A control method and control system of a home terminal

Abstract

The application provides a control method and a control system of a home terminal. The control method of the home terminal comprises the following steps: obtaining a voice control instruction of a user and converting the voice control instruction into a text control instruction; performing dependency syntax analysis on the text control instruction to obtain a plurality of groups of predicate phrases based on a preset form, wherein the predicate phrases based on the preset form at least include an operation element, a position element and an object element; performing clustering analysis on a first sentence vector of a first predicate phrase which lacks any of the elements and a second sentence vector of a second predicate phrase to determine an associated predicate phrase of the first predicate phrase; filling the elements lacking in the first predicate phrase by using corresponding elements of the associated predicate phrase; and controlling a corresponding home terminal according to the first predicate phrase filled with the elements.

Description

Technical Field

[0001] This invention relates to the technical field of smart homes, and more particularly to a control method for a furniture terminal, a control system for a furniture terminal, and a corresponding computer-readable storage medium. Background Technology

[0002] With the popularization of smart technology and the development of internet technology, smart homes are gradually entering ordinary households. Smart homes mainly rely on intelligent dialogue systems to interact with users via voice. The dialogue system first converts the voice into text using an ASR (Automatic Speech Recognition) module, then performs semantic understanding on the text to obtain the user's intent, thereby controlling home devices.

[0003] When determining a user's intent, it's necessary to extract multiple entities related to various home appliances from the text, such as room names, appliance names, and the desired operation (e.g., "turn on," "turn off," "raise," "set"). Different entity combinations correspond to different intents, so the accuracy of entity recognition directly affects the accuracy of intent recognition, thus impacting the feedback and user experience of the dialogue system. Current entity recognition technologies in this field primarily utilize deep learning or rule-based methods. However, deep learning requires a large amount of labeled training data, which is difficult to obtain, and its quality significantly affects the entity recognition model. Rule-based methods are only suitable for scenarios with small amounts of training data. Therefore, regardless of whether deep learning or rule-based methods are used for entity recognition, current performance in grouped entity recognition is not ideal, especially in smart home applications, where existing grouped entity recognition technologies exhibit low accuracy and poor reliability.

[0004] In order to overcome the above-mentioned defects in the existing technology, there is an urgent need in the field for a control method for home terminals, which can accurately identify the grouped entity parts in smart home control commands, thereby improving the accuracy and reliability of smart home voice control. Summary of the Invention

[0005] The following provides a brief overview of one or more aspects to offer a basic understanding of them. This overview is not an exhaustive summary of all conceived aspects, nor is it intended to identify key or decisive elements of all aspects, nor to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed descriptions that follow.

[0006] To overcome the aforementioned deficiencies in the existing technology, the present invention provides a control method and control system for a home terminal, which can accurately identify the grouped entity portion in smart home control commands, thereby improving the accuracy and reliability of smart home voice control.

[0007] Specifically, the control method for the home terminal provided by the first aspect of the present invention includes the following steps: acquiring a user's voice control command and converting it into a text control command; performing dependency parsing on the text control command to obtain multiple sets of predicate phrases based on a preset form, wherein the predicate phrases based on the preset form include at least an operation element, a position element, and an object element; performing cluster analysis on the first sentence vector of a first predicate phrase lacking any of the aforementioned elements and the second sentence vectors of the remaining second predicate phrases to determine the associated predicate phrases of the first predicate phrase; using the corresponding elements of the associated predicate phrases to fill in the missing elements of the first predicate phrase; and controlling the corresponding home terminal according to the first predicate phrase filled with the aforementioned elements.

[0008] Furthermore, in some embodiments of the present invention, the step of performing dependency parsing on the text control instructions to obtain multiple sets of predicate phrases based on a preset form includes: extracting entity elements from the text control instructions according to preset entity keywords, wherein the entity elements include the operation element, the position element, and the object element; and performing dependency parsing on the text control instructions according to the extracted entity elements to obtain the multiple sets of predicate phrases.

[0009] Furthermore, in some embodiments of the present invention, the step of extracting entity elements from the text control command based on preset entity keywords includes: using a combination model of Bi-LSTM model and rule model to extract entity elements from the text control command, wherein the rule model stores preset entity keywords to constrain the recognition rules of the Bi-LSTM model.

[0010] Furthermore, in some embodiments of the present invention, the step of performing cluster analysis on the first sentence vector of the first predicate phrase lacking any of the aforementioned elements and the second sentence vectors of the remaining second predicate phrases to determine the associated predicate phrases of the first predicate phrase includes: determining a number parameter k based on the number of elements involved in the preset form; supplementing the first sentence vector of the first predicate phrase with the directional elements of the second sentence vectors and assigning negative values ​​to its missing elements to determine the third sentence vector; and using the K-means model to perform cluster analysis on the third sentence vector and each of the second sentence vectors based on the number parameter k to determine the associated predicate phrases.

[0011] Furthermore, in some embodiments of the present invention, before performing cluster analysis on the first sentence vector of the first predicate phrase lacking any of the aforementioned elements and the second sentence vector of the remaining second predicate phrases, the control method further includes the following steps: calculating the sentence vector of each group of the predicate phrases; and performing recall processing analysis and / or similarity matching on the sentence vector of each group of the predicate phrases and a preset plurality of sentence vector samples to filter out predicate phrases with low similarity to each of the sentence vector samples.

[0012] Furthermore, in some embodiments of the present invention, the step of acquiring the user's voice control command and converting it into a text control command includes: acquiring and parsing the voice control command provided by the user to generate a first candidate text; replacing at least one word in the first candidate text with at least one set of pre-set near-homophones to generate a plurality of second candidate texts; calculating the probability of the first candidate text and the plurality of second candidate texts using a language model; and determining the candidate text with the highest probability as the text control command.

[0013] Furthermore, in some embodiments of the present invention, the control method further includes the following step: controlling the corresponding home terminal based on the second predicate phrase which does not lack any of the elements.

[0014] Furthermore, the control system for a home terminal provided according to a second aspect of the present invention includes: a microphone for acquiring voice control commands from a user; a communication module for communicatively connecting to at least one home terminal; a memory for storing computer instructions; and a processor communicatively connected to the microphone, the communication module, and the memory, and configured to execute the computer instructions to implement the aforementioned control method for the home terminal.

[0015] Furthermore, according to a third aspect of the present invention, a computer-readable storage medium is provided thereon storing computer instructions that, when executed by a processor, implement a control method for a home terminal. Attached Figure Description

[0016] The above-described features and advantages of the present invention will be better understood after reading the following detailed description of embodiments of the present disclosure in conjunction with the accompanying drawings. In the drawings, components are not necessarily drawn to scale, and components having similar related characteristics or features may have the same or similar reference numerals.

[0017] Figure 1 An architecture diagram of a control system for a furniture terminal according to some embodiments of the present invention is shown;

[0018] Figure 2 A flowchart of a control method for a furniture terminal according to some embodiments of the present invention is shown. Detailed Implementation

[0019] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Although the description of the present invention is presented in conjunction with preferred embodiments, this does not mean that the features of the invention are limited to these embodiments. On the contrary, the purpose of describing the invention in conjunction with embodiments is to cover other options or modifications that may be derived based on the claims of the present invention. To provide a thorough understanding of the invention, many specific details will be included in the following description. The invention may also be implemented without using these details. Furthermore, to avoid confusion or obscuring the focus of the invention, some specific details will be omitted in the description.

[0020] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0021] Furthermore, the terms "upper," "lower," "left," "right," "top," "bottom," "horizontal," and "vertical" used in the following description should be understood as the orientations shown in the relevant paragraphs and accompanying drawings. These relative terms are for illustrative purposes only and do not imply that the described apparatus must be manufactured or operated in a specific orientation, and therefore should not be construed as limiting the invention.

[0022] It is understood that although terms such as "first," "second," and "third" may be used herein to describe various components, regions, layers, and / or parts, these components, regions, layers, and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers, and / or parts. Therefore, the first components, regions, layers, and / or parts discussed below may be referred to as second components, regions, layers, and / or parts without departing from some embodiments of the present invention.

[0023] As mentioned above, current entity recognition technologies in this field mainly rely on deep learning or rule-based methods. However, deep learning-based entity recognition requires a large amount of labeled training data, which is difficult to obtain, and its quality significantly impacts the entity recognition model. Rule-based entity recognition is only suitable for scenarios with small amounts of training data. Therefore, neither deep learning nor rule-based methods currently perform ideally in grouped entity recognition.

[0024] Especially in smart home applications, entity recognition might encounter text like, "Hello, please turn on the air conditioner in the living room, and also the lights." In this case, the home control system needs to extract two sets of entities from this text: {[room=living room, device=air conditioner], [room=living room, device=lights]}. However, existing entity recognition methods are very likely to extract entities as {[room=living room, device=air conditioner, device=lights]}, leading to voice control failure or partial failure. Similarly, entity recognition might encounter the question, "Turn on all the air conditioners and lights in the living room and bedroom." In this case, the home control system needs to extract two sets of entities from this text: {[room=living room, device=air conditioner], [room=living room, device=lights], [room=bedroom, device=lights], [room=bedroom, device=lights]}. Existing entity recognition methods are very likely to extract entities as {[room=living room, room=bedroom, device=air conditioner, device=lights]}, which will also lead to voice control failure or partial failure. Therefore, the existing technology has a low accuracy rate and poor reliability in identifying grouped entities in smart home application scenarios, which hinders the further promotion and development of smart homes.

[0025] To overcome the aforementioned deficiencies in the existing technology, the present invention provides a control method for a home terminal, a control system for a home terminal, and a corresponding computer-readable storage medium, which can accurately identify the grouped entity portion in smart home control commands, thereby improving the accuracy and reliability of smart home voice control.

[0026] In some non-limiting embodiments, the control method for the home terminal provided in the first aspect of the present invention can be implemented via the control system for the home terminal provided in the second aspect of the present invention. Specifically, the control system for the home terminal is equipped with a memory and a processor. The memory includes, but is not limited to, the computer-readable storage medium provided in the third aspect of the present invention, on which computer instructions are stored. The processor is connected to the memory and configured to execute the computer instructions stored in the memory to implement the control method for the home terminal provided in the first aspect of the present invention.

[0027] Please refer to the following first. Figure 1 , Figure 1 An architecture diagram of a home terminal control system provided according to some embodiments of the present invention is shown.

[0028] like Figure 1As shown, the control system for a home terminal provided by the second aspect of the present invention may include a microphone (not shown), a communication module 105, a memory, and a processor 102. The memory may include a read-only memory (ROM) 103 and a random access memory (RAM) 104. An internal communication bus 101 enables data communication between the control system components of the home terminal. During the operation of the control system, the processor 102 can first acquire voice information provided by the user from the microphone and perform text recognition on the voice information. Then, the processor 102 can perform text correction on the text. After obtaining the text information after text correction, the processor 102 can also perform entity recognition on the text information and perform dependency parsing on the entities to determine at least one entity involved. Then, the processor 102 can group these entities and issue control commands based on the predicate phrases obtained from the grouped entities to control the corresponding home terminal to perform corresponding operations. Here, the communication bus 101 can transmit the above control commands to the home execution components to control the smart home.

[0029] The working principle of the control system of the aforementioned home appliance terminal will be described below with reference to some embodiments of control methods for home appliances. Those skilled in the art will understand that these embodiments of control methods for home appliances are merely non-limiting implementations provided by the present invention, intended to clearly demonstrate the main concepts of the invention and provide specific solutions convenient for public implementation, rather than limiting all functions or operating modes of the control system of the home appliance terminal. Similarly, the control system of the home appliance terminal is also only a non-limiting implementation provided by the present invention and does not limit the entities executing the steps in these control methods of home appliances.

[0030] Please refer to Figure 2 . Figure 2 A flowchart of a control method for a furniture terminal according to some embodiments of the present invention is shown.

[0031] like Figure 2As shown, during the control process of the furniture terminal, the present invention can first acquire the user's voice control commands and convert them into text control commands. In some embodiments, the step of converting the voice control commands into text control commands can be implemented based on offline processing of commonly used text corpora in the smart home field. Specifically, technicians can collect commonly used words in the smart home field and words with similar pronunciations into a text correction table in the database. Here, the primary key of the database field can be the word itself. For example, a database record can include a primary key (e.g., bedroom) and other keys with similar pronunciations. Then, the present invention can count the probability of the above-mentioned similar words and sort these similar words from high to low according to the probability. Then, the present invention can use the word with the highest probability as the primary key, and add the remaining words with lower probabilities as non-primary key elements into the primary key in order, so that the words with high probabilities in the text correction table are placed before the words with low probabilities. For example, a record in the database can be {bedroom, I four, I seem, lying down, bedroom, woshi, make me, I city, I things, I try}. Here, "bedroom" is the primary key of the above field. In other words, when the microphone detects any word in the above 10 fields, it can replace it with the primary key "bedroom" to accurately correct text errors.

[0032] After establishing the text correction database table, this invention can also use sentences as the basic unit to split the text corpus in the above-mentioned smart home scenario, and use algorithms such as Bert and Word2Vec to calculate the sentence vectors of these sentences to represent the degree of similarity and semantic similarity between sentences.

[0033] Those skilled in the art will understand that the above-described embodiments of calculating sentence vectors using natural language processing algorithms such as BERT and Word2Vec are merely non-limiting implementations provided by the present invention, intended to clearly demonstrate the main concept of the present invention and provide some specific solutions that are easy for the public to implement, rather than being used to limit the scope of protection of the present invention.

[0034] Alternatively, in other embodiments, those skilled in the art may use other methods to characterize the sentence vectors of each sentence to achieve the corresponding effect, which will not be elaborated here.

[0035] Furthermore, after offline processing of common text corpora in the smart home field, the present invention can perform automatic speech recognition processing on the above-mentioned voice control instructions obtained through the microphone to convert the voice signal into a first candidate text. After obtaining the above-mentioned first candidate text, the present invention can also replace at least one corresponding vocabulary in the first candidate text according to the above-mentioned preset at least one set of near-sound words to generate multiple second candidate texts. Specifically, the present invention can replace words with similar voices according to the records in the text error correction table in the above-mentioned database. For example, when the above-mentioned first candidate text is "Turn on my light", the present invention can generate second candidate texts such as "Turn on the bedroom light", "Turn on my four lights", "Turn on my similar lights", etc. based on the first candidate text {Turn on my light} according to the records of {bedroom, I four, I similar,卧式, bedroom,沃施, I make, I city, I thing, I try}. Here, the recording method of the database field is to generate non-primary key fields according to the above-mentioned frequency sorting. Specifically, in the record of {bedroom, I four, I similar,卧式,卧四,沃施, I make, I city, I thing, I try}, the primary key is "bedroom", and the probabilities of the appearance of other words are: "bedroom" > "I four" > "I similar" > "卧式" > "卧四" > "沃施" > "I make" > "I city" > "I thing" > "I try". After replacing the vocabulary and generating the above-mentioned second candidate texts, the present invention can also use language models such as Bert, Word2Vec, and NGram to calculate the probabilities of the above-mentioned first candidate text and the above-mentioned multiple second candidate texts, and select the sentence with the highest probability as the final result of text error correction. For example, when the first candidate text is "Turn on my light", after the above-mentioned text error correction, the finally generated result is "Turn on the bedroom light". In this way, the present invention can perform text error correction through the above-mentioned method.

[0036] As Figure 2 shown, after obtaining the voice control instruction of the user and converting it into a text control instruction, the present invention can perform dependency syntactic analysis on the obtained text control instruction to obtain multiple groups of predicate phrases in a preset form.

[0037] Specifically, the present invention can first use a Bi-LSTM model and a rule model to perform entity recognition on the obtained text control instruction to obtain multiple entity elements. Here, the Bi-LSTM is a mature open-source light deep learning model with the advantage of fast inference speed. The rule model stores preset entity keywords for constraining the recognition rules of the Bi-LSTM model and can retrieve and extract entity elements in the text according to the existing entity keywords.

[0038] Those skilled in the art will understand that the above-described embodiments of entity recognition using Bi-LSTM and rule-based models are merely non-limiting implementations of this invention, intended to perform entity recognition at a faster speed, and are not intended to limit the scope of protection of this invention.

[0039] Alternatively, in other embodiments, those skilled in the art can also use other deep learning models to achieve the same entity recognition effect, which will not be elaborated here.

[0040] Furthermore, after obtaining the aforementioned entity elements, the present invention can perform dependency parsing on the aforementioned text control instructions based on the extracted entity elements to obtain multiple sets of predicate phrases based on a preset form. Here, the predicate phrases based on the preset form at least include operation elements, position elements, and object elements. For predicate phrases lacking any elements, the present invention can first leave the missing elements empty, pending subsequent element filling correction.

[0041] Specifically, this invention can first use dependency parsing tools such as HanLP, LTP, and DDParser to perform dependency parsing, obtaining multiple sets of first predicate phrases lacking one or more elements and / or second predicate phrases without any missing elements. For second predicate phrases without any missing elements, this invention can directly use the second predicate phrase to formulate control commands and send the control commands to the corresponding home terminal to control the home terminal to perform the corresponding operation.

[0042] Conversely, for a first predicate phrase lacking any of the above elements, the present invention performs cluster analysis on its first sentence vector and the second sentence vectors of the remaining second predicate phrases to determine the associated predicate phrases of the first predicate phrase, and uses the corresponding elements of the associated predicate phrases to fill in the missing elements of the first predicate phrase, so as to complete the empty elements therein.

[0043] Specifically, the present invention can first determine the number parameter k based on the number of elements involved in the preset form. Here, the parameter k can be the number of predicate phrases without missing any entity elements. Then, the present invention can supplement the first sentence vector of the first predicate phrase with its direction element corresponding to the second sentence vector, and assign negative values ​​to the missing elements to determine the third sentence vector. Here, the direction element indicates the order of the predicate phrases. For example, if the first predicate phrase precedes the corresponding second predicate phrase, the value of the direction element of the first predicate phrase can be 1.0. Conversely, if the first predicate phrase follows the corresponding second predicate phrase, the value of the direction element of the first predicate phrase can be -1.0. Thus, the present invention can use the direction element to represent the order of the predicate phrases. Furthermore, regarding the assignment strategy for each element, the present invention can assign corresponding small positive values ​​to different entity elements and assign negative values ​​to empty elements to indicate whether there are empty slots in the predicate phrases. For example, when the entity element of a predicate phrase is "study room", the value of the dimension of this entity element can be 0.01. When the entity element is "living room", the value of its dimension can be 0.02. When the entity element is empty (i.e., the predicate phrase lacks an entity element), the value of the empty entity element's dimension can be -1.0. Subsequently, this invention can add the aforementioned directional element as a new dimensional feature to the sentence vector of the predicate phrase, and assign values ​​to each entity element in the sentence vector of each predicate phrase as described above.

[0044] Thus, even when the first predicate phrase lacks entity elements, its sentence vector can still accurately represent the missing entity elements and the order of the first predicate phrase in the sentence. For example, after clustering multiple predicate phrases such as "turn on the bedroom light," "turn off the air conditioner," and "turn off the living room light and air conditioner," the K-means clustering analysis model, based on the number parameter k, will group "turn on the bedroom light" and "turn off the air conditioner" into group A, and "turn off the living room light and air conditioner" into group B. Then, this invention can, based on the corresponding element "bedroom" in the second predicate phrase of "turn on the bedroom light" in group A, supplement the first predicate phrase "turn off the air conditioner" in group A that lacks entity elements to "turn off the bedroom air conditioner." Afterward, the control system of the home terminal can formulate control instructions based on the first predicate phrase filled with the missing elements and send these control instructions to the corresponding home terminal to control the home terminal to perform the corresponding operation.

[0045] Those skilled in the art will understand that the above-described embodiments of adding a directional dimension and assigning negative values ​​to missing elements are merely a non-limiting implementation of the present invention. This dimension is intended to provide a scheme that can accurately characterize the correlation between multiple sets of predicate phrases, rather than being used to limit the scope of protection of the present invention.

[0046] Alternatively, in other embodiments, those skilled in the art may also employ other assignment methods or use other features that can indicate the degree of relevance to achieve the same effect of determining the relevance predicate phrase.

[0047] Furthermore, in some embodiments of the present invention, after extracting entity elements and before performing clustering analysis, the present invention may preferably use models such as BERT and Word2Vec to calculate the sentence vectors of each predicate phrase, and perform similarity matching between the sentence vectors of each predicate phrase and the sentence vectors of the commonly used smart home corpus after offline processing, so as to delete predicate phrases with low similarity, thereby reducing the number of predicate phrases participating in clustering analysis, reducing the data processing load of the clustering analysis model, and improving the processing speed of the control method. Here, the similarity matching can be implemented based on one or more of cosine distance and Jaccard distance to accurately characterize the similarity of each sentence vector. The specific implementation method is a well-known technique in the art and will not be described in detail here.

[0048] Furthermore, before performing similarity matching, this invention can prioritize recall processing of the sentence vectors of each predicate phrase, eliminating some completely irrelevant sentence vectors first. Specifically, the recall method can employ multi-path recall, using different strategies and simple models to filter the sentence vectors in parallel, thereby deleting sentences with low similarity. Thus, by performing recall processing before the computationally intensive similarity matching, reducing a large number of sentence vectors with very low similarity, this invention can further reduce the exit processing load of the similarity matching unit, thereby further improving the processing speed of the aforementioned control method.

[0049] like Figure 1 and Figure 2 As shown, after determining that the second predicate phrase in the user command is complete and that the first predicate phrase with missing elements is filled in based on cluster analysis, the control system of the home terminal can, via the processor 102, formulate corresponding control commands based on these predicate phrases, and transmit each control command to the corresponding smart home terminal via the data bus 101, thereby enabling accurate and reliable control of the smart home terminal. In this way, the present invention can accurately identify the grouped entity portion in the smart home control command, thereby improving the accuracy and reliability of smart home voice control.

[0050] Although the methods described above are illustrated and depicted as a series of actions for the sake of simplicity, it should be understood and appreciated that these methods are not limited by the order of the actions, as some actions may occur in a different order and / or concurrently with other actions from the illustrations and descriptions herein or not illustrated and described herein but which may be understood by those skilled in the art, according to one or more embodiments.

[0051] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability between hardware and software, the various illustrative components, blocks, modules, circuits, and steps are described above in a generalized manner in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. Those skilled in the art may implement the described functionality in different ways for each specific application, but such implementation decisions should not be construed as departing from the scope of the invention.

[0052] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of both. The software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor such that the processor can read and write information to / from the storage medium. In an alternative, the storage medium may be integrated into the processor. The processor and storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and storage medium may reside as discrete components in the user terminal.

[0053] In one or more exemplary embodiments, the described functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functionality may be stored or transmitted as one or more instructions or code on or through a computer-readable medium. A computer-readable medium includes both computer storage media and communication media, encompassing any medium that facilitates the transfer of a computer program from one location to another. A storage medium may be any available medium accessible to a computer. By way of example and not limitation, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage, disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and is accessible to a computer. Any connection is also legitimately referred to as a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of a medium. As used in this article, disk and disc include compact discs (CDs), laser discs, optical discs, digital multi-purpose discs (DVDs), floppy disks, and Blu-ray discs. Disks typically reproduce data magnetically, while discs reproduce data optically using lasers. Combinations of these should also be included within the scope of computer-readable media.

[0054] The prior description of this disclosure is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to this disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not intended to be limited to the examples and designs described herein, but should be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for controlling a home terminal, characterized in that, Includes the following steps: Obtain the user's voice control commands and convert them into text control commands; The text control instructions are subjected to dependency parsing to obtain multiple sets of predicate phrases based on a preset form, wherein the predicate phrases based on the preset form include at least an operation element, a position element, and an object element; For the first sentence vector of the first predicate phrase lacking any of the aforementioned elements and the second sentence vector of the remaining second predicate phrases, the number parameter k is determined based on the number of elements involved in the preset form; The first sentence vector of the first predicate phrase is supplemented with the direction elements of the second sentence vector, and the missing elements are negatively assigned to determine the third sentence vector; Based on the numerical parameter k, the K-means model is used to perform cluster analysis on the third sentence vector and each of the second sentence vectors to determine the associated predicate phrases of the first predicate phrase; Use the corresponding elements of the associated predicate phrase to fill in the missing elements of the first predicate phrase; and Control the corresponding home terminal based on the first predicate phrase filled with the element.

2. The control method as described in claim 1, characterized in that, The step of performing dependency parsing on the text control instructions to obtain multiple sets of predicate phrases based on a preset form includes: Based on preset entity keywords, entity elements are extracted from the text control instructions, wherein the entity elements include the operation element, the position element, and the object element; and Based on the extracted entity elements, the text control instructions are subjected to dependency parsing to obtain the multiple sets of predicate phrases.

3. The control method as described in claim 2, characterized in that, The step of extracting entity elements from the text control command based on preset entity keywords includes: A combination of a Bi-LSTM model and a rule model is used to extract entity elements from the text control instructions. The rule model stores preset entity keywords to constrain the recognition rules of the Bi-LSTM model.

4. The control method as described in claim 2, characterized in that, Before performing cluster analysis on the first sentence vector of the first predicate phrase lacking any of the aforementioned elements and the second sentence vectors of the remaining second predicate phrases, the control method further includes the following steps: Calculate the sentence vector of each group of predicate phrases; and Recall analysis and / or similarity matching are performed on the sentence vectors of the predicate phrases in each group and multiple preset sentence vector samples to filter out predicate phrases with low similarity to each sentence vector sample.

5. The control method as described in claim 1, characterized in that, The step of acquiring the user's voice control commands and converting them into text control commands includes: The user-provided voice control command is acquired and parsed to generate a first candidate text; Replace at least one word in the first candidate text with at least one set of near-homophones (pre-set) to generate multiple second candidate texts. The probability of the first candidate text and the plurality of second candidate texts is calculated using a language model; and The candidate text with the highest probability is determined as the text control instruction.

6. The control method as described in claim 1, characterized in that, It also includes the following steps: Control the corresponding home terminal based on the second predicate phrase which does not lack any of the aforementioned elements.

7. A control system for a home terminal, characterized in that, include: A microphone is used to acquire the user's voice control commands; The communication module provides a communication connection to at least one home terminal. Memory, used to store computer instructions; as well as The processor is communicatively connected to the microphone, the communication module, and the memory, and is configured to execute the computer instructions to implement the control method of the home terminal as described in any one of claims 1 to 6.

8. A computer-readable storage medium storing computer instructions thereon, characterized in that, When the computer instructions are executed by the processor, the control method of the home terminal as described in any one of claims 1 to 6 is implemented.

Images