Inference of semantic label(s) for assistant device(s) based on device-specific signal(s)

Abstract

To provide a method for inferring a semantic label for an assistant device based on a device-specific signal, and a storage medium.SOLUTION: Implementations can identify a given assistant device from among a plurality of assistant devices in an ecosystem, obtain device-specific signal(s) that are generated by the given assistant device, process the device-specific signal(s) to generate candidate semantic label(s), select a semantic label for the given semantic device from among the candidate semantic label(s), and assign, in a device topology representation of the ecosystem, the given semantic label to the given assistant device. Implementations can optionally receive a spoken utterance that includes a query or command at the assistant device(s), determine a semantic property of the query or command matches the given semantic label to the given assistant device, and assign the given semantic label to the given assistant device.SELECTED DRAWING: Figure 3

Description

[Background technology]

[0001] A person can engage in human-to-computer interactions with an interactive software application referred to herein as an “automated assistant” (also referred to as a “chatbot,” “interactive personal assistant,” “intelligent personal assistant,” “personal voice assistant,” “conversational agent,” etc.). For example, a person (who may be referred to as a “user” when interacting with an automated assistant) may provide input (e.g., commands, queries, and / or requests) to the automated assistant, which may cause the automated assistant to generate and provide a responsive output, control one or more Internet of things (IoT) devices, and / or perform one or more other functions. The input provided by the user may be, for example, a spoken natural language input (i.e., an utterance) that may be converted into text (or other semantic representation) and then further processed, and / or a typed natural language input.

[0002] In some cases, the automated assistant may include an automated assistant client that runs locally on the assistant device and with which the user interacts directly, as well as a cloud-based counterpart that leverages the virtually limitless resources of the cloud to help the automated assistant respond to the user's input. For example, the automated assistant may provide its cloud-based counterpart with an audio recording of the user's speech (or a text transcription thereof), and optionally data indicative of the user's identity (e.g., a certificate). The cloud-based counterpart may perform various processing on the query and return results to the automated assistant client, which may then provide the corresponding output to the user.

[0003] Many users may engage with an automated assistant using multiple assistant devices. For example, some users may have a coordinated "ecosystem" of assistant devices that can receive user input directed to the automated assistant and / or be controlled by the automated assistant, such as one or more smartphones, one or more tablet computers, one or more vehicle computing systems, one or more wearable computing devices, one or more smart televisions, one or more interactive standalone speakers, and / or one or more IoT devices, among other assistant devices. A user may use any of these assistant devices (assuming the automated assistant client is installed and the assistant device is capable of receiving input) to engage in human-computer interactions with the automated assistant. In some cases, these assistant devices may be scattered around the user's home, vacation home, workplace, and / or other structures. For example, mobile assistant devices such as smartphones, tablets, smart watches, etc. may be on the user's person and / or wherever the user last left them. Other assistant devices such as traditional desktop computers, smart televisions, interactive standalone speakers, and IoT devices may be more stationary, but may still be located in various places (e.g., rooms) in the user's home or workplace. Summary of the Invention [Problem to be solved by the invention]

[0004] Techniques exist for allowing a user (e.g., a single user, multiple users in a household, coworkers, housemates, etc.) to manually assign labels to assistant devices in an ecosystem of assistant devices and subsequently interact with or control any one of the assistant devices utilizing an automated assistant client of any one of the assistant devices. For example, a user can issue a spoken command to the automated assistant client of an assistant device, "Show me chili recipes on kitchen device," to cause the assistant device (or another assistant device in the ecosystem) to pull search results for chili recipes and present the search results to the user via the kitchen device. However, such techniques require the user to specify a particular assistant device by a previously assigned label (e.g., "kitchen device") that the user may have forgotten, or require the automated assistant to infer the "best" device (e.g., the device closest to the user) to provide search results. Furthermore, if a particular assistant device is newly introduced to the ecosystem or moves within the ecosystem, the label assigned to a particular assistant device by the user may not represent the particular assistant device. [Means for solving the problem]

[0005] Implementations described herein relate to assigning semantic labels to each assistant device in a device topology representation of an ecosystem including multiple assistant devices. The semantic labels assigned to each assistant device may be inferred based on one or more device-specific signals associated with the respective assistant device. The one or more device-specific signals may include, for example, one or more queries (if any) previously received at the respective assistant device, one or more commands (if any) previously executed at the respective assistant device, instances of ambient noise previously detected at the respective assistant device (and optionally only when speech reception was active at the respective assistant device), unique identifiers (or labels) of any other assistant devices positionally proximate to the respective assistant device, and / or user preferences of a user associated with the ecosystem determined based on user interactions with multiple assistant devices in the ecosystem. Each of the one or more device-specific signals associated with the respective assistant device may be processed to classify each of them into one or more semantic categories of a plurality of heterogeneous semantic categories. One or more candidate semantic labels may be generated for each assistant device based on the semantic category into which one or more of the device-specific signals are classified. Further, a given semantic label from among one or more candidate semantic labels and for a given one of the respective assistant devices may be selected and assigned to the given one of the respective assistant devices in the device topology representation of the ecosystem.

[0006] For example, assume that a given assistant device is a two-way standalone speaker device with a display located at a primary residence of a user associated with the ecosystem. Further assume that queries related to searching for food recipes have been received and executed at the given assistant device and / or commands related to setting timers have been received and executed at the given assistant device, assume that instances of ambient noise have been detected at the client device, assume that a unique identifier (or indicator) associated with an additional assistant device in the ecosystem corresponding to a "smart oven" is detected at the given assistant device, and assume that user preferences of a user associated with the ecosystem indicate that the user prefers a fictional chef named Johnny Flay. In this example, assume that a query related to searching for a food recipe is classified into the "recipe", "kitchen", and / or "kitchen" categories, and that a command related to setting a timer is classified into the "timing" and / or "cooking" categories, assume further that an instance of ambient noise is classified into the "kitchen" and / or "cooking" categories based on ambient noise capturing cooking sounds (e.g., food frying in a frying pan, a knife cutting food, a microwave in use, etc.), assume further that a unique identifier (or indicator) of a "smart oven" associated with an additional assistant device is classified into the "kitchen" and / or "cooking" categories, and assume further that a fictional chef is classified into the "cooking" category (or into the more specific category of "Johnny Flay"). As a result, the following semantic indicator candidates may be generated for a two-way standalone speaker with a display: "recipe display device", "kitchen display device", "cooking display device", "timing display device", and "Johnny Flay device".Further, from among the candidate semantic labels, a given semantic label may be assigned to a two-way standalone speaker device having a display in the device topology representation of the ecosystem for the user's main residence.

[0007] In some implementations, a given semantic label may be automatically assigned to a given assistant device in the device topology representation of the ecosystem. For example, if the reliability level associated with the given semantic label meets a threshold reliability level, the given semantic label may be automatically assigned to a given assistant device in the device topology representation of the ecosystem. The reliability level associated with the given semantic label may be determined during processing of one or more device-specific signals associated with the given assistant device. For example, the reliability level associated with the given assistant device may be based on the amount of one or more device-specific signals that are classified into one or more of the semantic categories. For example, if nine queries related to searching for food recipes have been received at a given assistant device and only one query related to searching for weather information has been received at a given assistant device, the semantic label "cooking display device" or "recipe display device" may be automatically assigned to the given assistant device in the device topology representation of the ecosystem (even if the given assistant device is not located in the user's kitchen). For example, the reliability level associated with a given assistant device may be based on a measure determined based on the output generated using a semantic classifier and / or an ambient noise detection model to process one or more device-specific signals. For example, a previously received query or command (or its corresponding text) may be processed using a semantic classifier to classify each of the queries or commands into one or more of the semantic categories, the ambient noise instances may be processed using an ambient noise detection model to classify each of the instances of ambient noise into one or more of the semantic categories, and the unique identifiers (or indicators) may be processed using a semantic classifier to classify each of the unique identifiers (or indicators) into one or more of the semantic categories along with a respective measure.As another example, if a given semantic label is unique (with respect to other assistant devices that are positionally close to the given assistant device in the ecosystem), the given semantic label may be automatically assigned to the given assistant device in the device topology representation of the ecosystem.

[0008] In some additional or alternative implementations, in response to receiving a user input to assign a given semantic label to a given assistant device, the given semantic label may be assigned to the given assistant device in the device topology representation of the ecosystem. For example, a prompt may be generated to ask a user associated with the ecosystem to select a given semantic label from one or more of the semantic label candidates. The prompt may be rendered on the user's client device (e.g., the user's given assistant device or another client device (e.g., a mobile phone)), and in response to receiving the selection of the given semantic label, the given semantic label may be assigned to the given assistant device. For example, assume that one or more semantic label candidates include a "cooking display device", a "recipe display device", and a "weather display device". In this case, the prompt may include each of the semantic label candidates and request the user to select a given semantic label to be assigned to the given assistant device from these semantic label candidates (and optionally replace the existing semantic label). It should be understood that while the above examples are described with respect to a given assistant device being assigned a single semantic label, this is for illustrative purposes and is not intended to be limiting. For example, the assistant devices described herein may be assigned multiple semantic labels such that each assistant device is stored in association with a list of semantic labels.

[0009] In various implementations, after one or more of the semantic labels are assigned to each assistant device in the device topology representation of the ecosystem, the semantic labels assigned to the assistant device according to the techniques described herein may also be utilized in processing the speech received at one or more of the assistant devices in the ecosystem. For example, audio data corresponding to the speech may be processed to identify semantic features included in the speech. Further, an embedding (e.g., a word2vec representation) of the identified semantic features may be generated and compared with the multiple embeddings (e.g., the respective word2vec representations) of the respective semantic labels assigned to the assistant devices in the ecosystem. Further, based on the comparison, it may be determined that the semantic features match a given embedding of the multiple embeddings of the respective semantic labels. For example, assume that the embedding is a word2vec representation. In this example, the cosine distance between the word2vec of the semantic feature and each of the word2vec representations of the respective semantic indicators can be determined, and the given semantic indicators associated with the respective cosine distances that meet the threshold distance can be utilized to determine that the semantic feature of the utterance matches (e.g., an exact match or a rough match) with the given semantic indicator. As a result, a given assistant device associated with the given semantic indicator can be selected to satisfy the utterance. Additionally or alternatively, the user's proximity to the given assistant device and / or the device capabilities of the given assistant device can be considered in selecting the given assistant device to satisfy the utterance.

[0010] By inferring and assigning semantic labels to assistant devices in the ecosystem using the techniques described herein, the device topology representation of the ecosystem can be kept up to date without requiring diverse user interface inputs (or any user interface inputs) to do so. Furthermore, the semantic labels assigned to the assistant devices are semantically useful to the user in that the semantic labels assigned to each assistant device are selected based on the use of the respective assistant device and / or the respective part of the ecosystem in which the respective assistant device is located. Thus, when an utterance is received at one or more of the assistant devices in the ecosystem, the automated assistant can more robustly and / or accurately select one or more of the assistant devices that are best suited to satisfy the utterance. As a result, the amount and / or length of user inputs received by one or more of the assistant devices in the ecosystem can be reduced, thereby saving computational and / or network resources at the assistant device by reducing network traffic, because a user associated with the ecosystem does not need to specify a particular device to satisfy the utterance or to repeat the utterance when the wrong device is selected to satisfy the utterance. Additionally, when an assistant device is newly added to the ecosystem, when it is moved within the ecosystem, or when it is located in a part of the ecosystem that has been repurposed (e.g., a room in a user's main residence that has been converted from a study to a bedroom), the amount of user input received by one or more of the assistant devices in the ecosystem can be reduced because the user does not need to manually update the device topology representation via a software application associated with the ecosystem.

[0011] The above description is provided as a summary of only some implementations of the present disclosure. Further descriptions of these and other implementations are described in more detail herein. As one non-limiting example, various implementations are described in more detail in the claims contained herein.

[0012] In addition, some implementations include one or more processors of one or more computing devices, the one or more processors operable to execute instructions stored in associated memory, the instructions configured to cause performance of any of the methods described herein. Some implementations also include one or more non-transitory computer-readable storage media that store computer instructions executable by the one or more processors to perform any of the methods described herein.

[0013] It is to be understood that all combinations of the foregoing concepts, and additional concepts more fully described herein, are contemplated as part of the subject matter disclosed herein, for example, all combinations of claimed subject matter appearing at the end of this disclosure are contemplated as part of the subject matter disclosed herein. [Brief description of the drawings]

[0014] [Figure 1] FIG. 1 is a block diagram of an example environment in which implementations disclosed herein may be implemented. [Figure 2A] FIG. 13 illustrates examples related to assigning a given semantic indicator to a given assistant device that is newly added to an ecosystem of assistant devices and / or moves within an ecosystem of assistant devices, according to various implementations. [Figure 2B]FIG. 13 illustrates examples related to assigning a given semantic indicator to a given assistant device that is newly added to an ecosystem of assistant devices and / or moves within an ecosystem of assistant devices, according to various implementations. [Diagram 3] 1 is a flowchart illustrating an example method for assigning a given semantic indicator to a given assistant device in an ecosystem, according to various implementations. [Figure 4] 1 is a flowchart illustrating an example method of using assigned semantic indicators in satisfying a query or command received at an assistant device in an ecosystem, according to various implementations. [Diagram 5] FIG. 1 illustrates an exemplary architecture of a computing device, according to various implementations. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] There has been a proliferation of smart, multi-sensing, network-connected devices (also referred to herein as assistant devices), such as smartphones, tablet computers, vehicle computing systems, wearable computing devices, smart televisions, interactive standalone speakers (e.g., with or without displays), sound speakers, home alarms, door locks, cameras, lighting systems, treadmills, thermostats, weight scales, smart beds, watering systems, garage door openers, home appliances, baby monitors, fire alarms, moisture meters, etc. Often, multiple assistant devices are located within the boundaries of a structure such as a home, or within various associated structures, such as a user's primary residence and a user's secondary residence, a user's vehicle, and / or a user's workplace.

[0016] Furthermore, there is a proliferation of assistant devices (also referred to herein as assistant input devices) that each include an automated assistant client that can form a logical instance of an automated assistant. These assistant input devices may be dedicated to assistant functions (e.g., an interactive standalone speaker and / or a standalone audio / visual device that includes only an assistant client and related interfaces and is dedicated to assistant functions) or can perform assistant functions in addition to other functions (e.g., a mobile phone or tablet that includes an assistant client as one of a variety of applications). Moreover, some IoT devices may also be assistant input devices. For example, some IoT devices may include an automated assistant client and at least a speaker and / or microphone that serves (at least in part) as a user interface output and / or input device for the assistant interface of the automated assistant client. Some assistant devices may not implement an automated assistant client or may not have a means for interfacing with a user (e.g., a speaker and / or microphone), but they may still be controlled by an automated assistant (also referred to herein as assistant non-input devices). For example, a smart light bulb may not include an automation assistant client, a speaker, and / or a microphone, but commands and / or requests can be sent to the smart light bulb via the automation assistant to control functionality of the smart lighting (e.g., turn the light on / off, dim the light, change color, etc.).

[0017] Various techniques have been proposed for labeling and / or grouping assistant devices (including both assistant-input devices and assistant-non-input devices) in an ecosystem of assistant devices. For example, when a new assistant device is added to the ecosystem, a user associated with the ecosystem can manually assign a label (or a unique identifier) ​​to the new assistant device in the device topology representation of the ecosystem and / or manually add the new assistant device to a group of assistant devices in the ecosystem via a software application (e.g., via an automated assistant application, a software application associated with the ecosystem, a software application associated with the new assistant device, etc.). As described herein, the label originally assigned to the assistant device may be forgotten by the user or may not be semantically meaningful for how the assistant device is utilized or where the assistant device is located in the ecosystem. Furthermore, when an assistant device moves in the ecosystem, the user may be required to manually change the label assigned to the assistant device and / or manually change the group to which the assistant device is assigned via a software application. Otherwise, the label assigned to the assistant device and / or the group to which the assistant device is assigned may not accurately reflect the location or use of the assistant device and / or may not be semantically meaningful to the assistant device. For example, if a smart speaker labeled "living room speaker" is located in the living room of a user's main residence, but the smart speaker moves to the kitchen of the user's main residence, the smart speaker may still be labeled as "living room speaker" even though the label does not represent the location of the assistant device, unless the user manually changes the label in the device topology representation for the ecosystem of the user's main residence.

[0018] The device topology representation may include a label (or unique identifier) ​​associated with each assistant device. In addition, the device topology representation may specify a label (or unique identifier) ​​associated with each assistant device. The device attributes for a given assistant device may, for example, indicate one or more input and / or output modalities supported by the respective assistant device. For example, a device attribute for a standalone speaker-only assistant client device may indicate that it is capable of providing an auditory output but not a visual output. The device attributes for a given assistant device may additionally or alternatively, for example, identify one or more controllable states of the given assistant device, identify a party (e.g., a first party (1P) or a third party (3P)) that manufactures, distributes, and / or creates the firmware for the assistant device, and / or identify a unique identifier for the given assistant device, such as an immutable identifier provided by 1P or 3P, or a label assigned to the given assistant device by a user. According to various implementations disclosed herein, the device topology representation can optionally further specify which smart devices can be locally controlled by which assistant devices, local addresses of locally controllable assistant devices (or local addresses of hubs that can directly locally control those assistant devices), local signal strengths, and / or other priority indicators between the respective assistant devices. Furthermore, according to various implementations disclosed herein, the device topology representation (or a variation thereof) can be stored locally on each of the multiple assistant devices for use in locally controlling the assistant device and / or locally assigning an indicator thereto. Moreover, the device topology representation can specify groups associated with each assistant device, which can be defined at various levels of granularity.For example, multiple smart lights in the living room of a user's main residence may be considered to belong to the "living room lights" group. Additionally, if the living room of the main residence also includes a smart speaker, all of the assistant devices located in the living room may be considered to belong to the "living room assistant device" group.

[0019] The automation assistant can detect various events occurring in the ecosystem based on one or more signals generated by one or more of the assistant devices. For example, the automation assistant can process one or more of the signals to detect these events using an event detection model or an event detection rule. Furthermore, the automation assistant can cause one or more actions to be performed based on an output generated based on one or more of the signals for an event occurring in the ecosystem. In some implementations, the detected event can be a device-related event related to one or more of the assistant devices (e.g., assistant input devices and / or assistant non-input devices). For example, a given one of the assistant devices can detect when it is newly added to the ecosystem based on one or more wireless signals (and optionally a unique identifier related to the given one of the assistant devices included in one or more of the wireless signals) generated by the given one of the assistant devices. As another example, a given one of the assistant devices can detect when it moves within the ecosystem based on the given one of the assistant devices being surrounded by one or more different assistant devices that previously surrounded the given one of the assistant devices (and optionally determined based on the unique identifiers of each of the one or more different assistant devices). In these implementations, one or more of the actions performed by the automation assistant can include, for example, determining a semantic indicator for the given one of the assistant devices in response to determining that it is newly introduced to the ecosystem or has moved within the ecosystem, and causing the semantic indicator to be assigned to the given one of the assistant devices in a device topology representation of the ecosystem.

[0020] In some additional or alternative implementations, the detected event may be an acoustic event captured through a respective microphone of one or more assistant devices. The automated assistant may cause audio data capturing the acoustic event to be processed using an acoustic event model. The acoustic event detected by the acoustic event model may include, for example, using a hot word detection model to detect a hot word for invoking an automated assistant included in an utterance, using an ambient noise detection model to detect ambient noise in the ecosystem (and optionally while speech acceptance is active on a given one of the assistant devices), using a sound detection model to detect a particular sound in the ecosystem (e.g., glass breaking, a dog barking, a cat meowing, a doorbell ringing, a fire alarm going off, or a carbon monoxide detector going off), and / or other acoustic-related events that may be detected using the respective acoustic event detection model. For example, assume that audio data is detected through at least one respective microphone of the assistant device. In this example, the automated assistant may cause audio data to be processed by at least one hot word detection model of the assistant device to determine whether the audio data captures a hot word for invoking an automated assistant. Further, the automated assistant may additionally or alternatively cause the audio data to be processed by at least one ambient noise detection model of the assistant device to classify any ambient (or background) noise captured in the audio data into one or more disparate semantic categories of ambient noise (e.g., movie or television sounds, cooking sounds, and / or other disparate sound categories). Moreover, the automated assistant may additionally or alternatively cause the audio data to be processed by at least one sound detection model of the assistant device to determine whether any particular sound is captured in the audio data.

[0021] The implementations described herein relate to inferring semantic indicators for assistant devices based on one or more signals generated by each of the respective devices. The implementations further relate to assigning semantic indicators to assistant devices in a device topology representation of the ecosystem. The semantic indicators may be automatically assigned to the assistant devices or may be presented to a user associated with the ecosystem to select one or more of the semantic indicators to be assigned to the assistant devices. The implementations further relate to subsequently using the semantic indicators in processing an utterance to determine whether the utterance includes a word or phrase that matches any of the semantic indicators, and using an assistant device associated with the matching one of the semantic indicators to satisfy the utterance when it is determined that the utterance includes a word or phrase that matches one of the semantic indicators.

[0022] Turning now to FIG. 1, an example environment in which the techniques disclosed herein may be implemented is shown. The example environment includes multiple assistant input devices 106. 1-N (also referred to herein simply as “assistant input device 106”), one or more cloud-based automated assistant components 119, one or more assistant non-input systems 180, one or more assistant non-input devices 185 1-N (also referred to herein simply as “assistant non-input devices 185”), device activity database 191, machine learning (“ML”) model database, and device topology database 193. Assistant input devices 106 and assistant non-input devices 185 of FIG. 1 may also be collectively referred to herein as “assistant devices.”

[0023] One or more (e.g., all) of the assistant input devices 106 may be connected to a respective automated assistant client 118. 1-NHowever, in some implementations, one or more of the assistant input devices 106 may optionally run a respective automated assistant client 118. 1-N The automated assistant client 118 may lack an instance of the automated assistant client 118 and still include engines and hardware components (e.g., microphone, speaker, speech recognition engine, natural language processing engine, speech synthesis engine, etc.) for receiving and processing user input directed to the automated assistant. 1-N An instance of may be an application separate from (e.g., installed "on top of") the operating system of each assistant input device 106, or may alternatively be implemented directly by the operating system of each assistant input device 106. As described further below, the automated assistant client 118 1-N Each instance of is optionally connected to a respective user interface component 107 of any one of the respective assistant input devices 106. 1-N In responding to various requests provided by the cloud-based automated assistant component 119, the cloud-based automated assistant component 119 may interact with one or more of the cloud-based automated assistant components 119. In addition, as also described below, other engines of the assistant input device 106 may optionally interact with one or more of the cloud-based automated assistant components 119.

[0024] One or more cloud-based automation assistant components 119 may be implemented on one or more computing systems (e.g., servers, collectively referred to as "cloud" or "remote" computing systems) that are communicatively coupled to the respective assistant input devices 106 via one or more local area networks ("LANs" including Wi-Fi LANs, Bluetooth networks, near field communications networks, mesh networks, etc.) and / or wide area networks ("WANs" including the Internet, etc.). The communicative coupling of the cloud-based automation assistant components 119 with the assistant input devices 106 is indicated generally by 1101 in FIG. 1. Also, in some embodiments, the assistant input devices 106 may be communicatively coupled to one another via one or more networks (e.g., LANs and / or WANs), indicated generally by 1102 in FIG. 1.

[0025] One or more cloud-based automation assistant components 119 may also be communicatively coupled to one or more assistant non-input systems 180 via one or more networks (e.g., LAN and / or WAN). The communicative coupling of the cloud-based automation assistant components 119 with the assistant non-input systems 180 is generally indicated by 1103 in FIG. 1. In addition, the assistant non-input systems 180 may each be communicatively coupled to one or more (e.g., groups) of assistant non-input devices 185 via one or more networks (e.g., LAN and / or WAN). For example, a first assistant non-input system 180 may be communicatively coupled to and receive data from one or more first groups of assistant non-input devices 185, a second assistant non-input system 180 may be communicatively coupled to and receive data from one or more second groups of assistant non-input devices 185, and so on. The communicative coupling of the assistant non-input systems 180 with the assistant non-input devices 185 is generally indicated by 1104 in FIG. 1.

[0026] An instance of an automation assistant client 118, through its interaction with one or more of the cloud-based automation assistant components 119, may form what appears to a user to be a logical instance of an automation assistant 120 through which the user may engage in human-computer interactions. Two instances of such an automation assistant are shown in FIG. 1. A first automation assistant 120A, enclosed in dashed lines, includes an automation assistant client 1181 of an assistant input device 1061 and one or more cloud-based automation assistant components 119. A second automation assistant 120B, enclosed in dash-dash-dot lines, includes an automation assistant client 1181 of an assistant input device 1061 and one or more cloud-based automation assistant components 119. N Automation Assistant Client 118 Nand one or more cloud-based automated assistant components 119. Thus, it should be understood that each user engaging with an automation assistant client 118 running on one or more of the assistant input devices 106 may effectively engage with a user-specific logical instance of the automation assistant 120 (or a logical instance of the automation assistant 120 shared among a household or other group of users). For brevity and simplicity, the term "automation assistant" as used herein refers to a combination of an automation assistant client 118 running on each one of the assistant input devices 106 and one or more cloud-based automated assistant components 119 (which may be shared among the various automation assistant clients 118). Although only multiple assistant input devices 106 are shown in FIG. 1, it will be understood that the cloud-based automated assistant component 119 can additionally serve many additional groups of assistant input devices.

[0027] The assistant input devices 106 may include, for example, one or more of a desktop computing device, a laptop computing device, a tablet computing device, a mobile phone computing device, a computing device in the user's vehicle (e.g., an in-vehicle communication system, an in-vehicle entertainment system, an in-vehicle navigation system), an interactive standalone speaker (e.g., with or without a display), a smart appliance such as a smart television, a wearable device of the user that includes a computing device (e.g., a user's watch with a computing device, a user's glasses with a computing device, a virtual reality or augmented reality computing device), and / or any IoT device capable of receiving user input directed to the automation assistant 120. Additional and / or alternative assistant input devices may be provided. The assistant non-input devices 185 may include many of the same devices as the assistant input devices 106, but are not capable of receiving user input directed to the automation assistant 120 (e.g., do not include a user interface input component). Although the assistant non-input devices 185 do not receive user input directed to the automation assistant 120, the assistant non-input devices 185 may still be controlled by the automation assistant 120.

[0028] In some implementations, the multiple assistant input devices 106 and the assistant non-input devices 185 may be associated with each other in various ways to facilitate the execution of the techniques described herein. For example, in some implementations, the multiple assistant input devices 106 and the assistant non-input devices 185 may be associated with each other by being communicatively coupled via one or more networks (e.g., via the network 110 of FIG. 1). This may be the case, for example, when the multiple assistant input devices 106 and the assistant non-input devices 185 are deployed across a particular area or environment, such as a home, a building, etc. Additionally or alternatively, in some implementations, the multiple assistant input devices 106 and the assistant non-input devices 185 may be associated with each other by being members of a cooperative ecosystem that is at least selectively accessible by one or more users (e.g., an individual, a family member, an employee of an organization, other predefined groups, etc.). In some of those implementations, the ecosystem of the multiple assistant input devices 106 and the assistant non-input devices 185 may be manually and / or automatically associated with each other in the device topology representation of the ecosystem stored in the device topology database 193.

[0029] The assistant non-input system 180 may include one or more first party (1P) systems and / or one or more third party (3P) systems. A 1P system refers to a system controlled by a party that is the same as the party that controls the automated assistant 120 referred to herein. A 3P system, as used herein, refers to a system controlled by a party other than the party that controls the automated assistant 120 referred to herein.

[0030] The assistant non-input system 180 can receive data from the assistant non-input device 185 and / or one or more cloud-based automation assistant components 119 communicatively coupled thereto (e.g., via the network 110 of FIG. 1 ) and selectively transmit data (e.g., status, status changes, and / or other data) to the assistant non-input device 185 and / or one or more cloud-based automation assistant components 119. For example, assume that the assistant non-input device 1851 is a smart doorbell IoT device. In response to an individual pressing a button on the doorbell IoT device, the doorbell IoT device can transmit data corresponding to one of the assistant non-input systems 180 (e.g., one of the assistant non-input systems managed by the doorbell manufacturer, which may be a 1P system or a 3P system). One of the assistant non-input systems 180 can determine a change in the status of the doorbell IoT device based on such data. For example, one of the assistant non-input systems 180 can determine a change in a doorbell from an inactive state (e.g., no recent button press) to an active state (recent button press), and the change in doorbell state can be transmitted (e.g., via the network 110 of FIG. 1 ) to one or more cloud-based automation assistant components 119 and / or one or more of the assistant input devices 106. In particular, user input is received at the assistant non-input device 1851 (e.g., a button press on the doorbell), but the user input is not directed to the automation assistant 120 (hence the term "assistant non-input device"). As another example, assume that the assistant non-input device 1851 is a smart thermostat IoT device with a microphone, but the smart thermostat does not include the automation assistant client 118. An individual can manipulate the smart thermostat (e.g., using touch input or spoken input) to change the temperature, set a particular value as a set point for controlling an HVAC system via the smart thermostat, etc.However, unless the smart thermostat includes the automation assistant client 118, an individual cannot communicate directly with the automation assistant 120 via the smart thermostat.

[0031] In various implementations, the one or more cloud-based automation assistant components 119 may further include various engines. For example, as shown in FIG. 1, the one or more cloud-based automation assistant components 119 may further include an event detection engine 130, a device identification engine 140, an event processing engine 150, a semantic sign engine 160, and a query / command processing engine 170. It should be understood that these various engines are shown as the one or more cloud-based automation assistant components 119 in FIG. 1 for purposes of illustration and are not intended to be limiting. For example, the assistant input device 106 and / or the assistant non-input device 185 may include one or more of these various engines. As another example, these various engines may be distributed across the assistant input device 106, and the assistant non-input device 185 may include one or more of these various engines and / or one or more cloud-based automation assistant components 119.

[0032] In some implementations, the event detection engine 130 can detect various events occurring in the ecosystem. In some versions of those implementations, the event detection engine 130 can detect when a given one of the assistant input devices 106 and / or a given one of the assistant non-input devices 185 (e.g., a given one of the assistant devices) is newly added to the ecosystem or moves within the ecosystem. For example, the event detection engine 130 can determine when a given one of the assistant devices is newly added to the ecosystem based on one or more wireless signals detected via the network 110 and via the device identification engine 140. For example, when a given one of the assistant devices is newly connected to one or more of the networks 110, the given one of the assistant devices can broadcast a signal indicating that it is newly added to the network 110. As another example, the event detection engine 130 can determine when a given one of the assistant devices has moved within the ecosystem based on one or more wireless signals detected via the network 110. In these examples, the device identification engine 140 can process the signals to determine that a given one of the assistant devices is newly added to the network 110 and / or that a given one of the assistant devices has moved within the ecosystem. The one or more wireless signals detected by the device identification engine 140 can be, for example, network signals and / or acoustic signals that are imperceptible to humans and optionally include a unique identifier of the given one of the assistant devices and / or each of the other assistant devices that are positionally close to the given one of the assistant devices. For example, when a given one of the assistant devices moves within the ecosystem, the device identification engine 140 can detect one or more wireless signals being transmitted by other assistant devices that are positionally close to the given one of the assistant devices.These signals can be processed to determine that one or more other assistant devices that are positionally proximate to a given one of the assistant devices are different from one or more assistant devices that were previously positionally proximate to the given one of the assistant devices.

[0033] In some further versions of those implementations, the automation assistant 120 can cause a given one of the assistant devices that is newly added to the ecosystem or moved within the ecosystem to be assigned to a group of assistant devices (e.g., in the device topology representation of the ecosystem stored in the device topology database 193). For example, in an implementation in which a given one of the assistant devices is newly added to the ecosystem, the given one of the assistant devices may be added to an existing group of assistant devices, or a new group of assistant devices including the given one of the assistant devices may be created. For example, if the given one of the assistant devices is positionally close to multiple assistant devices that belong to a "kitchen" group (e.g., a smart oven, a smart coffee maker, a two-way standalone speaker associated with a unique identifier or indicator that indicates that it is located in the kitchen, and / or other assistant devices), the given one of the assistant devices may be added to the "kitchen" group, or a new group may be created. As another example, in an implementation in which a given one of the assistant devices moves within the ecosystem, the given one of the assistant devices may be added to an existing group of assistant devices, or a new group of assistant devices including the given one of the assistant devices may be created. For example, if a given one of the assistant devices was locationally close to multiple assistant devices belonging to the aforementioned “Kitchen” group, but is now locationally close to multiple assistant devices (e.g., smart garage doors, smart door locks, and / or other assistant devices) belonging to the “Garage” group, the given one of the assistant devices may be removed from the “Kitchen” group and added to the “Garage” group.

[0034] In some additional or alternative versions of these implementations, the event detection engine 130 can detect the occurrence of an acoustic event. The occurrence of the acoustic event can be detected based on audio data received at one or more of the assistant input devices 106 and / or one or more of the assistant non-input devices 185 (e.g., one or more of the assistant devices). The audio data received at one or more of the assistant devices can be processed by an event detection model stored in the ML model database 192. In these implementations, each of the one or more assistant devices that detects the occurrence of the acoustic event includes a respective microphone.

[0035] In some further versions of these implementations, the occurrence of the acoustic event may include ambient noise captured in the audio data at one or more of the assistant devices (and optionally may include only the occurrence of ambient noise detected when speech acceptance is active at one or more of the assistant devices). The ambient noise detected at each of the one or more assistant devices may be stored in the device activity database 191. In these implementations, the event processing engine 150 may process the ambient noise detected at one or more assistant devices using an ambient noise detection model (e.g., stored in the ML model database 192) that is trained to classify the ambient noise into one or more of a plurality of heterogeneous semantic categories based on a measure generated when processing the ambient noise using the ambient noise detection model. The plurality of heterogeneous categories may include, for example, a movie or television sound category, a cooking sound category, a music sound category, a garage or workshop sound category, a patio sound category, and / or other semantically meaningful heterogeneous sound categories. For example, if the event processing engine 150 determines that the ambient noise processed using the ambient noise detection model includes sounds corresponding to a microwave oven humming, food frying in a frying pan, a food processor processing food, etc., the event processing engine 150 can classify the ambient noise into a cooking sound category. As another example, if the event processing engine 150 determines that the ambient noise processed using the ambient noise detection model includes sounds corresponding to a circular saw operating, hammering, etc., the event processing engine 150 can classify the ambient noise into a garage or workshop category. The classification of the ambient noise detected at a particular device can also be used as a device-specific signal to be utilized in inferring a semantic indicator (e.g., as described with respect to the semantic indicator engine 160) for the assistant device.

[0036] In some additional or alternative versions of these further implementations, the occurrence of the acoustic event may include a hot word or a particular sound detected at one or more of the assistant devices. In these implementations, the event processing engine 150 may process the audio data detected at one or more assistant devices using a hot word detection model trained to determine whether the audio data includes a particular word or phrase that invokes the automated assistant 120 based on a measure generated when processing the audio data using the hot word detection model. For example, the event processing engine 150 may process the audio data to determine whether the audio data captures a user utterance that includes "Assistant," "Hey Assistant," "OK Assistant," and / or any other word or phrase that invokes an automated assistant. Additionally, the measure generated using the hot word detection model may include a respective confidence level or probability indicating whether the audio data includes a word or phrase that invokes the automated assistant 120. In some versions of these implementations, the event processing engine 150 may determine that the audio data captures a word or phrase if the measure meets a threshold value. For example, if the event processing engine 150 generates a measure of 0.70 associated with audio data that captures a word or phrase that invokes the automation assistant 120, and the threshold value is 0.65, the event processing engine 150 may determine that the audio data captures a word or phrase that invokes the automation assistant 120.

[0037] In these implementations, the event processing engine 150 can additionally or alternatively process audio data detected at one or more assistant devices using a sound detection model trained to determine whether the audio data includes a particular sound based on a measure generated when processing the audio data using the sound detection model. The particular sound may include, for example, glass breaking, a dog barking or a cat meowing, a doorbell ringing, a fire alarm going off, or a carbon monoxide detector going off. For example, the event processing engine 150 can process the audio data to determine whether the audio data captures any of these particular sounds. In this example, a single sound detection model may be trained to determine whether multiple particular sounds are captured in the audio data, or multiple sound detection models may be trained to determine whether a given particular sound is captured in the audio data. Furthermore, the measure generated using the sound detection model may include a respective confidence level or probability indicating whether the audio data includes a particular sound. In some versions of these implementations, the event processing engine 150 can determine that the audio data captures a particular sound if the measure meets a threshold value. For example, if the event processing engine 150 generates a measure of 0.70 associated with audio data capturing the sound of glass breaking, and the threshold value is 0.65, the event processing engine 150 may determine that the audio data captures the sound of glass breaking.

[0038] In various implementations, the occurrence of an acoustic event may be captured by various assistant devices in the ecosystem. For example, various assistant devices in an environment may capture time-corresponding audio data (e.g., time-corresponding in that the respective audio data is detected at the various assistant devices at the same time or within a threshold length of time). In these implementations, in response to a given assistant device detecting audio data in the ecosystem, the device identification engine 140 may identify one or more additional assistant devices that may have similarly detected time-corresponding audio data capturing the acoustic event. For example, the device identification engine 140 may identify one or more of the additional assistant devices that may have similarly detected time-corresponding audio data capturing the acoustic event based on the one or more additional assistant devices having previously detected time-corresponding audio data capturing the acoustic event. In other words, the device identification engine 140 may predict that the one or more additional assistant devices may similarly capture audio data including the acoustic event because the given assistant device and one or more additional assistant devices have previously captured time-corresponding audio data including the same sound.

[0039] In various implementations, one or more device-specific signals generated or detected by each assistant device may be stored in a device activity database 191. In some implementations, the device activity database 191 may correspond to a portion of memory dedicated to the device activity of that particular assistant device. In some additional or alternative implementations, the device activity database 191 may correspond to memory of a remote system in communication with the assistant device (e.g., via the network 110 of FIG. 1). This device activity may be utilized in generating candidate semantic indicators for a given one of the assistant devices (e.g., as described with respect to the semantic indicator engine 160). Device activity can include, for example, queries or requests received at each assistant device (and / or semantic categories associated with each of the multiple queries or requests), commands executed at each assistant device (and / or semantic categories associated with each of the multiple commands), ambient noise detected at each assistant device (and / or semantic categories associated with various instances of ambient noise), unique identifiers or indicators (e.g., identified via event detection engine 140) of any assistant devices positionally proximate to a given assistant device, user preferences of a user associated with the ecosystem determined based on user interactions with the multiple assistant devices in the ecosystem (e.g., browsing history, search history, purchase history, music history, movie or television history, and / or any other user interactions associated with the multiple assistant devices), and / or any other data received, generated, and / or executed by each assistant device.

[0040] In some implementations, the semantic label engine 160 can process one or more device-specific signals to generate a semantic label candidate for a given one of the assistant devices (e.g., a given one of the assistant input devices 106 and / or a given one of the assistant non-input devices 185) based on the one or more device-specific signals. In some versions of those implementations, in response to determining that a given assistant device is newly added to the ecosystem and / or moves within the ecosystem, a given assistant device for which a semantic label candidate is generated can be identified. In some additional or alternative versions of those implementations, a given assistant device for which a semantic label candidate is generated can be identified periodically (e.g., once a month, once every six months, once a year, etc.). In some additional or alternative versions of these implementations, a given assistant device for which a candidate semantic indicator is generated may be identified in response to determining that a portion of the ecosystem in which the given assistant device is located has been repurposed (e.g., a room in the main residence of the ecosystem has been repurposed from a study to a bedroom). In these implementations, the given assistant device may be identified utilizing event detection engine 130. Identifying a given assistant device in these and other manners is described with respect to FIGS. 2A and 2B.

[0041] In some implementations, the semantic indicator engine 160 can select a given semantic indicator for a given assistant device from among the candidate semantic indicators based on one or more of the device-specific signals. Generating semantic indicator candidates for a given assistant device and selecting a given semantic indicator from among the candidate semantic indicators based on one or more device-specific indicators is described below (e.g., in conjunction with FIGs. 2A and 2B).

[0042] In an implementation in which semantic indicator candidates for a given assistant device are generated based on queries, requests, and / or commands stored in the device activity database 191 (or text corresponding thereto), the queries, requests, and / or commands may be processed using a semantic classifier (e.g., stored in the ML model database 192) to index the device activity for the given assistant device into one or more different semantic categories corresponding to the disparate queries, requests, and / or commands. The semantic indicator candidates may be generated based on the semantic categories into which the queries, commands, and / or requests are classified, and a given semantic indicator selected for a given assistant device may be selected based on the amount of queries, requests, and / or commands classified in a given semantic category. For example, assume that a given assistant device has previously received nine queries related to obtaining cooking recipes and two commands related to controlling smart lights in an ecosystem. In this example, the candidate semantic indicators may include, for example, a first semantic indicator of “kitchen device” and a second semantic indicator of “smart lighting control device.” Additionally, the semantic indicator engine 160 may select the first semantic indicator of “kitchen device” as the given semantic indicator for the given assistant device because past usage of the given assistant device indicates that the given assistant device is primarily used for cooking-related activities.

[0043] In some implementations, the semantic classifier stored in the ML model database 192 may be a natural language understanding engine (e.g., implemented by the NLP module 122 described below). The intent determined based on processing a query, command, and / or request previously received at the assistant device may be mapped to one or more of the semantic categories. In particular, the heterogeneous semantic categories described herein may be defined at various levels of granularity. For example, a semantic category may be associated with a class category of smart device commands, such as a category of smart lighting commands, a category of smart thermostat commands, and / or a category of smart camera commands, and / or a type category of the class category. In other words, each category may have a unique set of intents associated with it as determined by the semantic classifier, although some intents of a category may also be associated with additional categories. In some additional or alternative implementations, semantic classifiers stored in the ML model database 192 may be utilized to generate text embeddings (e.g., low-dimensional representations, such as word2vec representations) corresponding to the query, command, and / or request text. These embeddings may be points in an embedding space such that semantically similar words or phrases are associated with the same or similar portions of the embedding space. Furthermore, these portions of the embedding space may be associated with one or more of a plurality of heterogeneous semantic categories, and a given one of the embeddings may be classified into a given one of the semantic categories if a distance measure between the given one of the embeddings and one or more of the portions of the embedding space meets a distance threshold. For example, a word or phrase related to cooking may be associated with a first portion of the embedding space associated with the semantic label "cooking," a word or phrase related to weather may be associated with a second portion of the embedding space associated with the semantic label "weather," and so on.

[0044] In implementations in which one or more device-specific signals additionally or alternatively include ambient noise activity, the instances of ambient noise may be processed using an ambient noise detection model (e.g., stored in the ML model database 192) to index device activity for a given assistant device into one or more different semantic categories corresponding to disparate ambient noise. Semantic label candidates may be generated based on the semantic categories into which the instances of ambient noise are classified, and a given semantic label selected for a given assistant device may be selected based on the amount of instances of ambient noise classified in a given semantic category. For example, assume that the ambient noise detected at a given assistant device (and optionally only when speech recognition is active) mainly includes ambient noise classified as cooking sounds. In this example, the semantic label engine 160 may select the semantic label "Kitchen device" as the given semantic label for the given assistant device because the ambient noise captured in the audio data indicates that the device is located near cooking-related activities.

[0045] In some implementations, an ambient noise detection model stored in the ML model database 192 can be trained to detect a particular sound, and whether an instance of ambient noise includes a particular sound can be determined based on an output generated across the ambient noise detection model. The ambient noise detection model can be trained, for example, using supervised learning techniques. For example, multiple training instances can be obtained. Each of the training instances can include a training instance input that includes ambient noise and a corresponding training instance output that includes an indication of whether the training instance input includes the particular sound that the ambient noise detection model is trained to detect. For example, if the ambient noise detection model is trained to detect the sound of glass breaking, a training instance that includes the sound of glass breaking can be assigned a sign (e.g., "yes") or value (e.g., "1"), and a training instance that does not include the sound of glass breaking can be assigned a different sign (e.g., "no") or value (e.g., "0"). In some additional or alternative implementations, the ambient noise detection models stored in the ML model database 192 may be utilized to generate audio embeddings (e.g., low-dimensional representations of the ambient noise instances) based on the ambient noise instances (or their acoustic features, such as Mel-frequency cepstral coefficients, raw audio waveforms, and / or other acoustic features). These embeddings may be points in the embedding space such that similar sounds (or acoustic features that capture the sounds) are associated with the same or similar portions of the embedding space. Furthermore, these portions of the embedding space may be associated with one or more of a plurality of heterogeneous semantic categories, and a given one of the embeddings may be classified into a given one of the semantic categories if a distance measure between the given one of the embeddings and one or more of the portions of the embedding space meets a distance threshold. For example, an instance of glass breaking may be associated with a first portion of the embedding space related to the sound of "glass breaking," an instance of doorbell ringing may be associated with a second portion of the embedding space related to the sound of "doorbell," and so on.

[0046] In an implementation in which the one or more device-specific signals additionally or alternatively include unique identifiers or indicators of additional assistant devices that are positionally close to the given assistant device, the semantic indicator candidates may be generated based on those unique identifiers or indicators, and the given semantic indicator selected for the given assistant device may be selected based on one or more of the unique identifiers or indicators of the additional assistant devices. For example, assume that a first indicator "smart oven" is associated with a first assistant device that is positionally close to the given assistant device, and a second indicator "smart coffee maker" is associated with a second assistant device that is positionally close to the given assistant device. In this example, the semantic indicator engine 160 may select the semantic indicator "kitchen device" as the given semantic indicator for the given assistant device, since the indicators associated with the additional assistant devices that are positionally close to the given assistant device are related to cooking. The unique identifier or indicator may be processed using a semantic classifier stored in the ML model database 192 in the same or similar manner as described above in connection with processing queries, commands, and / or requests.

[0047] In an implementation in which semantic label candidates for a given assistant device are generated based on user preferences, the user preferences can be processed using a semantic classifier (e.g., stored in the ML model database 192) to index the user preferences into one or more different semantic categories corresponding to the disparate user preferences. Semantic label candidates can be generated based on the semantic categories into which the user preferences are classified, and a given semantic label selected for a given assistant device can be selected based on the given semantic category into which the user preferences associated with the given assistant device are classified. For example, assume that the user preferences indicate that a user associated with the ecosystem likes cooking and likes a fictional chef named Johnny Flay. In this example, the semantic label candidates can include, for example, a first semantic label called "cooking device" and a second semantic label candidate called "Johnny Flay device". In some versions of those implementations, utilizing the user preferences as a device-specific signal to generate one or more candidate semantic indicators may be in response to receiving user input to assign semantic indicators to the assistant device based on the user preferences.

[0048] In some implementations, the semantic sign engine 160 can automatically assign a given semantic sign to a given assistant device in a device topology representation of the ecosystem (e.g., stored in the device topology database 193). In some additional or alternative implementations, the semantic sign engine 160 can cause the automated assistant 120 to generate a prompt that includes the semantic sign candidates. The prompt can ask a user associated with the ecosystem to select one of the sign candidates as the given semantic sign. Furthermore, the prompt can be visually and / or audibly rendered on a given one of the assistant devices (which may or may not be the given assistant device to which the given semantic sign is assigned) and / or on the user's client device (e.g., a mobile device). In response to receiving a selection of one of the sign candidates as the given semantic sign, the selected given semantic sign can be assigned to the given assistant device in the device topology representation of the ecosystem (e.g., stored in the device topology database 193). In some versions of these implementations, a given semantic indicator assigned to a given assistant device may be added to a list of semantic indicators for the given assistant device. In other words, multiple semantic indicators may be associated with a given assistant device. In other versions of these implementations, a given semantic indicator assigned to a given assistant device may replace any other semantic indicators for the given assistant device. In other words, only a single semantic indicator may be associated with a given assistant device.

[0049] In some implementations, the query / command processing engine 170 can process queries, requests, or commands directed to the automated assistant 120 and received via one or more of the assistant input devices 106. The query / command processing engine 170 can process the queries, requests, or commands to select one or more of the assistant devices to satisfy the query or command. In particular, the one or more of the assistant devices selected to satisfy the query or command can be different from the one or more of the assistant input devices 106 that received the query or command. The query / command processing engine 170 can select one or more assistant devices to satisfy the utterance based on one or more criteria. The one or more criteria can include, for example, one or more proximity of the device to the user who provided the utterance (e.g., determined using the presence sensor 105 described below), one or more device capabilities of the devices in the ecosystem, semantic indicators assigned to one or more assistant devices, and / or other criteria for selecting an assistant device to satisfy the utterance.

[0050] For example, assume that a display device is required to satisfy the utterance. In this example, the assistant device candidates considered when selecting a given assistant device to satisfy the utterance may be limited to those that include a display device. If multiple assistant devices in the ecosystem include a display device, the given assistant device that includes a display device and is closest to the user may be selected to satisfy the utterance. In contrast, in an implementation in which only a speaker is required to satisfy the utterance (e.g., a display device is not required to satisfy the utterance), the assistant device candidates considered when selecting a given assistant device to satisfy the utterance may include those that have a speaker, regardless of whether they include a display device.

[0051] As another example, assume that an utterance includes a semantic characteristic that matches a semantic indicator assigned to a given assistant device. The query / command processing engine 170 can determine that the semantic characteristic of the utterance matches the semantic indicator assigned to the given assistant device by generating a first embedding corresponding to one or more words of the utterance (or text corresponding thereto) and a second embedding corresponding to one or more words of the semantic indicator assigned to the given assistant device, comparing the embeddings, and determining whether a distance measure between the embeddings meets a distance threshold that indicates that the embeddings match (e.g., whether it is an exact match or a rough match). In this example, the query / command processing engine 170 can select a given assistant device to satisfy the utterance based on the utterance's match with the semantic indicator (optionally in addition to or instead of the proximity of the user who provided the utterance to the given assistant device). In this way, the selection of an assistant device to satisfy an utterance can be biased toward semantic indicators assigned to the assistant device as described herein.

[0052] In various implementations, one or more of the assistant input devices 106 include one or more respective presence sensors 105 configured to provide a signal indicative of a detected presence, in particular the presence of a person, upon acknowledgement from a corresponding user. 1-N(also referred to herein simply as "presence sensors 105"). In some of those implementations, the automated assistant 120 can identify one or more of the assistant input devices 106 based at least in part on the presence of the user at one or more of the assistant input devices 106 to satisfy an utterance from a user associated with the ecosystem. The utterance can be satisfied by rendering responsive content (e.g., audibly and / or visually) at one or more of the assistant input devices 106, by causing one or more of the assistant input devices 106 to be controlled based on the utterance, and / or by causing one or more of the assistant input devices 106 to perform any other action to satisfy the utterance. As described herein, the automated assistant 120 can leverage data determined based on the respective presence sensors 105 in determining those assistant input devices 106 based on where the user is or has recently been, and provide corresponding commands to only those assistant input devices 106. In some additional or alternative implementations, the automated assistant 120 can utilize data determined based on the respective presence sensors 105 in determining whether a user (any user or a specific user) is currently near any of the assistant input devices 106, and can optionally refrain from providing commands based on determining that a user (any user or a specific user) is not near any of the assistant input devices 106.

[0053] Each presence sensor 105 may take a variety of forms. Some assistant input devices 106 may be equipped with one or more digital cameras configured to capture and provide signals indicative of detected motion in their field of view. Additionally or alternatively, some assistant input devices 106 may be equipped with other types of light-based presence sensors 105, such as passive infrared ("PIR") sensors that measure infrared ("IR") light emanating from objects in their field of view. Additionally or alternatively, some assistant input devices 106 may be equipped with presence sensors 105 that detect sound waves (or pressure waves), such as one or more microphones. Moreover, in addition to the assistant input devices 106, one or more of the assistant non-input devices 185 may additionally or alternatively include a respective presence sensor 105 as described herein, and signals from such sensors may be additionally utilized by the automated assistant 120 when determining whether and / or how to satisfy an utterance according to the implementations described herein.

[0054] Additionally or alternatively, in some implementations, the presence sensor 105 may be configured to detect other phenomena related to the presence of a person or a device in the ecosystem. For example, in some embodiments, a given one of the assistant devices may be equipped with a presence sensor 105 that detects various types of wireless signals (e.g., radio waves, ultrasound, electromagnetic waves, etc. waves) emitted, for example, by other assistant devices (e.g., mobile devices, wearable computing devices, etc.) carried / operated by a particular user and / or other assistant devices in the ecosystem (e.g., as described with respect to the event detection engine 130). For example, some of the assistant devices may be configured to emit waves that are imperceptible to humans, such as ultrasound or infrared waves, that can be detected by one or more of the assistant input devices 106 (e.g., via an ultrasound / infrared receiver, such as an ultrasound-enabled microphone).

[0055] Additionally or alternatively, the various assistant devices may emit other types of waves imperceptible to humans, such as radio waves (e.g., Wi-Fi, Bluetooth, cellular, etc.), that may be detected by other assistant devices (e.g., mobile devices, wearable computing devices, etc.) carried / operated by a particular user and used to determine the specific location of the operating user. In some implementations, Wi-Fi triangulation may be used to detect the location of a person, for example, based on Wi-Fi signals to / from the assistant device. In other implementations, other wireless signal characteristics, such as time-of-flight, signal strength, etc., may be used by the various assistant devices, either alone or collectively, to determine the location of a particular person based on signals emitted by other assistant devices carried / operated by a particular user.

[0056] Additionally or alternatively, in some implementations, one or more of the assistant input devices 106 may perform voice recognition to recognize the user from the user's voice. For example, some instances of the automated assistant 120 may be configured to match the voice with a profile of the user, for example, for the purpose of providing / restricting access to various resources. In some implementations, the speaker's movements may then be determined, for example, by the presence sensor 105 of the assistant device. In some implementations, based on such detected movements, the user's location may be predicted, and this location may be considered to be the user's location when any content is rendered at the assistant devices based at least in part on the proximity of the assistant devices to the user's location. In some implementations, the user may simply be considered to be at the location where the user last engaged with the automated assistant 120, especially if not much time has passed since that last engagement.

[0057] Each of the assistant input devices 106 further includes a respective user interface component 107. 1-N (also referred to herein simply as "user interface component 107"), each of which may include one or more user interface input devices (e.g., microphone, touch screen, keyboard) and / or one or more user interface output devices (e.g., display, speaker, projector). As an example, user interface component 1071 of assistant input device 1061 may include only a speaker and microphone, while assistant input device 106 N User Interface Components 107 N may include a speaker, a touch screen, and a microphone. Additionally or alternatively, in some implementations, assistant non-input devices 185 may include one or more user interface input devices and / or one or more user interface output devices of user interface component 107, but the user input devices (if any) for assistant non-input devices 185 may not allow a user to directly interact with automation assistant 120.

[0058] Each of the assistant input devices 106 and / or any other computing devices operating one or more of the cloud-based automation assistant components 119 may include one or more memories for storage of data and software applications, one or more processors for accessing data and executing applications, and other components that facilitate communication over a network. Operations performed by one or more of the assistant input devices 106 and / or by the automation assistant 120 may be distributed among multiple computer systems. The automation assistant 120 may be implemented, for example, as a computer program running on one or more computers at one or more locations coupled to each other through a network (e.g., any of the networks 110 of FIG. 1).

[0059] As discussed above, in various implementations, each of the assistant input devices 106 can operate a respective automated assistant client 118. In various implementations, each automated assistant client 118 can operate a respective speech capture / text-to-speech (TTS) / speech-to-text (STT) module 114. 1-N (also referred to herein simply as “speech capture / TTS / STT module 114”). In other implementations, one or more aspects of each speech capture / TTS / STT module 114 may be implemented separately from each automated assistant client 118.

[0060] Each respective speech capture / TTS / STT module 114 may be configured to perform one or more functions, including, for example, capturing the user's speech (speech capture, e.g., via a respective microphone (which may in some cases comprise a presence sensor 105)), converting the captured audio to text and / or other representations or embeddings (STT) using speech recognition models stored in ML model database 192, and / or converting text to speech (TTS) using speech synthesis models stored in ML model database 192. Instances of these models may be stored locally on each respective assistant input device 106 and / or accessible by the assistant input device (e.g., via network 110 of FIG. 1). In some implementations, one or more of the assistant input devices 106 may have relatively limited computational resources (e.g., processor cycles, memory, batteries, etc.), so a respective speech capture / TTS / STT module 114 local to each of the assistant input devices 106 may be configured to use a speech recognition model to convert a finite number of different spoken phrases into text (or other forms, such as low-dimensional embeddings). Other speech inputs may be sent to one or more of the cloud-based automated assistant components 119, which may include a cloud-based TTS module 116 and / or a cloud-based STT module 117.

[0061] The cloud-based STT module 117 can be configured to leverage the virtually infinite resources of the cloud to convert audio data captured by the speech capture / TTS / STT module 114 into text (which can then be provided to the natural language processor module 122) using speech recognition models stored in the ML model database 192. The cloud-based TTS module 116 can be configured to leverage the virtually infinite resources of the cloud to convert text data (e.g., text compiled by the automated assistant 120) into computer-generated speech output using speech synthesis models stored in the ML model database 192. In some implementations, the cloud-based TTS module 116 can provide computer-generated speech output to one or more of the assistant devices, for example to be directly output using the respective speakers of the respective assistant devices. In other implementations, text data generated by the automated assistant 120 using the cloud-based TTS module 116 (e.g., a client device notification included in a command) may be provided to the speech capture / TTS / STT module 114 of the respective assistant device, which may then use a speech synthesis model to locally convert the text data into computer-generated speech that is rendered through a local speaker of the respective assistant device.

[0062] The automation assistant 120 (and in particular the one or more cloud-based automation assistant components 119) may include a natural language processing (NLP) module 122, the aforementioned cloud-based TTS module 116, the aforementioned cloud-based STT module 117, and other components, some of which are described in more detail below. In some implementations, one or more of the engines and / or modules of the automation assistant 120 may be omitted, combined, and / or implemented in a component separate from the automation assistant 120. An instance of the NLP module 122 may additionally or alternatively be implemented locally in the assistant input device 106.

[0063] In some implementations, the automation assistant 120 generates response content in response to various inputs generated by a user of one of the assistant input devices 106 during a human-computer interaction session with the automation assistant 120. The automation assistant 120 may provide the response content (e.g., via one or more of the networks 110 of FIG. 1 when separate from the assistant device) for presentation to a user as part of an interaction session via the assistant input device 106 and / or the assistant non-input device 185. For example, the automation assistant 120 may generate response content in response to free-form natural language input provided via one of the assistant input devices 106. As used herein, free-form input is input organized by a user that is not constrained to a group of options presented for selection by the user.

[0064] The NLP module 122 of the automation assistant 120 may process natural language input generated by a user via the assistant input devices 106 and generate annotated output for use by one or more other components of the automation assistant 120, the assistant input devices 106, and / or the assistant non-input devices 185. For example, the NLP module 122 may process free-form natural language input generated by a user via one or more respective user interface input devices of the assistant input devices 106. The annotated output generated based on processing the free-form natural language input may include one or more annotations of the natural language input and, optionally, one or more (e.g., all) of the terms of the natural language input.

[0065] In some implementations, the NLP module 122 is configured to identify and annotate various types of grammatical information in the natural language input. For example, the NLP module 122 may include a portion of an utterance tagger configured to annotate words with their grammatical roles. In some implementations, the NLP module 122 may additionally and / or alternatively include an entity tagger (not shown) configured to annotate references to entities in one or more segments, such as references to people (e.g., including literary characters, celebrities, public figures, etc.), organizations, locations (real and imaginary), etc. In some implementations, data about the entities may be stored in one or more databases, such as a knowledge graph (not shown). In some implementations, the knowledge graph may include nodes that represent known entities (and possibly entity attributes), as well as edges that connect the nodes and represent relationships between the entities.

[0066] The entity tagger of NLP module 122 may annotate references to an entity at a high level of granularity (e.g., to enable identification of all references to an entity class, such as people) and / or at a low level of granularity (e.g., to enable identification of all references to a particular entity, such as a particular person). The entity tagger may rely on the content of the natural language input to resolve for a particular entity, and / or may optionally communicate with a knowledge graph or other entity database to resolve for a particular entity.

[0067] In some implementations, the NLP module 122 may additionally and / or alternatively include a coreference resolver (not shown) configured to group or "cluster" mentions into the same entity based on one or more contextual cues. For example, the coreference resolver may be utilized to resolve the word "it" in the natural language input "lock it" to "front door lock" based on "front door lock" being mentioned in a client device notification rendered immediately prior to receiving the natural language input "lock it."

[0068] In some implementations, one or more components of the NLP module 122 may rely on annotations from one or more other components of the NLP module 122. For example, in some implementations, a designated entity tagger may rely on annotations from a coreference resolver and / or a dependency analyzer when annotating all references to a particular entity. Also, for example, in some implementations, a coreference resolver may rely on annotations from a dependency analyzer when clustering references to the same entity. In some implementations, when processing a particular natural language input, one or more components of the NLP module 122 may determine one or more annotations using relevant data outside of the particular natural language input, such as an assistant input device notification rendered immediately before receiving the natural language input on which the assistant input device notification is based.

[0069] It should be understood that while FIG. 1 is illustrated as having a particular configuration of components implemented by the assistant device and / or server and as having the assistant device and / or server communicate over a particular network, this is for illustrative purposes and is not intended to be limiting. For example, the assistant input device 106 and the assistant non-input device may be directly communicatively coupled to each other over one or more networks (not shown). As another example, the operation of one or more cloud-based automation assistant components 119 may be implemented locally in one or more of the assistant input devices 106 and / or one or more of the assistant non-input devices. As yet another example, instances of the various ML models stored in the ML model database 192 may be stored locally at the assistant device, and / or instances of the device topology representation of the ecosystem stored in the device topology database 193 may be stored locally at the assistant input device. Additionally, in implementations in which data (e.g., device activity, corresponding audio data or recognized text, device topology representations, and / or any other data described herein) is transmitted over any of the one or more networks 110 of FIG. 1, the data may be encrypted, filtered, or otherwise protected in any manner to ensure user privacy.

[0070] Using the techniques described herein, the device topology representation of the ecosystem can be kept up to date by inferring and assigning semantic labels to assistant devices in the ecosystem. Furthermore, the semantic labels assigned to the assistant devices are semantically meaningful to the user in that the semantic labels assigned to each assistant device are selected based on the use of the respective assistant device and / or the respective part of the ecosystem in which the respective assistant device is located. Thus, when an utterance is received by one or more of the assistant devices in the ecosystem, the automated assistant can more accurately select one or more of the assistant devices that are best suited to satisfy the utterance. As a result, the amount of user input received by one or more of the assistant devices in the ecosystem can be reduced because a user associated with the ecosystem does not need to specify a particular device to satisfy the utterance or to repeat the utterance when the wrong device is selected to satisfy the utterance, thereby saving computational and / or network resources at the assistant device by reducing network traffic. Furthermore, when an assistant device is newly added to the ecosystem or moved within the ecosystem, the amount of user input received by one or more of the assistant devices in the ecosystem can be reduced because a user does not need to manually update the device topology representation via a software application associated with the ecosystem.

[0071] Further description of the various components of FIG. 1 will now be provided with reference to FIGS. 2A and 2B. A floor plan of a house is illustrated in FIGS. 2A and 2B. The illustrated floor plan includes a number of rooms 250-262. A number of assistant input devices 106 1-5 are deployed throughout at least some of the rooms. 1-5Each of the may implement an instance of the automated assistant client 118 configured with selected aspects of the present disclosure and may include one or more input devices, such as a microphone, capable of capturing speech spoken by nearby people. For example, a first assistant input device 1061 in the form of an interactive standalone speaker and a display device (e.g., a display screen, a projector, etc.) is deployed in room 250 of FIG. 2A, which is a kitchen in this example, and in room 256 of FIG. 2B, which is a living room in this example. A second assistant input device 1062 in the form of a so-called "smart" television (e.g., a network-connected television with one or more processors implementing a respective instance of the automated assistant client 118) is deployed in room 252, which is a study in this example. A third assistant input device 1063 in the form of an interactive standalone speaker without a display is deployed in room 254, which is a bedroom in this example. A fourth assistant input device 1064 in the form of another interactive standalone speaker is deployed in room 256, which is a living room in this example. A fifth assistant input device 1065, also in the form of a smart television, is also located in room 250, in this example the kitchen.

[0072] Although not shown in FIGS. 2A and 2B, multiple assistant input devices 106 1-4may be communicatively coupled to each other and / or other resources (e.g., the Internet) via one or more wired or wireless WANs and / or LANs (e.g., via network 110 of FIG. 1). In addition, other assistant input devices may also be present, particularly certain mobile devices such as smartphones, tablets, laptops, wearable devices, etc., which may be carried by one or more people in the home, and which may or may not be connected to the same WANs and / or LANs. The configuration of assistant input devices illustrated in FIGS. 2A and 2B is only one example, and more or fewer and / or different assistant input devices 106 may be deployed throughout any number of other rooms and / or areas of the home and / or in locations other than the home (e.g., businesses, hotels, public places, airports, vehicles, and / or other locations or spaces).

[0073] Multiple Assistant Non-Input Devices 185 1-5 are further illustrated in FIG. 2A and FIG. 2B. For example, a first assistant non-input device 1851 in the form of a smart doorbell is deployed outside the house near the front door of the house. A second assistant non-input device 1852 in the form of a smart lock is deployed outside the house adjacent to the front door of the house. A third assistant non-input device 1853 in the form of a smart washing machine is deployed in the room 262, which is a laundry room in this example. A fourth assistant non-input device 1854 in the form of a door open / close sensor is deployed near the back door of the room 262 to detect whether the back door is open or closed. A fifth assistant non-input device 1855 in the form of a smart thermostat is deployed in the room 252, which is a study in this example.

[0074] Each of the assistant non-input devices 185 can communicate (e.g., via network 110 of FIG. 1) with a respective assistant non-input system 180 (shown in FIG. 1) to provide data to the respective assistant non-input system 180, and optionally have the data controlled based on commands provided by the respective assistant non-input system 180. One or more of the assistant non-input devices 185 can additionally or alternatively communicate directly (e.g., via network 110 of FIG. 1) with one or more of the assistant input devices 106 to provide data to one or more of the assistant input devices 106, and optionally have the data controlled based on commands provided by one or more of the assistant input devices 106. The configuration of assistant non-input devices 185 shown in FIGS. 2A and 2B is merely an example, and more or fewer and / or different assistant non-input devices 185 may be deployed throughout any number of other rooms and / or areas of the home and / or in locations other than the home (e.g., businesses, hotels, public places, airports, vehicles, and / or other locations or spaces).

[0075] In various implementations, a semantic indicator may be assigned to a given assistant device (e.g., a given one of the assistant input devices 106 or the assistant non-input devices 185) based on processing of one or more device-specific signals associated with the respective assistant device. The one or more device-specific signals may be detected by and / or generated by the given assistant device. The one or more device-specific signals may include, for example, one or more queries (if any) previously received at the given assistant device, one or more commands (if any) previously executed at the given assistant device, instances of ambient noise previously detected at the given assistant device (and optionally only when speech reception was active at the given assistant device), unique identifiers (or indicators) of respective assistant devices positionally proximate to the given assistant device, and / or user preferences of users associated with the ecosystem. Each of the one or more device-specific signals associated with the given assistant device may be processed to classify each of them into one or more semantic categories of a plurality of heterogeneous semantic categories.

[0076] Further, one or more candidate semantic labels may be generated based on one or more device-specific signals. The candidate semantic labels may be generated using one or more rules (optionally heuristically defined) or machine learning models (e.g., stored in the ML model database 192). For example, one or more heuristically defined rules may indicate that a candidate semantic label associated with each of the semantic categories into which one or more device-specific signals are classified should be generated. For example, assume that the device-specific signals are classified into a "kitchen" category, a "cooking" category, a "bedroom" category, and a "living room" category. In this example, the candidate semantic labels may include a first candidate semantic label "kitchen assistant device", a second candidate semantic label "cooking assistant device", a third candidate semantic label "bedroom assistant device", and a fourth candidate semantic label "living room assistant device". As another example, one or more device-specific signals (or one or more semantic categories corresponding thereto) may be processed using a machine learning model that is trained to generate candidate semantic labels. For example, a machine learning model may be trained based on multiple training instances. Each of the training instances may include a training instance input and a corresponding training instance output. The training instance input may include, for example, one or more device-specific signals and / or one or more semantic categories, and the corresponding training instance output may include, for example, a ground truth output corresponding to a semantic label to be assigned based on the training instance input.

[0077] Moreover, the semantic label to be assigned to a given assistant device may be selected from among one or more semantic label candidates. The semantic label to be assigned to a given assistant device may be selected from among one or more semantic label candidates based on a reliability level associated with each of the one or more semantic label candidates. In some implementations, the semantic label to be assigned to a given assistant device may be automatically assigned to the given assistant device, while in additional or alternative implementations, a user associated with the ecosystem of FIG. 2A and FIG. 2B may be prompted to select a semantic label to be assigned to the given assistant device from a list of one or more semantic label candidates (e.g., as described with respect to FIG. 3). In some additional or alternative implementations, a semantic label may be automatically assigned to a given assistant device if the given semantic label is unique (with respect to other assistant devices positionally close to the given assistant device in the ecosystem).

[0078] In some versions of those implementations, a given assistant device to which a semantic label should be assigned may be identified in response to determining (e.g., via event detection engine 130 and / or device identification engine 140 of FIG. 1) that the given assistant device is newly added to the ecosystem. For example, with particular reference to FIG. 2A , assume that a first assistant input device 1061 and device, taking the form of a two-way standalone speaker, are newly deployed in room 250, in this example a kitchen. Assume that no previous queries or commands have been received or performed by first assistant input device 1061 (other than configuring first assistant input device 1061) since the first assistant input device 1061 was newly added to the ecosystem in association with one or more device-specific signals associated with first assistant input device 1061 of FIG. 2A , and that no previous queries or commands have been received or performed by first assistant input device 1061 since the first assistant input device 1061 was newly added to the ecosystem while the first assistant input device 1061 was configured by a user of the ecosystem (including, for example, a name, a test utterance to establish a speech embedding for the user, etc., Assume that some instances of ambient noise have been captured (when speech reception is active as speech is provided by the user), a unique identifier (or indicator) associated with a fifth assistant input device 1065 in the form of a smart television in room 250 (e.g., "Kitchen TV") and a unique identifier (or indicator) associated with a fifth assistant non-input device 1855 in the form of a smart thermostat in room 252 (e.g., "Thermostat") are detected at the first assistant input device 1061, and assume that the user's user preferences associated with the ecosystem are known.

[0079] In this example, the instances of ambient noise (if any) may be processed using an ambient noise detection model to classify the ambient noise into one or more semantic categories. For example, assume that the instances of ambient noise capture water dripping in a sink in room 250, the rumble of a microwave or oven in room 250, food frying in a frying pan on a stove in room 250, etc. These instances of ambient noise may be classified into a "kitchen" semantic category, a "cooking" semantic category, and / or other semantic category associated with noises typically encountered in a kitchen. Additionally or alternatively, the ambient noise may capture a movie, television program, or advertisement being visually and / or audibly rendered via a fifth assistant input device 1065 in the form of a smart television in room 250. These instances of ambient noise may be classified into a "television" semantic category, a "movie" semantic category, and / or other semantic category associated with noises typically encountered from a smart television. Additionally or alternatively, the unique identifier (or indicator) of the fifth assistant input device 1065 (e.g., “kitchen TV”) and the unique identifier (or indicator) of the fifth assistant non-input device 1855 (e.g., thermostat) may be processed to generate one or more of the semantic indicators. These unique identifiers (or indicators) may be classified into a “kitchen” semantic category, a “smart device” semantic category, and / or other semantic categories associated with assistant devices that are positionally close to the first assistant input device 1061 in the ecosystem of FIG. 2A. Additionally or alternatively, assume that the user preferences indicate that the user is interested in cooking and a fictional chef named Johnny Flay. These user preferences may be identified as associated with the first assistant input device 1061 based on one or more of the other device-specific signals being classified into a “cooking” or “kitchen” category, and based on the user preferences of cooking and Johnny Flay also being classified into a “cooking” or “kitchen” category.As a result, the semantic label candidates in this example may include "kitchen speaker device", "cooking speaker device", "television speaker device", "movie speaker device", "Johnny Flay device", and / or other semantic label candidates based on one or more device-specific labels associated with the first assistant input device 1061. Further in this example, a given one of the semantic labels may be automatically assigned to the first assistant input device 1061, or a user associated with the ecosystem may be prompted to select one or more of the semantic label candidates to assign the given semantic label to the first assistant input device 1061 (e.g., while the first assistant input device 1061 is being configured).

[0080] In some additional or alternative implementations, a given assistant device to which a semantic label is to be assigned may be identified periodically (e.g., weekly, monthly, six months, and / or any other period via event detection engine 130 and / or device identification engine 140 of FIG. 1). For example, with particular reference to FIG. 2A, assume that a first assistant input device 1061 in the form of a two-way standalone speaker and a display device has been deployed in room 250, in this example the kitchen, for six months. With respect to one or more device-specific signals associated with the first assistant input device 1061 of FIG. 2A , assume that a previous query or command has been received at or executed by the first assistant input device 1061, assume that an instance of ambient noise has been captured during the first assistant input device 1061, and assume that a unique identifier (or indicator) associated with a fifth assistant input device 1065 in the form of a smart television in room 250 (e.g., “Kitchen TV”) and a unique identifier (or indicator) associated with a fifth assistant non-input device 1855 in the form of a smart thermostat in room 252 (e.g., “Thermostat”) are still detected at the first assistant input device 1061.

[0081] In this example, the queries and commands (or the text corresponding thereto) may be processed using a semantic classifier to classify the queries and commands into one or more semantic categories. For example, the queries and commands previously received at the first assistant input device 1061 may include a query for requesting a cooking recipe, a command for setting a timer, a command for controlling any smart devices in the kitchen, and / or other queries or commands. These instances of queries and commands may be classified into a "cooking" semantic category, a "smart device control" category, and / or other semantic categories based on the queries and commands received at the first assistant input device 1061. Semantic indicator candidates may additionally or alternatively be determined based on instances of ambient noise and / or unique identifiers (or indicators) positionally proximate to the first assistant input device 1061 as described above. As a result, the semantic label candidates in this example may include "kitchen device," "timer device," "thermostat display device," "cooking display device," "television device," "movie device," "Johnny Flay recipe device," and / or other semantic label candidates based on one or more device-specific labels associated with the first assistant input device 1061. Further, in this example, a given one of the semantic labels may be automatically assigned to the first assistant input device 1061, or a user associated with the ecosystem may be prompted to select one or more of the semantic label candidates to assign the given semantic label to the first assistant input device 1061 (e.g., while the first assistant input device 1061 is configured).

[0082] In some additional or alternative implementations, a given assistant device to which a semantic label is to be assigned may be identified in response to determining (e.g., via event detection engine 130 and / or device identification engine 140 of FIG. 1) that the given assistant device has moved within the ecosystem. For example, and referring now specifically to FIG. 2B, assume that a first assistant input device 1061 in the form of a two-way standalone speaker and a display device is moved from room 250, in this example the kitchen, to room 256, in this example the living room. 2B , assume that a query or command has previously been received at or executed by the first assistant input device 1061, assume that some instances of ambient noise have been captured, assume that a respective unique identifier (or indicator) associated with a fourth assistant input device 1064 in the form of another two-way standalone speaker in room 256 (e.g., "living room speaker device") is detected at the first assistant input device 1061, and assume that the user preferences of the user associated with the ecosystem are known. In this example, the one or more device-specific signals may be limited to those generated or received after the first assistant input device 1061 has been moved within the ecosystem (fewer user preferences).

[0083] In this example, the queries and commands (or their corresponding text) may be processed using a semantic classifier to classify the queries and commands into one or more semantic categories. For example, the queries and commands previously received at the first assistant input device 1061 may include queries related to requesting weather or traffic information, commands related to planning a vacation, and / or other queries or commands. These instances of queries and commands may be classified into an "information" semantic category (or more specifically, a "weather information" category and a "traffic information" category), a "schedule" category, and / or other semantic categories based on the queries and commands received at the first assistant input device 1061. Additionally or alternatively, instances of ambient noise may be processed using an ambient noise detection model to classify the ambient noise into one or more semantic categories. For example, the ambient noise may capture music or podcasts being aurally rendered by the fourth assistant input device 1064, people sitting on a couch conversing as illustrated in room 256, movies or television programs being aurally rendered by computing devices in the ecosystem, and the like. These instances of ambient noise may be classified into a "music" semantic category, a "conversation" semantic category, a "movie" semantic category, a "television program" semantic category, and / or other semantic categories associated with noises typically encountered in a kitchen. Additionally or alternatively, a unique identifier (or indicator) of the fourth assistant input device 1064 (e.g., "living room speaker device") may be processed to generate one or more of the semantic indicators. This unique identifier (or indicator) may be classified, for example, into the "living room" semantic category and / or other semantic categories associated with assistant devices positionally proximate to the first assistant input device 1061 in the ecosystem of FIG. 2B.Additionally or alternatively, assume that the user preferences indicate that the user is interested in fictional movies titled Cooking and Vehicles and a particular character in the movie named Thunder McKing. These user preferences may be identified as being associated with the first assistant input device 1061 based on one or more of the other device-specific signals being classified into "Movies" and the user preferences for the movies Vehicles and Thunder McKing also being classified into the "Movies" category. As a result, the semantic indicator candidates in this example may include "Living Room Device", "Schedule Device", "Vehicles Device", "Thunder McKing Device", and / or other semantic indicator candidates based on one or more device-specific indicators associated with the first assistant input device 1061. Further, in this example, a given one of the semantic indicators can be automatically assigned to the first assistant input device 1061, or a user associated with the ecosystem can be prompted to select one or more of the candidate semantic indicators to assign the given semantic indicator to the first assistant input device 1061.

[0084] In some additional or alternative implementations, a given assistant device to which a semantic label is to be assigned may be identified in response to determining (e.g., via event detection engine 130 and / or device identification engine 140 of FIG. 1) that the portion of the ecosystem in which the given assistant device is located has been repurposed. For example, with particular reference now to FIG. 2B, assume that a first assistant input device 1061 in the form of a two-way standalone speaker and a display device is located in room 256, which in this example is a living room, but that the living room is being converted into a bedroom. 2B , assume that a query or command has previously been received at or executed by the first assistant input device 1061, assume that some instances of ambient noise have been captured, and assume that a respective unique identifier (or indicia) associated with a fourth assistant input device 1064 in the form of another two-way standalone speaker in room 256 (e.g., “living room speaker device”) is detected at first assistant input device 1061. In this example, the one or more device-specific signals may be limited to those generated or received after room 256 has been repurposed.

[0085] In this example, the queries or commands (e.g., text corresponding thereto) may be processed using a semantic classifier to classify the queries and commands into one or more semantic categories. For example, the queries and commands previously received at the first assistant input device 1061 may include commands for setting an alarm, commands for wake-up or bedtime routines, and / or other queries or commands. These instances of commands may be classified into an "alarm" semantic category, a "routine" category (or more specifically a "wake-up routine" category or a "bedtime routine" category), and / or other semantic categories based on the queries and commands received at the first assistant input device 1061. Additionally or alternatively, the instances of ambient noise may be processed using an ambient noise detection model to classify the ambient noise into one or more semantic categories. For example, the ambient noise may capture one or more users' snoring, people's conversations, etc. These instances of ambient noise may be classified into a "bedroom" semantic category, a "conversation" semantic category, and / or other semantic categories associated with noises typically encountered in a bedroom. Additionally or alternatively, a unique identifier (or indicator) of the fourth assistant input device 1064 (e.g., "living room speaker device") may be processed to generate one or more of the semantic indicators. This unique identifier (indicator) may be classified, for example, into the "living room" semantic category and / or other semantic categories associated with assistant devices positionally close to the first assistant input device 1061 in the ecosystem of FIG. 2B. As a result, the semantic indicator candidates in this example may include "living room display device", "bedroom display device", and / or other semantic indicator candidates based on one or more device-specific indicators associated with the first assistant input device 1061.Further, in this example, a given one of the semantic indicators can be automatically assigned to the first assistant input device 1061, or a user associated with the ecosystem can be prompted to select one or more of the candidate semantic indicators to assign the given semantic indicator to the first assistant input device 1061. In particular, in this example, the system can determine that the given semantic indicator corresponds to a "bedroom display device" because use of the first assistant input device 1061 indicates that the fourth assistant input device 1064 is located in a bedroom, even though the unique identifier (or indicator) of the fourth assistant input device 1064 corresponds to a "living room" category.

[0086] 2A and 2B are described herein with respect to a given assistant device to which a semantic label is assigned being an assistant input device (e.g., the first assistant input device 1061), it should be understood that this is for illustration purposes and is not intended to be limiting. For example, the techniques described herein may also be utilized to assign respective semantic labels to assistant non-input devices 185. For example, assume that a smart light (perhaps an assistant non-input device without a microphone) is newly added to room 252, which in this example is a bedroom. In this example, a unique identifier or label (e.g., "bedroom speaker device") associated with a third assistant input device 1063, which takes the form of a two-way standalone speaker without a display, may be utilized to infer a semantic label of "bedroom smart light" for the newly added smart light using the techniques described herein. Further assume that a smart light is moved from room 254 to room 262, which in this example is a laundry room. In this example, a unique identifier or indicator associated with a third assistant non-input device 1853 in the form of smart clothing can be utilized to infer a semantic indicator of “laundry room smart lighting” for a recently moved smart lighting using techniques described herein.

[0087] Turning now to FIG. 3, a flow chart is shown illustrating an example method 300 of assigning a given semantic label to a given assistant device in an ecosystem. For convenience, the operations of the method 300 are described with reference to a system that performs the operations. The system of the method 300 includes one or more processors and / or other components of a computing device. For example, the system of the method 300 may be implemented by the assistant input device 106 of FIG. 1, FIG. 2A, or FIG. 2B, the assistant non-input device 185 of FIG. 1, FIG. 2A, or FIG. 2B, the computing device 510 of FIG. 5, one or more servers, other computing devices, and / or any combination thereof. Moreover, although the operations of the method 300 are shown in a particular order, this is not intended to be limiting. One or more operations may be reordered, omitted, and / or added.

[0088] In block 352, the system identifies a given assistant device from among the multiple assistant devices in the ecosystem. The given assistant device may be an assistant input device (e.g., one of the assistant input devices 106 in FIG. 1) or an assistant non-input device (e.g., one of the assistant non-input devices 185 in FIG. 1). In some implementations, the given assistant device may be identified in response to determining that it is newly added to the ecosystem, while in other implementations, the given assistant device may be identified in response to determining that it has moved within the ecosystem (e.g., as described with respect to the event detection engine 130 in FIG. 1). In some additional or alternative implementations, the given assistant device may be identified periodically (e.g., once a month, once every six months, once a year, etc.).

[0089] In block 354, the system obtains a device-specific signal associated with the given assistant device. The device-specific signal may be detected by and / or generated by the given assistant device. In some implementations, block 354 may include one or more of optional sub-blocks 354A, 354B, 354C, or 354D. If included, in sub-block 354A, the system obtains a number of queries or commands (if any) previously received at the given assistant device. If included, in sub-block 354B, the system additionally or alternatively obtains an instance of ambient noise previously detected at the given assistant device (and optionally only when speech acceptance was active at the given assistant device (e.g., after receiving a specific word or phrase that invokes an automated assistant) or when speech acceptance is not active via a digital signal processor (DSP)). In some implementations, the ambient noise obtained is constrained to the ambient noise detected when speech acceptance is active at the given assistant device. If included, in subblock 354C, the system additionally or alternatively obtains a unique identifier (or indicator) for each assistant device that is positionally close to the given assistant device (e.g., determined using device identification engine 140 of FIG. 1). If included, in subblock 354D, the system additionally or alternatively obtains the user preferences of the user associated with the ecosystem.

[0090] In block 356, the system processes the device-specific signals to generate candidate semantic indicators for the given assistant device. In an implementation in which one or more device-specific signals include multiple queries or commands previously received at the given assistant device, the multiple queries or commands (or the text corresponding thereto) can be processed using a semantic classifier to classify the multiple queries or commands into one or more heterogeneous semantic categories. For example, queries related to cooking recipes and commands related to controlling a smart oven or a smart coffee maker in the kitchen can be classified into a cooking category or a kitchen category, queries related to weather can be classified into a weather category, commands related to controlling lighting can be classified into a lighting category, and so on. In an implementation in which one or more device-specific signals additionally or alternatively include instances of ambient noise detected at the given assistant device, the instances of ambient noise can be processed using an ambient noise detection model to classify each of the instances of ambient noise into one or more heterogeneous semantic categories. For example, if an instance of ambient noise is determined to correspond to a microwave oven ringing, food frying in a frying pan, food processor processing food, etc., the instance of ambient noise may be classified into a cooking category. As another example, if an instance of ambient noise is determined to correspond to a circular saw operating, hammering, etc., the instance of ambient noise may be classified into a garage category and a workshop category. In an implementation in which one or more device-specific signals additionally or alternatively include a unique identifier (or indicator) of each assistant device positionally proximate to a given assistant device, the unique identifier (or indicator) of each assistant device may be classified into one or more disparate semantic categories. For example, if the unique identifier (or indicator) of each assistant device corresponds to "coffee maker", "oven", "microwave", the unique identifier may be classified into a kitchen category or a cooking category.As another example, if the unique identifier (or indicator) of each assistant device corresponds to a "bedroom light" and a "bedroom casting device," the unique identifier may be categorized into a bedroom category. In an implementation in which one or more device-specific signals additionally or alternatively include a user's user preferences related to the ecosystem, the user preferences may be categorized into one or more disparate semantic categories. For example, if the user preferences indicate that the user is interested in cooking, cooking shows, specific cooks, and / or has other cooking-related interests, the user preferences may be categorized into a kitchen category, a cooking category, or a category related to specific cooks.

[0091] The semantic indicator candidates may be generated based on processing of one or more device-specific signals. For example, assume that one or more device-specific signals indicate that a given assistant device is located in a kitchen or living room of a user's main residence associated with the ecosystem. Further assume that the given assistant device is a two-way standalone speaker device with a display. In this example, a first semantic indicator candidate "kitchen display device" and a second semantic indicator candidate "living room display device" may be generated. Generating semantic indicator candidates based on one or more device-specific signals is described in more detail herein (e.g., in connection with FIG. 2A and FIG. 2B).

[0092] In block 358, the system determines whether to prompt a user associated with the ecosystem to select a given semantic label from among the candidate semantic labels. The system can determine whether to prompt a user to select a given semantic label based on whether the respective confidence levels associated with the candidate semantic labels meet a threshold confidence level. The respective confidence levels can be determined, for example, based on the amount of one or more device-specific signals that are classified into a given semantic category. For example, assume that the given assistant device identified in block 352 is a two-way stand-alone speaker that implements an instance of an automated assistant. Further assume that each of the one or more device-specific signals indicates that the two-way stand-alone speaker is located in a bedroom. For example, based on previous queries or commands received at the two-way stand-alone speaker being related to an alarm or bedtime routine, based on instances of ambient noise including snoring, and / or based on other assistant devices having unique identifiers (or labels) of "bedroom lighting" and "bedroom casting device". In this example, the system may be confident in the semantic label "bedroom speaker device" because the two-way standalone speaker is associated with a bedroom in a primary residence of a user associated with the ecosystem. However, if some of the queries or commands received at the two-way standalone speaker are related to cooking recipes, the system may be less confident in the semantic label "bedroom speaker device."

[0093] If in the iteration of block 358, the system determines not to prompt the user to select the given semantic indicator, the system may proceed to block 360. In block 360, the system automatically assigns the given semantic indicator from among the semantic indicator candidates to the given assistant device in the device topology representation in the ecosystem. The device topology representation of the ecosystem may be stored locally in one or more of the assistant devices in the ecosystem and / or in a remote system in communication with one or more of the assistant devices in the ecosystem. In some implementations, the given semantic indicator may be the only semantic indicator associated with the given assistant device (and may optionally replace other unique identifiers or indicators assigned to the given assistant device), while in other implementations, the given semantic indicator may be added to a list of unique identifiers or indicators assigned to the given assistant device. The system may then return to block 352 to identify additional given assistant devices from among the multiple assistant devices in the ecosystem and generate and assign additional given semantic indicators to the additional given assistant devices.

[0094] If, in an iteration of block 358, the system determines to prompt the user to select a given semantic indicator, the system may proceed to block 362. In block 362, the system generates a prompt to prompt the user of the client device to select a given semantic indicator based on the semantic indicator candidates. In block 364, the system causes a prompt to be rendered at the user's client device. The prompt may be rendered visually and / or audibly at the user's client device, and may optionally be based on the capabilities of the client device on which the prompt is rendered. For example, the prompt may be rendered via a software application available at the client device (e.g., a software application associated with the ecosystem or one or more of the assistant devices included in the ecosystem). In this example, if the client device includes a display, the prompt may be rendered visually via the software application (or a home screen of the client device) and / or audibly via a speaker of the client device. However, if the client device does not include a display, the prompt may only be rendered audibly via a speaker of the client device. In block 366, the system receives a selection of the given semantic indicator in response to the prompt. The user's client device may be, for example, a given assistant identified in block 352, or a separate client device (e.g., the user's mobile device, or any other assistant device in the ecosystem capable of rendering prompts). For example, assume that the system generates the semantic label candidates "bedroom speaker device" and "kitchen speaker device."In this example, the system may generate a prompt that includes selectable elements associated with both of these semantic indicators and request the user to provide an input (e.g., touch or voice) to select one of the candidate semantic indicators to be assigned to the given assistant device as the given semantic indicator to be assigned to the given assistant device. In block 368, the system assigns the given semantic indicator to the given assistant device in a device topology representation of the ecosystem in a manner similar to that described above with respect to block 360.

[0095] Turning now to FIG. 4, a flow chart is shown illustrating an example method 400 of using assigned semantic indicators in satisfying a query or command received at an assistant device in an ecosystem. For convenience, the operations of method 400 are described with reference to a system that performs the operations. The system of method 400 includes one or more processors and / or other components of a computing device. For example, the system of method 400 may be implemented by assistant input device 106 of FIG. 1, FIG. 2A, or FIG. 2B, assistant non-input device 185 of FIG. 1, FIG. 2A, or FIG. 2B, computing device 510 of FIG. 5, one or more servers, other computing devices, and / or any combination thereof. Moreover, although the operations of method 400 are shown in a particular order, this is not intended to be limiting. One or more operations may be reordered, omitted, and / or added.

[0096] In block 452, the system receives audio data corresponding to a user's speech via each microphone of each assistant device in an ecosystem that includes multiple assistant devices. A user may be associated with the ecosystem.

[0097] At block 454, the system processes the audio data to identify semantic characteristics of the query or command contained in the utterance. The semantic characteristics of the query or command may correspond to a linguistic unit, such as a word or phrase, that defines a related field or related set of the word and / or phrase. In some implementations, the system can process the audio data using a speech recognition model to convert the utterance captured in the audio data into text and can use a semantic classifier to identify the semantic characteristics based on the recognized text. In additional or alternative implementations, the system can process the audio data using a semantic classifier and can directly identify the semantic characteristics based on the audio data. For example, assume that the utterance received at block 452 is "Show me the chili recipe." In this example, the utterance (or its corresponding text) can be processed using a semantic classifier to identify the semantic characteristics "chili," "food," "kitchen," and / or "cooking." In particular, the semantic characteristics identified in block 454 may include one or more terms or phrases contained in the utterance, or may include a given semantic category into which the utterance is classified.

[0098] In block 456, the system determines whether the utterance specifies a given assistant device to be utilized in satisfying the query or command contained in the utterance. If, in an iteration of block 456, the system determines that the utterance specifies a given assistant device to be utilized in satisfying the query or command, the system may proceed to block 466. Block 466 is described below. For example, assume that the utterance received in block 452 is "Show me the chili recipe on the kitchen display device." In this example, the system may utilize the "kitchen display device" because the user specified that the "kitchen display device" should be utilized to present the "chili recipe" to the user who provided the utterance. As a result, the system may select the "kitchen display device" to satisfy the utterance. If, in an iteration of block 456, the system determines that the utterance does not specify a given assistant device to be utilized in satisfying the query or command, the system may proceed to block 458. For example, assume that the utterance received in block 452 is simply "Show me the chili recipe" without specifying an assistant device that should satisfy the utterance. In this example, the system can determine that the utterance does not specify a given assistant device that should satisfy the utterance. As a result, the system needs to determine which assistant device in the ecosystem should be utilized to satisfy the utterance.

[0099] In block 458, the system determines whether the semantic characteristic identified in block 454 matches a given semantic label assigned to a given assistant device of the plurality of assistant devices. The system can generate an embedding corresponding to one or more terms of the semantic characteristic, and can compare the embedding of the semantic characteristic with a plurality of embeddings corresponding to one or more respective terms of the respective semantic labels assigned to one or more of the plurality of assistant devices in the ecosystem. Furthermore, the system can determine whether the embedding of the semantic characteristic matches any of a plurality of embeddings of the respective semantic labels assigned to one or more of the plurality of assistant devices in the ecosystem. For example, the system can determine a distance measure between the embedding of the semantic characteristic and each of a plurality of embeddings of the respective semantic labels assigned to one or more of the plurality of assistant devices in the ecosystem. Furthermore, the system can determine whether the distance measure meets a threshold distance (e.g., to identify an exact match or a rough match).

[0100] In the iteration of block 458, if the system determines that the semantic characteristic identified in block 454 matches the given semantic indicator assigned to the given assistant device, the system may proceed to block 460. In block 460, the system may cause the given client device associated with the given semantic indicator in the device topology representation of the ecosystem to satisfy the query or command included in the utterance. The system may then return to block 452 and monitor additional audio through the microphones of each of the multiple assistant devices in the ecosystem. More specifically, the system may cause the given assistant device to perform one or more actions to satisfy the utterance. For example, assume that the utterance received in block 452 is "Show me the chili recipe" and the identified semantic characteristic is "chili", "food", "kitchen", and / or "cooking". Further assume that the semantic indicator assigned to the given assistant device is "kitchen display device". In this example, the system can select a given assistant device that is assigned the semantic indicator "kitchen display device" even if the utterance was not received at this assistant device. Furthermore, the system can cause a given assistant device that is assigned the semantic indicator "kitchen display device" to visually render a chili recipe in response to the utterance being received by the respective microphone of the respective assistant device.

[0101] In various implementations, multiple assistant devices in the ecosystem may be assigned semantic labels that match the semantic characteristics identified based on the utterance. In particular, although not shown in FIG. 4 for clarity, the system may determine whether the semantic characteristics identified in block 454 match a given semantic label assigned to a given assistant device of the multiple assistant devices in addition to or instead of using proximity information (e.g., as described above with respect to the query / command processing engine 170 of FIG. 1) when selecting one or more assistant devices for satisfying the utterance. Continuing with the above example, assume that there are multiple assistant devices that are assigned the semantic label "kitchen display device". In this example, the assistant device of the multiple assistant devices that is assigned the semantic label "kitchen display device" that is closest to the user in the ecosystem may be utilized to satisfy the utterance. Moreover, although not shown in FIG. 4 for clarity, the system may determine whether the semantic characteristics identified in block 454 match a given semantic label assigned to a given assistant device of the plurality of assistant devices in addition to or instead of using device capability information (e.g., as described above with respect to the query / command processing engine 170 of FIG. 1) in selecting one or more assistant devices for satisfying the utterance. Continuing with the above example, assume that a first assistant device having a display device is assigned the semantic label "kitchen display device" and a second assistant device not having a display device is assigned the semantic label "kitchen speaker device".In this example, a first assistant device assigned the semantic marker "kitchen display device" can be selected to satisfy the utterance in preference to a second assistant device assigned the semantic marker "kitchen speaker device" because the utterance specifies "show me" a chili recipe and the first assistant device can display a chili recipe in response to the utterance, while the second assistant device cannot display a chili recipe.

[0102] If, in an iteration of block 458, the system determines that the semantic characteristic identified in block 454 does not match any semantic indicator assigned to any of the assistant devices in the ecosystem, the system may proceed to block 462. In block 462, the system may identify a given assistant device that is close to the user. For example, the system may identify a given assistant device that is closest to the user in the ecosystem (e.g., as described with respect to the presence sensor 105 of FIG. 1).

[0103] In block 464, the system determines whether the given assistant device identified in block 462 can satisfy the query or command. The capabilities of each of the assistant devices may be stored in the device topology representation of the ecosystem and in association with the respective assistant device (e.g., as a device attribute of the respective assistant device). If, in an iteration of block 464, the system determines that the given assistant device identified in block 462 cannot satisfy the query or command, the system may return to block 462 and identify additional given assistant devices that are also close to the user. The system may again proceed to block 464 and determine whether additional given assistant devices identified in subsequent iterations of block 462 can satisfy the query or command. The system may repeat this process until an assistant device that can satisfy the query or command is identified. For example, assume that the assistant device identified in block 462 is a standalone speaker device without a display, but a display is required to satisfy the speech. In this example, the system may determine that the assistant device identified in block 462 cannot satisfy the utterance in block 464 and may return to block 462 and identify additional assistant devices in the user's vicinity in block 462. If in the iteration of block 464 the system determines that the given assistant device identified in block 462 can satisfy the query or command, the system may proceed to block 462. Continuing with the above example, further assume that the given assistant device identified in the first iteration of block 462 is a standalone speaker device including a display, or that the additional assistant devices identified in the additional iterations of block 462 are standalone speaker devices with displays.In this example, the system may determine that the assistant device identified in a subsequent iteration of block 462 can satisfy the utterance of block 464, and the system may proceed to block 466.

[0104] In block 466, the system causes the given assistant device to satisfy the query or command contained in the utterance. The system may satisfy the utterance in a manner similar to that described above with respect to block 460. The system may then return to block 452 and monitor additional audio data via the microphones of each of the multiple assistant devices in the ecosystem.

[0105] 5 is a block diagram of an example computing device 510 that may be optionally utilized to perform one or more aspects of the techniques described herein. In some implementations, one or more of the assistant input devices, one or more of the cloud-based automated assistant components, one or more assistant non-input systems, one or more assistant non-input devices, and / or other components may comprise one or more components of the example computing device 510.

[0106] Computing device 510 typically includes at least one processor 514 that communicates with a number of peripheral devices via a bus subsystem 512. These peripheral devices may include, for example, a storage subsystem 524, including a memory subsystem 525 and a file storage subsystem 526, user interface output devices 520, user interface input devices 522, and a network interface subsystem 516. The input and output devices enable user interaction with the computing device 510. The network interface subsystem 516 provides an interface to external networks and is coupled to corresponding interface devices in other computing devices.

[0107] The user interface input devices 522 may include a keyboard, a pointing device such as a mouse, a trackball, a touchpad, or a graphics tablet, a scanner, a touch screen integrated into a display, a voice recognition system, an audio input device such as a microphone, and / or other types of input devices. In general, use of the term "input device" is intended to include all possible types of devices and methods for inputting information into the computing device 510 or a communications network.

[0108] The user interface output devices 520 may include a display subsystem, a printer, a fax machine, or a non-visual display such as an audio output device. The display subsystem may include a flat panel device such as a cathode ray tube (CRT), a liquid crystal display (LCD), a projection device, or some other mechanism for producing a visible image. The display subsystem may also provide a non-visual display, such as via an audio output device. In general, use of the term "output device" is intended to include all possible types of devices and methods for outputting information from the computing device 510 to a user or to another machine or computing device.

[0109] Storage subsystem 524 stores programming and data constructs that provide the functionality of some or all of the modules described herein. For example, storage subsystem 524 may include logic for performing selected aspects of the methods described herein, as well as for implementing the various components shown in FIG.

[0110] These software modules are generally executed by the processor 514, alone or in combination with other processors. The memory 525 used by the storage subsystem 524 may include several memories, including a main random access memory (RAM) 530 for storing instructions and data during program execution, and a read-only memory (ROM) 532 in which fixed instructions are stored. The file storage subsystem 526 may provide persistent storage for program and data files and may include a hard disk drive, a floppy disk drive with associated removable media, a CD-ROM drive, an optical drive, or a removable media cartridge. Modules implementing the functionality of some implementations may be stored by the file storage subsystem 526 in the storage subsystem 524, or on other machines accessible by the processor 514.

[0111] The bus subsystem 512 provides a mechanism for allowing the various components and subsystems of the computing device 510 to communicate with each other as intended. Although the bus subsystem 512 is shown generally as being a single bus, alternative implementations of the bus subsystem may use multiple buses.

[0112] The computing device 510 can be of various types, including a workstation, a server, a computing cluster, a blade server, a server farm, or any other data processing system or computing device. Due to the ever-changing nature of computers and networks, the description of the computing device 510 shown in Figure 5 is intended only as a specific example intended to illustrate some implementations. Many other configurations of the computing device 510 can have more or fewer components than the computing device shown in Figure 5.

[0113] In situations where some implementations discussed herein may collect or use personal information about a user (e.g., user data extracted from other electronic communications, information about the user's social network, the user's location, the user's time, the user's biometric information, and user activity and attribute information, relationships between users, etc.), the user is given one or more opportunities to control whether information is collected, whether the personal information is stored, whether the personal information is used, and how information about the user is collected, stored, and used. That is, the systems and methods discussed herein collect, store, and / or use a user's personal information only if they receive explicit authorization to do so from the associated user.

[0114] For example, a user may be given control over whether a program or feature collects user information about that particular user or other users associated with that program or feature. Each user about whom personal information is to be collected is presented with one or more options to allow control over the collection of information associated with that user, to provide permission or approval as to whether the information is collected and what portions of the information should be collected. For example, a user may be given one or more such control options over a communications network. In addition, certain data may be handled in one or more ways before it is stored or used, such that personally identifiable information is removed. As one example, a user's identity may be handled such that personally identifiable information cannot be determined. As another example, a user's geographic location may be generalized to a broader area such that a specific location of the user cannot be determined.

[0115] In some implementations, a method is provided that is implemented by one or more processors, the method including: identifying a given assistant device from among a plurality of assistant devices in an ecosystem; acquiring one or more device-specific signals associated with the given assistant device, the one or more device-specific signals being generated or received by the given assistant device; processing one or more of the device-specific signals to generate one or more candidate semantic indicators for the given assistant device; selecting a given semantic indicator for the given assistant device from among the one or more candidate semantic indicators; and assigning the given semantic indicator to the given assistant device in a device topology representation of the ecosystem. Assigning the given semantic indicator to the given assistant device includes automatically assigning the given semantic indicator to the given assistant device.

[0116] These and other implementations of the techniques disclosed herein may include one or more of the following features.

[0117] In some implementations, one or more of the device-specific signals may include at least device activity associated with the given assistant device, and the device activity associated with the given assistant device may include a plurality of queries or commands previously received at the given assistant device. In some versions of those implementations, processing one or more of the device-specific signals to generate one or more candidate semantic indicators for the given assistant device may include processing the device activity associated with the given assistant device using a semantic classifier to classify each of the plurality of queries or commands into one or more of a plurality of disparate categories, and generating one or more of the candidate semantic indicators based on one or more of the plurality of disparate categories into which each of the plurality of queries or commands is classified. In some further versions of those implementations, selecting the given semantic indicator for the given assistant device may include selecting the given semantic indicator for the given assistant device based on a quantity of queries or commands that fall into a given category of the plurality of disparate categories, the given semantic indicator being associated with the given semantic indicator.

[0118] In some implementations, one or more of the device-specific signals may include at least device activity associated with the given assistant device, the device activity associated with the given assistant device comprising ambient noise previously detected at the given assistant device, the ambient noise previously detected at the given assistant device when speech reception was active. In some versions of those implementations, the method may further include processing the ambient noise previously detected at the given assistant device using an ambient noise detection model to classify the ambient noise into one or more of a plurality of heterogeneous categories, and generating one or more candidate semantic indicators based on one or more of the plurality of heterogeneous categories into which the ambient noise is classified. In some further versions of those implementations, selecting the given semantic indicator for the given assistant device may include selecting the given semantic indicator for the given assistant device based on the ambient noise being classified into the given category of the plurality of heterogeneous categories, the given semantic indicator being associated with the given semantic indicator.

[0119] In some implementations, one or more of the device-specific signals may include a respective unique identifier for one or more of a plurality of assistant devices positionally proximate to the given assistant device in the ecosystem. In some versions of those implementations, processing one or more of the device-specific signals to generate one or more candidate semantic indicators for the given assistant device may include identifying one or more of a plurality of assistant devices positionally proximate to the given assistant device in the ecosystem based on one or more wireless signals, obtaining a respective unique identifier for one or more of the plurality of assistant devices positionally proximate to the given assistant device, and generating one or more of the candidate semantic indicators based on the respective unique identifier for one or more of the plurality of assistant devices positionally proximate to the given assistant device. In some further versions of those implementations, selecting a given semantic indicator for the given assistant device may include selecting a given semantic indicator for the given assistant device based on a property of the respective unique identifier for one or more of the plurality of assistant devices positionally proximate to the given assistant device.

[0120] In some implementations, a given assistant device may be identified in response to determining that the given assistant device is newly added to the ecosystem or in response to determining that the given assistant device has moved within the ecosystem.

[0121] In some versions of those implementations, the given assistant device may be identified in response to determining that the given assistant device has moved within the ecosystem. In some further versions of those implementations, assigning the given semantic indicator to the given assistant device may include adding the given semantic indicator to a list of semantic indicators associated with the given assistant device or replacing an existing semantic indicator associated with the given assistant device with the given semantic indicator. In some additional or alternative versions of those implementations, determining that the given assistant device has moved within the ecosystem may include identifying, based on one or more wireless signals, that a current subset of the plurality of assistant devices positionally proximate to the given assistant device in the ecosystem is different from a stored subset of the plurality of assistant devices associated with the given assistant device. In still further versions of those implementations, the method may further include switching the given assistant device from an existing group of assistant devices that includes one or more of the multiple assistant devices to an additional existing group of assistant devices that includes one or more of the multiple assistant devices, or creating a new group of assistant devices that includes at least the given assistant device.

[0122] In some versions of those implementations, the given assistant device may be identified in response to determining that the given assistant device is newly added to the ecosystem. In some further versions of those implementations, assigning the given semantic indicator to the given assistant device may include adding the given semantic indicator to a list of semantic indicators associated with the given assistant device. In some additional or alternative versions of those further implementations, determining that the given assistant device is newly added to the ecosystem may include identifying that the given assistant device is added to a wireless network associated with the ecosystem based on one or more wireless signals. In still further versions of those implementations, the method may further include adding the given assistant to an existing group of assistant devices including one or more of the multiple assistant devices, or creating a new group of assistant devices including at least the given assistant device.

[0123] In some implementations, a given assistant device may be periodically identified to detect whether existing semantic indicators assigned to the given assistant device are correct.

[0124] In some implementations, the method may further include, after assigning a given semantic indicator to the given assistant device in a device topology representation of the ecosystem, receiving audio data corresponding to an utterance from a user associated with the ecosystem via one or more respective microphones of one of the plurality of assistant devices in the ecosystem, the utterance including a query or a command; processing the audio data corresponding to the utterance to determine semantic characteristics of the query or command; determining that the semantic characteristics of the query or command match the given semantic indicator assigned to the given assistant device; and, in response to determining that the semantic indicator of the query or command matches the given semantic indicator assigned to the given assistant device, causing the given assistant device to satisfy the query or command.

[0125] In some implementations, one or more of the device-specific signals may include at least a user preference of a user associated with the ecosystem, and the user preference may be determined based on user interactions with multiple assistant devices in the ecosystem. In some versions of those implementations, processing one or more of the device-specific signals to generate one or more candidate semantic indicators for the given assistant device may include processing the user preferences using a semantic classifier to identify at least one semantic category of a plurality of disparate semantic categories associated with the user preferences, and generating one or more of the candidate semantic indicators based on the given semantic category. In yet further versions of those implementations, selecting a given semantic indicator for the given assistant device may include determining that a given semantic indicator of the plurality of candidate semantic indicators is associated with the given assistant device, and selecting the given semantic indicator for the given assistant device in response to determining that the given semantic indicator is associated with the given assistant device. Determining that a given semantic indicator may be associated with a given assistant device is based on one or more additional device-specific signals associated with the given assistant device. In some additional or alternative versions of these still further implementations, processing user preferences to identify at least one semantic category may be in response to receiving a user input to assign one or more semantic indicators to at least the given assistant device.

[0126] In some implementations, a method implemented by one or more processors is provided, the method including: identifying a given assistant device from among a plurality of assistant devices in an ecosystem; acquiring one or more device-specific signals associated with the assistant device, the one or more device-specific signals being generated or received by the given assistant device; processing one or more of the device-specific signals to generate one or more candidate semantic indicators for the given assistant device; generating a prompt for a user of a client device to select a given semantic indicator based on one or more of the candidate semantic indicators, the selection being from among one or more of the candidate semantic indicators; causing the prompt to be rendered at the user's client device; and, in response to receiving a selection of the given semantic indicator in response to the prompt, assigning the given semantic indicator to the given assistant device in a device topology representation of the ecosystem.

[0127] These and other implementations of the techniques disclosed herein may include one or more of the following features.

[0128] In some implementations, the one or more device-specific signals may include two or more of device activity associated with the given assistant device, where the device activity associated with the given assistant device comprises multiple queries or commands previously received at the given assistant device, ambient noise previously detected at the given assistant device when speech acceptance was active, or a respective unique identifier for one or more of multiple assistant devices positionally proximate to the given assistant device in the ecosystem.

[0129] In some implementations, a method implemented by one or more processors is provided, the method including steps of identifying a given assistant device from among a plurality of assistant devices in an ecosystem; obtaining one or more device-specific signals associated with the given assistant device, the one or more device-specific signals being generated or received by the given assistant device; determining a given semantic indicator for the given assistant device based on one or more of the device-specific signals; assigning the given semantic indicator to the given assistant device in a device topology representation of the ecosystem; and assigning the given semantic indicator to the given assistant device in a device topology representation of the ecosystem. After the allocation, the method includes receiving audio data corresponding to an utterance from a user associated with the ecosystem via one or more respective microphones of one of the plurality of assistant devices in the ecosystem, the utterance including a query or a command; processing the audio data corresponding to the utterance to determine semantic characteristics of the query or command; determining that the semantic characteristics of the query or command match a given semantic indicator assigned to the given assistant device; and, in response to determining that the semantic characteristics of the query or command match the given semantic indicator assigned to the given assistant device, causing the given assistant device to satisfy the query or command.

[0130] In addition, some implementations include one or more processors (e.g., a central processing unit (CPU), a graphics processing unit (GPU and / or a tensor processing unit (TPU))) of one or more computing devices, where the one or more processors are operable to execute instructions stored in associated memory, where the instructions are configured to cause performance of any of the aforementioned methods. Some implementations also include one or more non-transitory computer-readable storage media that store computer instructions executable by the one or more processors to perform any of the aforementioned methods. Some implementations also include a computer program product that includes instructions executable by the one or more processors to perform any of the aforementioned methods.

[0131] It is understood that all combinations of the foregoing concepts and additional concepts more fully described herein are contemplated as being part of the subject matter disclosed herein, for example, all combinations of claimed subject matter appearing at the end of this disclosure are contemplated as being part of the subject matter disclosed herein. [Explanation of symbols]

[0132] 105 Presence Sensor 106 Assistant Input Device 107 User Interface Components 110 Network 114 Speech Capture / TTS / STT Module 116 Cloud-based TTS module 117 Cloud-based STT module 118 Automation Assistant Client 119 Cloud-based automation assistant components 120 Automation Assistant 122 Natural Language Processing (NLP) Module, Natural Language Processor Module 130 Event Detection Engine 140 Device Specific Engine 150 Event Processing Engine 160 Semantic Marking Engine 170 Query / Command Processing Engine 180 Assistant Non-Input System 185 Assistant Non-Input Device 191 Device Activity Database 192 ML model database 193 Device Topology Database 250 rooms 252 rooms 254 rooms 256 rooms 258 rooms 260 rooms 262 rooms 510 Computing Devices 512 Bus Subsystem 514 processor 516 Network Interface Subsystem 520 User Interface Output Device 522 User Interface Input Devices 524 Storage Subsystem 525 Memory Subsystem 526 File Storage Subsystem 630 Main Random Access Memory (RAM) 632 Read-Only Memory (ROM)

Claims

1. 1. A method implemented by one or more processors, comprising: Identifying a given assistant device from among a plurality of assistant devices in the ecosystem; Obtaining one or more device-specific signals associated with the given assistant device, the one or more device-specific signals being generated or received by the given assistant device; Processing one or more of the device-specific signals to generate one or more candidate semantic indicators for the given assistant device; Selecting a given semantic indicator for the given assistant device from among the one or more candidate semantic indicators; and assigning the given semantic indicator to the given assistant device in a device topology representation of the ecosystem, wherein the step of assigning the given semantic indicator to the given assistant device comprises automatically assigning the given semantic indicator to the given assistant device.

2. 2. The method of claim 1, wherein one or more of the device-specific signals comprises at least a device activity associated with the given assistant device, and the device activity associated with the given assistant device comprises a plurality of queries or commands previously received at the given assistant device.

3. Processing the one or more of the device-specific signals to generate one or more of the candidate semantic indicators for the given assistant device, Using a semantic classifier, process device activity associated with the given assistant device to classify each of a plurality of queries or commands into one or more of a plurality of disparate categories; and generating one or more of the candidate semantic labels based on the one or more of the plurality of heterogeneous categories into which each of the plurality of queries or commands falls.

4. The step of selecting the given semantic indicator for the given assistant device includes:

4. The method of claim 3, further comprising: selecting the given semantic indicator for the given assistant device based on an amount of the plurality of queries or commands that are classified into a given category of the plurality of disparate categories, the given semantic indicator being associated with the given semantic indicator.

5. 5. The method of claim 1, wherein one or more of the device-specific signals comprises at least device activity associated with the given assistant device, the device activity associated with the given assistant device comprises ambient noise previously detected at the given assistant device, and the ambient noise was previously detected at the given assistant device when speech acceptance was active.

6. Using an ambient noise detection model, process the ambient noise previously detected at the given assistant device to classify the ambient noise into one or more of a plurality of disparate categories; and generating one or more of the candidate semantic labels based on the one or more of the plurality of heterogeneous categories into which the ambient noise is classified.

7. The step of selecting the given semantic indicator for the given assistant device includes:

7. The method of claim 6, further comprising: selecting the given semantic indicator for the given assistant device based on the ambient noise being classified into a given category of the plurality of heterogeneous categories, the given semantic indicator being associated with the given semantic indicator.

8. 8. The method of claim 1, wherein one or more of the device-specific signals comprise a respective unique identifier for one or more of the plurality of assistant devices that are positionally proximate to the given assistant device in the ecosystem.

9. Processing the one or more of the device-specific signals to generate one or more of the candidate semantic indicators for the given assistant device, Identifying one or more of the plurality of assistant devices that are positionally proximate to the given assistant device in the ecosystem based on one or more wireless signals; Obtaining the respective unique identifiers for one or more of the plurality of assistant devices that are positionally proximate to the given assistant device; and generating one or more of the candidate semantic signs based on the respective unique identifiers for one or more of the plurality of assistant devices that are positionally proximate to the given assistant device.

10. The step of selecting the given semantic indicator for the given assistant device includes:

10. The method of claim 9, further comprising: selecting the given semantic indicator for the given assistant device based on a property of the respective unique identifiers for one or more of the plurality of assistant devices that are positionally proximate to the given assistant device.

11. 11. The method of claim 1, wherein the given assistant device is identified in response to determining that the given assistant device is newly added to the ecosystem or in response to determining that the given assistant device has moved within the ecosystem.

12. 12. The method of claim 11, wherein the given assistant device is identified in response to determining that the given assistant device has moved within the ecosystem.

13. The step of assigning the given semantic indicator to the given assistant device includes: adding the given semantic indicator to a list of semantic indicators associated with the given assistant device; or The method of claim 12, comprising replacing an existing semantic indicator associated with the given assistant device with the given semantic indicator.

14. The step of determining that the given assistant device has moved within the ecosystem includes:

14. The method of claim 12 or 13, comprising: determining, based on one or more wireless signals, that a current subset of the plurality of assistant devices that are positionally proximate to the given assistant device in the ecosystem is different from a stored subset of the plurality of assistant devices stored in association with the given assistant device.

15. Switching the given assistant device from an existing group of assistant devices that includes one or more of the plurality of assistant devices to an additional existing group of assistant devices that includes one or more of the plurality of assistant devices; or The method of claim 14, further comprising creating a new group of assistant devices that includes at least the given assistant device.

16. The method of claim 11 , wherein the given assistant device is identified in response to determining that the given assistant device is newly added to the ecosystem.

17. The step of assigning the given semantic indicator to the given assistant device includes:

17. The method of claim 16, comprising adding the given semantic indicator to a list of semantic indicators associated with the given assistant device.

18. The step of determining that the given assistant device is newly added to the ecosystem includes:

18. The method of claim 16 or 17, comprising determining that the given assistant device has been added to a wireless network associated with the ecosystem based on one or more wireless signals.

19. Adding the given assistant device to an existing group of assistant devices that includes one or more of the plurality of assistant devices; or The method of claim 18, further comprising creating a new group of assistant devices that includes at least the given assistant device.

20. 20. The method of claim 1, wherein the given assistant device is periodically identified to verify whether existing semantic indicators assigned to the given assistant device are correct.

21. After assigning the given semantic indicator to the given assistant device in the device topology representation of the ecosystem, receiving, via one or more respective microphones of one of the plurality of assistant devices in the ecosystem, audio data corresponding to an utterance from a user associated with the ecosystem, the utterance including a query or a command; processing the audio data corresponding to the utterance to determine semantic characteristics of the query or command; determining that the semantic characteristics of the query or command match the given semantic indicator assigned to the given assistant device; In response to determining that the semantic characteristics of the query or command match the given semantic indicator assigned to the given assistant device, causing the given assistant device to satisfy the query or command; 21. The method of any one of claims 1 to 20, further comprising:

22. 22. The method of claim 1, wherein one or more of the device-specific signals comprise at least a user preference of a user associated with the ecosystem, the preference being determined based on user interaction with the plurality of assistant devices in the ecosystem.

23. Processing the one or more of the device-specific signals to generate one or more of the candidate semantic indicators for the given assistant device, processing the user preferences using a semantic classifier to identify at least one semantic category of a plurality of disparate semantic categories associated with the user preferences; and generating one or more of the candidate semantic labels based on the given semantic category.

24. The step of selecting the given semantic indicator for the given assistant device includes: determining that the given semantic indicator among the plurality of candidate semantic indicators is associated with the given assistant device, the determining that the given semantic indicator is associated with the given assistant device being based on one or more additional device-specific signals associated with the given assistant device; and selecting the given semantic indicator for the given assistant device in response to determining that the given semantic indicator is associated with the given assistant device.

25. 25. The method of claim 23 or 24, wherein the step of processing the user preferences to identify the at least one semantic category is in response to receiving user input for assigning one or more semantic indicators to at least the given assistant device.

26. Identifying a given assistant device from among a plurality of assistant devices in the ecosystem; Obtaining one or more device-specific signals associated with the assistant device, the one or more device-specific signals being generated or received by the given assistant device; Processing one or more of the device-specific signals to generate one or more candidate semantic indicators for the given assistant device; generating a prompt to ask a user of a client device to select a given semantic indicator based on one or more of the candidate semantic indicators, the selection being from one or more of the candidate semantic indicators; causing the prompt to be rendered at the client device of the user; In response to receiving the selection of the given semantic indicator in response to the prompt, assigning the given semantic indicator to the given assistant device in a device topology representation of the ecosystem.

27. the one or more device specific signals: device activity associated with the given assistant device, the device activity comprising a number of queries or commands previously received at the given assistant device; Ambient noise previously detected at the given assistant device when speech acceptance was active, or A respective unique identifier for one or more of the plurality of assistant devices that are geographically proximate to the given assistant device in the ecosystem.

27. The method of claim 26, comprising two or more of:

28. Identifying a given assistant device from among a plurality of assistant devices in the ecosystem; Obtaining one or more device-specific signals associated with the given assistant device, the one or more device-specific signals being generated or received by the given assistant device; Determining a given semantic indicator for the given assistant device based on one or more of the device-specific signals; Assigning the given semantic indicator to the given assistant device in a device topology representation of the ecosystem; After assigning the given semantic indicator to the given assistant device in the device topology representation of the ecosystem, receiving, via one or more respective microphones of one of the plurality of assistant devices in the ecosystem, audio data corresponding to an utterance from a user associated with the ecosystem, the utterance including a query or a command; processing the audio data corresponding to the utterance to determine semantic characteristics of the query or command; determining that the semantic characteristics of the query or command match the given semantic indicator assigned to the given assistant device; In response to determining that the semantic characteristics of the query or command match the given semantic indicator assigned to the given assistant device, causing the given assistant device to satisfy the query or command; 23. A method implemented by one or more processors, comprising:

29. At least one processor; and a memory storing instructions that, when executed, cause the at least one processor to perform the method of any one of claims 1 to 28.

30. 29. A non-transitory computer readable storage medium storing instructions that, when executed, cause the at least one processor to perform the method of any one of claims 1 to 28.

Images