System, method and computer-readable medium for data search
The present disclosure relates to a system, a method and a computer-readable medium for live streaming data search. The method includes receiving a search word, converting the search word into a search vector, receiving an attribute vector matching the search vector, and providing information of a first live streaming program corresponding to the attribute vector. The attribute vector varies according to contents of the first live streaming program while the first live streaming program is being distributed. The present disclosure can deliver contextual search results and text search results simultaneously in a real time manner.
1. A method for live streaming data search, comprising:
receiving a search word;
converting the search word into a search vector;
receiving tags of a first live streaming program, wherein the tags are dynamically generated in real time according to contents of the first live streaming program while the first live streaming program is being distributed;
receiving an attribute text corresponding to the first live streaming program;
combining the tags and the attribute text into a mix text;
converting the mix text to an attribute vector with a BERT model;
receiving the attribute vector matching the search vector;
providing information of the first live streaming program corresponding to the attribute vector;
matching the search word with an attribute text corresponding to a second live streaming program, wherein the second live streaming program is same as or different from the first streaming program;
determining a matching result of the matching between the search word and the attribute text of the second live streaming program to be greater than a threshold value; and
providing information of the second live streaming program simultaneously with the first live streaming program,
wherein the attribute vector varies according to contents of the first live streaming program while the first live streaming program is being distributed.
2. The method according to claim 1 , wherein the converting the search word into the search vector includes a BERT word embedding process.
3. The method according to claim 1 , wherein the matching between the search vector and the attribute vector includes a KNN search process.
4. The method according to claim 1 , further comprising:
generating the tags of the first live streaming program with an image recognizing process, a motion recognizing process or a sound recognizing process.
5. The method according to claim 1 , wherein the information of the first live streaming program and the information of the second live streaming program are configured to display the first live streaming program and the second live streaming program simultaneously on a user terminal, wherein the search word is received from the user terminal.
6. A system for live streaming data search, comprising one or a plurality of processors, wherein the one or plurality of processors execute a machine-readable instruction to perform:
receiving a search word;
converting the search word into a search vector;
receiving tags of a first live streaming program, wherein the tags are dynamically generated in real time according to contents of the first live streaming program while the first live streaming program is being distributed;
receiving an attribute text corresponding to the first live streamlining program;
combining the tags and the attribute text into a mix text;
converting the mix text to an attribute vector with a BERT model;
receiving the attribute vector matching the search vector;
providing information of the first live streaming program corresponding to the attribute vector;
matching the search word with an attribute text corresponding to a second live streaming program, wherein the second live streaming program is same as or different from the first streaming program;
determining a matching result of the matching between the search word and the attribute text of the second live streaming program to be greater than a threshold value; and
providing information of the second live streaming program simultaneously with the first live streaming program,
wherein the attribute vector varies according to contents of the first live streaming program while the first live streaming program is being distributed.
7. A non-transitory computer-readable medium including a program for live streaming data search, wherein the program causes one or a plurality of computers to execute:
receiving a search word;
converting the search word into a search vector;
receiving tags of a first live streaming program, wherein the tags are dynamically generated in real time according to contents of the first live streaming program while the first live streaming program is being distributed;
receiving an attribute text corresponding to the first live streaming program;
combining the tags and the attribute text into a mix text;
converting the mix text to an attribute vector with a BERT model;
receiving the attribute vector matching the search vector;
providing information of the first live streaming program corresponding to the attribute vector;
matching the search word with an attribute text corresponding to a second live streaming program, wherein the second live streaming program is same as or different from the first streaming program;
determining a matching result of the matching between the search word and the attribute text of the second live streaming program to be greater than a threshold value; and
providing information of the second live streaming program simultaneously with the first live streaming program,
wherein the attribute vector varies according to contents of the first live streaming program while the first live streaming program is being distributed.