Disclosure of Invention
The invention aims to provide a video live broadcast interface commodity display intelligent management system based on big data, the display key word set of various commodities is formed by classifying the commodity types displayed by live video broadcast and taking the display direction of various commodities as the display key word, collecting the display forms of different display key words of various commodities and storing the display key word sets in a commodity resource database, meanwhile, the voice input of the anchor is combined to carry out voice recognition matching to extract the commodity types and commodity display keywords, matching with each display keyword in the category of commodity display keyword set, calling and pushing the display form corresponding to the commodity display keyword which is successfully matched to a video live broadcast interface through a background, and the watching satisfaction of the consumer is counted according to the watching time length of the consumer entering the live broadcast room, and the problems in the background technology are solved.
The purpose of the invention can be realized by the following technical scheme:
a video live broadcast interface commodity display intelligent management system based on big data comprises a commodity collection and classification module, a commodity resource database, a voice input module, a voice recognition and extraction module, a management server, a background calling and pushing module, a live broadcast watching duration counting module for consumers and a background display terminal;
the system comprises a commodity collection and classification module, a voice recognition and extraction module, a management server, a background calling and pushing module and a background display terminal, wherein the commodity collection and classification module is connected with a commodity resource database, the voice recognition and extraction module is respectively connected with a voice input module and the commodity resource database, the management server is respectively connected with the voice recognition module, a consumer live broadcast watching duration counting module and the commodity resource database, the background calling and pushing module is connected with the management server, and the background display terminal is connected with the management server;
the commodity collecting and classifying module is used for collecting and classifying data related to commodities displayed in live video broadcast, specifically classifying the types of the commodities displayed in live video broadcast, and meanwhile collecting display forms of different display keywords by taking a display direction of a certain commodity as a display keyword to form a commodity display keyword set G
i(g
i1,g
i2,…,g
il,...,g
i6),g
il is the showing form of the l-th showing keyword, l is 1,2,3, …,6, l is the commodity keyword, i is the showing form, i is i1, i2, i3, i1, i2, i3 is respectively shown as picture, GIF animation and short video, the showing directions comprise the upper surface, the lower surface, the front surface, the rear surface, the left surface and the right surface of the commodity, the showing form comprises picture, GIF animation and short video, the showing forms of different showing keywords of each commodity displayed by video live broadcast are collected according to the above mode, and the showing forms of different showing keywords of each kind of commodity showing keyword set are formed
g
kil represents a display form of the ith keyword of the kth commodity, k is 1,2,3, …, n, k represents the type of the commodity, n represents the number of the commodity types, i represents the display form, i is i1, i2, i3, i1, i2 and i3 respectively represent pictures, GIF animations and short videos, and the commodity collection and classification module sends the collected and classified various types of commodity display keyword sets to the commodity resource database;
the commodity resource database is used for receiving various commodity display keyword sets sent by the commodity collection and classification module, storing the various commodity display keyword sets and storing a preset voice template library;
the voice input module is used for receiving the original voice information of the anchor and sending the original voice information to the voice recognition and extraction module;
the voice recognition extraction module comprises a front-end preprocessing unit, a voice recognition matching unit and a text keyword extraction unit. The front-end preprocessing unit is used for carrying out endpoint detection and voice enhancement processing on the received original voice information to obtain primary voice and sending the primary voice to the voice recognition matching unit; the voice recognition matching unit receives the primary voice sent by the front-end preprocessing unit, captures feature vectors in the primary voice, meanwhile extracts a voice template base prestored in a commodity resource database, analyzes the captured voice feature vectors with each template in the voice template base in sequence, counts the analysis similarity between the captured voice feature vectors and each template in the voice template base, screens the voice template with the maximum similarity, outputs the voice template with the maximum similarity when the screened maximum similarity is larger than a set similarity threshold, does not process when the screened maximum similarity is smaller than the set similarity threshold, and then obtains a text recognition result of the computer through table lookup according to the definition of the output voice template and sends the text recognition result to the text keyword extraction unit. The text keyword extraction unit receives the text recognition result sent by the voice recognition matching unit and carries out word segmentation on the text recognition result to obtain each word group, and then extracts commodity types and commodity display keywords from each word group and sends the commodity types and the commodity display keywords to the management server;
the management server is used for receiving the commodity types and the commodity display keywords sent by the voice recognition extraction module, extracting various commodity display keyword sets in the commodity resource database, screening the commodity display keyword sets according to the commodity types, matching the commodity display keywords extracted by the voice recognition with each display keyword in the commodity display keyword sets, terminating the matching if one of the commodity display keywords is successfully matched with each display keyword in the commodity display keyword sets, and sending the display forms corresponding to the matched commodity display keywords to the background calling and pushing module; if the related information is not matched, the related information does not exist in the category commodity display keyword set, and the commodity display keyword set extracted by the current voice is automatically sent to a commodity collection and classification module for subsequent collection and updating to the category commodity display keyword set;
the background calling and pushing module is used for receiving the display form corresponding to the commodity display keyword sent by the management server, calling out the display form and pushing the display form to a video live broadcast interface;
the consumer watches live broadcast duration statistics module, count and number the account names of each consumer entering the live broadcast room, mark as 1,2, …, i, …, n in sequence, record the time of each account name entering the live broadcast room for the first time and the time of each account name exiting the live broadcast room for the last time, count the total duration of each account name watching live broadcast according to the time of each account name entering the live broadcast room for the first time and the time of each account name exiting the live broadcast room for the last time, and form a total duration set T (T) of each account name watching live broadcast
1,T
2,…,T
i,...,T
n),T
iThe method comprises the steps of representing the total time length of watching the live broadcast for the ith account name, simultaneously checking the frequency of each account name entering a live broadcast room in the process of watching the total time length of the live broadcast, recording the time of each account name entering the live broadcast room and the time of exiting the live broadcast room, calculating the time length of watching the live broadcast once, counting the frequency of each account name entering the live broadcast room in the process of watching the total time length of the live broadcast and the time length of watching the live broadcast once, and forming a single account name live broadcast watching frequency time length set T
i(t
i1,t
i2,...,t
ij,...,t
im),t
ij represents the live broadcast watching time length of the ith account name entering the live broadcast room for the jth time, m represents the total frequency of the ith account name entering the live broadcast room in the process of watching the total time length of the live broadcast, and
sending the counted live broadcast watching frequency and duration set of each account name to a management server;
the management server receives the live broadcast watching frequency and duration set of each account name sent by the live broadcast watching duration counting module of the consumer, counts the watching satisfaction degree of the consumer according to the received live broadcast watching frequency and duration set of each account name, and sends the counting result to the background display terminal;
and the background display terminal receives the customer watching satisfaction sent by the management server and displays the customer watching satisfaction.
Preferably, the speech template library further includes an additional template library, where the additional template library is a template library automatically extracted from the search engine to obtain an optimal speech analysis mode when corresponding information does not exist in the speech template library, and when the extracted speech feature vector does not analyze a related template in the speech template library, automatically transferring the extracted speech feature vector to the additional template library for analysis, and if the related template is not also analyzed in the additional template library, automatically updating the additional template library.
Further, the automatic updating of the additional template library is to link data input by a user to a search engine interface, automatically acquire analysis information, obtain a result with the highest relevance of each search engine, then perform integration analysis on the results of the search engines, obtain an optimal voice analysis mode through a weighting algorithm, and finally store the voice analysis mode into the additional template library.
Furthermore, an asynchronous concurrency mechanism is adopted in the management server for matching the commodity display keyword extracted by voice recognition with the commodity display keyword set of the type, and the asynchronous concurrency mechanism matches each display keyword in the commodity display keyword set at the same time.
Further, the total time length of each account name watching the live broadcast is calculated by subtracting the time of the account name entering the live broadcast room for the first time from the time of the account name exiting the live broadcast room for the last time, and the time length of the account name watching the live broadcast for the single time is calculated by subtracting the time of the account name entering the live broadcast room from the time of the account name exiting the live broadcast room for the single time.
Further, the calculation formula of the customer viewing satisfaction is as follows:
t
ij represents the live broadcast watching time length of the jth account name entering the live broadcast room for the jth time
iExpressed as the total length of time that the live view is viewed for the ith account name.
Has the advantages that:
(1) the invention provides a big data-based intelligent management system for commodity display of a video live broadcast interface, which is characterized in that the commodity types of a video live broadcast display place are classified, the display directions of various commodities are taken as display keywords, the display forms of different display keywords of various commodities are collected to form a display keyword set of various commodities, the display keyword set of various commodities is stored in a commodity resource database, voice input of a main broadcast is combined, the commodity types and the commodity display keywords are extracted through voice recognition matching, the display keywords are matched with each display keyword in the display keyword set of various commodities, the display forms corresponding to the successfully matched commodity display keywords are called and pushed to the video live broadcast interface through a background, the video live broadcast commodity display without real commodities is met, the picture feeling of watching live broadcast by people is improved, and more intuitional and intuitive effects are brought to consumers, The live broadcast method has the advantages that the live broadcast experience is realized, the watching satisfaction degree of a consumer is counted according to the watching time length when the consumer enters a live broadcast room, the live broadcast background staff can visually know the live broadcast effect conveniently, and relevant reference basis is provided for subsequent similar live broadcast.
(2) According to the video live broadcast interface commodity display intelligent management system based on the big data, the endpoint detection and the voice enhancement front-end processing are carried out before voice recognition matching, the influence caused by noise and different speakers is eliminated, the processed signals can reflect the essential characteristics of voice, and the system plays an important role in the accuracy and the recognition accuracy of subsequent voice recognition matching.
(3) According to the video live broadcast interface commodity display intelligent management system based on big data, the commodity display keywords extracted by voice recognition are asynchronously and concurrently matched with each display keyword in the commodity display keyword set, a fast and efficient data analysis result can be obtained, and meanwhile when the result is not matched, the commodity display keywords extracted by current voice are automatically sent to the commodity collection and classification module for subsequent collection and updating to the commodity display keyword set, so that the intelligence of the system is embodied.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, an intelligent management system for commodity display of a big data live video interface includes a commodity collection and classification module, a commodity resource database, a voice input module, a voice recognition and extraction module, a management server, a background retrieval and push module, a live watching duration counting module for consumers, and a background display terminal.
The commodity collection and classification module is connected with a commodity resource database, the voice recognition and extraction module is respectively connected with the voice input module and the commodity resource database, the management server is respectively connected with the voice recognition module, the live broadcast watching duration counting module of a consumer and the commodity resource database, the background calling and pushing module is connected with the management server, and the background display terminal is connected with the management server.
The commodity collecting and classifying module is used for collecting and classifying data related to commodities displayed in live video broadcast, specifically classifying the types of the commodities displayed in live video broadcast, and meanwhile collecting display forms of different display keywords by taking a display direction of a certain commodity as a display keyword to form a commodity display keyword set G
i(g
i1,g
i2,...,g
il,...,g
i6),g
il is the showing form of the l-th showing keyword, l is 1,2,3, …,6, l is the commodity keyword, i is the showing form, i is i1, i2, i3, i1, i2, i3 is respectively shown as picture, GIF animation and short video, the showing directions comprise the upper surface, the lower surface, the front surface, the rear surface, the left surface and the right surface of the commodity, the showing form comprises picture, GIF animation and short video, the showing forms of different showing keywords of each commodity displayed by video live broadcast are collected according to the above mode, and the showing forms of different showing keywords of each kind of commodity showing keyword set are formed
g
kil is a display form of the ith keyword of the kth commodity, k is 1,2,3, …, n, k is a commodity type, n is a number of commodity types, i is a display form, i is i1, i2, i3, i1, i2 and i3 are respectively pictures, GIF animations and short videos, and the commodity collection and classification module sends the collected and classified various commodity display keyword sets to the commodity resource database.
The commodity resource database is used for receiving the sets of the display keywords of various commodities sent by the commodity collection and classification module, storing the sets of the display keywords and storing the voice template database.
The voice input module is used for receiving the original voice information of the anchor and sending the original voice information to the voice recognition extraction module, and in order to improve the accuracy of voice recognition, the expression of the anchor voice is preferably expressed by Mandarin.
The voice recognition extraction module comprises a front-end preprocessing unit, a voice recognition matching unit and a text keyword extraction unit. The front-end preprocessing unit is used for carrying out end point detection and voice enhancement processing on received original voice information to obtain primary voice and sending the primary voice to the voice recognition matching unit, the end point detection is to distinguish voice signal time intervals from non-voice signal time intervals in a voice signal to accurately determine the initial point of the voice signal, and after the end point detection, subsequent processing can be carried out on the voice signal only; the voice enhancement is to eliminate the influence of environmental noise on voice, so that the processed signal can reflect the essential characteristics of voice. The voice recognition matching unit receives the primary voice sent by the front-end preprocessing unit, captures feature vectors in the primary voice, extracts a voice template library in a commodity resource database, analyzes the captured voice feature vectors with each template in the voice template library in sequence, counts the similarity between the captured voice feature vectors and each template in the voice template library, screens the voice template with the maximum similarity, outputs the voice template with the maximum similarity when the screened maximum similarity is larger than a set similarity threshold, does not process the voice template when the screened maximum similarity is smaller than the set similarity threshold, and obtains a text recognition result of the computer by looking up a table according to the definition of the output voice template and sends the text recognition result to the text keyword extraction unit. The text keyword extraction unit receives the text recognition result sent by the voice recognition matching unit and carries out word segmentation on the text recognition result to obtain each phrase, Python3.0 word segmentation software is adopted in the word segmentation process, the word segmentation is based on a Chinese corpus dictionary, open source HanLP natural language processing is adopted, and then commodity types and commodity display keywords are extracted from each phrase and sent to the management server.
Furthermore, the voice template library also comprises an additional template library, when corresponding information does not exist in the voice template library, the additional template library is a template library which automatically extracts the optimal voice analysis mode from the search engine, when the captured voice feature vector can not analyze the related template in the voice template library, automatically transferring the captured voice feature vector to the additional template library for analysis, if the related template can not be analyzed in the additional template library, automatically updating the additional template library, the automatic updating additional template library links the data input by the user to the search engine interface, automatically acquires the analysis information, obtains the result with the highest relevancy of each search engine, and then, integrating and analyzing the results of the search engine, obtaining an optimal voice analysis mode through a weighting algorithm, and finally storing the voice analysis mode into an additional template library.
The management server is used for receiving the commodity types and the commodity display keywords sent by the voice recognition extraction module, extracting various commodity display keyword sets in a commodity resource database, screening the commodity display keyword sets according to the commodity types, matching each display keyword in the commodity display keyword sets by using an asynchronous concurrency mechanism for the commodity display keywords extracted by voice recognition, matching each display keyword in the commodity display keyword sets by using the asynchronous concurrency mechanism, terminating the matching if one of the display keywords is successfully matched, and sending the display forms corresponding to the matched commodity display keywords to the background retrieval and pushing module; if the related information is not matched, checking whether all the display keywords in the category of commodity display keyword set are matched completely or not, if the matching is not completed, continuing to use an asynchronous concurrency mechanism for matching until the matching is completed, and if the matching is completed and the related information is not matched, indicating that the information related to the commodity display keywords extracted by the current voice does not exist in the category of commodity display keyword set, sending the commodity display keywords extracted by the current voice to a commodity collection and classification module for subsequent collection and updating to the category of commodity display keyword set.
The background calling and pushing module is used for receiving the display form corresponding to the commodity display keyword sent by the management server, calling out the display form and pushing the display form to a video live broadcast interface.
The consumer watches the direct broadcast time length statistic module, and accounts and numbers the account names of all consumers entering the direct broadcast room, and the account names are marked as1,2, …, i, …, n, and recording the time of each account name entering the live broadcasting room for the first time and the time of exiting the live broadcasting room for the last time, counting the total time of each account name watching the live broadcasting according to the time of each account name entering the live broadcasting room for the first time and the time of exiting the live broadcasting room for the last time, wherein the total time of each account name watching the live broadcasting is calculated by subtracting the time of entering the live broadcasting room for the first time from the time of each account name exiting the live broadcasting room for the last time, and forming a total time set T (T) of each account name watching the live broadcasting
1,T
2,...,T
i,...,T
n),T
iThe method comprises the steps of representing the total time length of watching the live broadcast for the ith account name, simultaneously checking the frequency of each account name entering a live broadcast room in the process of watching the total time length of the live broadcast, recording the time of each account name entering the live broadcast room and the time of exiting the live broadcast room, calculating the time length of watching the live broadcast once by subtracting the time of entering the live broadcast room from the time of exiting the live broadcast room, counting the frequency of each account name entering the live broadcast room in the process of watching the total time length of the live broadcast and the time length of watching the live broadcast once, and forming a time length set T of watching the live broadcast frequency time length by the single account name
i(t
i1,t
i2,...,t
ij,...,t
im),t
ij represents the live broadcast watching time length of the ith account name entering the live broadcast room for the jth time, m represents the total frequency of the ith account name entering the live broadcast room in the process of watching the total time length of the live broadcast, and
and sending the counted live broadcast watching frequency and duration set of each account name to a management server.
The management server receives the live broadcast watching frequency time length set of each account name sent by the live broadcast watching time length counting module of the consumer, and counts the watching satisfaction degree of the consumer according to the received live broadcast watching frequency time length set of each account name
t
ij represents the live broadcast watching time length of the jth account name entering the live broadcast room for the jth time
iShows the total time for watching the live broadcast for the ith account nameThe larger the watching satisfaction degree value of the consumer is, the more satisfied the consumer is with the live video, and the statistical watching satisfaction degree of the consumer is sent to the background display terminal by the management server.
And the background display terminal receives the customer watching satisfaction sent by the management server and displays the satisfaction, so that the live background staff can visually know the live video effect conveniently.
The method comprises the steps of classifying the types of commodities displayed by live video broadcast, taking the display direction of each type of commodity as a display keyword, collecting the display forms of different display keywords of each type of commodity to form a display keyword set of each type of commodity, storing the display keyword set in a commodity resource database, carrying out voice recognition matching to extract the types of commodities and the commodity display keywords by combining with the voice input of a main broadcast, matching with each display keyword in the display keyword set of each type of commodity, calling and pushing the display form corresponding to the successfully matched commodity display keyword to a live video broadcast interface through a background, meeting the requirement of live video broadcast commodity display without a real commodity, improving the picture feeling of live broadcast watching of people, and bringing more visual and vivid watching experience to consumers.
The foregoing is merely exemplary and illustrative of the principles of the present invention and various modifications, additions and substitutions of the specific embodiments described herein may be made by those skilled in the art without departing from the principles of the present invention or exceeding the scope of the claims set forth herein.