpretation can be determined based on the first vector corresponding to the word to be analyzed in the sentence to be analyzed and the second vector corresponding to the word to be analyzed in the candidate sentence, and when the score value is greater than the preset score value, the value of the classification label is determined to be the first preset value; when the score value is less than or equal to the preset score value, the value of the classification label is determined to be the second preset value. The score value represents the accuracy of the second interpretation.

[0088] Step S608, model training is performed according to the labels corresponding to the sentence to be analyzed, the words to be analyzed and the concatenation results to obtain a preset model, so that the preset model outputs the target interpretation corresponding to the words to be analyzed.

[0089] Optionally, when the value of the classification label is the first preset value and the number of candidate interpretations is one, the candidate interpretation corresponding to the candidate sentence is determined to be the target interpretation. When the value of the classification label is the first preset value and the number of candidate interpretations is multiple, the candidate score values ​​corresponding to the first vector and the second vector under each candidate interpretation are obtained, and the candidate interpretation with the largest candidate score value is determined to be the target interpretation. When the values ​​of the classification labels corresponding to all candidate sentences are the second preset value, the candidate score values ​​corresponding to the first vector and the second vector under each candidate interpretation are obtained, and the candidate interpretation with the largest candidate score value is determined to be the target interpretation.

[0090] Based on the scheme defined by the above steps S602 to S608, it can be known that in an embodiment of the present invention, a preset model is obtained by sentence concatenation, so that the preset model predicts the target interpretation of the word to be analyzed in the sentence to be analyzed. After obtaining the sentence to be analyzed and the word to be analyzed in the sentence to be analyzed, the sentence to be analyzed and the candidate sentence are concatenated to obtain a concatenation result, and the label corresponding to the concatenation result is determined according to the first interpretation of the word to be analyzed in the sentence to be analyzed and the second interpretation of the word to be analyzed in the candidate sentence. Finally, model training is performed according to the sentence to be analyzed, the word to be analyzed and the label corresponding to the concatenation result to obtain the preset model, so that the preset model outputs the target interpretation corresponding to the word to be analyzed.

[0091] It is easy to notice that in the above process, the preset model is trained based on sentence concatenation, so that only one model needs to be trained for all words, and there is no need to train models for all words separately. Moreover, training the preset model by sentence concatenation reduces the data volume requirement for model training and simplifies the data sampling process, thereby improving the training efficiency of the preset model.

[0092] It can be seen that the solution provided in the present application achieves the purpose of translating words in a sentence, thereby realizing the technical effect of improving the translation efficiency of words, and further solving the technical problem of low translation efficiency when translating words in a sentence in the prior art.

[0093] In an optional embodiment, after constructing the data sample, positive samples and negative samples can be generated, and the number of positive samples and negative samples can be balanced by data enhancement, and training, testing, and validation data sets can be generated at the same time. Then, the above data sets are input into the model for training to obtain a preset model.

[0094] Optionally, in the process of constructing a positive data sample, a first preset sentence is first obtained, and word segmentation is performed on the first preset sentence, and the words to be replaced in the first preset sentence are determined. Then, synonyms corresponding to the words to be replaced are obtained, and the words to be replaced are replaced with synonyms according to preset grammatical rules to obtain a second preset sentence. Finally, a positive data sample is constructed based on the first preset sentence and the second preset sentence.

[0095] It should be noted that the process of training the preset model has been described in detail in Example 1 and will not be repeated here.

[0096] Example 3

[0097] According to an embodiment of the present invention, there is also provided an embodiment of a device for analyzing the interpretation of an object, wherein: Figure 7 is a schematic diagram of a device for analyzing object interpretation according to an embodiment of the present invention, such as Figure 7 As shown, the device includes: an acquisition module 701, an analysis module 703, a determination module 705 and a processing module 707.

[0098] Among them, the acquisition module 701 is used to obtain the target sentence and the words to be analyzed in the target sentence; the analysis module 703 is used to perform syntactic analysis on the target sentence and determine the target part of speech of the words to be analyzed in the target sentence; the determination module 705 is used to determine the candidate interpretations corresponding to the words to be analyzed according to the target part of speech; the processing module 707 is used to process the target sentence, the words to be analyzed and the target part of speech based on a preset model to obtain the target interpretation of the words to be analyzed in the target sentence, wherein the preset model is trained by sentence splicing.

[0099] It should be noted that the above-mentioned acquisition module 701, analysis module 703, determination module 705 and processing module 707 correspond to steps S102 to S108 in the above-mentioned embodiment, and the examples and application scenarios implemented by the four modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment 1.

[0100] Optionally, the processing module includes: a first acquisition module, a first processing module and a first determination module. The first acquisition module is used to acquire a candidate sentence corresponding to the candidate interpretation in a preset model, wherein the candidate sentence at least includes the word to be analyzed; the first processing module is used to concatenate the candidate sentence with the target sentence to obtain a concatenation result; and the first determination module is used to determine the target interpretation of the word to be analyzed in the target sentence from the candidate interpretation according to the concatenation result.

[0101] Optionally, the first processing module includes: a second processing module, a second acquisition module, a second determination module and a third determination module. Among them, the second processing module is used to set a separator between the candidate sentence and the target sentence, and splice the target sentence and the candidate sentence in the order of classification label, target sentence, separator, and candidate sentence to obtain a first splicing result; the second acquisition module is used to obtain the first vector corresponding to the words to be analyzed in the target sentence and the second vector corresponding to the words to be analyzed in the candidate sentence from the first splicing result; the second determination module is used to determine the value of the classification label corresponding to the candidate sentence according to the first vector and the second vector; the third determination module is used to determine whether the first interpretation of the words to be analyzed in the target sentence and the second interpretation of the words to be analyzed in the candidate sentence are the same according to the value of the classification label to obtain the splicing result.

[0102] Optionally, the second determination module includes: a fourth determination module, a fifth determination module and a sixth determination module. The fourth determination module is used to determine the score value corresponding to the second interpretation according to the first vector and the second vector, wherein the score value represents the accuracy of the second interpretation; the fifth determination module is used to determine that the value of the classification label is a first preset value when the score value is greater than a preset score value; the sixth determination module is used to determine that the value of the classification label is a second preset value when the score value is less than or equal to the preset score value.

[0103] Optionally, the first determination module includes: a seventh determination module, used to determine the candidate interpretation corresponding to the candidate sentence as the target interpretation when the value of the classification label is a first preset value and the number of candidate interpretations is one.

[0104] Optionally, the first determination module includes: a third acquisition module and an eighth determination module. The third acquisition module is used to obtain the candidate score value corresponding to the first vector and the second vector under each candidate interpretation when the value of the classification label is the first preset value and the number of candidate interpretations is multiple; the eighth determination module is used to determine the candidate interpretation with the largest candidate score value as the target interpretation.

[0105] Optionally, the first determination module includes: a fourth acquisition module and a ninth determination module. The fourth acquisition module is used to obtain the candidate score value corresponding to the first vector and the second vector under each candidate interpretation when the values ​​of the classification labels corresponding to all candidate sentences are the second preset values; the ninth determination module is used to determine the candidate interpretation with the largest candidate score value as the target interpretation.

[0106] Optionally, the device for analyzing the interpretation of an object further includes: a first construction module and a training module. The first construction module is used to construct a data sample before processing the target sentence, the words to be analyzed and the candidate interpretations based on the preset model to obtain the target interpretation of the words to be analyzed in the target sentence, wherein the data sample includes at least a positive data sample and a negative data sample; the training module is used to train the data sample to obtain the preset model.

[0107] Optionally, the first construction module includes: a fifth acquisition module, a tenth determination module, a sixth acquisition module, a replacement module, and a second construction module. The fifth acquisition module is used to acquire a first preset sentence; the tenth determination module is used to perform word segmentation processing on the first preset sentence and determine the words to be replaced in the first preset sentence; the sixth acquisition module is used to acquire synonyms corresponding to the words to be replaced; the replacement module is used to replace the words to be replaced with synonyms according to preset grammatical rules to obtain a second preset sentence; the second construction module is used to construct a data positive sample based on the first preset sentence and the second preset sentence.

[0108] Example 4

[0109] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the method for interpreting an analysis object in the above-mentioned embodiment 1 or embodiment 2 when running.

[0110] Example 5

[0111] According to another aspect of an embodiment of the present invention, a processor is provided, which is used to run a program, wherein the program is configured to execute the method for analyzing the interpretation of an object in the above-mentioned embodiment 1 or embodiment 2 when running.

[0112] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0113] In the above 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 can be made to the relevant descriptions of other embodiments.

[0114] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0115] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0116] 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 separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0117] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, 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 a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment 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.

[0118] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for analyzing the interpretation of an object, characterized in that: include: Obtaining a target sentence and words to be analyzed in the target sentence; Performing syntactic analysis on the target sentence to determine the target part of speech of the word to be analyzed in the target sentence; Determining candidate interpretations corresponding to the word to be analyzed according to the target part of speech; Processing the target sentence, the word to be analyzed, and the candidate interpretations based on a preset model to obtain a target interpretation of the word to be analyzed in the target sentence, wherein the preset model is trained by sentence splicing; The target sentence, the word to be analyzed, and the candidate interpretations are processed based on a preset model to obtain a target interpretation of the word to be analyzed in the target sentence, including: In the preset model, a candidate sentence corresponding to the candidate interpretation is obtained, wherein the candidate sentence at least includes the word to be analyzed; Concatenate the candidate sentence with the target sentence to obtain a concatenation result; Determining a target interpretation of the word to be analyzed in the target sentence from the candidate interpretations according to the splicing result; The candidate sentence is concatenated with the target sentence to obtain a concatenation result, including: Setting a separator between the candidate sentence and the target sentence, and concatenating the target sentence and the candidate sentence in the order of the classification label, the target sentence, the separator, and the candidate sentence to obtain a first concatenation result; Obtaining, from the first concatenation result, a first vector corresponding to the word to be analyzed in the target sentence and a second vector corresponding to the word to be analyzed in the candidate sentence; Determining a value of a classification label corresponding to the candidate sentence based on the first vector and the second vector; Determine whether the first interpretation of the word to be analyzed in the target sentence is the same as the second interpretation of the word to be analyzed in the candidate sentence according to the value of the classification label, and obtain the splicing result.

2. The method according to claim 1, characterized in that Determining a value of the classification label according to the first vector and the second vector includes: determining a score corresponding to the second interpretation based on the first vector and the second vector, wherein the score represents the accuracy of the second interpretation; When the score value is greater than the preset score value, determining the value of the classification label to be a first preset value; When the score value is less than or equal to the preset score value, the value of the classification label is determined to be a second preset value.

3. The method according to claim 2, characterized in that Determining a target interpretation of the word to be analyzed in the target sentence from the candidate interpretations according to the splicing result includes: When the value of the classification label is the first preset value and the number of the candidate interpretations is one, the candidate interpretation corresponding to the candidate sentence is determined as the target interpretation.

4. The method according to claim 2, characterized in that Determining a target interpretation of the word to be analyzed in the target sentence from the candidate interpretations according to the splicing result includes: When the value of the classification label is the first preset value and there are multiple candidate interpretations, obtaining candidate scoring values ​​corresponding to the first vector and the second vector under each candidate interpretation; The candidate interpretation with the largest candidate score is determined as the target interpretation.

5. The method according to claim 2, characterized in that Determining a target interpretation of the word to be analyzed in the target sentence from the candidate interpretations according to the splicing result includes: When the values ​​of the classification labels corresponding to all candidate sentences are the second preset values, obtaining the candidate scoring values ​​corresponding to the first vector and the second vector under each candidate interpretation; The candidate interpretation with the largest candidate score is determined as the target interpretation.

6. The method according to claim 1, wherein Before processing the target sentence, the word to be analyzed, and the candidate interpretations based on a preset model to obtain a target interpretation of the word to be analyzed in the target sentence, the method further includes: Constructing a data sample, wherein the data sample includes at least a positive data sample and a negative data sample; The data samples are trained to obtain the preset model.

7. The method according to claim 6, characterized in that Constructing data positive samples, including: Get the first preset statement; Performing word segmentation processing on the first preset sentence and determining words to be replaced in the first preset sentence; Obtaining synonyms corresponding to the word to be replaced; Replacing the word to be replaced with the synonymous word according to the preset grammatical rules to obtain a second preset sentence; The data positive sample is constructed based on the first preset sentence and the second preset sentence.

8. A method for analyzing the interpretation of an object, characterized in that: include: Obtaining a sentence to be analyzed and words to be analyzed in the sentence to be analyzed; Concatenating the sentence to be analyzed with the candidate sentence to obtain a concatenation result, wherein the candidate sentence at least includes the word to be analyzed; Determining a label corresponding to the concatenation result according to a first interpretation of the word to be analyzed in the sentence to be analyzed and a second interpretation of the word to be analyzed in the candidate sentence; Performing model training based on the sentence to be analyzed, the words to be analyzed, and the labels corresponding to the splicing results to obtain a preset model, so that the preset model outputs the target interpretation corresponding to the words to be analyzed; The sentence to be analyzed is concatenated with the candidate sentence to obtain a concatenation result, including: Setting a separator between the candidate sentence and the sentence to be analyzed, and concatenating the sentence to be analyzed and the candidate sentence in the order of classification label, the sentence to be analyzed, the separator, and the candidate sentence to obtain a first concatenation result; Obtaining, from the first concatenation result, a first vector corresponding to the word to be analyzed in the sentence to be analyzed and a second vector corresponding to the word to be analyzed in the candidate sentence; Determining a value of a classification label corresponding to the candidate sentence based on the first vector and the second vector; Determine whether the first interpretation of the word to be analyzed in the sentence to be analyzed is the same as the second interpretation of the word to be analyzed in the candidate sentence according to the value of the classification label, and obtain the splicing result.

9. A device for analyzing the interpretation of an object, characterized in that: include: An acquisition module, configured to acquire a target sentence and words to be analyzed in the target sentence; An analysis module, configured to perform syntactic analysis on the target sentence and determine a target part of speech of the word to be analyzed in the target sentence; A determination module, configured to determine a candidate interpretation corresponding to the word to be analyzed according to the target part of speech; a processing module, configured to process the target sentence, the word to be analyzed, and the candidate interpretations based on a preset model to obtain a target interpretation of the word to be analyzed in the target sentence, wherein the preset model is trained by sentence splicing; Wherein, the processing module includes: a first acquisition module, a first processing module and a first determination module; The first acquisition module is used to obtain a candidate sentence corresponding to the candidate interpretation in a preset model, wherein the candidate sentence includes at least the word to be analyzed; the first processing module is used to splice the candidate sentence with the target sentence to obtain a splicing result; the first determination module is used to determine the target interpretation of the word to be analyzed in the target sentence from the candidate interpretations based on the splicing result; The first processing module includes: a second processing module, a second acquisition module, a second determination module and a third determination module; wherein the second processing module is used to set a separator between the candidate sentence and the target sentence, and splice the target sentence and the candidate sentence in the order of classification label, target sentence, separator, and candidate sentence to obtain a first splicing result; the second acquisition module is used to obtain the first vector corresponding to the to-be-analyzed word in the target sentence and the second vector corresponding to the to-be-analyzed word in the candidate sentence from the first splicing result; the second determination module is used to determine the value of the classification label corresponding to the candidate sentence based on the first vector and the second vector; the third determination module is used to determine whether the first interpretation of the to-be-analyzed word in the target sentence and the second interpretation of the to-be-analyzed word in the candidate sentence are the same based on the value of the classification label to obtain a splicing result.

10. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a computer program, wherein the computer program is configured to execute the method for analyzing the interpretation of an object as claimed in any one of claims 1 to 8 when running.

11. A processor, characterized in that: The processor is used to run a program, wherein the program is configured to execute the method for analyzing object interpretation as claimed in any one of claims 1 to 8 when run.