A positioning method for voice-controlled IoT device

Abstract

The present invention provides a method for positioning a voice-controlled IoT device, comprising the following steps: obtaining all device information of the user through user information and performing deduplication processing; identifying and replacing custom names; identifying intents and entities; and selecting and positioning devices according to intents and custom names. By constructing a knowledge graph of statements and device search rules to identify devices in user voice commands, the problem of inaccurate device positioning when the user's expression is not standard is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and in particular to a positioning method for a voice-controlled Internet of Things device. Background Art

[0002] Smart home is the embodiment of Internet of Things. Smart home connects various devices in the home (such as audio and video equipment, lighting system, curtain control, air conditioning control, security system, digital cinema system, network home appliances and three-meter copy, etc.) through the Internet of Things technology, providing home appliance control, lighting control, curtain control, telephone remote control, indoor and outdoor remote control, anti-theft alarm, environmental monitoring, HVAC control, infrared forwarding and programmable timing control and other functions and means. Compared with ordinary home, smart home not only has traditional residential functions, but also combines architecture, network communication, information home appliances, equipment automation, and integrates system, structure, service and management into an efficient, comfortable, safe, convenient and environmentally friendly living environment. It provides a full range of information interaction functions, helps families keep information exchanges with the outside world, optimizes people's lifestyles, helps people arrange time effectively, enhances the safety of home life, and even saves money for various energy costs.

[0003] In today's era, smart homes have gradually entered people's lives, and it is convenient and fast to control various home appliances through voice. In order to enhance the user's smart experience, it is particularly important to accurately recognize the devices that users want to control with voice. At present, the technology for directly identifying standard device names or custom names is relatively mature, but if the user's voice commands use non-standard expressions such as synonyms of these words, the recognition effect is often poor. In addition, when registering a smart scene with multiple mixed devices, distinguishing whether the user's voice command is to control an ordinary device or a smart scene is a difficulty that needs to be solved urgently. Summary of the invention

[0004] The purpose of the present invention is to provide a positioning method for voice-controlled IoT devices, which aims to solve the problem of inaccurate device positioning when the user's voice control of IoT devices is not standard. The essence of this problem is that the text referring to the device in the user's voice command and its reference relationship cannot be fully recognized.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for positioning a voice-controlled IoT device comprises the following steps:

[0007] Obtain all device information of the user through user information and perform deduplication processing;

[0008] Custom name recognition and replacement;

[0009] Intent and entity recognition;

[0010] Device selection and positioning based on intent and custom name.

[0011] In some embodiments, the custom name recognition and replacement includes:

[0012] Check whether the user text contains a custom device name. If so, replace the custom name in the user's statement with the standard device name. The found custom device name is used to locate the device, and the replaced text is used for intent recognition.

[0013] In some embodiments, searching whether the user text contains a custom device name includes: direct text matching and knowledge graph text matching, followed by information fusion.

[0014] In some embodiments, the direct text matching includes: looping through all user devices, performing text processing on each device custom name, including removing "-", converting numeric expressions, converting uppercase and lowercase expressions, etc., to obtain all custom name expressions, and if a name is in the user text, it is a candidate custom name;

[0015] The names with the longest text of common devices and smart scenes in the candidate custom names are selected as their respective custom names. If there are none, no comparison is made.

[0016] In some embodiments, the knowledge graph text matching includes: storing user statement data in the knowledge graph, including synonyms and approximate word data; aggregating the custom names of all devices with custom names and standard device names in the user device list, and uniformly searching for synonyms and near-synonyms data in the knowledge graph;

[0017] First, synonym data is used for direct matching of user statements after text processing. The matching method is the same as direct text matching. The longest text names of ordinary devices and smart scenes are found as the custom names expressed by the user. If neither is found, synonym data is used for direct matching of user statements after text processing.

[0018] In some embodiments, if any, the custom name in the user's statement is replaced with a standard device name, the found custom device name is used to locate the device, and the replaced text is used for intent recognition, including:

[0019] Each method first finds the common device and smart scene with the longest matching text, and then compares and selects the common device and smart scene with the longest text; the rules for determining the final custom name are:

[0020] 1) When common devices and smart scenes exist at the same time, if the scene name contains common devices and is longer than the common device, the custom name is the scene, otherwise it is the common device;

[0021] 2) When only common devices exist, the custom name is the corresponding name;

[0022] 3) When there is only a smart scene, there are two cases: when the user text contains a standard common device name and the found smart scene name does not contain this name, there is no custom name; in other cases, the custom name is the found smart scene name;

[0023] 4) When neither ordinary devices nor smart scenes exist, there is no custom name;

[0024] Finally, if it is determined that there is a custom device name in the user text, the corresponding name is replaced.

[0025] In some embodiments, the intent and entity recognition is to perform intent and entity recognition using the user instructions replaced by this article.

[0026] In some embodiments, the selecting and locating a device according to the intent and the custom name includes:

[0027] 1) Multiple rounds of indexing intend to select the index-th device in the previous round;

[0028] 2) Multiple rounds of pure action intentions select the device in the previous round;

[0029] 3) For the rest, when the custom name and the standard device name are the same, the device is selected according to the intent; when the custom name and the standard device name are different, the device is selected according to the custom name; if there is no custom name, there is no device for the smart scene-related intent, and the rest of the intents select devices according to the intent.

[0030] In some embodiments, selecting a device by custom name refers to selecting a device whose custom name is consistent and whose standard device name is the same as the intended device.

[0031] In some embodiments, selecting a device according to intent refers to selecting a device based on the device indicated by intent, first selecting a device with the same standard device name as the device indicated by intent, and if there is none, selecting a device with a custom name containing the device indicated by intent, and if there is none, then there is no device.

[0032] The beneficial effects of the voice-controlled positioning method of the IoT device provided in this application include but are not limited to:

[0033] The technical solution of this patent is to identify the device referred to in the user's voice command by constructing a knowledge graph of synonyms and near-synonyms and device search rules to obtain the desired device of the user's command. This solves the problem of inaccurate device positioning caused by non-standard expression when the user controls the IoT device by voice, which is conducive to improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] AttachedFigure 1 It is a flow chart of the present invention.

[0035] Attached Figure 2 It is a flow chart of custom name replacement of the present invention.

[0036] Attached Figure 3 It is a flow chart of equipment selection of the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0038] On the contrary, the present application covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present application as defined by the claims. Further, in order to make the public have a better understanding of the present application, some specific details are described in detail in the detailed description of the present application below. Those skilled in the art can fully understand the present application without the description of these details.

[0039] The following will be combined Figure 1-3 A method for positioning a voice-controlled IoT device according to an embodiment of the present application is described in detail. It should be noted that the following embodiments are only used to explain the present application and do not constitute a limitation on the present application.

[0040] The technical essence of this patent to solve this problem is to construct a knowledge graph of sayings and device search rules to locate the device. The unique technical feature is to identify the text referring to the device through the knowledge graph of synonyms and near-synonyms and device search rules. The implementation of the method for locating the voice-controlled IoT device is mainly divided into the following steps:

[0041] S1: User equipment information acquisition

[0042] Obtain all device information of the user through user information and perform deduplication processing.

[0043] S2: Custom name recognition and replacement

[0044] Find whether the user text contains a custom device name. If so, replace the custom name in the user statement with the standard device name. For example, if a user customizes an air conditioner and names it "Xiaobai" and the user commands "turn on Xiaobai", the custom name "Xiaobai" is found and the user command is replaced with "turn on the air conditioner".

[0045] There are two ways to find custom names: direct text matching and knowledge graph text matching. Then information fusion is performed.

[0046] Direct text matching: loop all user devices, and process the text of each device custom name, including removing "-", converting numbers, converting uppercase and lowercase, etc., to obtain all custom name descriptions. If there is a name in the user text, it is a candidate custom name. Select the longest name of the common device and the smart scene in the candidate custom name as their respective custom names, otherwise no comparison will be made.

[0047] Knowledge graph text matching: Store the statement data in the knowledge graph offline, including synonyms and approximate word data. Synonym data include "dining room light" and "dining room light", and near-synonym data include "master bedroom light" and "bedroom light". Aggregate the custom names of all devices with custom names and standard device names in the user device list, and uniformly search for synonyms and near-synonym data with a depth of 1 in the knowledge graph. First, use the synonym data for direct matching of user statements after text processing. The matching method is the same as that described in the direct text matching method. Find the longest text name of each ordinary device and smart scene as the custom name expressed by the user. If neither is found, use the near-synonym data for direct matching of user statements after text processing. The method is the same as above.

[0048] Each method first finds the common device and smart scene with the longest matching text, and then compares and selects the common device and smart scene with the longest text. The rules for determining the final custom name are: 1) When common devices and smart scenes exist at the same time, if the scene name contains common devices and is longer than common devices, the custom name is the scene, otherwise it is a common device. 2) When only common devices exist, the custom name is the corresponding name. 3) When there are only smart scenes, there are two cases. When the user text contains the standard common device name and the found smart scene name does not contain the name, there is no custom name. In other cases, the custom name is the found smart scene name. 4) When neither common devices nor smart scenes exist, there is no custom name.

[0049] Finally, if it is determined that there is a custom device name in the user text, the corresponding name is replaced.

[0050] S3: Intent Recognition

[0051] The text processed by S2 is recognized through multiple rounds of recognition and intent recognition to obtain the text intent.

[0052] S4: Equipment selection

[0053] Device selection is targeted based on intent and custom name.

[0054] The multi-round index intention selects the index-th device in the previous round. The multi-round index intention refers to the expression "which one" and so on, and the index is the recognized number. For example, if the user has multiple air conditioners, the first round of instructions is "turn on the air conditioner", and the system provides feedback "There are multiple air conditioners, which one do you want to turn on?" The user's instruction is "the first one", and the first device in the previous round of devices is selected.

[0055] Multiple rounds of pure action intents select the device in the previous round. Multiple rounds of pure action intents refer to the situation where the user does not mention the device name in this round, but only says the relevant action. For example, if the user's first round command is "turn on the air conditioner", and the second round command is "adjust the temperature to 26 degrees", the device in the previous round is selected.

[0056] If the above conditions are not met, a single round of device selection is performed. Determine whether the custom name has been replaced. If it has been replaced, there are two cases. When the custom name and the standard device name are the same, the device is selected according to the intent. When the custom name and the standard device name are different, the device is selected according to the custom name. If it has not been replaced, determine whether it is a smart scene-related intent. If so, there is no device, otherwise the device is selected according to the intent. The above "selecting devices according to intent" means selecting devices based on the devices indicated by the intent, and giving priority to devices with the same standard device name and the device indicated by the intent. If not, select devices with the intent-indicated device in the custom name. If not, select devices. "Selecting devices by custom name" means selecting devices with consistent custom names and the same standard device name and intent-indicated device based on the custom name.

[0057] S5: Result output

[0058] Print multiple or a single device if found, or a default virtual device if not found.

[0059] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for positioning a voice-controlled IoT device, characterized in that: The following steps are involved: Obtain all device information of the user through user information and perform deduplication processing; Custom name recognition and replacement; Intent and entity recognition; Device selection and positioning based on intent and custom name; The custom name identification and replacement includes: Find whether the user text contains a custom device name. If so, replace the custom name in the user's statement with the standard device name. The found custom device name is used to locate the device, and the replaced text is used for intent recognition. Finding whether the user text contains a custom device name includes: direct text matching and knowledge graph text matching, followed by information fusion; If there is one, the custom name in the user statement is replaced with the standard device name. The found custom device name is used to locate the device, and the replaced text is used for intent recognition, including: Each method first finds the common device and smart scene with the longest matching text, and then compares and selects the common device and smart scene with the longest text; the rules for determining the final custom name are: 1) When common devices and smart scenes exist at the same time, if the scene name contains common devices and is longer than the common device, the custom name is the scene, otherwise it is the common device; 2) When only common devices exist, the custom name is the corresponding name; 3) When there is only a smart scene, there are two cases: when the user text contains a standard common device name and the found smart scene name does not contain a standard common device name, there is no custom name; in other cases, the custom name is the found smart scene name; 4) When neither ordinary devices nor smart scenes exist, there is no custom name; Finally, if it is determined that there is a custom device name in the user text, the corresponding name is replaced; The device selection and positioning according to the intent and the custom name includes: 1) Multiple rounds of indexing intent select the index-th device in the previous round; 2) Multiple rounds of pure action intentions select the device in the previous round; 3) For the rest, when the custom name and the standard device name are the same, the device is selected according to the intent; when the custom name and the standard device name are different, the device is selected according to the custom name; if there is no custom name, there is no device for the smart scene-related intent, and the rest of the intents select devices according to the intent.

2. The method for positioning a voice-controlled IoT device according to claim 1, characterized in that: The direct text matching includes: looping through all user devices, performing text processing on each device custom name, including removing "-", converting numeric expressions, and converting uppercase and lowercase expressions, to obtain all custom name expressions, and if a name exists in the user text, it is a candidate custom name; The names with the longest text of common devices and smart scenes in the candidate custom names are selected as their respective custom names. If there are none, no comparison is made.

3. The method for positioning a voice-controlled IoT device according to claim 1, characterized in that: The knowledge graph text matching includes: storing user statement data in the knowledge graph, including synonyms and approximate word data; aggregating the custom names of all devices with custom names and standard device names in the user device list, and uniformly searching for synonyms and near-synonyms data in the knowledge graph; First, synonym data is used for direct matching of user statements after text processing. The matching method is the same as direct text matching. The longest text names of ordinary devices and smart scenes are found as the custom names expressed by the user. If neither is found, synonym data is used for direct matching of user statements after text processing.

4. The method for positioning a voice-controlled IoT device according to claim 1, characterized in that: The intent and entity recognition is to perform intent and entity recognition using the user instructions replaced by this article.

5. The method for positioning a voice-controlled IoT device according to claim 1, characterized in that: Selecting a device by custom name refers to selecting a device whose custom name is consistent and whose standard device name is the same as the intended device.

6. The method for positioning a voice-controlled IoT device according to claim 1, characterized in that: The selecting device according to the intent refers to selecting the device according to the device indicated by the intent, first selecting the device with the same standard device name as the device indicated by the intent, and if there is no device, then selecting the device with the custom name containing the device indicated by the intent, and if there is no device, then there is no device.

Images