# US 2020/0401765 A1 · Man-machine conversation method, electronic device, and computer-readable medium > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2020/0401765 A1 - Document ID: US20200401765A1 - Status: Published application - Kind: A1 - Application number: 17/013,705 - Filed: 2020-09-07 - Published: 2020-12-24 - Earliest claimed: 2018-08-29 - Assignees: Tencent Technology (Shenzhen) Company Limited - Applicants: Tencent Technology (Shenzhen) Company Limited - Inventors: Qiu RAN; Xiao ZHOU; Cheng NIU; Jie ZHOU - Web page: https://patentgrep.com/patent/US20200401765A1 ## CPC - G06F 40/279: Recognition of textual entities - G06F 40/289: Phrasal analysis, e.g. finite state techniques or chunking - G06F 16/3329: Natural language query formulation - G06F 40/284: Lexical analysis, e.g. tokenisation or collocates - G06F 40/30: Semantic analysis - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract A man-machine conversation method includes: extracting a keyword from input information of a user; determining a candidate article according to the keyword; obtaining a plurality of sentences in the candidate article, and determining a score of each of the plurality of sentences; extracting part of the plurality of sentences based on the score of each of the plurality of sentences; and generating return information that includes the extracted part of the plurality of sentences. ## Claims ### Claim 1 A man-machine conversation method, performed by an electronic device, the method comprising: extracting a keyword from input information of a user; determining a candidate article according to the keyword; obtaining a plurality of sentences in the candidate article, and determining a score of each of the plurality of sentences; extracting part of the plurality of sentences based on the score of each of the plurality of sentences; and generating return information that includes the extracted part of the plurality of sentences. ### Claim 2 (depends on 1) The method according to claim 1, wherein the extracting a keyword from input information of a user comprises: extracting the keyword from voice input information of the user; or extracting the keyword from text input information of the user. ### Claim 3 (depends on 1) The method according to claim 1, wherein the extracting a keyword from input information of a user comprises: performing word segmentation on the input information, to obtain a plurality of input words; determining word vectors of the plurality of input words; determining word similarities between the plurality of input words based on cosine similarities and the word vectors; determining importance levels of the plurality of input words according to the word similarities; and determining the keyword according to a sorting order of the importance levels. ### Claim 4 (depends on 1) The method according to claim 1, wherein the determining a candidate article according to the keyword comprises: searching in a predetermined platform according to the keyword, to determine the candidate article. ### Claim 5 (depends on 1) The method according to claim 1, wherein the determining a score of each of the plurality of sentences comprises: determining a first score according to an importance level of each of the plurality of sentences in the article; determining a second score according to a similarity between each of the plurality of sentences and the input information; and determining the score of each sentence according to the first score and the second score. ### Claim 6 (depends on 5) The method according to claim 5, wherein the determining a first score according to an importance level of each of the plurality of sentences in the article comprises: determining a sentence vector of each of the plurality of sentences; determining sentence similarities between sentence vectors of the plurality of sentences; and obtaining the importance level of each sentence in the article based on the sentence similarities, to determine the first score. ### Claim 7 (depends on 6) The method according to claim 6, wherein the determining a sentence vector of each of the plurality of sentences comprises: performing word segmentation on a sentence, to obtain a plurality of words; determining word vectors of the plurality of words; determining term frequency-inverse document frequencies (TF-IDFs) of the plurality of words; and determining the sentence vector of the sentence according to the TF-IDFs and the word vectors. ### Claim 8 (depends on 6) The method according to claim 6, wherein the obtaining the importance level of each sentence in the article based on the sentence similarities, to determine the first score comprises: constructing a similarity matrix based on the sentence similarities; constructing a graphical model based on the similarity matrix by using each sentence as a node and using each sentence similarity as a weight of an edge; obtaining the importance level of each sentence in the article based on the graphical model; and determining the first score through the importance level of each of the plurality of sentences. ### Claim 9 (depends on 5) The method according to claim 5, wherein the determining a second score according to a similarity between each of the plurality of sentences and the input information comprises: determining an information vector of the input information of the user; determining a sentence vector of each of the plurality of sentences; determining the similarity between the information vector and the sentence vector of each sentence based on a cosine similarity; and determining the second score according to the similarity between the information vector and the sentence vector of each sentence. ### Claim 10 (depends on 1) The method according to claim 1, wherein the extracting part of the plurality of sentences comprises: sorting the plurality of sentences according to the score of each of the plurality of sentences, to obtain a sorting result; and extracting the part of the plurality of sentences according to the sorting result. ### Claim 11 (depends on 10) The method according to claim 10, wherein generating the return information comprises: performing natural language processing on the extracted part of sentences, to generate the return information. ### Claim 12 (depends on 3) The method according to claim 3, further comprising: obtaining the word vectors by training word data through a word vector training algorithm. ### Claim 13 An electronic device, comprising: one or more processors; and a storage apparatus, configured to store one or more programs; the one or more processors are, when executing the one or more programs, configured to perform: extracting a keyword from input information of a user; determining a candidate article according to the keyword; obtaining a plurality of sentences in the candidate article, and determining a score of each of the plurality of sentences; extracting part of the plurality of sentences based on the score of each of the plurality of sentences; and generating return information that includes the extracted part of the plurality of sentences. ### Claim 14 (depends on 13) The device according to claim 13, wherein the extracting a keyword from input information of a user comprises: performing word segmentation on the input information, to obtain a plurality of input words; determining word vectors of the plurality of input words; determining word similarities between the plurality of input words based on cosine similarities and the word vectors; determining importance levels of the plurality of input words according to the word similarities; and determining the keyword according to a sorting order of the importance levels. ### Claim 15 (depends on 13) The device according to claim 13, wherein the determining a candidate article according to the keyword comprises: searching in a predetermined platform according to the keyword, to determine the candidate article. ### Claim 16 (depends on 13) The device according to claim 13, wherein the determining a score of each of the plurality of sentences comprises: determining a first score according to an importance level of each of the plurality of sentences in the article; determining a second score according to a similarity between each of the plurality of sentences and the input information; and determining the score of each sentence according to the first score and the second score. ### Claim 17 (depends on 16) The device according to claim 16, wherein the determining a first score according to an importance level of each of the plurality of sentences in the article comprises: determining a sentence vector of each of the plurality of sentences; determining sentence similarities between sentence vectors of the plurality of sentences; and obtaining the importance level of each sentence in the article based on the sentence similarities, to determine the first score. ### Claim 18 (depends on 17) The device according to claim 17, wherein the determining a sentence vector of each of the plurality of sentences comprises: performing word segmentation on a sentence, to obtain a plurality of words; determining word vectors of the plurality of words; determining term frequency-inverse document frequencies (TF-IDFs) of the plurality of words; and determining the sentence vector of the sentence according to the TF-IDFs and the word vectors. ### Claim 19 (depends on 17) The device according to claim 17, wherein the obtaining the importance level of each sentence in the article based on the sentence similarities, to determine the first score comprises: constructing a similarity matrix based on the sentence similarities; constructing a graphical model based on the similarity matrix by using each sentence as a node and using each sentence similarity as a weight of an edge; obtaining the importance level of each sentence in the article based on the graphical model; and determining the first score through the importance level of each of the plurality of sentences. ### Claim 20 A non-transitory computer-readable medium, storing a computer program, the program, when executed by a processor, cause the processor to perform: extracting a keyword from input information of a user; determining a candidate article according to the keyword; obtaining a plurality of sentences in the candidate article, and determining a score of each of the plurality of sentences; extracting part of the plurality of sentences based on the score of each of the plurality of sentences; and generating return information that includes the extracted part of the plurality of sentences.