WO2018045646A1 - Procédé et dispositif à base d'intelligence artificielle pour interaction humain-machine - Google Patents

Procédé et dispositif à base d'intelligence artificielle pour interaction humain-machine Download PDF

Info

Publication number
WO2018045646A1
WO2018045646A1 PCT/CN2016/108418 CN2016108418W WO2018045646A1 WO 2018045646 A1 WO2018045646 A1 WO 2018045646A1 CN 2016108418 W CN2016108418 W CN 2016108418W WO 2018045646 A1 WO2018045646 A1 WO 2018045646A1
Authority
WO
WIPO (PCT)
Prior art keywords
type
preset
keyword
score
information
Prior art date
Application number
PCT/CN2016/108418
Other languages
English (en)
Chinese (zh)
Inventor
林英展
凌光
梁一川
王文敏
周超
Original Assignee
北京百度网讯科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京百度网讯科技有限公司 filed Critical 北京百度网讯科技有限公司
Publication of WO2018045646A1 publication Critical patent/WO2018045646A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/903—Querying
    • G06F16/9032—Query formulation
    • G06F16/90332—Natural language query formulation or dialogue systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/903—Querying
    • G06F16/90335—Query processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation

Definitions

  • the present application relates to the field of computers, and in particular to the field of human-computer interaction, and more particularly to a human-computer interaction method and apparatus based on artificial intelligence.
  • Artificial Intelligence is a new technical science that studies and develops theories, methods, techniques, and application systems for simulating, extending, and extending human intelligence.
  • Artificial intelligence is a branch of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that responds in a manner similar to human intelligence. Research in this area includes robotics, speech recognition, image recognition, Natural language processing and expert systems. Artificial intelligence is increasingly integrated into human-computer interaction. Human-computer interaction combined with artificial intelligence can analyze user needs and feed back the answers that users expect.
  • the commonly used method is: adopting the rule matching method, continuously fetching the latest content from the massive information, and configuring the rule matching template corresponding to the requirement type, when the user inputs the problem and When the rule matches the template match, the requirement type is determined, and the answer corresponding to the requirement type is generated.
  • the rule matching template fails to timely cover the keywords in the latest content, the user cannot be recognized.
  • the input question is such as the movie name of the movie to be released, and thus the answer to the demand for a group purchase ticket such as the user's upcoming movie is not recognized.
  • users often enter questions that are often Now, for example, in the case of a colloquial expression or a typo, since the keyword in the rule matching is a word conforming to the expression specification, it is impossible to recognize the name of the movie input by the user, such as a colloquial expression or a typo, and thus cannot be recognized. Such as the need for a group purchase coupon for a movie that the user needs to be released.
  • the present application provides a human-computer interaction method and apparatus based on artificial intelligence, which is used to solve the technical problems existing in the above background art.
  • the present application provides a human-computer interaction method based on artificial intelligence, the method comprising: receiving an interaction statement input by a user; acquiring a search result corresponding to the interaction statement in the search engine, and searching from a preset type in the search result.
  • the type keyword is extracted, and the type keyword includes: a subject name of the preset type of search result; whether the type keyword is associated with the acquisition requirement of the preset type information, and the judgment result is obtained; Generate an answer corresponding to the interactive statement.
  • the present application provides a human-computer interaction device based on artificial intelligence, the device includes: a receiving unit configured to receive an interaction statement input by a user; and a processing unit configured to acquire a search result corresponding to the interaction statement in the search engine And extracting a type keyword from the preset type of search results in the search result, the type keyword includes: a subject name of the preset type of search result; and a determining unit configured to determine whether the type keyword is related to the preset type
  • the acquisition request of the information is associated with the judgment result; and the generation unit is configured to generate an answer corresponding to the interaction statement in a manner corresponding to the judgment result.
  • the artificial intelligence-based human-computer interaction method and apparatus receives an interaction statement input by a user, acquires a search result corresponding to an interaction statement in a search engine, and extracts a type from a preset type of search result in the search result.
  • the keyword is determined whether the type keyword is associated with the acquisition requirement of the preset type of information, and the judgment result is obtained; and the answer corresponding to the interaction statement is generated in a manner corresponding to the judgment result.
  • the search engine is used to update the query result corresponding to the query and the error correction and optimization of the query, and the type keyword of the movie name is extracted from the search result corresponding to the interactive sentence input by the user in the human-computer interaction. .
  • the question input by the user includes a type keyword of a movie name such as the latest release movie, a colloquial expression in the interactive sentence such as a movie name or a typo in the movie name, the correct type keyword can be extracted, Rear Continued protection of the need for identification of interactive statements.
  • the rule matching determines whether the interactive sentence input by the user has an acquisition requirement of information of a preset type such as a movie type, and it is determined to adopt a strong probability threshold or a standard probability threshold. Comparing the extracted common word probability of a type keyword such as a movie name with a strong probability threshold or a standard probability threshold according to a pre-calculated common word probability of a type keyword such as a mass movie name, and finally determining the extracted movie. Whether there is a demand for movies. The manner in which the answer is generated is determined based on whether the extracted type keyword such as the movie name is related to the acquisition demand of the information of the preset type such as the movie type. Improve the accuracy of the need identification of interactive statements entered by the user.
  • FIG. 1 illustrates an exemplary system architecture of an artificial intelligence based human-computer interaction method or apparatus that can be applied to the present application
  • FIG. 2 shows a flow chart of one embodiment of an artificial intelligence based human-computer interaction method according to the present application
  • FIG. 3 shows an exemplary flowchart of a score extraction type keyword according to a preset type of search result
  • 4A shows an exemplary effect diagram of extracting a type keyword from search results in a search engine corresponding to a PC side
  • 4B shows an exemplary effect diagram of extracting a type keyword from search results in a search engine corresponding to a mobile terminal
  • FIG. 5 is a flow chart showing another embodiment of an artificial intelligence based human-computer interaction method according to the present application.
  • FIG. 6 illustrates an exemplary flowchart for identifying whether an interaction statement is associated with an acquisition requirement of a preset type of information by rule matching
  • FIG. 7 shows an exemplary flowchart of determining whether to filter a movie name
  • FIG. 8 illustrates an exemplary flowchart of an artificial intelligence based human-computer interaction method according to the present application
  • FIG. 9 is a schematic structural diagram of an embodiment of an artificial intelligence based human-machine interaction apparatus according to the present application.
  • FIG. 10 is a schematic structural diagram of a computer system suitable for implementing an artificial intelligence based human-machine interaction apparatus according to an embodiment of the present application.
  • FIG. 1 illustrates an exemplary system architecture 100 of an embodiment of an artificial intelligence based human-computer interaction method or apparatus that can be applied to the present application.
  • system architecture 100 can include terminal devices 101, 102, 103, network 104, and server 105.
  • the network 104 is used to provide a medium for the transmission link between the terminal devices 101, 102, 103 and the server 105.
  • Network 104 may include various types of connections, such as wired, wireless transmission links, or fiber optic cables, to name a few.
  • the user can interact with the server 105 over the network 104 using the terminal devices 101, 102, 103 to receive or transmit messages and the like.
  • Various communication applications such as an input method application, a browser application, a search application, a word processing application, and the like, may be installed on the terminal devices 101, 102, and 103.
  • the terminal devices 101, 102, 103 may be various electronic devices having a display screen and supporting network communication, including but not limited to smart phones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III, dynamic The video specialist compresses the standard audio layer 3), MP4 (Moving Picture Experts Group Audio Layer IV) player, laptop portable computer and desktop computer, and the like.
  • MP3 players Motion Picture Experts Group Audio Layer III, dynamic The video specialist compresses the standard audio layer 3
  • MP4 Moving Picture Experts Group Audio Layer IV
  • the terminal device 101, 102, 103 can be configured with artificial intelligence based human-computer interaction
  • the artificial intelligence-based human-computer interaction application can perform semantic recognition on the content input by the user, such as voice, recognize the user's intention, and generate an answer corresponding to the content input by the user according to the recognized intention.
  • the server 105 can receive the search request sent by the terminal devices 101, 102, and 103, find out the search result, and send the search result to the terminal devices 101, 102, and 103.
  • FIG. 2 illustrates a flow 200 of one embodiment of an artificial intelligence based human-computer interaction method in accordance with the present application.
  • the method for human-computer interaction based on artificial intelligence provided by the embodiments of the present application may be performed by the terminal devices 101, 102, and 103 in FIG. 1, for example, artificial intelligence configured by the terminal devices 101, 102, and 103.
  • the human-computer interaction application is executed, and accordingly, the artificial intelligence-based human-computer interaction device may be disposed in the terminal devices 101, 102, and 103.
  • the method includes the following steps:
  • Step 201 Receive an interactive statement input by a user.
  • the human-computer interaction application based on the artificial intelligence on the terminal device receives the human-computer interaction
  • the user inputs the interactive statement by using a keyboard, a voice, or the like.
  • a keyboard a voice
  • the voice may be first recognized to obtain a statement corresponding to the voice.
  • Step 202 Extract a type keyword from a preset type of search result in a search result corresponding to the interaction statement in the search engine.
  • the type keyword includes: a subject name of a preset type of search result.
  • the preset type includes: a movie type, a music type.
  • the search result corresponding to the interactive sentence input by the user in the search engine may be first obtained.
  • the interaction statement input by the user may be sent to the server, and the search engine of the server may use the interactive statement as a search type, ie, query, and use the inverted index to find the search result corresponding to the interactive statement.
  • the type keyword may be extracted from the search result belonging to the movie type of the search result corresponding to the interactive sentence in the search engine, and the type keyword includes the topic name of the search result.
  • the subject name of the movie type search result can be the movie name.
  • the search result when the search result is presented in the html page, the information is defined in the html tag.
  • the search result When the content in the tag of the html page is detected as a link, and the link contains a keyword of the type of the video, the search result may be determined as movie types.
  • the title of the search result can be defined in the title tag, and the content in the title tag can be extracted.
  • extracting the type keyword from the preset type of search results in the search result includes: performing a preset type search in a preset number of search results corresponding to the interaction statement a position of the result, a score of a preset type of search result is calculated, the score indicates a popularity level of the preset type of search result; a judgment score is greater than a score threshold; and when the score is greater than the score threshold, from a preset type of search result Extract the type keywords.
  • FIG. 3 shows an exemplary flowchart of a score extraction type keyword according to a preset type of search result.
  • the preset type keyword as the movie name example, firstly obtain the first 10 search results in the search engine when the query is searched on the pc end and the mobile terminal, and the first 10 search results and the movie "Forrest Gump" The name of the movie in the relevant search results.
  • the position of the movie result in the first 10 search results on the PC side and the position of the movie result in the first 10 search results on the mobile terminal are respectively taken as input parameters of the nonlinear function, and the score on the PC end and the score of the mobile terminal are obtained. Then, with different weights, the PC-side score and the mobile terminal's score are fitted to a final score final_score.
  • a threshold threshold can be divided according to the scores of the search results corresponding to the queries of the plurality of strong demands.
  • a score threshold threshold can be divided according to the scores of the search results corresponding to the query of the plurality of pre-identified strong movie requirements.
  • final_score is greater than the fractional threshold threshold, the power parsed from the search results can be adopted.
  • the first 10 search results in the search engine when searching for the "Gump" and the search engine in the mobile terminal are searched separately.
  • Top 10 search results The positions of the top 10 search results in the search engine and the search results of the movie type in the top 10 search results in the search engine when the mobile terminal performs the search may be separately determined.
  • the "A-Gump" can be searched on the pc side, and the position of the first search result belonging to the movie type in the first 10 search results in the search engine is used as the input parameter of the preset nonlinear function, and the PC end is obtained. Score.
  • the position of the first one of the first 10 search results in the search engine belonging to the movie type may be used as the input parameter of the preset nonlinear function when the "Forrest Gump" searches in the mobile terminal, and the mobile terminal score is obtained.
  • the sum of the product of the pc end score and the preset pc end weight value and the product of the mobile terminal score and the preset mobile terminal weight value is used as an input parameter of the preset linear function, thereby obtaining a score of the movie type search result.
  • FIG. 4A shows an exemplary effect diagram of extracting type keywords from search results in a search engine corresponding to the PC side.
  • the default type of search results are movie type search results, such as the website of the movie "Forrest Gump” and the "Gump of Forrest Gump".
  • the type keyword can be extracted from the search result of the movie type, for example, the movie name "A” is extracted from the theme name of the search result corresponding to the website of the movie playing "Forrest Gump”. Gan Zhengchuan.”
  • FIG. 4B an exemplary effect diagram of extracting a type keyword from search results in a search engine corresponding to a mobile terminal is shown.
  • the search engine has the error correction function, the search result associated with the correct movie name "Forrest Gump” can be obtained.
  • the type keyword can be extracted from the search result of the movie type as the movie name "Forrest Gump".
  • the search engine is used to update the search result corresponding to the query and the error correction and optimization of the query are performed, and a type keyword such as a movie name is extracted from the search result.
  • a type keyword such as the movie name of the latest release movie, or a colloquial expression in the interactive sentence, such as a movie name or a movie name
  • a typo can be searched for from the search engine.
  • the correct movie name is extracted from the result, which provides guarantee for the subsequent identification of the requirement of the interactive statement, such as the requirement of the movie type.
  • the word does not exist in the rule matching template, resulting in the interaction statement.
  • the word is not recognized in the requirement identification, and the requirement cannot be identified.
  • the rule matching template only contains the correct word "Forrest Gump” As a result, the word cannot be recognized in the requirement identification of the interactive statement, and the requirement cannot be identified.
  • the search result corresponding to the obtained interactive statement includes the upcoming release.
  • the movie's movie name will be indefinite", so that the movie name of the upcoming movie can be extracted from the search result corresponding to the interactive sentence.
  • the search result corresponding to the obtained interactive sentence is the correct word corrected by the search engine.
  • Gan Zhengchuan search results obtained by searching, so that the correct movie name "Forrest Gump” can be extracted from the search results.
  • Step 203 Determine whether the type keyword is associated with the acquisition requirement of the preset type information, and obtain the judgment result.
  • the type keyword is extracted from the search result of the preset type in the search result
  • the type keyword is a movie name
  • Step 204 Generate an answer corresponding to the interaction statement in a manner corresponding to the determination result.
  • the answer corresponding to the interaction statement may be generated in a manner corresponding to the determination result.
  • the artificial intelligence based human-computer interaction application can generate the movie name as a common word. answer.
  • the artificial intelligence based human-computer interaction application can generate an answer based on the acquisition requirement of the movie type information. For example, to get a link to the movie coupon for the movie, then generate the answer "I recommend you a few cinemas + links that are close to you and have discounts.”
  • FIG. 5 illustrates a flow 500 of another embodiment of an artificial intelligence based human-computer interaction method in accordance with the present application.
  • the method for human-computer interaction based on artificial intelligence provided by the embodiments of the present application may be performed by the terminal devices 101, 102, and 103 in FIG. 1, for example, artificial intelligence configured by the terminal devices 101, 102, and 103.
  • Step 501 Receive an interactive statement input by a user.
  • the human-computer interaction application based on the artificial intelligence on the terminal device can be used to receive an interactive statement input by the user through a keyboard, a voice, or the like when the human-computer interaction is received. For example, when a user wants to know what a good movie to watch, he can input "What good movie have you recently?" by voice. After receiving the interactive statement for input, the voice may be first recognized to obtain a statement corresponding to the voice.
  • Step 502 a preset type in the search result corresponding to the interaction statement in the search engine Type keywords are extracted from the search results.
  • the search result corresponding to the interaction statement input by the user in the search engine may be first acquired. Then, the score of the preset type of search result may be calculated based on the preset type of search result at the position of the search result corresponding to the interactive sentence. The score of the preset type of search result indicates the popularity of the preset type of search result. When the score is greater than the score threshold, the type keyword can be extracted from the preset type of search result.
  • the first 10 search results in the search engine when searching for the "Gump” and the search engine in the mobile terminal are searched separately.
  • Top 10 search results The positions of the top 10 search results in the search engine and the top 10 search results in the search engine when searching on the mobile terminal can be separately determined by "Forrest Gump". Then, the "A-Gump" can be searched on the pc side, and the position of the first search result belonging to the movie type in the first 10 search results in the search engine is used as the input parameter of the preset nonlinear function, and the PC end is obtained. Score.
  • the position of the first one of the first 10 search results in the search engine belonging to the movie type may be used as the input parameter of the preset nonlinear function when the "Forrest Gump" searches in the mobile terminal, and the mobile terminal score is obtained.
  • the sum of the product of the pc end score and the preset pc end weight value and the product of the mobile terminal score and the preset mobile terminal weight value is used as an input parameter of the preset linear function, thereby obtaining a score of the movie type search result.
  • Step 503 Determine, according to the probability of the common word, whether the type keyword is associated with the acquisition requirement of the preset type of information, and obtain the judgment result.
  • the web crawler type keyword may be pre-selected, and the common word probability of the captured type keyword being a common word may be calculated. Taking the type keyword as the movie name, you can use the web crawler to capture a large number of movie names from the website in advance. Then, the ordinary word probability of the movie named ordinary word is calculated separately.
  • the common word probability of the type keyword extracted by step 502 can be compared with the determined probability threshold strong probability threshold or standard probability threshold according to the common word probability of the pre-calculated massive type keyword.
  • the type keyword is Whether it is associated with the acquisition requirement of the preset type of information, the judgment result is obtained.
  • the common word probability of the movie name extracted by step 502 and the determined probability threshold value may be based on the common word probability of the pre-calculated movie name. Compare. When the common word probability corresponding to the movie name is greater than the determined probability threshold, it may be determined that the movie name is not associated with the acquisition requirement of the movie type information. When the common word probability corresponding to the movie name is less than the determined probability threshold, it may be determined that the movie name is associated with the acquisition requirement of the information of the movie type. Thereby, it is finally determined whether or not the movie name extracted by step 502 has a request for acquisition of information of a movie type.
  • whether the strong threshold or the standard threshold is adopted is determined by the rule matching whether the interaction statement input by the user received in step 501 has a preset type of information.
  • the rule matching template containing the preset requirement keyword matches the interactive sentence input by the user, and determines whether the interactive keyword input by the user includes the demand keyword.
  • the preset demand keywords are “have a sequel?” and “when a release”.
  • the rule matching template matches the interaction statement input by the user, and it is determined that the interaction statement input by the user includes the requirement keyword, it may be determined that the interaction statement input by the user is associated with the acquisition requirement of the information of the movie type.
  • the rule matching template matches the interactive sentence input by the user, and it is determined that the interactive keyword input by the user does not include the demand keyword, it may be determined that the interactive sentence input by the user is not associated with the acquisition requirement of the movie type information.
  • the strong probability threshold may be determined as a probability threshold for comparison with the common word probability corresponding to the type keyword.
  • the standard probability threshold may be determined as a probability threshold for comparison with the common word probability corresponding to the type keyword.
  • FIG. 6 illustrates an exemplary flowchart for identifying whether an interaction statement is associated with an acquisition requirement of a preset type of information by rule matching.
  • the intent of the query and the contents of the slot under the intent of the parsing can be identified by rule matching.
  • Demand for film For example, when the query has a movie intent, that is, the query is associated with the acquisition requirement of the information of the movie type, the slot content of each dimension obtained by the parsing, such as a movie name, an actor, and a type, may be recorded.
  • the query does not have a movie intent, that is, the query is not associated with the acquisition requirement of the movie type information, the slot content can be emptied.
  • FIG. 7 shows an exemplary flowchart for determining whether to filter a movie name.
  • the common word probability is calculated by the massive movie name obtained by the web crawler, and a common word probability dictionary is obtained.
  • the training is divided according to the probability value, and a strong common word set, a general common word set, and a movie demand word set are obtained. Strong probability threshold, standard probability threshold.
  • the current movie name is subjected to different degrees of probability threshold detection, that is, a strong probability threshold corresponding to the result obtained by matching the movie name and the rule matched in the search result corresponding to the interactive sentence input by the user or Standard probability thresholds are compared. Only when the common word probability of the current movie name is lower than the threshold, the movie name is finally retained, that is, the answer is generated based on the movie name, otherwise the movie name is filtered, that is, the movie name is generated as an ordinary word.
  • degrees of probability threshold detection that is, a strong probability threshold corresponding to the result obtained by matching the movie name and the rule matched in the search result corresponding to the interactive sentence input by the user or Standard probability thresholds are compared. Only when the common word probability of the current movie name is lower than the threshold, the movie name is finally retained, that is, the answer is generated based on the movie name, otherwise the movie name is filtered, that is, the movie name is generated as an ordinary word.
  • the strong probability threshold is compared with the common word probability of the current movie name. It is judged whether the type keyword is associated with the acquisition requirement of the preset type of information. Thus, the movie name is more carefully filtered to improve the accuracy of the demand identification.
  • Step 504 Determine, according to the determination result, a manner of generating an answer corresponding to the question.
  • the step 503 is based on the probability that the type keyword is a common word, it is determined whether the type keyword is associated with the acquisition requirement of the preset type information, and after the judgment result is obtained, the generation problem may be determined according to the judgment result. The way the answer is.
  • the answer is generated based on the type keyword as a common word; when the type keyword is associated with the acquisition requirement of the preset type of information, based on the preset type The information is obtained by the need to generate an answer.
  • the artificial intelligence based human-computer interaction should be Use the movie name as a common word to generate an answer.
  • the artificial intelligence based human-computer interaction application can generate an answer based on the acquisition requirement of the movie type information. For example, to get a link to the movie coupon for the movie, then generate the answer "I recommend you a few cinemas + links that are close to you and have discounts.”
  • FIG. 8 illustrates an exemplary flowchart of an artificial intelligence based human-computer interaction method according to the present application.
  • the user can input the interactive statement through the keyboard, voice, etc. when the human-computer interaction application based on the artificial intelligence on the terminal device receives the human-computer interaction. For example, when a user wants to know what a good movie to watch, he can input "What good movie have you recently?" by voice.
  • the voice After receiving the interactive statement for input, the voice may be first recognized to obtain a statement corresponding to the voice.
  • the search engine-based parsing system can obtain the search result of the query, that is, the content input by the user in the search engine, and extract the movie related content.
  • the score of the search result of the movie type may be calculated based on the search result belonging to the movie type at the position of the preset number of search results corresponding to the interactive sentence, and whether the score is greater than the score threshold. When the score is greater than the score threshold, a type keyword such as a movie name is extracted from the search result of the movie type.
  • the rule matching based parsing system can perform rule matching on the query, that is, the content input by the user, and extract the movie related content. Then, the rule matching method can be used to identify whether the interactive sentence input by the user is associated with the acquisition request of the movie type information.
  • the strong probability threshold may be used as a probability threshold for comparison with the common word probability corresponding to the type keyword such as the movie name.
  • the standard probability threshold may be used as a probability threshold for comparison with a common word probability corresponding to a type keyword such as a movie name.
  • the filtering system based on the common word probability can adopt the web crawler crawling type keyword in advance, and calculate the common word probability that the captured type keyword is a common word. Then, it can be judged whether the type of the keyword, for example, the common word probability corresponding to the movie name is greater than the comparison probability threshold for the comparison. When the movie name corresponds to the common word probability is less than the sum used to compare it When the threshold is used, it can be determined that the movie name is not associated with the acquisition requirement of the movie type information, that is, the current demand of the user is non-movie, then the current movie name is filtered, and the human-computer interaction application based on artificial intelligence can use the movie name as a common word. To generate an answer.
  • the probability of the common word corresponding to the movie name is greater than the probability threshold for comparison with the movie name, it may be determined that the movie name is associated with the acquisition requirement of the information of the movie type, that is, the current demand of the user is a movie, and the current movie name is retained, which may be based on artificial intelligence.
  • the human-computer interaction application can generate an answer based on the acquisition requirements of the information of the movie type. For example, to get a link to the movie coupon for the movie, then generate the answer "I recommend you a few cinemas + links that are close to you and have discounts.”
  • FIG. 9 is a schematic structural diagram of an embodiment of an artificial intelligence based human-machine interaction apparatus according to the present application.
  • the artificial intelligence based human-machine interaction apparatus 900 includes a receiving unit 901, a processing unit 902, a determining unit 903, and a generating unit 904.
  • the receiving unit 901 is configured to receive an interaction statement input by the user;
  • the processing unit 902 is configured to acquire a search result corresponding to the interaction statement in the search engine, and extract the type keyword from the preset type of search result in the search result.
  • the type keyword includes: a subject name of the search result of the preset type;
  • the determining unit 903 is configured to determine whether the type keyword is associated with the acquisition requirement of the preset type of information, and obtain a determination result;
  • the generating unit 904 is configured to The method corresponding to the judgment result generates an answer corresponding to the interaction statement.
  • the preset type includes: a movie type, a music type.
  • the processing unit 902 includes: a calculation subunit (not shown) configured to preset a search result of a preset type based on a preset number of search results corresponding to the interaction statement. a location, a score of a preset type of search result is calculated, the score indicates a popularity level of the preset type of search result; a score judgment subunit (not shown) configured to determine whether the score is greater than a score threshold; the extraction subunit (not shown) And configured to extract the type keyword from the preset type of search result when the score is greater than the score threshold.
  • the calculating subunit is further configured to: search result of the first preset type in the preset number of search results corresponding to the interaction statement when searching on the pc end
  • the position of the first preset type of search result among the preset number of search results corresponding to the interactive sentence when searching on the mobile terminal is used as an input parameter of the preset nonlinear function, and the mobile terminal score is obtained.
  • the sum of the product of the pc end score and the preset pc end weight value and the product of the mobile terminal score and the preset mobile terminal weight value is used as an input parameter of the preset linear function to obtain a score of the preset type search result.
  • the apparatus 900 further includes: a crawling unit (not shown) configured to use a web crawler to capture a type keyword before receiving the interactive statement input by the user;
  • a probability calculation unit (not shown) is configured to calculate a common word probability that the captured type keyword is a common word.
  • the determining unit 903 includes: an identifying subunit (not shown) configured to identify, by using a rule matching manner, whether the interactive statement is associated with an acquisition requirement of a preset type of information, Obtaining a recognition result; a probability threshold determining subunit (not shown) configured to determine, according to the recognition result, a probability threshold for comparing a common word probability corresponding to the type keyword, wherein the probability threshold comprises: a standard probability threshold, which is greater than a standard a strong probability threshold of the probability threshold; a requirement determining subunit (not shown) configured to determine that the type keyword is not related to the preset type of information when the common word probability corresponding to the type keyword is greater than the determined probability threshold The requirement is associated; when the common word probability corresponding to the type keyword is less than the determined probability threshold, determining the type keyword is associated with the acquisition requirement of the preset type of information.
  • the probability threshold determining subunit is further configured to: when the recognition result is that the interaction statement is associated with an acquisition requirement of the preset type of information, use the strong probability threshold as a The probability threshold for comparing the common word probabilities corresponding to the type keywords; when the recognition result is that the interaction statement is not associated with the acquisition requirement of the preset type of information, the standard probability threshold is used as the common word probability corresponding to the type keyword The probability threshold for comparison.
  • the generating unit 904 includes: a first answer generating subunit (not shown) configured to be used when the type keyword is not associated with the acquiring requirement of the preset type of information. Generating an answer based on the type keyword as a common word; a second answer generating subunit (not shown) configured to be based on the preset type information when the type keyword is associated with the acquisition requirement of the preset type of information Generate an answer by getting the demand.
  • FIG. 10 is a block diagram showing the structure of a computer system suitable for implementing the artificial intelligence based human-machine interaction device of the embodiment of the present application. .
  • computer system 1000 includes a central processing unit (CPU) 1001 that can be loaded into a program in random access memory (RAM) 1003 according to a program stored in read only memory (ROM) 1002 or from storage portion 908. And perform various appropriate actions and processes.
  • RAM random access memory
  • ROM read only memory
  • RAM random access memory
  • various programs and data required for the operation of the system 1000 are also stored.
  • the CPU 1001, the ROM 1002, and the RAM 1003 are connected to each other through a bus 1004.
  • An input/output (I/O) interface 1005 is also coupled to bus 1004.
  • the following components are connected to the I/O interface 1005: an input portion 1006 including a keyboard, a mouse, etc.; an output portion 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), and the like, and a speaker; a storage portion 1008 including a hard disk or the like And a communication portion 1009 including a network interface card such as a LAN card, a modem, or the like.
  • the communication section 1009 performs communication processing via a network such as the Internet.
  • Driver 1010 is also coupled to I/O interface 1005 as needed.
  • a removable medium 1011 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory or the like is mounted on the drive 1010 as needed so that a computer program read therefrom is installed into the storage portion 1008 as needed.
  • an embodiment of the present disclosure includes a computer program product comprising a computer program tangibly embodied on a machine readable medium, the computer program comprising program code for executing the method illustrated in the flowchart.
  • the computer program can be downloaded and installed from the network via the communication portion 1009, and/or installed from the removable medium 1011.
  • each block of the flowchart or block diagrams can represent a module, a program segment, or a portion of code that includes one or more logic for implementing the specified.
  • Functional executable instructions can also occur in a different order than that illustrated in the drawings. For example, two successively represented blocks may in fact be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowcharts, and combinations of blocks in the block diagrams and/or flowcharts can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be used A combination of dedicated hardware and computer instructions is implemented.
  • the present application further provides a non-volatile computer storage medium, which may be a non-volatile computer storage medium included in the device described in the above embodiments; It may be a non-volatile computer storage medium that exists alone and is not assembled into the terminal.
  • the non-volatile computer storage medium stores one or more programs, when the one or more programs are executed by a device, causing the device to: receive an interaction statement input by a user; and acquire the interaction statement in the search engine Corresponding search results, and extracting a type keyword from the preset type of search results in the search result, the type keyword includes: a theme name of the preset type of search result; determining the type key Whether the word is associated with the acquisition requirement of the preset type of information, and the judgment result is obtained; and the answer corresponding to the interaction statement is generated in a manner corresponding to the judgment result.

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

L'invention concerne un procédé et un dispositif à base d'intelligence artificielle pour une interaction humain-machine. Un mode de réalisation du procédé comprend les étapes suivantes : recevoir une déclaration interactive saisie par un utilisateur (201) ; extraire un mot-clé de type à partir d'un type prédéfini de résultats de recherche de résultats de recherche correspondant à la déclaration interactive dans un moteur de recherche (202) ; déterminer si le mot-clé de type est associé à une demande d'acquisition pour des informations du type prédéfini, ce qui produit le résultat de détermination (203) ; et, d'une manière correspondant au résultat de détermination, produire une réponse correspondant à la déclaration interactive (204). D'une part, la vitesse rapide de mise à jour et une fonction de correction d'erreur du moteur de recherche sont utilisées pour extraire le mot-clé de type, ce qui assure une reconnaissance de demande d'une déclaration interactive ultérieure. D'autre part, la probabilité de terme commun du mot-clé de type est comparée à un seuil élevé ou à un seuil standard en fonction de la probabilité de terme commun du mot-clé de type, et il est finalement déterminé si le mot-clé de type est associé à une demande d'acquisition pour les informations du type prédéfini. La précision de la reconnaissance de demande est augmentée par rapport à la déclaration interactive saisie par l'utilisateur.
PCT/CN2016/108418 2016-09-09 2016-12-02 Procédé et dispositif à base d'intelligence artificielle pour interaction humain-machine WO2018045646A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201610812567.8 2016-09-09
CN201610812567.8A CN106383875B (zh) 2016-09-09 2016-09-09 基于人工智能的人机交互方法和装置

Publications (1)

Publication Number Publication Date
WO2018045646A1 true WO2018045646A1 (fr) 2018-03-15

Family

ID=57935611

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/108418 WO2018045646A1 (fr) 2016-09-09 2016-12-02 Procédé et dispositif à base d'intelligence artificielle pour interaction humain-machine

Country Status (2)

Country Link
CN (1) CN106383875B (fr)
WO (1) WO2018045646A1 (fr)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110415828A (zh) * 2019-06-21 2019-11-05 深圳壹账通智能科技有限公司 一种基于数据分析的预检信息交互方法及相关设备
CN111984763A (zh) * 2020-08-28 2020-11-24 海信电子科技(武汉)有限公司 一种答问处理方法及智能设备
CN112364128A (zh) * 2020-11-06 2021-02-12 北京乐学帮网络技术有限公司 一种信息处理的方法、装置、计算机设备和存储介质
CN114021560A (zh) * 2021-11-10 2022-02-08 竹间智能科技（上海）有限公司 文本纠错方法及装置、电子设备、存储介质

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11100144B2 (en) 2017-06-15 2021-08-24 Oracle International Corporation Data loss prevention system for cloud security based on document discourse analysis
CN107688950B (zh) * 2017-08-31 2022-03-11 广东美的制冷设备有限公司 家电售后咨询方法、电子设备和计算机可读存储介质
CN109726387A (zh) * 2017-10-31 2019-05-07 科沃斯商用机器人有限公司 人机交互方法和系统
WO2019084810A1 (fr) 2017-10-31 2019-05-09 腾讯科技（深圳）有限公司 Procédé et terminal de traitement d'informations et support d'informations pour ordinateur
CN108536852B (zh) * 2018-04-16 2021-07-23 上海智臻智能网络科技股份有限公司 问答交互方法和装置、计算机设备及计算机可读存储介质
JP7258047B2 (ja) * 2018-05-09 2023-04-14 オラクル・インターナショナル・コーポレイション 収束質問に対する回答を改善するための仮想談話ツリーの構築
CN111414760B (zh) * 2018-12-18 2023-06-16 广东美的白色家电技术创新中心有限公司 自然语言处理方法及相关设备、系统和存储装置
CN110581772B (zh) * 2019-09-06 2020-10-13 腾讯科技（深圳）有限公司 即时通讯消息的交互方法、装置以及计算机可读存储介质
CN110929014B (zh) * 2019-12-09 2023-05-23 联想(北京)有限公司 信息处理方法、装置、电子设备及存储介质
CN111737972A (zh) * 2020-05-20 2020-10-02 华为技术有限公司 人机交互系统中实现自然语言理解的方法和装置
CN112487164A (zh) * 2020-12-01 2021-03-12 中译语通科技（青岛）有限公司 一种人工智能交互方法
CN114124860A (zh) * 2021-11-26 2022-03-01 中国联合网络通信集团有限公司 会话管理方法、装置、设备及存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6665658B1 (en) * 2000-01-13 2003-12-16 International Business Machines Corporation System and method for automatically gathering dynamic content and resources on the world wide web by stimulating user interaction and managing session information
CN104199810A (zh) * 2014-08-29 2014-12-10 科大讯飞股份有限公司 一种基于自然语言交互的智能服务方法及系统
CN104899285A (zh) * 2015-06-04 2015-09-09 百度在线网络技术（北京）有限公司 搜索结果展示方法和装置
CN105260459A (zh) * 2015-10-13 2016-01-20 百度在线网络技术（北京）有限公司 搜索方法和装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140149378A1 (en) * 2006-06-22 2014-05-29 Rohit Chandra Method and apparatus for determining rank of web pages based upon past content portion selections
US8700592B2 (en) * 2010-04-09 2014-04-15 Microsoft Corporation Shopping search engines
CN102955798B (zh) * 2011-08-25 2018-04-17 深圳市世纪光速信息技术有限公司 一种基于搜索引擎的搜索方法及搜索服务器

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6665658B1 (en) * 2000-01-13 2003-12-16 International Business Machines Corporation System and method for automatically gathering dynamic content and resources on the world wide web by stimulating user interaction and managing session information
CN104199810A (zh) * 2014-08-29 2014-12-10 科大讯飞股份有限公司 一种基于自然语言交互的智能服务方法及系统
CN104899285A (zh) * 2015-06-04 2015-09-09 百度在线网络技术（北京）有限公司 搜索结果展示方法和装置
CN105260459A (zh) * 2015-10-13 2016-01-20 百度在线网络技术（北京）有限公司 搜索方法和装置

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110415828A (zh) * 2019-06-21 2019-11-05 深圳壹账通智能科技有限公司 一种基于数据分析的预检信息交互方法及相关设备
CN111984763A (zh) * 2020-08-28 2020-11-24 海信电子科技(武汉)有限公司 一种答问处理方法及智能设备
CN111984763B (zh) * 2020-08-28 2023-09-19 海信电子科技(武汉)有限公司 一种答问处理方法及智能设备
CN112364128A (zh) * 2020-11-06 2021-02-12 北京乐学帮网络技术有限公司 一种信息处理的方法、装置、计算机设备和存储介质
CN112364128B (zh) * 2020-11-06 2024-05-24 北京乐学帮网络技术有限公司 一种信息处理的方法、装置、计算机设备和存储介质
CN114021560A (zh) * 2021-11-10 2022-02-08 竹间智能科技（上海）有限公司 文本纠错方法及装置、电子设备、存储介质

Also Published As

Publication number Publication date
CN106383875B (zh) 2019-10-15
CN106383875A (zh) 2017-02-08

Similar Documents

Publication Publication Date Title
WO2018045646A1 (fr) Procédé et dispositif à base d'intelligence artificielle pour interaction humain-machine
US12147732B2 (en) Analyzing graphical user interfaces to facilitate automatic interaction
US20220214775A1 (en) Method for extracting salient dialog usage from live data
AU2019200437B2 (en) A method to build an enterprise-specific knowledge graph
US11157490B2 (en) Conversational virtual assistant
US9805718B2 (en) Clarifying natural language input using targeted questions
EP3183728B1 (fr) Système et procédé de détection d'énoncé orphelin
US10831796B2 (en) Tone optimization for digital content
US12026194B1 (en) Query modification based on non-textual resource context
US9626622B2 (en) Training a question/answer system using answer keys based on forum content
US9734193B2 (en) Determining domain salience ranking from ambiguous words in natural speech
US20160328467A1 (en) Natural language question answering method and apparatus
US10997223B1 (en) Subject-specific data set for named entity resolution
US10108698B2 (en) Common data repository for improving transactional efficiencies of user interactions with a computing device
US9613093B2 (en) Using question answering (QA) systems to identify answers and evidence of different medium types
US10896222B1 (en) Subject-specific data set for named entity resolution
JP6361351B2 (ja) 発話ワードをランク付けする方法、プログラム及び計算処理システム
CN106663123B (zh) 以评论为中心的新闻阅读器
JP2023002690A (ja) セマンティックス認識方法、装置、電子機器及び記憶媒体
WO2020052060A1 (fr) Procédé et appareil permettant de générer une instruction de correction
WO2020052059A1 (fr) Procédé et appareil de génération d'informations
CN116187341A (zh) 语义识别方法及其装置
CN116978028A (zh) 视频处理方法、装置、电子设备及存储介质
CN116010571A (zh) 知识库构建方法、信息查询方法、装置以及设备
CN115062136A (zh) 基于图神经网络的事件消歧方法及其相关设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16915580

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16915580

Country of ref document: EP

Kind code of ref document: A1