A method and system for tracking network live broadcast freezes

Abstract

An embodiment of the present invention provides a method for tracking freezes in live streaming, relating to the field of live streaming. The method comprises: obtaining, in real time, comment information posted by each audience user of the live streaming during a target user's live streaming, and constructing a user corpus based on the comment information of each audience user; determining first comment information within the user corpus that contains freeze semantics, and obtaining a freeze information library based on each first comment information; performing vector processing on each first comment information within the freeze information library to obtain a corresponding freeze vector set, inputting the freeze vector set into a naive Bayes classifier, and outputting the freeze type corresponding to the target user. Comment information is an existing resource, and fully utilizing existing resources does not require consuming large amounts of data or high storage costs, resulting in low overall costs.

Description

Technical Field

[0001] The present invention relates to the field of live broadcasting, and in particular to a method and system for tracking freezes in online live broadcasting. Background Art

[0002] With the rapid development of network communications technology, the widespread adoption of 4G, and the anticipated application of 5G, short videos and live streaming are poised for rapid growth. The rapid growth of online short videos on mobile devices has significantly boosted the internet economy. In recent years, at many major events and emergencies, media outlets have utilized short videos, aerial photography, and other tools to create new media content, enhancing the impact of their reporting and dissemination. Mobile networks and mobile phones have become the primary means of online video. Users are developing the habit of watching live streams over 4G, and live streaming will have a vast market for traffic. More and more users are shifting from images and text to short videos and live streaming.

[0003] The characteristics of live streaming scenarios are mainly reflected in the following aspects: First, with the continuous development of live streaming services, users' requirements for the live streaming experience are becoming increasingly higher, requiring detailed crowd optimization. Second, many live streaming services are currently mainly aimed at ordinary people. While the rich and realistic scenes bring some problems, such as complex user network conditions. Third, the large user base and huge live streaming traffic require the use of multiple CDN providers to ensure business stability, which also brings management and business complexity. Fourth, the requirements for live streaming vary from scenario to scenario, often facing conflicting choices such as clarity or smoothness, instant first screen opening or low latency. These business characteristics lead to diverse experience issues, long demand coordination cycles between different CDNs, and complex and changing network environments.

[0004] Conventional live streaming quality detection systems rely on a large number of user data points to collect and report DC traffic problems. This method is too simple and the cost is relatively high. The collection and storage of a large amount of end-point data consumes a lot of server costs and requires the creation of hundreds of GB of storage indexes every day, which greatly increases the stability requirements of the quality detection system. Summary of the Invention

[0005] The embodiments of the present invention provide a method and system for tracking freezes in online live broadcasts, which can solve the above-mentioned technical problems existing in the prior art.

[0006] To achieve the above-mentioned objectives, on the one hand, an embodiment of the present invention provides a method for tracking freezes in a live broadcast, comprising:

[0007] During the live broadcast of the target user, the comment information posted by each audience user of the live broadcast is obtained in real time, and a user corpus is constructed based on the comment information of each audience user;

[0008] Determining first comment information containing lag semantics in the user corpus, and obtaining a lag information database based on each of the first comment information;

[0009] Each first comment information in the jam information library is vectorized to obtain a corresponding jam vector set, the jam vector set is input into a naive Bayes classifier, and the jam type corresponding to the target user is output.

[0010] On the other hand, an embodiment of the present invention provides a system for tracking network live broadcast freezes, comprising:

[0011] A user corpus construction unit is used to obtain, in real time, comment information posted by each audience user of the live broadcast during the live broadcast of the target user, and to construct a user corpus based on the comment information of each audience user;

[0012] a jam information database construction unit, configured to determine first comment information containing jam semantics in the user corpus, and obtain a jam information database based on each of the first comment information;

[0013] The jam type determination unit is used to perform vector processing on each first comment information in the jam information library to obtain a corresponding jam vector set, input the jam vector set into a naive Bayes classifier, and output the jam type corresponding to the target user.

[0014] The above technical solution has the following beneficial effects: during the live broadcast of the target user, the comment information posted by each audience user of the live broadcast is obtained in real time, and a user corpus is constructed based on the comment information of each audience user; the comment information is an existing resource, and making full use of the existing resources does not require a large amount of data or high storage costs, and the total cost is low. Determine the first comment information containing the semantics of lag in the user corpus, and obtain a lag information library based on each of the first comment information; perform vector processing on each of the first comment information in the lag information library to obtain a corresponding lag vector set, input the lag vector set into the naive Bayes classifier, and output the lag type corresponding to the target user. Through the comment information of the audience of the live broadcast, the lag problems of different target users can be discovered in the first place, and relatively accurate lag types can be obtained, so as to timely discover the playback quality problems in the live stream, so that efficient responses can be made to solve the problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 any creative work.

[0016] Figure 1 This is a flow chart of a method for tracking freezes in a live network broadcast according to an embodiment of the present invention;

[0017] Figure 2 It is a structural diagram of a system for tracking network live broadcast freezes according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, in combination with an embodiment of the present invention, a method for tracking network live broadcast freezes is provided, comprising:

[0020] S101: During a live broadcast of a target user, obtaining comment information posted by each audience user of the live broadcast in real time, and constructing a user corpus based on the comment information of each audience user;

[0021] S102: Determine first comment information containing lag semantics in the user corpus, and obtain a lag information database based on each of the first comment information;

[0022] S103: Perform vector processing on each first comment information in the jam information database to obtain a corresponding jam vector set, input the jam vector set into a naive Bayes classifier, and output the jam type corresponding to the target user.

[0023] Preferably, S101: during the live broadcast of the target user, the comment information posted by each audience user of the live broadcast is obtained in real time, and a user corpus is constructed based on the comment information of each audience user, specifically including:

[0024] Accessing the real-time comment information stream of the target user who initiated the live broadcast through a message queue, obtaining the comment information posted by each audience user of the live broadcast from the real-time comment information stream, and constructing a user corpus based on the comment information of each audience user;

[0025] S102: Determining first comment information containing lag semantics in the user corpus, and obtaining a lag information database based on each of the first comment information, specifically includes:

[0026] For each comment information in the user corpus, based on the text content of the comment information and the characteristics of the target user, determine whether there are words containing lag semantics in the text content of the comment information, and determine the comment information containing words containing lag semantics in the text content as the first comment information. The lag information library is obtained based on each of the first comment information.

[0027] Preferably, S103: performing vector processing on each first comment information in the jam information library to obtain a corresponding jam vector set specifically includes:

[0028] S1031: For each first comment information in the jam information library, segment the text content of each first comment information using a text semantic segmentation method based on a preset proprietary dictionary set to obtain jam keywords corresponding to each first comment information, and construct the jam keywords corresponding to each first comment information into a text word vector, and form all the text word vectors into a jam vector set; the proprietary dictionary set includes at least one jam type and jam keywords containing jam semantics corresponding to each jam type.

[0029] Preferably, it also includes:

[0030] S104: Matching the jamming type corresponding to the target user with jamming keywords of the jamming type in a preset proprietary dictionary, and generating jamming recommendation information corresponding to the target user based on the matched jamming keywords;

[0031] S105: Displaying the jam recommendation information corresponding to the target user in the information display area provided by the live broadcast, so as to be displayed to each audience user of the target user; and

[0032] S106: Adding the jam recommendation information corresponding to the target user to the information recommendation vocabulary of the target user.

[0033] Preferably, S107: the method for constructing the preset proprietary dictionary set includes:

[0034] Capture search information of a specified online live broadcast platform and use the search information as keywords;

[0035] Determine the jam keywords containing jam semantics in each keyword, and classify the jam keywords according to jam types to obtain the proprietary dictionary set.

[0036] like Figure 2 As shown, in combination with an embodiment of the present invention, a system for tracking network live broadcast freezes is provided, comprising:

[0037] The user corpus construction unit 21 is used to obtain the comment information posted by each audience user of the live broadcast in real time during the live broadcast of the target user, and construct a user corpus based on the comment information of each audience user;

[0038] A jam information database construction unit 22 is configured to determine first comment information containing jam semantics in the user corpus, and obtain a jam information database based on each of the first comment information;

[0039] The jam type determination unit 23 performs vector processing on each first comment information in the jam information library to obtain a corresponding jam vector set, inputs the jam vector set into a naive Bayes classifier, and outputs the jam type corresponding to the target user.

[0040] Preferably, the user corpus construction unit 21 is specifically configured to:

[0041] Accessing the real-time comment information stream of the target user who initiated the live broadcast through a message queue, obtaining the comment information posted by each audience user of the live broadcast from the real-time comment information stream, and constructing a user corpus based on the comment information of each audience user;

[0042] The Caton information library construction unit 22 is specifically used to:

[0043] For each comment information in the user corpus, based on the text content of the comment information and the characteristics of the target user, determine whether there are words containing lag semantics in the text content of the comment information, and determine the comment information containing words containing lag semantics in the text content as the first comment information. The lag information library is obtained based on each of the first comment information.

[0044] Preferably, the freeze type determination unit 23 includes:

[0045] A jam vector set construction subunit is used to segment the text content of each first comment information in the jam information library using a text semantic segmentation method based on a preset proprietary dictionary set, obtain the jam keywords corresponding to each first comment information, and construct the jam keywords corresponding to each first comment information into a text word vector, and form all the text word vectors into a jam vector set; the proprietary dictionary set includes at least one jam type and jam keywords with jam semantics corresponding to each jam type.

[0046] Preferably, it also includes:

[0047] a jam recommendation information generating unit, configured to match the jam type corresponding to the target user with jam keywords of the jam type in a preset proprietary dictionary set, and generate jam recommendation information corresponding to the target user based on the matched jam keywords;

[0048] A jam recommendation information display unit, configured to display the jam recommendation information corresponding to the target user in the information display area provided by the live broadcast, so as to display it to each audience user of the target user;

[0049] An information recommendation vocabulary library is used to receive and store the jam recommendation information generated by the jam recommendation information generation unit; the information recommendation vocabulary library is set on the target user side.

[0050] Preferably, the tracking system for network live broadcast freezes also includes:

[0051] The proprietary dictionary set construction unit is used to: capture the search information of the specified online live broadcast platform and use the search information as keywords; determine the jamming keywords containing jamming semantics in each keyword, and classify the jamming keywords according to the jamming type to obtain a preset proprietary dictionary set.

[0052] The beneficial technical effects achieved by the embodiments of the present invention are as follows:

[0053] By leveraging livestream audience comments, we can immediately identify lag issues for different target users and accurately derive the type of lag. This allows for timely detection of quality issues within livestreams, enabling efficient response and resolution, while covering a wide range of lag scenarios. Furthermore, the commentary information leverages existing resources, eliminating the need for large amounts of data or high storage costs, resulting in low overall costs and the ability to reuse them.

[0054] The above technical solutions of the embodiments of the present invention are described in detail below with reference to specific application examples. For technical details not introduced during the implementation process, please refer to the relevant description above.

[0055] The present invention provides a method and system for tracking live streaming freezes. Based on the viewer's end-user experience, this method obtains commentary from the target user's live streaming audience, solving the problem of determining the live streaming quality of the target user in real time. The detailed scheme based on the CDN network is designed as follows:

[0056] 1. User Review Information Collection

[0057] During the live broadcast of the target user, the comment information posted by each audience user of the live broadcast is obtained in real time, and a user corpus is constructed based on the comment information of each audience user; specifically, the real-time comment information stream of the target user who initiates the live broadcast is accessed through a message queue, that is, the target user information stream is accessed, and the comment information (comment text) posted by each audience user of the live broadcast is obtained from the real-time comment information stream, and a user corpus is constructed based on the comment information of each audience user; wherein the real-time comment information stream is in an information stream user comment recommendation system based on text recommendation.

[0058] 2. Collection of Stuttering Text Content

[0059] Determine the first comment information containing the semantics of lag in the user corpus, and obtain the lag information library based on each of the first comment information; specifically: for each comment information in the user corpus, determine whether the text content of the comment information contains words containing the semantics of lag based on the text content of the comment information and the characteristics of the target user, determine the comment information containing words containing the semantics of lag in the text content as the first comment information, and obtain the lag information library based on each of the first comment information. For example, extract the first comment information from the corpus to obtain the lag information library; or retain the first comment information in the user corpus and delete the others to obtain the lag information library; wherein, the words containing the semantics of lag refer to words that describe the semantics of lag, such as lag, delay, and jitter.

[0060] 3. Constructing a Specialized Dictionary

[0061] The search information of a designated online live broadcast platform is captured and used as a keyword; wherein the online live broadcast platform may be a mainstream live broadcast platform. A lag keyword containing a lag semantic is determined in each keyword, and the lag keywords are classified according to lag type to obtain a preset dedicated dictionary set.

[0062] 4. Determine the type of jam

[0063] Each first comment information in the jam information library is vectorized to obtain a corresponding jam vector set, specifically: for each first comment information in the jam information library, based on a preset proprietary dictionary set, the text content of each first comment information is segmented using a text semantic segmentation method to obtain the jam keywords corresponding to each first comment information, and the jam keywords corresponding to each first comment information are constructed into a text word vector, and all the text word vectors form a jam vector set; the proprietary dictionary set includes at least one jam type, and jam keywords containing jam semantics corresponding to each jam type.

[0064] The lag vector set is input into a naive Bayes classifier, which classifies the lag vector set and outputs the lag type corresponding to the target user. The lag types include: mild lag, moderate lag, severe lag, and blue screen.

[0065] Combined with the term frequency-inverse document frequency algorithm TF-IDF, the jam type corresponding to the target user is matched with the jam keywords of the jam type in a preset proprietary dictionary set, and jam recommendation information corresponding to the target user is generated based on the matched jam keywords. The jam recommendation information corresponding to the target user is added to the information recommendation vocabulary of the target user, so that each audience member can select the corresponding jam recommendation information from the information recommendation vocabulary as comment information or a component of comment information when posting comment information.

[0066] 5. Append the jam recommendation information corresponding to the target user to the information display area provided by the live broadcast, thereby displaying the jam recommendation information to the target user's audience users.

[0067] 6. You can also combine the push-stream end, pull-stream end and CDN to analyze the frame rate, bit rate and resolution, track the distribution and causes of the jamming, and achieve low-cost and differentiated discovery of live broadcast jamming quality issues.

[0068] Before using the naive Bayes algorithm classifier, the naive Bayes algorithm is trained to obtain the naive Bayes algorithm classifier.

[0069] First, we access the target user's information stream through a message queue, using a text-based recommendation system for user comments. This real-time information stream serves as a sample. We perform a preliminary data screening based on the text content of the real-time information and the characteristics of the target user, retaining comments that describe lag, latency, jitter, and other issues. We manually annotate the comments based on their content and assign them to the corresponding lag type.

[0070] Secondly, we used the text semantic segmentation method to pre-process the text content of the comments marked as lag. We constructed a text word vector through word segmentation, and input the text word vector into the Naive Bayes algorithm for training. Finally, we obtained a Naive Bayes classification model (Naive Bayes classifier), which classified the text word vectors and obtained a preliminary classification of the lag type of the text content. Based on the classification results, we combined TF-IDF to calculate the comments with the highest similarity to the lag type in each category, and obtained a set of lag categories. The details are as follows:

[0071] (1) Use the text semantic segmentation method to pre-process the text content of the comments marked as stuck, and construct text word vectors through word segmentation. Use the TF-IDF algorithm to calculate the text word vectors to obtain training samples.

[0072] (2) Use the text word vector as a training sample and perform preliminary classification using the Naive Bayes algorithm.

[0073] The text word vector is represented as

[0074] X=(x1,x2,...,x m ) T (1-1)

[0075] Among them, x j is the jth eigenvalue of the text content.

[0076] The corresponding jam category is represented as

[0077] C=(y1,y2,...,y m ) (1-2)

[0078] Among them, y j is the jam category corresponding to the j-th text content of the text.

[0079] According to the manual labeling of the text content, we can get the jam category when the jam category is y j Under the condition that feature X j is x j (x j is the index in the dictionary) j =x j |C=y j ); At the same time, the probability P(C=y j ), thus generating a Bayesian classifier. According to the conditional independence hypothesis, when the jam category is y j Under the condition, the probability that the input is a text word vector X

[0080]

[0081] For a given input x j , calculate the posterior probability distribution P(C=y j |X=x), and the jamming category corresponding to the maximum posterior probability is taken as x j The category output can be expressed as

[0082]

[0083] The Bayesian classifier generated by formula 1-4 is used to predict the text word vector X and obtain the pre-classification result: the jam category Y of each keyword.

[0084] (3) According to the pre-classification result Y, the word segmentation under each corresponding category is used as a training sample to construct a word vector set S. For the text word vector X to be classified, statistics x j Each eigenvalue x i Inverse frame rate in set S:

[0085]

[0086] And calculate each eigenvalue x i Frequency of occurrence in text word vector X:

[0087] TF=T(x i ) / Tx (1-6)

[0088] Combining Equations 1-5 and 1-6, we can calculate the weights of each feature of the text word vector X:

[0089] f(x i )=TF(x i )*IDF(x i ) (1-7)

[0090] The weight vector corresponding to the text word vector X is expressed as:

[0091] F=(f(x1),f(x2),...,f(x n )) (1-8)

[0092] Finally, for the text word vector X, retain the word segmentation containing the Caton keyword in the Caton proprietary dictionary set to obtain the Caton word vector matrix:

[0093] P=(P1,P2,…,P k ) (1-9)

[0094] in

[0095] P q =(p1, p2, ..., p k ) T (1-10)

[0096] P q The unit vector representing the qth stuck keyword, q = 1, 2, ..., k, where k is the number of texts containing the stuck keyword.

[0097] The final Bayesian classifier is obtained by training the Naive Bayesian algorithm on the Caton word vector matrix.

[0098] The beneficial technical effects achieved by the embodiments of the present invention are as follows:

[0099] By leveraging livestream audience comments, we can immediately identify lag issues for different target users and accurately derive the type of lag. This allows for timely detection of quality issues within livestreams, enabling efficient response and resolution, while covering a wide range of lag scenarios. Furthermore, the commentary information leverages existing resources, eliminating the need for large amounts of data or high storage costs, resulting in low overall costs and the ability to reuse them.

[0100] It should be understood that the specific order or hierarchy of steps in the disclosed processes is an example of an exemplary method. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the process can be rearranged without departing from the scope of the present disclosure. The accompanying method claims present elements of the various steps in an exemplary order and are not intended to be limited to the specific order or hierarchy described.

[0101] In the foregoing detailed description, various features are grouped together in a single embodiment to simplify the disclosure. This method of disclosure should not be interpreted as reflecting an intention that embodiments of the claimed subject matter require more features than are expressly recited in each claim. On the contrary, as reflected in the appended claims, the invention comprises less than all the features of any individual disclosed embodiment. The appended claims are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate preferred embodiment of the invention.

[0102] The above description of the disclosed embodiments is intended to enable any person skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the spirit and scope of the present disclosure. Therefore, the present disclosure is not limited to the embodiments presented herein but is intended to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0103] The foregoing description includes examples of one or more embodiments. Of course, it is not possible to describe all possible combinations of components or methods for the purposes of describing the above embodiments, but one of ordinary skill in the art will recognize that the various embodiments may be further combined and arranged. Therefore, the embodiments described herein are intended to encompass all such changes, modifications and variations that fall within the scope of the appended claims. Furthermore, to the extent the term "comprising" is used in the specification or claims, the term is intended to be encompassed in a manner similar to the term "including," as explained in terms of "including," used as a transitional word in the claims. Furthermore, any use of the term "or" in the specification of the claims is intended to mean a "non-exclusive or."

[0104] Those skilled in the art will also appreciate that the various illustrative logical blocks, units, and steps listed in the embodiments of the present invention can be implemented by electronic hardware, computer software, or a combination of the two. To clearly demonstrate the interchangeability of hardware and software, the various illustrative components, units, and steps described above have generally described their functions. Whether such functions are implemented by hardware or software depends on the specific application and the design requirements of the entire system. Those skilled in the art may use various methods to implement the described functions for each specific application, but such implementation should not be understood as exceeding the scope of protection of the embodiments of the present invention.

[0105] The various illustrative logic blocks or units described in the embodiments of the present invention can be implemented or operated by a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field programmable gate array or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof. The general-purpose processor can be a microprocessor, and optionally, the general-purpose processor can also be any conventional processor, controller, microcontroller or state machine. The processor can also be implemented by a combination of computing devices, such as a digital signal processor and a microprocessor, a plurality of microprocessors, one or more microprocessors combined with a digital signal processor core, or any other similar configuration.

[0106] The steps of the methods or algorithms described in the embodiments of the present invention may be directly embedded in hardware, a software module executed by a processor, or a combination of the two. The software module may be stored in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. For example, the storage medium may be connected to the processor so that the processor can read information from the storage medium and write information to the storage medium. Alternatively, the storage medium may also be integrated into the processor. The processor and storage medium may be provided in an ASIC, which may be provided in a user terminal. Alternatively, the processor and storage medium may also be provided in different components in the user terminal.

[0107] In one or more exemplary designs, the above-mentioned functions described in the embodiments of the present invention can be implemented in hardware, software, firmware, or any combination of the three. If implemented in software, these functions can be stored on a computer-readable medium or transmitted in the form of one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media that facilitate the transfer of computer programs from one location to another. Storage media can be any available medium that can be accessed by a general or special computer. For example, such computer-readable media can include but are not limited to RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store program code in the form of instructions or data structures and other forms that can be read by a general or special computer, or a general or special processor. In addition, any connection can be appropriately defined as a computer-readable medium. For example, if the software is transmitted from a website, server or other remote resource via a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless methods such as infrared, wireless, and microwave, it is also included in the definition of computer-readable media. The disks and discs mentioned above include compact disks, laser disks, optical disks, DVDs, floppy disks, and Blu-ray discs. Disks typically reproduce data magnetically, while discs typically reproduce data optically with lasers. Combinations of the above may also be included in computer-readable media.

[0108] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for tracking freezes in live streaming, characterized in that: include: During the live broadcast of the target user, the comment information posted by each audience user of the live broadcast is obtained in real time, and a user corpus is constructed based on the comment information of each audience user; Determining first comment information containing lag semantics in the user corpus, and obtaining a lag information database based on each of the first comment information; Performing vector processing on each first comment information in the jam information database to obtain a corresponding jam vector set, inputting the jam vector set into a naive Bayes classifier, and outputting the jam type corresponding to the target user; The method for tracking network live broadcast freezes also includes: According to the jam type corresponding to the target user, match the jam keywords of the jam type in a preset proprietary dictionary set, and generate jam recommendation information corresponding to the target user according to the matched jam keywords; Displaying the jam recommendation information corresponding to the target user in the information display area provided by the live broadcast, so as to be displayed to each audience user of the target user; and The jam recommendation information corresponding to the target user is added to the information recommendation vocabulary of the target user.

2. The method for tracking network live broadcast freezes according to claim 1, characterized in that: During the live broadcast of the target user, the comment information posted by each audience user of the live broadcast is obtained in real time, and a user corpus is constructed based on the comment information of each audience user, specifically including: Accessing the real-time comment information stream of the target user who initiated the live broadcast through a message queue, obtaining the comment information posted by each audience user of the live broadcast from the real-time comment information stream, and constructing a user corpus based on the comment information of each audience user; The determining of first comment information containing stuck semantics in the user corpus and obtaining a stuck information database based on each of the first comment information specifically includes: For each comment information in the user corpus, based on the text content of the comment information and the characteristics of the target user, determine whether there are words containing lag semantics in the text content of the comment information, and determine the comment information containing words containing lag semantics in the text content as the first comment information. The lag information library is obtained based on each of the first comment information.

3. The method for tracking network live broadcast freezes according to claim 2, characterized in that: The step of performing vector processing on each first comment information in the jam information library to obtain a corresponding jam vector set specifically includes: For each first comment information in the jam information library, the text content of each first comment information is segmented using a text semantic segmentation method based on a preset proprietary dictionary set to obtain jam keywords corresponding to each first comment information, and the jam keywords corresponding to each first comment information are constructed into text word vectors, and all the text word vectors form a jam vector set; the proprietary dictionary set includes at least one jam type and jam keywords containing jam semantics corresponding to each jam type.

4. The method for tracking network live broadcast freezes according to claim 1 or 3, characterized in that: The method for constructing the preset proprietary dictionary set includes: Capture search information of a specified online live broadcast platform and use the search information as keywords; Determine the jam keywords containing jam semantics in each keyword, and classify the jam keywords according to jam types to obtain the proprietary dictionary set.

5. A tracking system for network live broadcast freezes, characterized in that: include: A user corpus construction unit is used to obtain, in real time, comment information posted by each audience user of the live broadcast during the live broadcast of the target user, and to construct a user corpus based on the comment information of each audience user; a jam information database construction unit, configured to determine first comment information containing jam semantics in the user corpus, and obtain a jam information database based on each of the first comment information; The jam type determination unit is used to perform vector processing on each first comment information in the jam information library to obtain a corresponding jam vector set, input the jam vector set into a naive Bayes classifier, and output the jam type corresponding to the target user.

6. The network live broadcast freeze tracking system according to claim 5, characterized in that: The user corpus construction unit is specifically configured to access the real-time comment information stream of the target user who initiated the live broadcast through a message queue, obtain the comment information posted by each audience user of the live broadcast from the real-time comment information stream, and construct a user corpus based on the comment information of each audience user; The jam information library construction unit is specifically configured to determine, for each comment in the user corpus, whether the text of the comment contains words with jam semantics based on the text content of the comment and the characteristics of the target user, determine the comment information containing words with jam semantics in the text as first comment information, and obtain the jam information library based on each first comment information; The network live broadcast freeze tracking system also includes: a jam recommendation information generating unit, configured to match the jam type corresponding to the target user with jam keywords of the jam type in a preset proprietary dictionary set, and generate jam recommendation information corresponding to the target user based on the matched jam keywords; A jam recommendation information display unit, configured to display the jam recommendation information corresponding to the target user in the information display area provided by the live broadcast, so as to display it to each audience user of the target user; An information recommendation vocabulary library is used to receive and store the jam recommendation information generated by the jam recommendation information generation unit; the information recommendation vocabulary library is set on the target user side.

7. The network live broadcast freeze tracking system according to claim 6, characterized in that: The jam type determination unit includes: A jam vector set construction subunit is used to segment the text content of each first comment information in the jam information library using a text semantic segmentation method based on a preset proprietary dictionary set, obtain the jam keywords corresponding to each first comment information, and construct the jam keywords corresponding to each first comment information into a text word vector, and form all the text word vectors into a jam vector set; the proprietary dictionary set includes at least one jam type and jam keywords with jam semantics corresponding to each jam type.

8. The network live broadcast freeze tracking system according to claim 5 or 7, characterized in that: Also includes: A proprietary dictionary set construction unit is used to capture search information of a specified online live broadcast platform and use the search information as a keyword; Determine the jam keywords containing jam semantics in each keyword, and classify the jam keywords according to jam types to obtain a preset proprietary dictionary set.

Images