A content recommendation method, device and storage medium based on in-vehicle conversations

Abstract

The present invention provides a content recommendation method based on in-vehicle conversations, comprising the following steps: S1. Acquiring voice information from a user. S2. Analyzing the user's emotional tendencies and / or user preferences contained in the voice information through a language understanding module; when the emotional tendencies or user preferences are analyzed as negative, a recommendation module performs a recommendation operation; otherwise, the method proceeds to a conversation strategy module. S3. The conversation strategy module determines whether to proceed to the next round of interaction. When the conversation strategy module determines that a conversation interaction is not necessary, the method proceeds to the recommendation module to perform a recommendation operation; otherwise, the method proceeds to S1 to proceed to the next round of interaction. The present invention gradually clarifies user preferences through conversational interactions without affecting the user experience.

Description

Technical Field

[0001] The present invention relates to a content recommendation method, device and storage medium, and in particular to a content recommendation method, device and storage medium based on in-vehicle conversation. Background Art

[0002] Speech recognition is a high-tech technology that allows machines to identify and understand speech signals and convert them into corresponding text files or commands. As a specialized research field, speech recognition is also an interdisciplinary subject, closely connected with many other disciplines, including acoustics, phonetics, linguistics, digital signal processing theory, information theory, and computer science.

[0003] Speech recognition is a key component of the human-computer interface and a crucial application technology in speech signal processing worldwide. The goal of speech recognition is to enable machines to "understand" human speech, a crucial aspect of machine intelligence.

[0004] The difficulty of getting machines to recognize speech is, in some ways, like someone who isn't proficient in a foreign language trying to understand a foreign speaker. It depends on the speaker, speaking speed, content, and environmental conditions. The inherent characteristics of speech signals contribute to the difficulty in speech recognition. These characteristics include variability, dynamism, temporality, and continuity.

[0005] Currently, voice control technology is most widely used in home appliances. Examples include Sony's voice-activated car stereos, JVC's KD-LXSO voice-activated car box, InVoca's omnidirectional voice remote control, and Delonghi's microwave ovens, all of which can operate various functions using voice commands. In December 2001, Panasonic Corporation launched the "BS Digital High-Definition TV with Built-in AV Hard Drive," a 36-inch television equipped with voice recognition. This model features a microphone remote control that allows for voice input. Users can easily select a specific station, either BS digital broadcasting or terrestrial analog broadcasting, by simply stating the station name or channel number.

[0006] Traditional remote controls required stepping through menus to search for program information or schedule recordings. Voice control, however, eliminates the need for intermediate steps. Voice control is finding widespread application not only in home appliances but also in communications, automated control, and home automation.

[0007] In car scenarios, voice assistants often passively play various roles, such as navigation, music playback, video playback, etc.

[0008] When it comes to music, user preferences are not static. For example, a user may have previously enjoyed listening to Bach, but recently switched to Jay Chou. This conflict between the user profile and their recent needs can lead to multiple rounds of conversations. Users can often develop aversion and resistance to these interactions, impacting the user experience.

[0009] Problems with existing car recommendations:

[0010] 1. Current in-car scenarios do not fully utilize the proactive conversational capabilities of voice assistants to explore user preferences. They can only push notifications based on user profiles or extract keywords based on user voice input.

[0011] 2. When there is no user data for a certain scenario, we can only refer to the preferences of other users, which cannot represent the user's real needs. Summary of the Invention

[0012] In order to solve the above problems in the prior art, the present invention provides a content recommendation method based on in-vehicle conversation, comprising the following steps:

[0013] S1. Obtain the voice message sent by the user.

[0014] S2. The language understanding module analyzes the user's emotional tendencies and / or user preferences contained in the voice information. When the emotional tendencies or user preferences are analyzed as negative, the recommendation module performs the recommendation operation; otherwise, the process switches to the dialogue strategy module.

[0015] S3. The dialogue strategy module decides whether to enter the next round of interaction. When the dialogue strategy module determines that there is no need for the next round of interaction, it enters the recommendation module to perform the recommendation operation; otherwise, it goes to S1 to enter the next round of interaction.

[0016] Through the above method, when the user's emotional tendency and user preference are positive or other, the dialogue strategy module determines whether it is necessary to enter the next round of interaction. If necessary, the user is guided to continue inputting voice information. If not, the module switches to the recommendation module to recommend products or services, such as songs, to the user.

[0017] The content of the conversation is based on a specific scenario. By analyzing the conversation content, building a knowledge graph for the corresponding scenario, and guiding the conversation based on prior knowledge, we can clarify the user's preferences in the specific scenario, making the conversation recommendation process more in line with the business scenario.

[0018] In the dialogue strategy, by improving the model structure in dialogue management, introducing more prior knowledge bases, and improving the structure of dialogue management, the dialogue strategy focuses more on the exploration process of user preferences, so as to obtain user preferences in as few dialogue rounds as possible.

[0019] Without affecting the user experience, gradually clarify the user's preferences through dialogue interaction.

[0020] When the user's emotional tendency or user preference is negative, it will directly go to the recommendation module to recommend products or services to the user. When the user expresses a negative emotional tendency, it can stop exploring user preferences in time and recommend items and services in time.

[0021] Furthermore, the recommendation module uses the user profile constructed based on the user's historical log as the user feature matrix and item matrix to build a two-stage recommendation system based on recall and ranking.

[0022] Furthermore, the recommendation module also constructs a feature matrix based on the preferences in the user's features, taking the current user input and short-term user logs as short-term preferences, and the preferences in the user portrait as long-term preferences.

[0023] Furthermore, the recommendation module also uses the current user input and short-term user log as short-term preferences, and the preferences in the user portrait as long-term preferences based on the preferences in the user's characteristics, and performs weighted calculation of the short-term preferences and long-term preferences to obtain the user preferences.

[0024] Furthermore, the short-term user log includes the types of songs that the user has listened to in a short period of time, the duration of listening to the songs, and the types of songs that the user has not finished listening to.

[0025] Furthermore, the short-term duration does not exceed 7 days.

[0026] Furthermore, the recommendation module uses attention weighted calculation.

[0027] Furthermore, the dialogue strategy module determines the next action based on the DQN network of each time state.

[0028] A content recommendation device based on in-vehicle conversation is also provided, comprising a processor and a memory, wherein the memory stores executable code. When the executable code is executed by the processor, the processor executes the above method.

[0029] A non-transitory machine-readable storage medium is also provided, on which executable code is stored. When the executable code is executed by a processor of an electronic device, the processor is caused to perform the above method.

[0030] This invention gradually identifies user preferences through interactive dialogue without compromising user experience. When a user's sentiment or preferences are negative, the system directly switches to the recommendation module to recommend products or services. When a user expresses negative sentiment, the system promptly stops exploring user preferences and recommends products and services.

[0031] When faced with a new scenario, the present invention can integrate the user's long-term and short-term preferences to promptly determine the user's true preferences. At the same time, it fully utilizes the active dialogue capability of the voice assistant to explore the user's preferences. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 Schematic diagram of the present invention. DETAILED DESCRIPTION

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It will be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all structures.

[0034] Example 1:

[0035] The content recommendation method based on in-vehicle conversation of this embodiment includes the following steps:

[0036] S1. Obtain the voice message sent by the user.

[0037] S2. The language understanding module analyzes the user's emotional tendencies and / or user preferences contained in the voice information. When the emotional tendencies or user preferences are analyzed as negative, the recommendation module performs the recommendation operation; otherwise, the process switches to the dialogue strategy module.

[0038] Using natural language processing technology, the text content expressed by the user is passed through a text classification model to determine whether its emotional state is positive, negative or other.

[0039] S3. The dialogue strategy module decides whether to enter the next round of interaction. When the dialogue strategy module determines that there is no need for the next round of interaction, it enters the recommendation module to perform the recommendation operation; otherwise, it goes to S1 to enter the next round of interaction.

[0040] Through the above method, when the user's emotional tendency and user preference are positive or other, the dialogue strategy module determines whether it is necessary to enter the next round of interaction. If necessary, the user is guided to continue inputting voice information. If not, the module switches to the recommendation module to recommend products or services, such as songs, to the user.

[0041] The content of the conversation is based on a specific scenario. By analyzing the conversation content, building a knowledge graph for the corresponding scenario, and guiding the conversation based on prior knowledge, we can clarify the user's preferences in the specific scenario, making the conversation recommendation process more in line with the business scenario.

[0042] In the dialogue strategy, by improving the model structure in dialogue management, introducing more prior knowledge bases, and improving the structure of dialogue management, the dialogue strategy focuses more on the exploration process of user preferences, so as to obtain user preferences in as few dialogue rounds as possible.

[0043] Without affecting the user experience, gradually clarify the user's preferences through dialogue interaction.

[0044] When the user's emotional tendency or user preference is negative, it will directly go to the recommendation module to recommend products or services to the user. When the user expresses a negative emotional tendency, it can stop exploring user preferences in time and recommend items and services in time.

[0045] Example 2:

[0046] The content recommendation method based on in-vehicle conversation of this embodiment includes the following steps:

[0047] S1. Obtain the voice message sent by the user.

[0048] S2: The language understanding module analyzes the user's emotional tendencies and / or preferences contained in the voice information. If the emotional tendencies or user preferences are negative, the recommendation module makes a recommendation; otherwise, the process switches to the dialogue strategy module. The recommendation module uses the user profile constructed from the user's historical logs as a user feature matrix and an item matrix to construct a two-stage recommendation system based on recall and ranking.

[0049] S3. The dialogue strategy module decides whether to enter the next round of interaction. When the dialogue strategy module determines that there is no need for the next round of interaction, it enters the recommendation module to perform the recommendation operation; otherwise, it goes to S1 to enter the next round of interaction.

[0050] Without affecting the user experience, gradually clarify the user's preferences through dialogue interaction.

[0051] Through the above method, when the user's emotional tendency and user preference are positive or other, the dialogue strategy module determines whether it is necessary to enter the next round of interaction. If necessary, the user is guided to continue inputting voice information. If not, the module switches to the recommendation module to recommend products or services, such as songs, to the user.

[0052] Example 3:

[0053] The content recommendation method based on in-vehicle conversation of this embodiment includes the following steps:

[0054] S1. Obtain the voice message sent by the user.

[0055] S2. The language understanding module analyzes the user's emotional tendencies and / or preferences contained in the voice information. If the emotional tendencies or user preferences are negative, the recommendation module makes a recommendation; otherwise, the process switches to the dialogue strategy module. The recommendation module uses the user profile constructed from the user's historical logs as the user feature matrix and item matrix to construct a two-stage recommendation system based on recall and ranking. The recommendation module also constructs a feature matrix based on the user's preferences, using the current user input and short-term user logs as short-term preferences and the preferences in the user profile as long-term preferences.

[0056] S3. The dialogue strategy module decides whether to enter the next round of interaction. When the dialogue strategy module determines that there is no need for the next round of interaction, it enters the recommendation module to perform the recommendation operation; otherwise, it goes to S1 to enter the next round of interaction.

[0057] Without affecting the user experience, gradually clarify the user's preferences through dialogue interaction.

[0058] Through the above method, when the user's emotional tendency and user preference are positive or other, the dialogue strategy module determines whether it is necessary to enter the next round of interaction. If necessary, the user is guided to continue inputting voice information. If not, the module switches to the recommendation module to recommend products or services, such as songs, to the user.

[0059] Example 4:

[0060] The content recommendation method based on in-vehicle conversation of this embodiment includes the following steps:

[0061] S1. Obtain the voice message sent by the user.

[0062] S2. The language understanding module analyzes the user's emotional tendencies and / or preferences contained in the voice information. If the emotional tendencies or user preferences are negative, the recommendation module performs the recommendation; otherwise, the process switches to the dialogue strategy module. The recommendation module uses the user profile constructed from the user's historical logs as the user feature matrix and item matrix to construct a two-stage recommendation system based on recall and ranking. The recommendation module also uses the current user input and short-term user logs as short-term preferences, and the preferences in the user profile as long-term preferences. This weighted calculation calculates the user's preferences based on the short-term and long-term preferences.

[0063] S3. The dialogue strategy module decides whether to enter the next round of interaction. When the dialogue strategy module determines that there is no need for the next round of interaction, it enters the recommendation module to perform the recommendation operation; otherwise, it goes to S1 to enter the next round of interaction.

[0064] Without affecting the user experience, gradually clarify the user's preferences through dialogue interaction.

[0065] Through the above method, when the user's emotional tendency and user preference are positive or other, the dialogue strategy module determines whether it is necessary to enter the next round of interaction. If necessary, the user is guided to continue inputting voice information. If not, the module switches to the recommendation module to recommend products or services, such as songs, to the user.

[0066] Example 5:

[0067] The content recommendation method based on in-vehicle conversation of this embodiment includes the following steps:

[0068] S1. Obtain the voice message sent by the user.

[0069] S2: The language understanding module analyzes the user's emotional tendencies and / or preferences contained in the voice information. If the emotional tendencies or user preferences are negative, the recommendation module makes a recommendation; otherwise, the process switches to the dialogue strategy module. The recommendation module uses the user profile constructed from the user's historical logs as a user feature matrix and an item matrix to construct a two-stage recommendation system based on recall and ranking.

[0070] The recommendation module also uses the current user input and short-term user log as short-term preferences based on the preferences in the user's characteristics, and uses the preferences in the user portrait as long-term preferences, and performs weighted calculation of short-term preferences and long-term preferences to obtain user preferences.

[0071] Short-term user logs contain the types of songs a user has listened to, the duration of their listening, and the types of songs they haven't finished listening to. Preferably, the short-term period is no longer than seven days, for example, three days. This short-term user log can help us understand a user's recent preferences and facilitate targeted product or service recommendations.

[0072] S3. The dialogue strategy module decides whether to enter the next round of interaction. When the dialogue strategy module determines that there is no need for the next round of interaction, it enters the recommendation module to perform the recommendation operation; otherwise, it goes to S1 to enter the next round of interaction.

[0073] Without affecting the user experience, gradually clarify the user's preferences through dialogue interaction.

[0074] Through the above method, when the user's emotional tendency and user preference are positive or other, the dialogue strategy module determines whether it is necessary to enter the next round of interaction. If necessary, the user is guided to continue inputting voice information. If not, the module switches to the recommendation module to recommend products or services, such as songs, to the user.

[0075] Example 6:

[0076] The content recommendation method based on in-vehicle conversation of this embodiment includes the following steps:

[0077] S1. Obtain the voice message sent by the user.

[0078] S2: The language understanding module analyzes the user's emotional tendencies and / or preferences contained in the voice information. If the emotional tendencies or user preferences are negative, the recommendation module makes a recommendation; otherwise, the process switches to the dialogue strategy module. The recommendation module uses the user profile constructed from the user's historical logs as a user feature matrix and an item matrix to construct a two-stage recommendation system based on recall and ranking.

[0079] The recommendation module also uses the current user input and short-term user log as short-term preferences based on the preferences in the user's characteristics, and uses the preferences in the user portrait as long-term preferences, and performs weighted calculation of short-term preferences and long-term preferences to obtain user preferences.

[0080] Short-term user logs contain the types of songs a user has listened to, the duration of their listening, and the types of songs they haven't finished listening to. Preferably, the short-term period is no longer than seven days, for example, three days. This short-term user log can help us understand a user's recent preferences and facilitate targeted product or service recommendations.

[0081] As a preferred solution, the recommendation module uses attention weighted calculation.

[0082] S3. The dialogue strategy module decides whether to enter the next round of interaction. When the dialogue strategy module determines that there is no need for the next round of interaction, it enters the recommendation module to perform the recommendation operation; otherwise, it goes to S1 to enter the next round of interaction.

[0083] Through the above method, when the user's emotional tendency and user preference are positive or other, the dialogue strategy module determines whether it is necessary to enter the next round of interaction. If necessary, the user is guided to continue inputting voice information. If not, the module switches to the recommendation module to recommend products or services, such as songs, to the user.

[0084] Example 7:

[0085] See also Figure 1 The content recommendation method based on in-vehicle conversation of this embodiment includes the following steps:

[0086] S1. Obtain the voice message sent by the user.

[0087] S2. The language understanding module analyzes the user's emotional tendencies and / or user preferences contained in the voice information. When the emotional tendencies or user preferences are analyzed as negative, the recommendation module performs the recommendation operation; otherwise, the process switches to the dialogue strategy module.

[0088] S3. The dialogue strategy module decides whether to enter the next round of interaction. When the dialogue strategy module determines that there is no need for the next round of interaction, it enters the recommendation module to perform the recommendation operation; otherwise, it goes to S1 to enter the next round of interaction.

[0089] Through the above method, when the user's emotional tendency and user preference are positive or other, the dialogue strategy module determines whether it is necessary to enter the next round of interaction. If necessary, the user is guided to continue inputting voice information. If not, the module switches to the recommendation module to recommend products or services, such as songs, to the user.

[0090] The content of the conversation is based on a specific scenario. By analyzing the conversation content, building a knowledge graph for the corresponding scenario, and guiding the conversation based on prior knowledge, we can clarify the user's preferences in the specific scenario, making the conversation recommendation process more in line with the business scenario.

[0091] In the dialogue strategy, by improving the model structure in dialogue management, introducing more prior knowledge bases, and improving the structure of dialogue management, the dialogue strategy focuses more on the exploration process of user preferences, so as to obtain user preferences in as few dialogue rounds as possible.

[0092] The dialogue strategy module determines the next action based on the DQN network of each time state.

[0093] Without affecting the user experience, gradually clarify the user's preferences through dialogue interaction.

[0094] When the user's emotional tendency or user preference is negative, it will directly go to the recommendation module to recommend products or services to the user. When the user expresses a negative emotional tendency, it can stop exploring user preferences in time and recommend items and services in time.

[0095] When faced with a new scenario, the present invention can integrate the user's long-term and short-term preferences to promptly determine the user's true preferences. At the same time, it fully utilizes the active dialogue capability of the voice assistant to explore the user's preferences.

[0096] Example 8:

[0097] This embodiment provides a content recommendation device based on in-vehicle conversation, including a processor and a memory.

[0098] The processor may be a multi-core processor or may include multiple processors. In some embodiments, the processor may include a general-purpose main processor and one or more special coprocessors, such as a graphics processing unit (GPU), a digital signal processor (DSP), etc. In some embodiments, the processor may be implemented using customized circuits, such as application-specific integrated circuits (ASICs) or field-programmable gate arrays (FPGAs).

[0099] The memory may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage devices. ROM may store static data or instructions required by the processor or other modules of the computer. The permanent storage device may be a readable and writable storage device. The permanent storage device may be a non-volatile storage device that does not lose stored instructions and data even if the computer is powered off. In some embodiments, the permanent storage device uses a large-capacity storage device (such as a magnetic or optical disk, flash memory) as the permanent storage device. In other embodiments, the permanent storage device may be a removable storage device (such as a floppy disk, optical drive). The system memory may be a readable and writable storage device or a volatile readable and writable storage device, such as dynamic random access memory. The system memory may store some or all instructions and data required by the processor during operation. In addition, the memory may include any combination of computer-readable storage media, including various types of semiconductor memory chips (DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and magnetic disks and / or optical disks may also be used. In some embodiments, the memory may include a readable and / or writable removable storage device, such as a compact disc (CD), a read-only digital versatile disc (e.g., DVD-ROM, dual-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (e.g., SD card, mini SD card, Micro-SD card, etc.), a magnetic floppy disk, etc. Computer-readable storage media do not include carrier waves and transient electronic signals transmitted wirelessly or wired.

[0100] The memory stores executable codes, and when the executable codes are executed by the processor, the processor can execute the method described in the above embodiment.

[0101] Example 9:

[0102] This embodiment provides a non-transitory machine-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor is caused to execute the above method.

[0103] A non-transitory machine-readable storage medium (or computer-readable storage medium, or machine-readable storage medium) having executable code (or computer program, or computer instruction code) stored thereon, which, when executed by a processor of an electronic device (or computing device, server, etc.), enables the processor to perform the various steps of the above-mentioned method according to the present invention.

[0104] Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the disclosure herein may be implemented as electronic hardware, computer software, or combinations of both.

[0105] The flowcharts and block diagrams in the accompanying drawings show the possible implementation architecture, functions and operations of the systems and methods according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment or part of code, and the part of the module, program segment or code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0106] For those skilled in the art, any modifications and changes made according to the above-mentioned embodiments of the present invention should be included in the scope of protection of the present invention without departing from the spirit of the present invention.

Claims

1. A content recommendation method based on in-vehicle conversation, characterized in that The following steps are involved: S1, obtain the voice information sent by the user; S2, performing a user sentiment tendency analysis and a user preference analysis on the voice information through a language understanding module; when the analysis result shows that the voice information contains a negative sentiment bias, a recommendation module performs a recommendation operation; Otherwise, go to the dialogue strategy module; S3, the dialogue strategy module decides whether to enter the next round of interaction. When the dialogue strategy module determines that there is no need for further interaction, it enters the recommendation module to perform the recommendation operation; otherwise, it goes to S1 to enter the next round of interaction.

2. The method for content recommendation based on in-vehicle conversation according to claim 1, characterized in that: The recommendation module uses the user profile constructed based on the user's historical log as the user feature matrix and item matrix to build a two-stage recommendation system based on recall and ranking.

3. The method for content recommendation based on in-vehicle conversation according to claim 2, characterized in that: The recommendation module also constructs a feature matrix based on the preferences in the user's features, taking the current user input and short-term user logs as short-term preferences, and the preferences in the user portrait as long-term preferences.

4. The method for content recommendation based on in-vehicle conversation according to claim 2, characterized in that: The recommendation module also uses the current user input and short-term user log as short-term preferences based on the preferences in the user's characteristics, and uses the preferences in the user portrait as long-term preferences, and performs weighted calculation of short-term preferences and long-term preferences to obtain user preferences.

5. The method for content recommendation based on in-vehicle conversation according to claim 4, characterized in that: The short-term user log includes the types of songs that the user has listened to in a short period of time, the duration of listening to the songs, and the types of songs that the user has not finished listening to.

6. The method for content recommendation based on in-vehicle conversation according to claim 5, characterized in that: The short-term duration shall not exceed 7 days.

7. The method for content recommendation based on in-vehicle conversation according to claim 6, characterized in that: The recommendation module uses attention weighted calculation.

8. The method for content recommendation based on in-vehicle conversation according to claim 1, characterized in that: The dialogue strategy module determines the next action based on the DQN network of each time state.

9. A content recommendation device based on in-vehicle conversation, characterized by: The method comprises a processor and a memory, wherein the memory stores executable codes, and when the executable codes are executed by the processor, the processor is caused to perform the method according to any one of claims 1 to 8.

10. A non-transitory machine-readable storage medium, characterized in that: An executable code is stored thereon, and when the executable code is executed by a processor of an electronic device, the processor is caused to execute the method according to any one of claims 1 to 8.

Images