Text information completion method and device and refrigeration equipment

Abstract

The invention provides a text information completion method and device and refrigeration equipment, and the method comprises the steps: analyzing first text data, and at least obtaining first sentence information; obtaining reference sentence information; and matching the reference sentence information with the first sentence information, and complementing the first text data according to a matching result to obtain complemented text data. According to the text information completion method provided by the invention, the text key points can be quickly retrieved and understood, a basis is provided for subsequent matching with the reference sentence information, and the relevance and difference between the first sentence information and the reference sentence information are obtained according to the matching result, so that the to-be-completed information can be accurately obtained from the reference sentence information; the accuracy and the intelligent degree of text completion data are improved, and the user experience effect is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a text information completion method, device and refrigeration equipment. Background Art

[0002] Text completion is an important task in natural language processing, also known as text prediction. Its purpose is to predict the next word, phrase, sentence or paragraph based on the existing text content, thereby obtaining some key information from the existing text content to complete the current text content. Text completion is widely used, especially in smart home appliances such as refrigerators, for example, the multi-round dialogue information completion problem of smart voice assistants.

[0003] At present, there are two ways to complete text information: one is to complete information completion through model training, but this method requires a lot of manpower and material resources to train the model, and it cannot solve the slot correspondence problem. The second is to use a multi-round dialogue mechanism to complete text information. For example, the first sentence: "Add ingredients", due to the lack of the necessary slot information "specific ingredients", the intelligent voice assistant triggers the follow-up slot filling process, and then answers: "What ingredients do you want to add?"; the second sentence: "Apple", the intelligent voice assistant answers: "Okay, the apple has been added to the refrigerator", and the necessary information required for an instruction operation is supplemented through multi-round dialogue. However, the multi-round dialogue system designed based on the instruction slot logic can only handle some specific problems related to the pre-edited slot templates or speech templates; or it can only realize some simple text information completion, such as reverse search for missing slot information or key slot information replacement. The text information processed by this method is simple, the text sentence structure is single, the coverage is narrow, and even requires user cooperation (such as requiring users to fill in the slot), and the degree of intelligence and work efficiency are low. Summary of the invention

[0004] One of the purposes of the present invention is to provide a text information completion method to solve the technical problems in the prior art that the text information completion method has a single text sentence structure, a small coverage, a low intelligence level and work efficiency, large errors, and cannot accurately and efficiently restore the original meaning of the text data.

[0005] One of the purposes of the present invention is to provide a text information completion device.

[0006] One of the objects of the present invention is to provide a refrigeration device.

[0007] In order to achieve one of the above-mentioned purposes of the invention, the present invention provides a text information completion method, including: analyzing first text data to obtain at least first sentence information; obtaining reference sentence information; matching the reference sentence information and the first sentence information, and completing the first text data according to the matching result to obtain completed text data.

[0008] As a further improvement of an embodiment of the present invention, the method further includes: updating reference text data according to the completion text data; wherein the reference text data corresponds to the reference sentence information.

[0009] As a further improvement of an embodiment of the present invention, the first text data includes at least one of original text data, voice text data and video text data.

[0010] As a further improvement of an embodiment of the present invention, the first sentence information represents the sentence characteristics of the first text data in units of sentences; the reference sentence information represents the sentence characteristics of the reference text data in units of sentences.

[0011] As a further improvement of an embodiment of the present invention, the "analyzing the first text data to obtain at least the first sentence information" specifically includes: obtaining the first text data and extracting its sentence features; inputting the sentence features into a sentence classification model to obtain a first sentence classification as the first sentence information; the "obtaining reference sentence information" specifically includes: obtaining a reference sentence classification as the reference sentence information.

[0012] As a further improvement of an embodiment of the present invention, the "analyzing the first text data to obtain at least the first sentence information" specifically includes: according to the first text data, using an information extraction method to obtain a first sentence feature matrix as the first sentence information; the "obtaining reference sentence information" specifically includes: obtaining a reference sentence feature matrix as the reference sentence information.

[0013] As a further improvement of an embodiment of the present invention, the "according to the first text data, using the information extraction method to obtain the first sentence feature matrix, and using it as the first sentence information" specifically includes: determining whether the first text data contains annotated text data; if not, performing a labeling operation on the first text data including at least one of feature word labeling, entity labeling, slot labeling and relationship labeling, and based on the labeling results, extracting text feature information using the information extraction method to obtain the first sentence feature matrix; wherein the text feature information includes at least one of feature words and their position information, slot extraction result information and relationship extraction result information.

[0014] As a further improvement of an embodiment of the present invention, the method also includes: the "determining whether the first text data contains annotated text data" specifically includes: if so, based on the current text annotation, using the information extraction method to extract text feature information to obtain the first sentence feature matrix.

[0015] As a further improvement of an embodiment of the present invention, the reference sentence information includes a reference sentence pattern classification; the first sentence information includes a first sentence pattern classification; the "matching the reference sentence information and the first sentence information, completing the first text data according to the matching result, and obtaining the completed text data" specifically includes: according to the reference sentence pattern classification and the first sentence pattern combination of the first sentence pattern classification, searching a preset completion strategy mapping table to obtain a first completion strategy corresponding to the first sentence pattern combination; according to the first completion strategy, completing the first text data to obtain the completed text data.

[0016] As a further improvement of an embodiment of the present invention, the reference sentence information includes a reference sentence feature matrix; the first sentence information includes a first sentence feature matrix; the "completing the first text data according to the first completion strategy to obtain the completed text data" specifically includes: determining the position to be completed of the first sentence feature matrix according to the first position to be completed indicated by the first completion strategy; determining the content to be completed corresponding to the first text data according to the reference sentence feature matrix and the position to be completed; completing the first text data according to the position to be completed and the content to be completed to obtain the completed text data.

[0017] As a further improvement of an embodiment of the present invention, the "determining the position to be completed in the first sentence feature matrix according to the first position to be completed indicated by the first completion strategy" specifically includes: judging whether there is a preset keyword in the first text data; if so, traversing and searching the first position to be completed in the first completion strategy corresponding to the preset keyword; according to the first position to be completed, traversing and searching the first sentence feature matrix, determining the second position to be completed and using it as the position to be completed in the first sentence feature matrix.

[0018] As a further improvement of an embodiment of the present invention, the "determining the content to be completed corresponding to the first text data based on the reference sentence feature matrix and the position to be completed" specifically includes: traversing and searching for the content corresponding to the position to be completed in the reference sentence feature matrix as the content to be completed.

[0019] As a further improvement of an embodiment of the present invention, the "completing the first text data according to the position to be completed and the content to be completed to obtain the completed text data" specifically includes: filling the content to be completed into the position to be completed in the first sentence feature matrix to refit the first sentence feature matrix; generating the completed first text data according to the first sentence feature matrix to obtain the completed text data.

[0020] In order to achieve one of the above-mentioned invention purposes, the present invention also provides a text information completion device, including: a sentence information extraction module, used for sentence information extraction, extracting first sentence information corresponding to first text data and / or reference sentence information corresponding to reference text data; a text information completion module, used for text information completion, obtaining a first completion strategy according to the sentence information and a preset completion strategy mapping table, and completing the first text data according to the first completion strategy to obtain completed text data; a reference text information recording module, used for updating reference text data, updating the reference text data according to the completed text data; wherein, the reference text data corresponds to the reference sentence information.

[0021] To achieve one of the above-mentioned purposes of the invention, the present invention further provides a refrigeration device, comprising: a memory for storing executable instructions; and a processor for implementing the steps of any of the above-mentioned text information completion methods when running the executable instructions stored in the memory.

[0022] Compared with the prior art, the embodiments of the present invention have at least one of the following beneficial effects:

[0023] The present invention adopts a text information completion method. By analyzing the sentence information of the first text data, the main points of the text can be quickly retrieved and understood, providing a basis for subsequent matching with the reference sentence information, and obtaining the correlation and difference between the first sentence information and the reference sentence information according to the matching result, so as to accurately obtain the information to be completed from the reference sentence information and improve the accuracy of the text completion data; in addition, by matching the results of the sentence information, the corresponding completion strategy is obtained and automatically completed, so that the processed text data structure is diversified, the coverage is wide, and the range is large, and it is not limited to a single text structure, which is conducive to improving the processing efficiency and intelligence of text information completion and enhancing the user experience effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the steps of a text information completion method in one embodiment of the present invention.

[0025] Figure 2 It is a schematic diagram of some steps of the text information completion method in the first embodiment of the present invention.

[0026] Figure 3 It is a schematic diagram of some steps of the text information completion method in the second embodiment of the present invention.

[0027] Figure 4 It is a partial step schematic diagram of step S3 of the text information completion method in one embodiment of the present invention.

[0028] Figure 5 It is a partial step schematic diagram of step S3 of the text information completion method in one embodiment of the present invention.

[0029] FIG. 6 is a partial schematic diagram of step S3 of the text information completion method in one embodiment of the present invention.

[0030] Figure 7 It is a structural schematic diagram of a text information completion device in one embodiment of the present invention. DETAILED DESCRIPTION

[0031] The present invention will be described in detail below in conjunction with the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional changes made by a person skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0032] It should be noted that the term "comprises" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In addition, the terms "first", "second", etc. are used for descriptive purposes only and cannot be understood as indicating or implying relative importance.

[0033] The so-called text information completion aims to automatically complete the missing parts in the text data to generate a complete sentence or paragraph. However, in the process of using natural language, it may omit expressions for the sake of simplicity and clarity, or rely on context, or have unclear expressions, which may lead to missing information and inaccurate semantic expression in the process of expressing or understanding text information. Therefore, in the field of natural language processing, text information completion has a relatively important practical significance.

[0034] Based on this, the present invention provides a text information completion method, such as Figure 1 As shown, the specific steps include:

[0035] Step S1, analyzing first text data to obtain at least first sentence information.

[0036] Step S2, obtaining reference sentence information.

[0037] Step S3, matching the reference sentence information with the first sentence information, and completing the first text data according to the matching result to obtain completed text data.

[0038] In this way, a completion strategy is obtained based on the relevance and difference between the first sentence information and the reference sentence information, which can not only cover diverse text data and expand the scope of text information completion, but also improve the accuracy of text completion data.

[0039] Furthermore, after step S3, the text information completion method may further include:

[0040] Step S4, updating the reference text data according to the completed text data; wherein the reference text data corresponds to the reference sentence information.

[0041] In this way, the updated reference text data can provide richer and more accurate judgment basis and options, allowing the completion information to be passed on, expanding the text types covered by the reference text, and solving various text completion problems.

[0042] It should be emphasized that the reference sentence information or the reference text data can be empty. Specifically, in an application scenario of multiple rounds of dialogue, if in the first round of dialogue, the first text data is the first sentence text data, the reference sentence information is empty, and the reference text data can be updated to the first text data by executing steps S1 to S4. That is, after executing the above steps for the first time, the content of the reference text data and the first text data is the same. In other words, the reference text data can be understood as text data after information completion. In this way, the information completion method of the present invention can realize the inheritance or transmission of completed information.

[0043] For example, taking the information completion scenario of multiple rounds of dialogue as an example, the updating process of the completed information of the present invention is described. For example, the first sentence is "Are there any apples in the refrigerator?"; the second sentence is "When were they deposited?" and the third sentence is "When did they expire?". For the third sentence, its previous sentence is not "When were they deposited?", but "When were the apples deposited", that is, its previous sentence is the text data after the original second sentence is completed; and the text data after the second sentence is completed is completed based on the original data of the first sentence, thereby realizing the transfer of the completion information "apple" of the first sentence to the second and third sentences. Even if the first sentence is missing now, the above third sentence can also realize information completion. It can be seen that this method can convert a multi-round dialogue problem into several two-round dialogue problems.

[0044] Optionally, for step S1, the first text data includes original text data, voice text data, and video text data.

[0045] In this way, by obtaining text data of different structural types, the first text data is made more complete, comprehensive and diversified; in addition, by connecting different types of text information together, the text content can be enriched from the aspects of multi-structure and multi-type of data.

[0046] In addition, the first sentence information can be used to characterize the sentence characteristics of the first text data in units of sentences. The reference sentence information can be used to characterize the sentence characteristics of the reference text data in units of sentences. In this way, the main points of the text can be quickly retrieved and understood based on the sentence information, providing a basis for subsequent sentence matching.

[0047] Specifically, in one embodiment, the first sentence information can represent the first sentence classification; in one embodiment, the first sentence information can represent the first sentence feature matrix. In a preferred embodiment, the above two embodiments can be included at the same time, that is, the first sentence information can represent the first sentence classification and the first sentence feature matrix at the same time.

[0048] Similarly, in one embodiment, the reference sentence information can represent the reference sentence classification; in one embodiment, the reference sentence information can represent the reference sentence feature matrix. In a preferred embodiment, the above two embodiments can be included at the same time, that is, the reference sentence information can represent the reference sentence classification and the reference sentence feature matrix at the same time.

[0049] Based on this, Figure 2 As shown, in one implementation, step S1 may specifically include:

[0050] Step S111, obtaining the first text data and extracting its sentence features.

[0051] Step S112: input the sentence features into a sentence classification model to obtain a first sentence classification as the first sentence information.

[0052] Based on this, step S2 may specifically include: step S2', obtaining a reference sentence classification and using it as the reference sentence information.

[0053] In this way, by obtaining the first sentence classification and the reference sentence classification, it is easy to know the structural type of the first text data and the reference text data to guide the acquisition of subsequent completion strategies. This method is simple and easy to operate, has strong flexibility, can cover text data of different structural types, and has a wide processing range. For other steps, refer to the aforementioned steps S1 to S3, which will not be repeated here.

[0054] In addition, step S1 may further specifically include:

[0055] Step S121: According to the first text data, use the information extraction method to obtain the first sentence feature matrix and use it as the first sentence information;

[0056] Based on this, step S2 can specifically include: step S2": Obtain the reference sentence feature matrix as the reference sentence information. For other steps, refer to the aforementioned steps S1 to S3, which will not be elaborated here.

[0057] In this way, by obtaining the sentence feature matrix, it is not only convenient to understand the structural framework of the first text data and the reference text data, and the constituent elements of the text data; but also the information to be complemented can be determined by judging the differences and correlations between the first sentence feature matrix and the reference sentence feature matrix, improving the accuracy of information complementation.

[0058] Among them, the first sentence pattern classification can be understood as being used to analyze and judge the functional type of the first text data, reflecting the macroscopic structural framework of the text data, indicating the tone or purpose expressed by the text data. It can ignore the specific constituent elements of the text data and the relationships between the constituent elements, and can be a certain predetermined structural type. For example, the sentence pattern classification corresponding to "Are there apples in the refrigerator?" is "existential yes-no question", and structural frameworks similar to this can be considered the same sentence pattern classification.

[0059] In addition, the first sentence feature matrix can be understood as being used to judge the internal constituent elements of the text data and the logical relationships between the constituent elements, reflecting the microscopic structural composition of the text data. It can be seen that the first sentence feature matrix needs to judge the element categories and element relationships in the sentence according to the specific sentence structure, and is the result driven by data, and the elements and relationships need to be determined according to the specific composition of the text data.

[0060] The sentence pattern feature in step S111 can be understood as the language feature used to judge the structural type of the text data, and can specifically include the attributes of the text data in terms of order, word formation, function, etc. For example, in the above example "Are there apples in the refrigerator?", the subject "refrigerator" is placed at the beginning of the sentence as the position; ending with "吗" makes it an interrogative sentence; using "有" to inquire about existence can be understood as inquiring about the existence of something, being an "existential interrogative sentence"; except for "吗", the word order is a declarative sentence, being a yes-no question sentence. Through the above sentence pattern features, it can be judged that this sentence is of the sentence pattern classification of "existential yes-no question".

[0061] Among them, step S2' and step S2" can be understood as derived steps of step S2; step S121 can be understood as a derived step of step S1; there is no execution order between step S1 and step S2, and step S2 can be set before step S1; steps S111 to S112 can be set as a whole after step S2' or step S2"; step S2' and step S2" can be set as a whole between step S111 and step S112, or can be interspersed therein; step S121 can be set after step S2' or step S2". The present invention does not make any specific restrictions on this.

[0062] In a preferred embodiment, the above two implementations can be used in combination to obtain two sets of sentence classifications and two sets of sentence feature matrices. In this way, the two sets of data can be combined with each other, so that the final completed data has a higher accuracy rate.

[0063] Specifically, if Figure 3 The step S121 may specifically include:

[0064] Step S1211, determining whether the first text data includes annotated text data.

[0065] If not, jump to step S1212A, perform at least one of feature word tagging, entity tagging, slot tagging and relationship tagging on the first text data, and extract text feature information using the information extraction method based on the tagging results to obtain the first sentence feature matrix.

[0066] If so, jump to step S1212B, extract text feature information using the information extraction method according to the current text annotation, and obtain the first sentence feature matrix.

[0067] The text feature information includes at least one of feature words and their position information, slot extraction result information, and relationship extraction result information.

[0068] In this way, by annotating text data and information extraction methods, valuable information can be extracted from massive text data and expressed in the form of sentence feature matrices, further optimizing the efficiency of data processing and improving the precision and accuracy of data mining.

[0069] like Figure 4 As shown, in a preferred embodiment, the reference sentence information includes reference sentence classification; the first sentence information includes first sentence classification. Based on words, step S3 may specifically include:

[0070] Step S31, searching a preset completion strategy mapping table according to the reference sentence classification and the first sentence combination of the first sentence classification, and obtaining a first completion strategy corresponding to the first sentence combination.

[0071] Step S32: completing the first text data according to the first completion strategy to obtain the completed text data.

[0072] In this way, the automatic selection and completion strategy based on the mapping table realizes the automatic processing of text data completion, which reduces manual intervention to a certain extent and improves processing efficiency; in addition, based on the configurable mapping table, the processing object scope of text completion can be improved, so that text completion is no longer limited to a specific type or structure, and its processing scope can be expanded, which can realize information supplementation for more extensive and complex language expressions.

[0073] The first completion strategy can be understood as any two groups of sentence classification combinations corresponding to an information completion strategy. The first completion strategy or the first sentence combination can be understood as a general term, specifically referring to the preset completion strategy mapping table that can contain several sentence combinations and several corresponding completion strategies. According to the first completion strategy corresponding to the first text data, information completion can be achieved for the first text data.

[0074] Furthermore, in order to clarify the first completion strategy and improve the accuracy of the completion information, Figure 5 As shown, in a preferred embodiment, the present invention provides detailed steps for step S32, which may specifically include:

[0075] Step S321: Determine the position to be completed in the first sentence feature matrix according to the first position to be completed indicated by the first completion strategy.

[0076] Step S322: Determine the to-be-completed content corresponding to the first text data according to the reference sentence feature matrix and the to-be-completed position.

[0077] Step S323: completing the first text data according to the position to be completed and the content to be completed to obtain the completed text data.

[0078] In this way, the position to be completed is determined according to the first completion strategy, and the sentence feature matrix can be used to select the content to be completed that is closest to the data structure and content of the text to be completed, avoiding completion errors caused by subjective assumptions and ensuring the accuracy, coherence and comprehensiveness of the completion information.

[0079] Specifically, as shown in FIG. 6( a ), in a specific implementation, step S321 may specifically include:

[0080] Step S3211, determining whether there are preset keywords in the first text data.

[0081] If so, jump to step S3212 to traverse and search for the first position to be completed in the first completion strategy corresponding to the preset keyword.

[0082] Step S3213: According to the first position to be completed, the first sentence feature matrix is ​​traversed and searched, and a second position to be completed is determined and used as the position to be completed in the first sentence feature matrix.

[0083] In this way, the completion position is automatically determined based on keywords, so that the position to be completed can be accurately located. At the same time, the position to be completed also provides important clues for the subsequent selection of the content to be completed, thereby realizing automatic processing of text completion and improving processing efficiency.

[0084] It should be noted that the preset keywords may include several ones, which can be understood as some modal particles or verbs, prepositions to determine the sentence type of the current text data. For example, the keywords of "Are the apples in the refrigerator?" can be set to "in" or "?". It can also be some connecting words. For example, the keyword in "What about the apples?" can be set to "then". The selection of these keywords can more conveniently determine the sentence type of the current text data, or the missing sentence components or sentence positions. In this regard, adaptive settings can be made according to the corresponding text data type without specific restrictions.

[0085] Optionally, when solving the instruction completion problem, information completion can be achieved without using preset keywords and sentence feature matrices, and the first text data and reference text data can be directly spliced ​​to complete the information. For example, "add ingredients" and "tomatoes", the completed text data after splicing is "add tomatoes".

[0086] In addition, the first position to be completed can be understood as the corresponding position determined in the first compensation strategy according to the preset keywords. For example, the first completion strategy is "there is still xxx time (expired|spoiled|inedible)", and the preset keyword is "there is still", then the first position to be completed can be set to the position immediately before "there is still". If the text data corresponding to the first sentence feature matrix is ​​"I want to know how long it will expire", the first sentence feature matrix is ​​subsequently traversed and searched according to this first position to be completed (the position immediately before "there is still") to determine the position to be completed corresponding to the first sentence feature matrix. It can be seen that the first position to be completed and the second position to be completed may be the same or different, and may change dynamically according to the specific text data.

[0087] Specifically, the present invention further provides detailed steps for step S322, which may specifically include:

[0088] Step S322', traverse and search for the content corresponding to the position to be completed in the reference sentence feature matrix as the content to be completed.

[0089] In this way, by automatically searching for the content corresponding to the position to be completed in the sentence feature matrix, automatic completion of information can be achieved, thereby improving work efficiency and intelligence.

[0090] Specifically, as shown in FIG. 6( b ), in a specific implementation manner, the present invention further provides detailed steps for step S323 , which may specifically include:

[0091] Step S3231, filling the to-be-completed content into the to-be-completed position in the first sentence feature matrix to refit the first sentence feature matrix.

[0092] Step S3232: Generate the completed first text data according to the first sentence feature matrix to obtain the completed text data.

[0093] In this way, by using the sentence feature matrix and refitting, a more accurate judgment can be made in selecting the content to be completed and generating the completed text, ensuring the accuracy and rationality of the completion result. Among them, step S322' can be understood as a derivative step of step S322.

[0094] For example, different sentence combinations can correspond to different completion strategies. For example, the combination of "existence is a yes or no question" and "status is a yes or no question" corresponds to the first completion strategy. Specifically, the two sentence combinations of "Are there apples in the refrigerator?" and "Is it expired?" correspond to the completion process: first, according to the first completion strategy, determine that the position to be completed of "Is it expired?" (corresponding to the first text data) is the first position; secondly, according to the sentence feature matrix corresponding to "Are there apples in the refrigerator?" (corresponding to the reference text data) and the position to be completed, determine that the content to be completed is the part "apple" between "yes" and "?"; finally, complete "apple" (content to be completed) to the first position (position to be completed) of the second sentence text to obtain the completion data of the second sentence "Are apples expired?"

[0095] The above-mentioned various implementation modes, embodiments or specific examples provided by the present invention can be combined with each other to finally form multiple better implementation modes.

[0096] like Figure 7 As shown, the present invention also provides a text information completion device, comprising:

[0097] The sentence information extraction module 701 is used for extracting sentence information, and extracting first sentence information corresponding to the first text data and / or reference sentence information corresponding to the reference text data.

[0098] The text information completion module 702 is used for completing text information, obtaining a first completion strategy according to the sentence information and a preset completion strategy mapping table, and completing the first text data according to the first completion strategy to obtain completed text data.

[0099] The reference text information recording module 703 is used to update the reference text data according to the completed text data; wherein the reference text data corresponds to the reference sentence information.

[0100] Specifically, the first text data is input, and the current reference text data is obtained from the reference text information recording module 703; it is input into the sentence information extraction module 701 to analyze and extract sentence information; based on the extracted sentence information, the text information completion module 702 determines the position to be completed and the content to be completed, and completes the first text data, and sends the completed text data to the required module, and at the same time updates the reference text data in the reference text information recording module 703 so that the information of the next text data to be completed can be completed.

[0101] The reference text information recording module enables the completion information to be continuously transferred to the next text data to be completed, and can transform a multi-round dialogue question into a number of two-round dialogue questions. For a specific discussion, please refer to the description of the update of the reference text data in step S4 above, which will not be repeated here.

[0102] The present invention also provides a refrigeration device, comprising: a memory for storing executable instructions; and a processor for implementing any of the steps of the above-mentioned text information completion method when running the executable instructions stored in the memory.

[0103] In summary, the present invention obtains sentence information of the first text data and the reference text data by analyzing, and obtains the corresponding text completion strategy by matching the two sets of sentence information. The text information that should be transmitted in the reference text data is transmitted to the first text data, so that the user's real semantics can be accurately restored, and the accuracy of the completed text data is improved; in addition, the mapping relationship between the reference sentence classification and the first sentence combination of the first sentence classification and the first completion strategy is configurable, with good flexibility and strong generalization ability, and at the same time, the requirement for the amount of data is relatively small, and it can cover completion problems of different ranges.

[0104] In addition, based on the sentence information of the first text data and the reference text data, a corresponding completion strategy is obtained. According to the completion strategy, the corresponding position to be completed and the content to be completed are automatically searched in the sentence feature matrix, so that automatic completion of information can be achieved, which reduces manual intervention to a certain extent and improves processing efficiency and intelligence.

[0105] It should be understood that although this specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each implementation mode may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

[0106] The series of detailed descriptions listed above are only specific descriptions of feasible implementation methods of the present invention. They are not intended to limit the scope of protection of the present invention. Any equivalent implementation methods or changes that do not deviate from the technical spirit of the present invention should be included in the scope of protection of the present invention.

Claims

1. A text information completion method, characterized in that: include: Analyze the first text data to obtain at least first sentence information; Get reference sentence information; The reference sentence information and the first sentence information are matched, and the first text data is completed according to the matching result to obtain completed text data.

2. The text information completion method according to claim 1, characterized in that: The method further comprises: The reference text data is updated according to the completed text data; wherein the reference text data corresponds to the reference sentence information.

3. The text information completion method according to claim 1, characterized in that: The first text data includes at least one of original text data, voice text data and video text data.

4. The text information completion method according to claim 1, characterized in that: The first sentence information represents the sentence characteristics of the first text data in units of sentences; the reference sentence information represents the sentence characteristics of the reference text data in units of sentences.

5. The text information completion method according to claim 1, characterized in that: The “analyzing the first text data to obtain at least the first sentence information” specifically includes: Acquire the first text data and extract its sentence features; Inputting the sentence features into a sentence classification model to obtain a first sentence classification as the first sentence information; The “obtaining reference sentence information” specifically includes: A reference sentence classification is obtained and used as the reference sentence information.

6. The text information completion method according to claim 1, characterized in that: The “analyzing the first text data to obtain at least the first sentence information” specifically includes: According to the first text data, a first sentence feature matrix is ​​obtained by using an information extraction method and used as the first sentence information; The “obtaining reference sentence information” specifically includes: A reference sentence feature matrix is ​​obtained as the reference sentence information.

7. The text information completion method according to claim 6, characterized in that: The “according to the first text data, using an information extraction method to obtain a first sentence feature matrix, and using it as the first sentence information” specifically includes: Determining whether the first text data includes annotated text data; If not, performing a tagging operation including at least one of feature word tagging, entity tagging, slot tagging and relationship tagging on the first text data, and extracting text feature information using the information extraction method according to the tagging result to obtain the first sentence feature matrix; The text feature information includes at least one of feature words and their position information, slot extraction result information, and relationship extraction result information.

8. The text information completion method according to claim 7, characterized in that: The “determining whether the first text data contains annotated text data” specifically includes: If so, the text feature information is extracted using the information extraction method according to the current text annotation to obtain the first sentence feature matrix.

9. The text information completion method according to claim 1, characterized in that: The reference sentence information includes a reference sentence classification; the first sentence information includes a first sentence classification; the "matching the reference sentence information and the first sentence information, completing the first text data according to the matching result, and obtaining the completed text data" specifically includes: According to the reference sentence classification and the first sentence combination of the first sentence classification, searching a preset completion strategy mapping table to obtain a first completion strategy corresponding to the first sentence combination; The first text data is completed according to the first completion strategy to obtain the completed text data.

10. The text information completion method according to claim 9, characterized in that: The reference sentence information includes a reference sentence feature matrix; the first sentence information includes a first sentence feature matrix; the “completing the first text data according to the first completion strategy to obtain the completed text data” specifically includes: Determining the position to be completed in the first sentence feature matrix according to the first position to be completed indicated by the first completion strategy; Determine the to-be-completed content corresponding to the first text data according to the reference sentence feature matrix and the to-be-completed position; The first text data is completed according to the position to be completed and the content to be completed to obtain the completed text data.

11. The text information completion method according to claim 10, characterized in that: The “determining the position to be completed in the first sentence feature matrix according to the first position to be completed indicated by the first completion strategy” specifically includes: Determining whether there is a preset keyword in the first text data; If so, traverse and search for the first position to be completed in the first completion strategy corresponding to the preset keyword; According to the first position to be completed, the first sentence characteristic matrix is ​​traversed and searched, and a second position to be completed is determined and used as the position to be completed in the first sentence characteristic matrix.

12. The text information completion method according to claim 10, characterized in that: The “determining the to-be-completed content corresponding to the first text data according to the reference sentence feature matrix and the to-be-completed position” specifically includes: The reference sentence feature matrix is ​​traversed to find the content corresponding to the position to be completed as the content to be completed.

13. The text information completion method according to claim 10, characterized in that: The “completing the first text data according to the position to be completed and the content to be completed to obtain the completed text data” specifically includes: Filling the to-be-completed content into the to-be-completed position in the first sentence feature matrix to refit the first sentence feature matrix; Generate the completed first text data according to the first sentence feature matrix to obtain the completed text data.

14. A text information completion device, characterized in that: include: A sentence information extraction module, used for extracting sentence information, extracting first sentence information corresponding to the first text data and / or reference sentence information corresponding to the reference text data; A text information completion module, used for completing text information, obtaining a first completion strategy according to the sentence information and a preset completion strategy mapping table, and completing the first text data according to the first completion strategy to obtain completed text data; A reference text information recording module is used to update reference text data, and update the reference text data according to the completed text data; wherein the reference text data corresponds to the reference sentence information.

15. A refrigeration device, characterized in that: include: A memory for storing executable instructions; The processor is used to implement the steps of the text information completion method described in any one of claims 1 to 13 when running the executable instructions stored in the memory.

Images