Interaction method and system of live broadcast room

Abstract

The invention discloses a live broadcast room interaction method and system. The method comprises the following steps: obtaining display information from a target live broadcast room; obtaining a consultation comment model essay; generating a corresponding consultation type virtual comment based on the display information and the consultation comment model essay; judging the authenticity of the virtual comments, and issuing the virtual comments which are judged to be true; the display information comprises effective explanation sentences, the effective explanation sentences are explanation sentences reflecting explanation types, and when the effective explanation sentences are recognized, corresponding consultation type virtual comments are generated based on the effective explanation sentences and the consultation comment model essays. Aiming at the content explained by the anchor, the consultation type virtual comments matched with the content are generated in time and published, and further explanation is performed for the user who has consultation intention but is inconvenient for active consultation.

Description

Technical Field

[0001] The present invention relates to the field of electrical communications, and in particular to an interactive method and system for a live broadcast room. Background Art

[0002] Online shopping in live broadcast rooms has become one of the main ways for the public to shop. The anchor displays the products in the live broadcast room and gives targeted explanations based on the comments of the audience;

[0003] However, there are cases where users have the intention to consult but do not actively consult. In a scenario with fewer viewers and less interaction, the anchor cannot give targeted explanations and can only give a general display of the products, resulting in poor live broadcast effects. Summary of the invention

[0004] In view of the situation that there are users in the live broadcast scene who have the intention to consult but do not actively consult, the present invention provides an interactive method and system for a live broadcast room that automatically publishes consulting virtual comments according to the display information corresponding to the current live broadcast room. The method can actively start from the scene that the user wants to know about and make comments and ask questions, which is convenient for the anchor to give targeted explanations, thereby facilitating users who have the intention to consult but are inconvenient to actively consult to further understand product information.

[0005] In order to solve the above technical problems, the present invention is solved by the following technical solutions:

[0006] A live broadcast room interaction method comprises the following steps:

[0007] Get display information from the target live broadcast room;

[0008] Get consulting review sample;

[0009] Generate corresponding consulting virtual reviews based on the display information and the consulting review sample;

[0010] Determining the authenticity of the virtual review and publishing the virtual review that is determined to be true;

[0011] The display information includes effective explanation sentences, which are explanation sentences that reflect the explanation type. When an effective explanation sentence is identified, a corresponding consulting virtual comment is generated based on the effective explanation sentence and the consulting comment model.

[0012] As an implementable method:

[0013] Pre-configure a number of type tags, wherein the type tags correspond one to one with the explanation types;

[0014] Perform voice conversion on the audio data of the target live broadcast room to obtain several explanation sentences;

[0015] Perform label matching on the explanation sentence and each type of label;

[0016] When a type label matching the explanation sentence is obtained, the explanation sentence is taken as a valid explanation sentence.

[0017] As an implementable method, the explanation types include product introduction, discount information, brand endorsement, benefit announcement, specification description, price analysis and attention to membership.

[0018] As an implementable method, the display information also includes product information of the product currently being explained;

[0019] According to a preset pulling cycle, the product information is pulled, and corresponding consulting virtual reviews are generated based on the product information and the consulting review template.

[0020] The present invention can also generate corresponding virtual comments based on the product information, which can not only actively ask questions about the content explained by the anchor, but also actively ask questions based on the product information.

[0021] As an implementable method, the method of obtaining consulting review sample includes:

[0022] Obtain comments posted by users in the target live broadcast room on the currently explained product, obtain candidate comments, and screen the candidate comments to obtain a first comment sample;

[0023] When the number of the first comment samples is greater than or equal to a preset text quantity threshold, extracting the consulting comment model text from the first comment samples;

[0024] When the number of the first comment samples is less than a preset text quantity threshold, a second comment sample is obtained, and the consulting comment model is extracted from the first comment sample and the second comment sample, wherein the second comment sample is the first comment sample corresponding to the previously explained product.

[0025] By judging the amount of text, the present invention can introduce historical comments when there are fewer comments published on the current product, thereby ensuring that rich virtual comments can still be generated when the first comment sample is small or even missing.

[0026] As an implementable method, the method of obtaining candidate reviews includes:

[0027] Pulling comments in the target live broadcast room to obtain a number of comment data, wherein the comment data includes comment time and comment text;

[0028] Get the start time of the explanation corresponding to the current explanation product;

[0029] Based on the explanation start time and the comment time, the comment text corresponding to the currently explained product is screened to obtain corresponding candidate comments.

[0030] As an implementable method, the method for obtaining the first comment sample includes:

[0031] The comments with consultation intention are extracted from the obtained candidate comments to obtain the first comment sample.

[0032] Further:

[0033] Screening the candidate comments based on a preset first keyword library to obtain valid comments, wherein the first keyword library includes a plurality of screening words;

[0034] The comments with consultation intention are extracted from the obtained valid comments to obtain the first comment sample.

[0035] The present invention also provides an interactive system for a live broadcast room, comprising:

[0036] The extraction module is used to obtain display information from the target live broadcast room and also used to obtain consulting review samples;

[0037] A generator, used for generating corresponding virtual consulting reviews based on the display information and the consulting review sample;

[0038] A discriminator, used for judging the authenticity of the virtual review and publishing the virtual review judged to be true;

[0039] The display information includes effective explanation sentences, which are explanation sentences that reflect the explanation type. The generator is used to generate corresponding consulting virtual comments based on the effective explanation sentences and the consulting comment model when the extraction module recognizes the effective explanation sentences.

[0040] As an implementable method:

[0041] The display information also includes product information of the product currently being explained;

[0042] When the extraction module obtains the product information based on a preset pulling cycle, the generator generates corresponding consulting virtual reviews based on the product information and the consulting review template.

[0043] The present invention has significant technical effects due to the adoption of the above technical solution:

[0044] The present invention identifies effective explanation sentences during the host's explanation process. When an effective explanation sentence is identified, a consulting virtual comment is generated based on the effective explanation sentence and a consulting comment model, and the authenticity of the generated virtual comment is judged. Only the virtual comments judged to be true are released, so as to automatically release virtual comments that fit the current scene to the target live broadcast room, and actively make comments and ask questions based on the scene that the user wants to know, so as to facilitate users who have the intention to consult but are inconvenient to actively consult to further understand the product information. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 It is a flow chart of an interactive method for a live broadcast room of the present invention;

[0047] Figure 2 It is a flowchart for obtaining display information;

[0048] Figure 3 It is a flowchart for obtaining consulting review samples. DETAILED DESCRIPTION

[0049] The present invention is further described in detail below in conjunction with embodiments. The following embodiments are for explanation of the present invention but the present invention is not limited to the following embodiments.

[0050] A method for interactive broadcast room, referring to Figure 1 , including the following steps:

[0051] S100, obtaining display information from the target live broadcast room;

[0052] Reference Figure 2 The display information includes a valid explanation sentence or product information of the product currently being explained.

[0053] S110. The method for obtaining an effective explanation sentence is specifically as follows:

[0054] S111, performing voice conversion on the audio data of the target live broadcast room to obtain a number of explanation sentences;

[0055] In this embodiment, the audio stream of the target live broadcast room is used as audio data; voice conversion is performed based on the audio stream to obtain the text information of the host's explanation in real time, thereby obtaining the corresponding explanation sentences.

[0056] S112, performing tag matching on the explanation sentence and various pre-configured type tags, and when a type tag matching the explanation sentence is obtained, taking the explanation sentence as a valid explanation sentence.

[0057] In this embodiment, a number of type tags are preconfigured, and the type tags correspond to the explanation types one by one;

[0058] For example, explanation types may include product introduction, discount information, brand endorsement, benefit notice, specification description, price analysis and membership attention, etc. Relevant staff can configure explanation types and type labels according to actual needs.

[0059] Those skilled in the art can use any publicly available label classification technology to identify the type of each explanation sentence. For example, a corresponding keyword thesaurus can be set for each type of label, and the explanation sentence can be matched with the corresponding keyword thesaurus to identify its explanation type.

[0060] Most of the explanation sentences cannot reflect the explanation type, that is, the explanation sentence has no type label matching it. In this embodiment, such explanation sentences are regarded as ordinary explanation sentences;

[0061] For an explanation sentence that can reflect the type of explanation, that is, the explanation sentence successfully matches the type label corresponding to it, in this embodiment, such an explanation sentence is regarded as a valid explanation sentence.

[0062] S120: The method for obtaining product information is as follows:

[0063] Based on the pre-configured pulling cycle, directly pull the relevant information of the current product from the target live broadcast room to obtain the corresponding product information;

[0064] Those skilled in the art may determine the scope of product information to be pulled according to actual needs, for example, the delivery information, parameter information, and / or introduction information corresponding to the product currently being explained may be pulled as product information;

[0065] Shipping information includes the place of shipment, whether shipping is free, delivery time, etc.

[0066] Parameter information, i.e. the product parameters corresponding to the product, such as model, specification, material, etc.;

[0067] Introduction information is used to reflect the product's functions, effects and other features. For example, the product details can be pulled to obtain the corresponding introduction information.

[0068] The relevant staff configures the pulling cycle based on the desired frequency of publishing of virtual reviews, so as to automatically generate and publish corresponding virtual reviews based on the product information according to the pulling cycle.

[0069] S200, obtain consulting review sample;

[0070] The consulting review template is a basic template used to expand virtual reviews based on AI technology.

[0071] When the display information is obtained, that is, when the corresponding valid explanation sentence is obtained based on the host's explanation voice recognition or the product information of the current explained product is obtained based on the preset pull cycle, the step of obtaining the consultation review sample is triggered, refer to Figure 3 The specific acquisition method includes the following steps:

[0072] S210, obtaining comments posted by users in the target live broadcast room on the currently explained product, obtaining candidate comments, and screening the candidate comments to obtain a first comment sample;

[0073] Specifically:

[0074] S211, pulling comments in the target live broadcast room to obtain a number of comment data, wherein the comment data includes comment time and comment text;

[0075] S212, obtaining the explanation start time corresponding to the currently explained product;

[0076] S213, based on the explanation start time and the comment time, filter and obtain the comment text corresponding to the currently explained product to obtain corresponding candidate comments;

[0077] That is, the comment texts whose comment time is within the time range from the start time of the explanation to the current time are regarded as candidate comments;

[0078] S214: Filter and obtain a first comment sample from the candidate comments.

[0079] For example, if the user is having a meal, or taking a bus or subway, it is inconvenient for him to operate his mobile phone and he is unable to post comments or inquiries in time. In this embodiment, the first sample comments screened and obtained are consulting comments, such as inquiries on product specifications, discount information, delivery time, etc., so that when the corresponding consulting virtual comments are generated later, they can actively start from the scenarios that the users want to know and actively ask comments and questions, so that such viewers who are not convenient to comment or inquire can further understand the relevant information of the products, retain the users who intend to watch and continue to watch the product explanation, and increase the product transaction rate.

[0080] As an example:

[0081] The comments with consultation intention are extracted from the obtained candidate comments to obtain the first comment sample.

[0082] That is, the comments with consulting intentions are screened out as the first comment sample.

[0083] Those skilled in the art may use existing publicly available intent recognition technologies to screen the candidate comments. For example, a second keyword library with several intention words may be preset, and the obtained candidate comments may be screened based on the preset second keyword library. A neural network model may also be used for intent recognition, and this specification does not impose any detailed restrictions on this.

[0084] As another example:

[0085] Screening the candidate comments based on a preset first keyword library to obtain valid comments, wherein the first keyword library includes a plurality of screening words;

[0086] The comments with consultation intention are extracted from the obtained valid comments to obtain the first comment sample.

[0087] Those skilled in the art may set the filter words according to actual needs, for example, the filter words may be set based on negative reviews, and this specification does not limit this in detail.

[0088] S220: when the number of the first comment samples is greater than or equal to a preset text quantity threshold, extracting the consulting comment model text from the first comment samples;

[0089] That is, a corpus is constructed based on the first comment sample, and one or more comments are extracted from the corpus based on pre-configured data as consulting comment model articles to provide a simulation basis for the generation of subsequent virtual comments.

[0090] S230. When the number of the first comment samples is less than a preset text quantity threshold, obtain a second comment sample, and extract the consulting comment model from the first comment sample and the second comment sample, wherein the second comment sample is the first comment sample corresponding to the previously explained product.

[0091] That is, when the amount of data of the first comment sample is small, the historical comment samples are supplemented to jointly construct the corpus. When the current product is just being explained or the number of comments is small, there are enough samples to ensure that the generated virtual comments are diversified, not monotonous and rigid, and more realistic.

[0092] S300, generating corresponding virtual consulting reviews based on the display information and the consulting review sample;

[0093] That is, after obtaining the effective explanation sentence or product information, a corresponding virtual review is generated based on the effective explanation sentence or product information and the consulting review sample;

[0094] S310: When a valid explanation sentence is identified, a corresponding consulting virtual comment is generated based on the valid explanation sentence and the consulting comment model.

[0095] Based on GPT technology (natural language processing technology based on artificial intelligence), a number of virtual comments corresponding to the effective explanation sentences and consulting comment model essays are generated.

[0096] For example, when the explanation type corresponding to the effective explanation sentence is discount information, virtual comments such as asking about discounts can be generated based on the language style of the consultation review sample and GPT technology;

[0097] Since the content explained by the anchor focuses on the selling points, user groups, usage scenarios, specifications, prices and other aspects of the product, it covers a wide range of user problem scenarios and is easier to generate consulting virtual comments that fit the audience's real thoughts;

[0098] This embodiment generates virtual comments based on effective explanation sentences, which is applicable to various live broadcast scenarios. Even if there are few users in the live broadcast room, the user comments are single, or there are few comments, rich and diverse consulting virtual comments can be generated.

[0099] S320: When the product information corresponding to the product currently being explained is extracted, a corresponding consulting virtual review is generated based on the product information and the consulting review sample.

[0100] Based on the product information, it can automatically expand user comments on product parameter inquiries, and can also expand comments on product delivery time inquiries, etc.

[0101] S400, determining the authenticity of the virtual review, and publishing the virtual review determined to be true;

[0102] This embodiment verifies the rationality of the generated virtual reviews based on the generative adversarial network to eliminate unreasonable virtual reviews and make the published virtual reviews more realistic.

[0103] Generative adversarial networks include generative models and discriminative models;

[0104] This embodiment adopts the GPT model as the generation model, that is, the display message and the consultation comment sample are used as the input of the generation network, and the virtual comment is the output of the generation model.

[0105] The input of the discriminant model is the virtual review output by the generative model, and the output is the probability that the virtual review is true. When the probability reaches a preset discrimination threshold, the virtual review is judged to be true, otherwise it is false. Technical personnel in this field can select an existing publicly available discriminant model according to actual needs, and this specification does not impose detailed restrictions on the discriminant model.

[0106] In practical applications, model training is performed based on the existing publicly available training method of generative adversarial networks, specifically:

[0107] First, the discriminant model is preliminarily trained using the collected real review samples to learn the characteristics of the real review samples;

[0108] After that, adversarial training is performed, that is, the discriminant model is trained based on real review samples and virtual reviews generated by the generative model, and the parameters of the generative model are updated based on the discriminator's judgment of the virtual reviews. This adversarial process will be repeated continuously, and the generative model and the discriminative model will continue to improve in the competition. As the training progresses, the virtual reviews generated by the generator become more and more realistic and reasonable, and the discriminator becomes more sensitive and accurate.

[0109] In practical applications, several interactive accounts can be pre-built, and when a virtual comment that is judged to be true is obtained, the virtual comment is published by a random interactive account in the target live broadcast room.

[0110] As another embodiment, a live broadcast room interactive system is also provided, including:

[0111] The extraction module is used to obtain display information from the target live broadcast room and also used to obtain consulting review samples;

[0112] A generator, used for generating corresponding virtual consulting reviews based on the display information and the consulting review sample;

[0113] A discriminator, used for judging the authenticity of the virtual review and publishing the virtual review judged to be true;

[0114] The display information includes effective explanation sentences, which are explanation sentences that reflect the explanation type. The generator is used to generate corresponding consulting virtual comments based on the effective explanation sentences and the consulting comment model when the extraction module recognizes the effective explanation sentences.

[0115] Further:

[0116] The display information also includes product information of the product currently being explained;

[0117] When the extraction module obtains the product information based on a preset pulling cycle, the generator generates corresponding consulting virtual reviews based on the product information and the consulting review template.

[0118] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0119] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0120] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0121] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, and the processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions.

[0122] These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the process Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0123] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0124] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0125] It should be noted that:

[0126] The "one embodiment" or "embodiment" mentioned in the specification means that the specific features, structures or characteristics described in conjunction with the embodiment are included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment" or "embodiment" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0127] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0128] In addition, it should be noted that the shapes and names of the parts and components of the specific embodiments described in this specification may be different. Any equivalent or simple changes made based on the structure, features and principles described in the patent concept of the present invention are included in the protection scope of the patent of the present invention. The technicians in the technical field of the present invention can make various modifications or supplements to the specific embodiments described or replace them in a similar manner, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A method for interactive broadcasting, characterized in that: The following steps are involved: Get display information from the target live broadcast room; Get consulting review sample; Generate corresponding consulting virtual reviews based on the display information and the consulting review sample; Determining the authenticity of the virtual review and publishing the virtual review that is determined to be true; The display information includes effective explanation sentences, which are explanation sentences that reflect the explanation type. When an effective explanation sentence is identified, a corresponding consulting virtual comment is generated based on the effective explanation sentence and the consulting comment model.

2. The interactive method of the live broadcast room according to claim 1, characterized in that: Pre-configure a number of type tags, wherein the type tags correspond one to one with the explanation types; Perform voice conversion on the audio data of the target live broadcast room to obtain several explanation sentences; Perform label matching on the explanation sentence and each type of label; When a type label matching the explanation sentence is obtained, the explanation sentence is taken as a valid explanation sentence.

3. The interactive method of the live broadcast room according to claim 2, characterized in that: The types of explanations include product introduction, discount information, brand endorsement, benefit preview, specification description, price analysis and membership interest.

4. The interactive method of the live broadcast room according to claim 1, characterized in that: The display information also includes product information of the product currently being explained; According to a preset pulling cycle, the product information is pulled, and corresponding consulting virtual reviews are generated based on the product information and the consulting review template.

5. The interactive method of a live broadcast room according to any one of claims 1 to 4, characterized in that: Methods for obtaining consulting review samples include: Obtain comments posted by users in the target live broadcast room on the currently explained product, obtain candidate comments, and screen the candidate comments to obtain a first comment sample; When the number of the first comment samples is greater than or equal to a preset text quantity threshold, extracting the consulting comment model text from the first comment samples; When the number of the first comment samples is less than a preset text quantity threshold, a second comment sample is obtained, and the consulting comment model is extracted from the first comment sample and the second comment sample, wherein the second comment sample is the first comment sample corresponding to the previously explained product.

6. The interactive method of the live broadcast room according to claim 5, characterized in that: Pulling comments in the target live broadcast room to obtain a number of comment data, wherein the comment data includes comment time and comment text; Get the start time of the explanation corresponding to the current explanation product; Based on the explanation start time and the comment time, the comment text corresponding to the currently explained product is screened to obtain corresponding candidate comments.

7. The interactive method of the live broadcast room according to claim 6, characterized in that: The comments with consultation intention are extracted from the obtained candidate comments to obtain the first comment sample.

8. The interactive method of the live broadcast room according to claim 7, characterized in that: Screening the candidate comments based on a preset first keyword library to obtain valid comments, wherein the first keyword library includes a plurality of screening words; The comments with consultation intention are extracted from the obtained valid comments to obtain the first comment sample.

9. An interactive system for a live broadcast room, characterized in that: include: The extraction module is used to obtain display information from the target live broadcast room and also used to obtain consulting review samples; A generator, used for generating corresponding virtual consulting reviews based on the display information and the consulting review sample, wherein the virtual reviews are consulting reviews; A discriminator, used for judging the authenticity of the virtual review and publishing the virtual review judged to be true; The display information includes effective explanation sentences, which are explanation sentences that reflect the explanation type. The generator is used to generate corresponding consulting virtual comments based on the effective explanation sentences and the consulting comment model when the extraction module recognizes the effective explanation sentences.

10. The interactive system of the live broadcast room according to claim 9, characterized in that: The display information also includes product information of the product currently being explained; When the extraction module obtains the product information based on a preset pulling cycle, the generator generates corresponding consulting virtual reviews based on the product information and the consulting review template.

Images