WO2020019745A1 - 一种基于知识库生成机器人幽默性格信息的方法及系统 - Google Patents

一种基于知识库生成机器人幽默性格信息的方法及系统 Download PDF

Info

Publication number
WO2020019745A1
WO2020019745A1 PCT/CN2019/080348 CN2019080348W WO2020019745A1 WO 2020019745 A1 WO2020019745 A1 WO 2020019745A1 CN 2019080348 W CN2019080348 W CN 2019080348W WO 2020019745 A1 WO2020019745 A1 WO 2020019745A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
knowledge base
audience
context
group
Prior art date
Application number
PCT/CN2019/080348
Other languages
English (en)
French (fr)
Inventor
张建军
张汀苒
Original Assignee
张建军
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 张建军 filed Critical 张建军
Priority to JP2020573335A priority Critical patent/JP7079530B2/ja
Priority to US16/965,308 priority patent/US20210117813A1/en
Publication of WO2020019745A1 publication Critical patent/WO2020019745A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/02Knowledge representation; Symbolic representation
    • G06N5/022Knowledge engineering; Knowledge acquisition
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/004Artificial life, i.e. computing arrangements simulating life
    • G06N3/006Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/02Knowledge representation; Symbolic representation
    • G06N5/022Knowledge engineering; Knowledge acquisition
    • G06N5/025Extracting rules from data
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/04Inference or reasoning models
    • G06N5/041Abduction
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/903Querying
    • G06F16/9035Filtering based on additional data, e.g. user or group profiles
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/906Clustering; Classification

Definitions

  • the present invention relates to the field of artificial intelligence technology, and in particular, to a method and system for generating humorous character information of a robot based on a knowledge base.
  • Humor makes people feel fun, enjoyable, and laughter, interweaving cognitive, psychological, physical, and social activities.
  • Humor has the following theories: sense of superiority, release, dissonance, and minor violations.
  • the theory of superiority holds that humor originates from the belief that one has a sense of superiority due to other people's abnormal, unfortunate, or shortcomings.
  • the release theory holds that humor originates from the sudden disappearance of psychological pressure.
  • the theory of inconsistency believes that humor comes from the resolution of contradictions, abnormalities, inconsistencies or inconsistencies. Minor violations of the theory suggest that humor stems from minor violations of some conventions without bad consequences.
  • the superiority theory finds the psychological consequences of humor, the release theory focuses on the psychological motivation of humor, and the inconsistency theory finds inconsistency and tries to dispel it.
  • a large category of existing technologies for generating humorous characters of robots is searching and matching ready-made humorous materials from the human humor knowledge base, such as Microsoft's chatbot Xiaobing; the other category is based on the belief that humor comes from superiority, The release of pressure, minor violations, inconsistencies, or resolution of inconsistencies, etc., is mainly to generate word games in specific languages, with weak humorous effects.
  • the present invention provides a method and system for generating robot humorous personality information based on a knowledge base, which can generate robot humorous personality information more effectively than the prior art.
  • the invention provides a method for generating humorous character information of a robot based on a knowledge base.
  • the method includes:
  • the generated information, hypothetical information, or development of the change is displayed.
  • the invention provides a system for generating robot humorous character information based on a knowledge base, including:
  • the first establishment module is used to establish a group knowledge base using various types of group prediction as information classification standards;
  • a judging module for judging a group category to which an audience belongs
  • a second establishment module configured to establish an audience knowledge base using the audience prediction as an information classification criterion from the group knowledge base
  • a generation module which is used to generate one piece of information or develop from this piece of information to another piece of information;
  • the third establishment module is used to establish a context knowledge base
  • a processing module for assuming that changing a piece of information in the context knowledge base causes the generated information to conflict with the information in the audience knowledge base, changing the meaning of the generated information, or changing the development mode Changing the generated information or changing the development according to changed information in the context knowledge base, and displaying the changed generated information, hypothetical information, or changed development.
  • the present invention provides a method and system for generating robot humorous character information based on a knowledge base.
  • the method includes firstly dividing groups, then establishing a group knowledge base, judging the category of the audience, establishing an audience knowledge base, generating a piece of information or development, establishing a context knowledge base, assuming a change in context, changing the generated information or development, and finally Shows changed information, hypothetical information, or changed development.
  • the present invention can more effectively generate humorous character information of the robot, and it is a negation of reasonable negation.
  • the rationality is judged by the contradiction between the displayed information and the information of the audience's knowledge base, so that the audience can understand and Nervous to experience intellectual pride, identity, tacit understanding, relaxation, happiness, and laughter through common background knowledge and unspoken understanding and sharing of secrets, and robot humor.
  • FIG. 1 is a method flowchart of Embodiment 1 of a method for generating humorous character information of a robot based on a knowledge base disclosed in the present invention
  • FIG. 2 is a schematic structural diagram of Embodiment 1 of a system for generating humorous character information of a robot based on a knowledge base disclosed in the present invention.
  • Embodiment 1 it is a flowchart of Embodiment 1 of a method for generating humorous character information of a robot based on a knowledge base disclosed in the present invention.
  • the method may include the following steps:
  • groups can be divided into “Chinese”, “American” and “Japanese” according to nationality.
  • groups can be divided into “doctors”, “teachers”, “peasants”, etc. according to occupations.
  • the group knowledge base that uses various types of group predictions as information classification criteria, instead of organizing the group knowledge base based on the content containing certain keywords as criteria.
  • the information likely to be known by a certain group of people is predictive information.
  • the group prediction information constitutes the group knowledge unit, and the group knowledge unit constitutes the group knowledge base.
  • the simplest group knowledge base can consist of only one group knowledge unit, and the complex one can consist of multiple group knowledge units.
  • a group knowledge base composed of only one group knowledge unit can only face this group; a group knowledge base composed of many group knowledge units can face many groups.
  • the group knowledge base provides a source of information for the audience knowledge base.
  • a group knowledge base consisting of only one "Chinese” group knowledge unit can only make the Chinese feel the robot's humor.
  • “Chinese” group prediction as the standard, "Beijing is a city”, “Beijing is in the north of China”, “China has a long history”, “Vietnam is to the south of China”, and “the Yangtze River is longer than the Yellow River”, etc.
  • the information that the "people” group is likely to know is organized together to establish a "Chinese” group knowledge unit, and then a group knowledge base.
  • the group knowledge base can include both the “Chinese” group knowledge unit, the "American” group knowledge unit, and the “German” group knowledge unit; both the “worker” group knowledge unit and " There is a “knowledge” group knowledge unit, and a “doctor” group knowledge unit.
  • the “junior high school students” group prediction is used as a standard to organize the “junior high school students” group knowledge unit by combining information such as “addition, subtraction, multiplication, and division operations” that the “junior high school students” group is likely to know.
  • information that may be known to the "Chinese junior high school students” group is organized together to establish a "Chinese junior high school students” group knowledge unit.
  • the information that the "Chinese junior high school students” are likely to know includes both the information that the "Chinese” group as a whole is likely to know and the information that the "junior high school” group as a whole is likely to know. ", That is, information that the" Chinese "part of the" junior high school students "group is likely to know, such as information about the Chinese entrance examination.
  • the criteria are used to determine the categories of the audience.
  • the information the audience is likely to know is predictive information.
  • the information in the audience knowledge base directly adopts the group knowledge unit of the group to which the audience belongs.
  • the group knowledge base contains the "Chinese junior high school student” group knowledge unit
  • the "Chinese junior high school student” audience knowledge base directly adopts the "Chinese junior high school student” group knowledge unit.
  • the audience knowledge base using multiple group characteristics as intersections adopts the union of group knowledge units having these characteristics.
  • group knowledge base contains the “Chinese” group knowledge unit and the "Football fan” group knowledge unit and there is no “Chinese football fan” group knowledge unit
  • the "Chinese football fan” audience knowledge base uses the “Chinese” group knowledge unit Union of knowledge units with the "Football Fans” group.
  • the more audience characteristics the more information in the audience's knowledge base. Without the characteristics of the audience, the information in the audience's knowledge base will only be known to all humans.
  • the audience knowledge base using multiple group characteristics as the union audience adopts the intersection of group knowledge units having these characteristics.
  • the audience of "Chinese and football fans” is based on the characteristics of "Chinese” and "football fans”.
  • the group knowledge base contains the “Chinese” group knowledge unit and the "football fan” group knowledge unit, but there is no “Chinese and football fan” group knowledge unit, the "Chinese and football fans” audience knowledge base uses "Chinese The intersection of the group knowledge unit and the "football fan” group knowledge unit.
  • the audience knowledge base makes the information exchange between the audience and the robot have a common context.
  • Information can be in the form of words, phrases, sentences or other languages or words, or in the form of expressions, expressions, actions, behaviors, events, etc. in the form of words or phrases such as words or phrases or sentences.
  • the form of the same word, phrase, or sentence can mean different things in different contexts.
  • the same word, phrase, sentence, tone, expression, action, behavior, event, etc. can mean different meanings in different contexts.
  • a single word can convey information.
  • a famous artist in Hong Kong is called Leslie Cheung, Leslie Cheung, Leslie, Rong Shao, elder brother, Zhang Shu or Zhang Monkey, etc. to express different meanings.
  • the weather is good today
  • the physical state of the weather is expressed in the context of weather forecast
  • the friendly and kindness is mainly expressed in the context of social.
  • information A When information A develops to information B, information A can be either generated information or received information.
  • the information searched from the audience knowledge base and the information just received by the audience using the words constituting the sentence of information A or information B as the keywords are the contents of the context knowledge base. Place the searched keywords that are closely related in content, time, or space to the information to be displayed, or arrange the information obtained by the audience at the most recent time to the front to establish a contextual knowledge base.
  • the contextual knowledge base gives precise meaning to the displayed information form.
  • horse ’s information obtained by the audience in the last minute horse ’s information obtained by the audience in the last few minutes
  • horse ’s information obtained by the audience in the last few minutes Information horse information obtained by the audience within the last few hours, horse information obtained by the audience within the last few days, and horse information obtained by the audience within the last few months are listed in order; the information obtained by the searched audience within the last minute , The information obtained by the audience within the last few minutes, the information obtained by the audience within the last few minutes, and the information obtained by the audience within the last few hours, in order.
  • Message M is incompatible with message N.
  • Information M is not inconsistent with all information in the context knowledge base.
  • the information N is not inconsistent with all information in the context knowledge base except the information M and the information caused by the information M.
  • Information M and all information in the audience's knowledge base may not conflict.
  • Information N and all information in the audience's knowledge base may not contradict each other.
  • the former is more likely than the latter.
  • the general is male.
  • the general in the audience knowledge base may be male or female. The two may not conflict with the information in the audience knowledge base, but the former is generally more likely than the latter.
  • the information A is generated "a person said to be the son of the general, and the general also admitted to be the father of the person.”
  • a piece of information M "General is male” in the context knowledge base is changed to information N "General is female”, and the information A contradicts the information in the audience knowledge base in the hypothetical context.
  • a piece of contextual information is generated: "The child says: I want to eat an apple.”
  • the information A is “It is now time for the child to sleep but not the time for the adult to sleep”, and it is developed to the information B "Mom: Apple is asleep”.
  • a message M in the context is "the similarity between the apple and the child plays a decisive role”, so the analogy goes to message B.
  • the information M is changed to the information N "the same point between the big apple and the adult plays a decisive role”, and the development takes a different way. Big apples develop like adults. From message A to message D, "Child: younger may be asleep, and big one must not be asleep.”
  • Information A and information B can be either simple judgments or composite judgments.
  • a piece of contextual information M is "There is no telephone line as long as the radio is used.” This is a fully conditional development. Suppose that the information M is changed to the information N "As long as the radio is used, there is no telephone line and only the radio is used, there is no telephone line”. This is a development based on the premise as a necessary and sufficient condition. In the hypothetical context, from message A to message B, "My hometown has no telephone line, so my hometown uses radio.”
  • the information A becomes contradictory to the information in the audience's knowledge base in the hypothetical context
  • the information A is changed into another information (set as the information C).
  • Information C contradicts the information in the audience's knowledge base in the default context, and does not conflict with the information in the audience's knowledge base in the hypothetical context.
  • the information M in the context knowledge base is changed to information N, and the information A is changed to information C in the hypothetical context.
  • the meaning of information C in the default context is the same as that in the hypothetical context, but the information C and the information in the audience knowledge base have changed from contradiction in the default context to non-contradiction in the hypothetical context.
  • information A is generated "a person is said to be the son of a general, and the general is acknowledged to be the father of that person.”
  • Information A is reasonable in the default context; suppose a piece of information in the context knowledge base M "General is male” is changed to message N "General is female”, message A is unreasonable in the hypothetical context; information A is changed to message C in the hypothetical context. "A person is said to be the son of the general, but the general Do n’t admit it is the father of this person ”, message C is unreasonable in the default context and reasonable in the hypothetical context; message C is displayed.
  • the information M in the context knowledge base is changed to information N
  • the information A is changed to information C in the hypothetical context.
  • the meaning of information C in the default context is different from that in the hypothetical context, and the information C and the information in the audience's knowledge base have changed from contradiction in the default context to non-contradiction in the hypothetical context.
  • a piece of contextual information is generated: "The child says: I want to eat an apple.”
  • the information A is "It is now time for the child to sleep but not the time for the adult to sleep”, and it is developed to the information B "Mom: Apple is asleep”.
  • a message M in the context is "the similarity between the apple and the child plays a decisive role", so the analogy goes to message B.
  • the information M is changed to the information N "the same point of the big apple and the adult plays a decisive role”, and the development takes a different way. Big apples develop like adults. From message A to message D, "Child: younger may be asleep, and big one must not be asleep.” Contextual information can be displayed, information A to information B can be displayed, and information A to information D can be displayed.
  • the resulting generated information is displayed.
  • the information C is displayed, the information E is displayed first and then the information N is displayed, or the information A and D are displayed first.
  • Information A and information E have different meanings and the same form, and displaying information A and E has the same effect.
  • the unique theoretical basis of the present invention is that humor is a negation of a reasonable subconscious negation.
  • the invention simulates human psychological activities based on the characteristics of a large amount of robot information storage and fast calculation speed, and makes the robot's sense of humor reflected in the robot's clever handling of rationality through comprehensive technical means.
  • the robot first consciously negates the reasonable negation, and the audience subconsciously traces this process back to the robot.
  • Rationality is judged by the contradiction between the information displayed and all the information in the fields of nature, society, and society, which are known and recognized by the audience. Contradictions are unreasonable, and non-contradictions are reasonable. For example, according to the same law of thinking, a certain form of expression is reasonable in a certain context.
  • the technical means of the present invention is divided into three levels.
  • the first layer of technical means is to share the displayed content from contradiction to non-contradiction by unconsciously sharing the purpose of humor, which mainly involves the fields of psychology and philosophy.
  • the second level of technical means is to achieve the purpose of sharing indiscriminately through the common context of the audience and the robot, and to achieve the purpose of displaying the content from contradiction to non-contradiction by changing from the default context to the hypothetical context, mainly involving logic And the field of linguistics.
  • the third level of technical means is to establish a contextual knowledge base to achieve the purpose of having a common context, and to change the database to achieve the purpose of changing the context.
  • the contextual knowledge base is mainly established through the audience knowledge base, and the audience knowledge is established through the group knowledge base. Libraries, mainly related to the computer field.
  • the present invention generates humor from general information, is rich in content, and has obvious effects. It overcomes the limitations of other technologies for generating humorous characters of robots, which rely on ready-made human humorous materials and mainly generate word games.
  • the effect of generating robot humorous information can be tested.
  • One test method is to make the robot being tested answer the question in the Turing test.
  • the success rate of the robot to deceive humans is set to S.
  • some people with a strong sense of humor will answer the question with the robot humorously.
  • the success rate of robots deceiving humans is set to T; then the success ratio of robots to imitate human humor and other human characteristics is T / S. If T / S is close to or greater than 1, it indicates that the robot is more successful in imitating human humor.
  • Another test method is to let some people with a strong sense of humor answer the question anonymously and humorously with the robot, and the human audience judges the humorous score of each respondent; the ranking of the robot among all the respondents is the performance of the robot's sense of humor . If the ranking is in the top 50 percent, the robot has a stronger sense of humor than humans.
  • FIG. 2 it is a schematic structural diagram of Embodiment 1 of a system for generating humorous character information of a robot based on a knowledge base disclosed in the present invention.
  • the system may include:
  • a dividing module 201 configured to divide a group
  • groups can be divided into “Chinese”, “American” and “Japanese” according to nationality.
  • groups can be divided into “doctors”, “teachers”, “peasants”, etc. according to occupations.
  • the first establishing module 202 is configured to establish a group knowledge base using various types of group predictions as information classification standards;
  • the group knowledge base that uses various types of group predictions as information classification criteria, instead of organizing the group knowledge base based on the content containing certain keywords as criteria.
  • the information likely to be known by a certain group of people is predictive information.
  • the group prediction information constitutes the group knowledge unit, and the group knowledge unit constitutes the group knowledge base.
  • the simplest group knowledge base can consist of only one group knowledge unit, and the complex one can consist of multiple group knowledge units.
  • a group knowledge base composed of only one group knowledge unit can only face this group; a group knowledge base composed of many group knowledge units can face many groups.
  • the group knowledge base provides a source of information for the audience knowledge base.
  • a group knowledge base consisting of only one "Chinese” group knowledge unit can only make the Chinese feel the robot's humor.
  • “Chinese” group prediction as the standard, "Beijing is a city”, “Beijing is in the north of China”, “China has a long history”, “Vietnam is to the south of China”, and “the Yangtze River is longer than the Yellow River”, etc.
  • the information that the "people” group is likely to know is organized together to establish a "Chinese” group knowledge unit, and then a group knowledge base.
  • the group knowledge base can include both the “Chinese” group knowledge unit, the "American” group knowledge unit, and the “German” group knowledge unit; both the “worker” group knowledge unit and " There is a “knowledge” group knowledge unit, and a “doctor” group knowledge unit.
  • the “junior high school students” group prediction is used as a standard to organize the “junior high school students” group knowledge unit by combining information such as “addition, subtraction, multiplication, and division operations” that the “junior high school students” group is likely to know.
  • information that may be known to the "Chinese junior high school students” group is organized together to establish a "Chinese junior high school students” group knowledge unit.
  • the information that the "Chinese junior high school students” are likely to know includes both the information that the "Chinese” group as a whole is likely to know and the information that the "junior high school” group as a whole is likely to know. ", That is, information that the" Chinese "part of the" junior high school students "group is likely to know, such as information about the Chinese entrance examination.
  • the information of the "Chinese male junior high school students” group prediction is organized together to form a "Chinese male junior high school students” group knowledge unit.
  • a judging module 203 configured to judge a group category to which an audience belongs
  • the criteria are used to determine the categories of the audience.
  • a second establishing module 204 configured to establish an audience knowledge base using the audience prediction as an information classification criterion from the group knowledge base;
  • the information the audience is likely to know is predictive information.
  • the information in the audience knowledge base directly adopts the group knowledge unit of the group to which the audience belongs.
  • the group knowledge base contains the "Chinese junior high school student” group knowledge unit
  • the "Chinese junior high school student” audience knowledge base directly adopts the "Chinese junior high school student” group knowledge unit.
  • the audience knowledge base using multiple group characteristics as intersections adopts the union of group knowledge units having these characteristics.
  • group knowledge base contains the “Chinese” group knowledge unit and the "Football fan” group knowledge unit and there is no “Chinese football fan” group knowledge unit
  • the "Chinese football fan” audience knowledge base uses the “Chinese” group knowledge unit Union of knowledge units with the "Football Fans” group.
  • the more audience characteristics the more information in the audience's knowledge base. Without the characteristics of the audience, the information in the audience's knowledge base will only be known to all humans.
  • the audience knowledge base using multiple group characteristics as the union audience adopts the intersection of group knowledge units having these characteristics.
  • the audience of "Chinese and football fans” is based on the characteristics of "Chinese” and "football fans”.
  • the group knowledge base contains the “Chinese” group knowledge unit and the "football fan” group knowledge unit, but there is no “Chinese and football fan” group knowledge unit, the "Chinese and football fans” audience knowledge base uses "Chinese The intersection of the group knowledge unit and the "football fan” group knowledge unit.
  • the audience knowledge base makes the information exchange between the audience and the robot have a common context.
  • a generating module 205 configured to generate a piece of information or develop from this piece of information to another piece of information;
  • Information can be in the form of words, phrases, sentences or other languages or words, or in the form of expressions, expressions, actions, behaviors, events, etc. in the form of words or phrases such as words or phrases or sentences.
  • the form of the same word, phrase, or sentence can mean different things in different contexts.
  • the same word, phrase, sentence, tone, expression, action, behavior, event, etc. can mean different meanings in different contexts.
  • a single word can convey information.
  • a famous artist in Hong Kong is called Leslie Cheung, Leslie Cheung, Leslie, Rong Shao, elder brother, Zhang Shu or Zhang Monkey, etc. to express different meanings.
  • the weather is good today
  • the physical state of the weather is expressed in the context of weather forecast
  • the friendly and kindness is mainly expressed in the context of social.
  • information A When information A develops to information B, information A can be either generated information or received information.
  • a third establishing module 206 configured to establish a context knowledge base
  • the information searched from the audience knowledge base and the information just received by the audience using the words constituting the sentence of information A or information B as the keywords are the contents of the context knowledge base. Place the searched keywords that are closely related in content, time, or space to the information to be displayed, or arrange the information obtained by the audience at the most recent time to the front to establish a contextual knowledge base.
  • the contextual knowledge base gives precise meaning to the displayed information form.
  • horse ’s information obtained by the audience in the last minute horse ’s information obtained by the audience in the last few minutes
  • horse ’s information obtained by the audience in the last few minutes Information horse information obtained by the audience in the last few hours, horse information obtained by the audience in the last few days, and horse information obtained by the audience in the last few months are listed in order; the information obtained by the searched audience within the last minute , The information obtained by the audience within the last few minutes, the information obtained by the audience within the last few minutes, and the information obtained by the audience within the last few hours, in order.
  • a processing module 207 configured to assume that changing a piece of information in the context knowledge base causes the generated information to conflict with the information in the audience knowledge base, changing the meaning of the generated information, or changing the development mode Changing, changing the generated information or changing the development according to the changed information in the context knowledge base, and displaying the changed generated information, hypothetical information, or changed development.
  • Message M is incompatible with message N.
  • Information M is not inconsistent with all information in the context knowledge base.
  • the information N is not inconsistent with all information in the context knowledge base except the information M and the information caused by the information M.
  • Information M and all information in the audience's knowledge base may not conflict.
  • Information N and all information in the audience's knowledge base may not contradict each other.
  • the former is more likely than the latter.
  • the general is male.
  • the general in the audience knowledge base may be male or female. The two may not conflict with the information in the audience knowledge base, but the former is generally more likely than the latter.
  • the information A is generated "a person said to be the son of the general, and the general also admitted to be the father of the person.”
  • a piece of information M "General is male” in the context knowledge base is changed to information N "General is female”, and the information A contradicts the information in the audience knowledge base in the hypothetical context.
  • a piece of contextual information is generated: "The child says: I want to eat an apple.”
  • the information A is “It is now time for the child to sleep but not the time for the adult to sleep”, and it is developed to the information B "Mom: Apple is asleep”.
  • a message M in the context is "the similarity between the apple and the child plays a decisive role”, so the analogy goes to message B.
  • the information M is changed to the information N "the same point between the big apple and the adult plays a decisive role”, and the development takes a different way. Big apples develop like adults. From message A to message D, "Child: younger may be asleep, and big one must not be asleep.”
  • Information A and information B can be either simple judgments or composite judgments.
  • a piece of contextual information M is "There is no telephone line as long as the radio is used.” This was developed in a fully conditional manner. Suppose that the information M is changed to the information N "As long as the radio is used, there is no telephone line and only the radio is used, there is no telephone line”. This is a development based on the premise as a necessary and sufficient condition. In the hypothetical context, from message A to message B, "My hometown has no telephone line, so my hometown uses radio.”
  • the information A becomes contradictory to the information in the audience's knowledge base in the hypothetical context
  • the information A is changed into another information (set as the information C).
  • Information C contradicts the information in the audience's knowledge base in the default context, and does not conflict with the information in the audience's knowledge base in the hypothetical context.
  • the information M in the context knowledge base is changed to information N, and the information A is changed to information C in the hypothetical context.
  • the meaning of information C in the default context is the same as that in the hypothetical context, but the information C and the information in the audience knowledge base have changed from contradiction in the default context to non-contradiction in the hypothetical context.
  • information A is generated "a person is said to be the son of a general, and the general is acknowledged to be the father of that person.”
  • Information A is reasonable in the default context; suppose a piece of information in the context knowledge base M "General is male” is changed to message N "General is female”, message A is unreasonable in the hypothetical context; information A is changed to message C in the hypothetical context. "A person is said to be the son of the general, but the general Do n’t admit it is the father of this person ”, message C is unreasonable in the default context and reasonable in the hypothetical context; message C is displayed.
  • the information M in the context knowledge base is changed to information N
  • the information A is changed to information C in the hypothetical context.
  • the meaning of information C in the default context is different from that in the hypothetical context, and the information C and the information in the audience's knowledge base have changed from contradiction in the default context to non-contradiction in the hypothetical context.
  • a piece of contextual information is generated: "The child says: I want to eat an apple.”
  • the information A is "It is now time for the child to sleep but not the time for the adult to sleep”, and it is developed to the information B "Mom: Apple is asleep”.
  • a message M in the context is "the similarity between the apple and the child plays a decisive role", so the analogy goes to message B.
  • the information M is changed to the information N "the same point between the big apple and the adult plays a decisive role”, and the development takes a different way. Big apples develop like adults. From message A to message D, "Child: younger may be asleep, and big one must not be asleep.” Contextual information can be displayed, information A to information B can be displayed, and information A to information D can be displayed.
  • the resulting generated information is displayed.
  • the information C is displayed, the information E is displayed first and then the information N is displayed, or the information A and D are displayed first.
  • Information A and information E have different meanings and the same form, and displaying information A and E has the same effect.
  • the unique theoretical basis of the present invention is that humor is a negation of a reasonable subconscious negation.
  • the invention simulates human psychological activities based on the characteristics of large amount of robot information storage and fast calculation speed, and makes the robot's sense of humor reflected in the robot's ingenious processing of rationality through comprehensive technical means.
  • the robot first consciously negates the reasonable negation, and the audience subconsciously traces this process back to the robot.
  • Rationality is judged by the contradiction between the information displayed and all the information in the fields of nature, society, and society, which are known and recognized by the audience. Contradictions are unreasonable, and non-contradictions are reasonable. For example, according to the same law of thinking, a certain form of expression is reasonable in a certain context.
  • the technical means of the present invention is divided into three levels.
  • the first layer of technical means is to share the displayed content from contradiction to non-contradiction by unconsciously sharing the purpose of humor, which mainly involves the fields of psychology and philosophy.
  • the second level of technical means is to achieve the purpose of sharing indiscriminately through the common context of the audience and the robot, and to achieve the purpose of displaying the content from contradiction to non-contradiction by changing from the default context to the hypothetical context, mainly involving logic And the field of linguistics.
  • the third level of technical means is to establish a contextual knowledge base to achieve the purpose of having a common context, and to change the database to achieve the purpose of changing the context.
  • the contextual knowledge base is mainly established through the audience knowledge base, and the audience knowledge is established through the group knowledge base. Libraries, mainly related to the computer field.
  • the present invention generates humor from general information, is rich in content, and has obvious effects. It overcomes the limitations of other technologies for generating humorous characters of robots, which rely on ready-made human humorous materials and mainly generate word games.
  • the effect of generating robot humorous information can be tested.
  • One test method is to make the robot being tested answer the question in the Turing test.
  • the success rate of the robot to deceive humans is set to S.
  • some people with a strong sense of humor will answer the question with the robot humorously.
  • the success rate of robots deceiving humans is set to T; then the success ratio of robots to imitate human humor and other human characteristics is T / S. If T / S is close to or greater than 1, it indicates that the robot is more successful in imitating human humor.
  • Another test method is to let some people with a strong sense of humor answer the question anonymously and humorously with the robot, and the human audience judges the humorous score of each respondent; the ranking of the robot among all the respondents is the performance of the robot's sense of humor . If the ranking is in the top 50 percent, the robot has a stronger sense of humor than humans.
  • RAM random access memory
  • ROM read-only memory
  • electrically programmable ROM electrically erasable programmable ROM
  • registers hard disks, removable disks, CD-ROMs, or in technical fields Any other form of storage medium known in the art.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明是一种基于知识库生成机器人幽默性格信息的方法及系统。方法包括:划分群体,建立以各类群体预知为信息分类标准的群体知识库,判断受众所属的群体类别,从群体知识库中建立以受众预知为信息分类标准的受众知识库,生成一条信息或者由这条信息发展到另一条信息,建立语境知识库,假设改变语境知识库中的一条信息,根据语境知识库中改变的信息改变生成的信息或改变发展,显示改变的生成的信息、假设的信息或改变的发展。本发明相对于现有技术能够更加有效的生成机器人幽默性格信息,使受众能够由不理解和紧张到通过共同的背景知识心照不宣地理解和共享秘密而体验到智力自豪感、认同感、默契、放松、愉快和发笑,体验到机器人的幽默。

Description

一种基于知识库生成机器人幽默性格信息的方法及系统
本申请要求于2018年07月27日提交中国专利局、申请号为201810843762.6、发明名称为“一种基于知识库生成机器人幽默性格信息的方法及系统”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本发明涉及人工智能技术领域,尤其涉及一种基于知识库生成机器人幽默性格信息的方法及系统。
背景技术
幽默使人感到有趣、愉快和发笑,交织着认知、心理、生理和社会活动。幽默主要有下面的几种理论:优越感、释放、不协调和轻微违反等。其中,优越感理论认为幽默来源于认为由于别人的不正常、不幸或缺点等负面特点而使自己有优越感。释放理论认为幽默来源于原有的心理压力突然消失。不协调理论认为幽默来源于矛盾、不正常、不协调或不协调的消解。轻微违反理论认为幽默来源于轻微违反了有些常规而没有恶果。优越感理论发现了幽默的心理结果,释放理论注重幽默的心理动机,不协调理论发现了不协调而且试图消解它。
生成机器人幽默性格的现有的技术中的一大类是从人类幽默知识库中搜索和匹配现成的幽默材料,例如微软的聊天机器人小冰;另一大类是基 于认为幽默来源于优越感、压力释放、轻微违反、不协调或者解决不协调等,主要是生成些特定语种的文字游戏,幽默效果微弱。
因此,如何更加有效的生成机器人幽默性格信息是一项亟待解决的问题。
发明内容
有鉴于此,本发明根据人类幽默的本质及机器人的特点,提供了一种基于知识库生成机器人幽默性格信息的方法及系统,相对于现有技术能够更加有效的生成机器人幽默性格信息。
本发明提供了一种基于知识库生成机器人幽默性格信息的方法,所述方法包括:
划分群体;
建立以各类群体预知为信息分类标准的群体知识库;
判断受众所属的群体类别;
从所述群体知识库中建立以受众预知为信息分类标准的受众知识库;
生成一条信息或者由这条信息发展到另一条信息;
建立语境知识库;
假设改变语境知识库中的一条信息使所述的生成的信息变得与受众知识库里的信息矛盾、使所述的生成的信息意思改变或使所述的发展方式改变;
根据所述的语境知识库中改变的信息改变所述的生成的信息或改变所述的发展;
显示所述的改变的生成的信息、假设的信息或改变的发展。
本发明提供了一种基于知识库生成机器人幽默性格信息的系统,包括:
划分模块,用于划分群体;
第一建立模块,用于建立以各类群体预知为信息分类标准的群体知识库;
判断模块,用于判断受众所属的群体类别;
第二建立模块,用于从所述群体知识库中建立以受众预知为信息分类标准的受众知识库;
生成模块,用于生成一条信息或者由这条信息发展到另一条信息;
第三建立模块,用于建立语境知识库;
处理模块,用于假设改变语境知识库中的一条信息使所述的生成的信息变得与受众知识库里的信息矛盾、使所述的生成的信息意思改变或使所述的发展方式改变,根据所述的语境知识库中改变的信息改变所述的生成的信息或改变所述的发展,和显示所述的改变的生成的信息、假设的信息或改变的发展。
从上述技术方案可以看出,本发明提供了一种基于知识库生成机器人幽默性格信息的方法和系统。方法包括首先划分群体,然后建立群体知识库,判断受众所属的群体类别,建立以受众知识库,生成一条信息或发展,建立语境知识库,假设改变语境,改变生成的信息或发展,最后显示改变的生成的信息、假设的信息或改变的发展。本发明相对于现有技术能够更加有效的生成机器人幽默性格信息,是对合理的否定之否定,合理性通过显示的信息与受众知识库的信息的矛盾性来判断,使受众能够由不理解和 紧张到通过共同的背景知识心照不宣地理解和共享秘密而体验到智力自豪感、认同感、默契、放松、愉快和发笑,体验到机器人的幽默。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1为本发明公开的一种基于知识库生成机器人幽默性格信息的方法实施例1的方法流程图;
图2为本发明公开的一种基于知识库生成机器人幽默性格信息的系统实施例1的结构示意图。
具体实施方式
下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本发明保护的范围。
如图1所示,为本发明公开的一种基于知识库生成机器人幽默性格信息的方法实施例1的流程图,所述方法可以包括以下步骤:
S101、划分群体;
当需要基于知识库生成机器人幽默性格信息时,首先分别以年龄、性别、教育、专业、职业、阶层、语言、民族、种族、兴趣、爱好、信仰、宗教、地区、国家或文化等群体特点作为标准划分群体。划分群体使受众有针对性。
具体地,群体可以根据国籍分为“中国人”、“美国人”和“日本人”等。
具体地,群体可以根据职业可以分为“医生”、“教师”、“农民”等。
S102、建立以各类群体预知为信息分类标准的群体知识库;
然后,建立以各类群体预知为信息分类标准的群体知识库,而不是以内容里含有某些关键词为标准来组织建立群体知识库。某类群体很可能知道的信息为预知信息。群体预知信息构成群体知识单元,群体知识单元构成群体知识库。群体知识库最简单的可仅由一个群体知识单元构成,复杂的可由多个群体知识单元构成。仅有一个群体知识单元构成的群体知识库只能面向这一个群体;有许多群体知识单元构成的群体知识库能面向许多群体。群体知识库为受众知识库提供信息来源。
例如,仅有一个“中国人”群体知识单元构成的群体知识库仅能使中国人感受到机器人的幽默。以“中国人”群体预知为标准把“北京是个城市”、“北京在中国的北方”、“中国有悠久历史”、“越南在中国的南面”和“长江比黄河长”等等这些“中国人”群体很可能知道的信息组织在一起建立“中国人”群体知识单元,进而建立群体知识库。
具体地,群体知识库里可以既有“中国人”群体知识单元,又有“美 国人”群体知识单元,又有“德国人”群体知识单元;既有“工人”群体知识单元,又有“农民”群体知识单元,又有“律师”群体知识单元,又有“医生”群体知识单元等。
具体地,以“初中生”群体预知为标准把“初中生”群体很可能知道的信息“加减乘除运算规则”等信息组织在一起建立“初中生”群体知识单元。
具体地,以“中国人初中生”预知为标准把“中国人初中生”群体可能知道的信息组织在一起建立“中国人初中生”群体知识单元。“中国人初中生”很可能知道的信息既包括“中国人”群体的整体很可能知道的信息和“初中生”群体整体很可能知道的信息,也包括只有“中国人”群体的“初中生”部分也即“初中生”群体的“中国人”部分才很可能知道的信息,例如关于中国中考的信息。
具体地,以“中国人男性初中生”预知为标准把“中国人男性初中生”群体预知的信息组织在一起形成“中国男性初中生”群体知识单元。
S103、判断受众所属的群体类别;
分别以所属年龄、性别、教育、国家或专业等受众特点为标准分别判断受众所属的群体类别。
例如,判断某受众属于“中国人”群体。
例如,判断某受众属于“男性”、“中国人”、“初中生”群体。
S104、从群体知识库中建立以受众预知为信息分类标准的受众知识库;
从群体知识库中建立以受众预知为信息分类标准的受众知识库。受众很可能知道的信息为预知信息。
具体地,受众所属群体在群体知识库中有现成的群体知识单元时,受众知识库的信息直接采用受众所属群体的群体知识单元。例如,群体知识库里有“中国人初中生”群体知识单元时,“中国人初中生”的受众知识库直接采用“中国人初中生”群体知识单元。
具体地,受众所属群体在群体知识库中没有现成的群体知识单元时,以多个群体特点作为交集的受众的受众知识库采用具有这些特点的群体知识单元的并集。例如“中国人足球迷”是以“中国人”特点和“足球迷”特点作为交集的受众。群体知识库里有“中国人”群体知识单元和“足球迷”群体知识单元而没有“中国人足球迷”群体知识单元时,“中国人足球迷”受众知识库采用“中国人”群体知识单元和“足球迷”群体知识单元的并集。受众的特点越多,受众知识库中的信息就越多。如果无受众的特点,受众知识库的信息就只有全人类都知道的内容。
具体地,受众所属群体在群体知识库中没有现成的群体知识单元时,以多个群体特点作为并集的受众的受众知识库采用具有这些特点的群体知识单元的交集。例如“有中国人也有足球迷”的受众,是以“中国人”特点和“足球迷”特点作为并集的。群体知识库里有“中国人”群体知识单元和“足球迷”群体知识单元而没有“有中国人也有足球迷”群体知识单元时,“有中国人也有足球迷”受众知识库采用“中国人”群体知识单元和“足球迷”群体知识单元的交集。
受众知识库使受众和机器人的信息交流有共同的语境。
S105、生成一条信息或者由这条信息发展到另一条信息;
生成一条信息(设为信息A)或者由这条信息发展到另一条信息(设 为信息B)。信息的形式可以是词语、词组、句子等语言或文字,也可以是词语、词组、句子等语言或文字等形式表达的语气、表情、动作、行为、事件等。
在特定语种中,同一个词、词组或句子的形式在不同的语境下可以表示不同的意思。同一个词、词组、句子、语气、表情、动作、行为、事件等在不同的语境下可以表示不同的意思。
具体地,单个的词就能表达信息。例如称呼香港的一位著名艺人为张国荣、国荣、Leslie、荣少、哥哥、张叔或张猴子等就表达不同的意思。
例如,“他突然站了起来”,既是这句子形式表示了腿快速地由弯曲到伸直的动作的浅层意思,也是这动作在不同的语境中表示生气、高兴、激动、冲动、服从、反对、屈服、赞同、支持、示威、离开、接受指示或者采取行动等不同的深层意思。
例如,“今天天气不错”,在天气预报的语境中表达的是天气的物理状态,在社交的语境中表达的主要是友好和善意。
信息A发展到信息B时,信息A既可以是生成的信息,也可以是接收的信息。
例如,“我有点饿了,我想去吃饭了。”由前一句发展到后一句。两句都是新生成的信息。
例如,对话“你的房子经常漏雨吗?”“不,只有下雨时才漏。”也是由前一句信息发展到后一句。前一句可以是别人现成的信息。
S106、建立语境知识库;
把以构成信息A或信息B句子的词语为关键词从受众知识库里搜索到 的信息和受众刚刚接收到的信息作为语境知识库的内容。把搜索到的含有的关键词在内容上、时间上或空间上与将要显示的信息的关系密切的信息排列在前面,或把受众最近时间获得的信息排列在前面,建立语境知识库。语境知识库使显示的信息形式有准确的意思。
例如:如想要建立“李明在草原上骑马”的语境知识库时,把“李明”、“草原”、“骑”和“马”作为关键词在受众知识库中进行搜索;把搜索到的关于马的信息、关于大型哺乳动物的信息、关于哺乳动物的信息、关于动物的信息和关于生物的信息依先后顺序排列;把搜索到的关于受众所在地点的马的信息、关于其它地点的马的信息、关于其它城市的马信息、关于其它地区的马的信息和关于其它国家的马的信息依先后顺序排序;把搜索到的关于最近数分钟内的马的信息、关于最近数小时内的马的信息、关于最近数天内的马的信息、关于最近数月内马的信息、关于最近数年内马的信息和关于最近数十年内的马的信息依先后顺序排列;把搜索到的受众最近一分钟内获得的马的信息、受众最近数分钟内获得的马的信息、受众最近数十分钟内获得的马的信息、受众最近数小时内获得的马的信息、受众最近数天内获得的马的信息和受众最近数月内获得的马的信息依先后顺序排列;把搜索到的受众最近一分钟内获得的信息、受众最近数分钟内获得的信息、受众最近数十分钟内获得的信息、受众最近数小时内获得的信息依先后顺序排列。
S107、假设改变语境知识库中的一条信息使所述的生成的信息变得与受众知识库里的信息矛盾、使所述的生成的信息意思改变或使所述的发展方式改变;
假设使语境知识库中的一条信息(设为信息M)变化成另一条信息(设为信息N)使所述的生成的信息A变得与受众知识库里的信息矛盾、使所述的生成的信息A意思改变或使所述的信息A发展到信息B的方式改变。例如:由信息M“他今天晚上在家。”假设改为信息N“他今天晚上不在家。”假设改变语境知识库中的一条信息使默认语境变成了假设语境。
信息M与信息N不相容。
信息M与语境知识库里的所有信息不矛盾。信息N与语境知识库里的除了信息M及由信息M引起的信息以外的所有信息不矛盾。
信息M和受众知识库里的所有信息可能不矛盾。信息N和受众知识库里的所有信息可能不矛盾,前者的可能性一般大于后者的可能性。例如一般在默认语境知识库中,将军是男性。事实上在受众知识库中将军可能是男性,也可能是女性,两者与受众知识库中的信息可能不矛盾,但前者的可能性一般大于后者。
想幽默被感知的快些时,使信息M在语境知识库排序中靠前些;想幽默被感知的慢些时,使信息M在语境知识库排序中靠后些。
想有些讽刺的效果时,使信息N与受众知识库里的信息矛盾,而且显示真实、合规、赞赏或友善等正面信息;想有些开玩笑的效果时,使信息N与受众知识库里的信息矛盾,而且显示虚假、违规、贬低或攻击等负面信息。例如:假如李明有两辆自行车,就会分给朋友一辆;假如李明有两辆汽车,就会分给朋友一辆;假如李明有两艘游艇,就会分给朋友一艘。假如李明有两艘宇宙飞船,就会分给朋友一艘。假设信息与客观现实矛盾的可能性越来越大,合理性越来越小。对李明由赞扬到讽刺越来越明显。
想有些滑稽的效果时,使信息N与受众知识库里的信息不矛盾而且把生成的信息的意思变成现实呈现在受众面前。当假设信息变为真实存在时,意思变为现实直接呈现在受众面前,受众无经过需思考就可以理解在默认语境中不合理的信息在假设语境中的合理,浅显易懂,显得滑稽。例如信息“老虎和羊和平地待在一起”在默认的语境下是不合理的,但假设语境中是否合理就需要经过受众的思考。如果假设信息“老虎吃很饱时不会去骚扰猎物”真实存在,把老虎喂得很饱,使得“老虎和羊和平地待在一起”直接在展现在受众面前就显得滑稽。
想有些诡辩效果时,把信息N存在时显示的信息与受众知识库里的信息不矛盾当成信息M存在时显示的信息与受众知识库里的信息不矛盾,即把假设条件下的合理当成默认条件下的合理,也即把在假设语境中的合理当做默认语境中的合理。
想增强幽默的效果时,使信息N的意思或者在信息N存在时显示的信息的意思违背受众不愿意接受的社会规则或不愿认可的事实等。受众因得到虚拟的满足而且难以受到惩罚而产生强烈的智力自豪感。例如对其它类别受众攻击的笑话。
假设使语境知识库中的信息M变成信息N,在假设语境里使信息A变得与受众知识库里的信息矛盾。
例如,生成信息A“一个人说是将军的儿子,将军也承认是这人的父亲”。假设把语境知识库中的一条信息M“将军是男性”改为信息N“将军是女性”,信息A在假设语境里与受众知识库里的信息矛盾。
假设使语境知识库中的信息M变成信息N,在假设语境里使信息A 的意思改变。
例如,生成信息A“如果老板不收回他今天上午对我说的话,我就离开这家公司”,假设把语境知识库中的一条信息M“老板和我的地位是平等的”改为信息N“我被老板解雇了”,信息A的意思由“我强烈地要求老板道歉”变为“我无奈地服从”。
假设使语境知识库中的信息M变成信息N,在假设语境里使信息A发展方式改变。
例如,由信息A“今年降水和温度适当”发展到信息B“今年的庄稼可能有好收成”。语境信息中有一条信息M“降水和温度合适是庄稼有好收成的必要条件”。发展是有前提可能有结论的方式。假设改变信息M为信息N“降水和温度合适是庄稼有好收成的充分条件”,发展是有前提必然有结论的方式。在假设语境中由信息A发展到信息D“今年的庄稼肯定有好收成”。
例如,生成一条语境信息“小孩说:我要吃苹果”。生成信息A“现在是到小孩睡觉而还没到大人睡觉时间”,发展到信息B“妈妈:苹果睡了”。语境中的一条信息M是“苹果和小孩的相同点起决定的作用”,因此类比发展到了信息B。假设把信息M改变为信息N“大苹果和大人的相同点起决定的作用”,发展采用不同的方式。大苹果类比大人发展。由信息A发展到信息D“小孩:小的可能睡了,大的肯定还没睡”。
信息A和信息B既可以是简单判断,也可以是复合判断。
例如,生成A“我的家乡用无线电,所以我的家乡没有电话线”。一条语境信息M是“只要用无线电,就没有电话线”这是以充分条件方式的发 展。假设把信息M改变为信息N“只要用无线电,就没有电话线而且只有用无线电,才没有电话线”,这是以前提为充分必要条件的发展。在假设语境中由信息A发展到信息B“我的家乡没电话线,所以我的家乡用无线电”。
语境和发展的改变可以不只一次。
例如,由“小孩冰淇淋掉地上了”发展到“小孩哭了”,进一步发展到“路人替小孩买了一个冰淇淋”,“进一步发展到小孩仍有一个冰淇淋吃”,进一步发展到“小孩不哭了。
有一条语境信息“只有冰淇淋掉地上了小孩才哭”,“小孩冰淇淋掉地上了”以前提为必要条件的方式,发展到“小孩哭了”。另有一条语境信息“只有小孩哭了路人才会替小孩买冰淇淋”,“小孩哭了”以前提为必要条件方式发展到“路人替小孩买了一个冰淇淋”。
假设这两条语境信息分别改为“没有冰淇淋掉地上小孩也会哭”或者“没有小孩哭路人也会替小孩买冰淇淋”。冰淇淋掉地上不再是小孩哭的必要条件或者小孩哭不再是路人替小孩买冰淇淋的必要条件。由“小孩冰淇淋没掉地上”发展到“小孩哭了”或者发展到“小孩没哭”,由“小孩没哭”进一步发展到“路人替小孩买了一个冰淇淋”,进一步发展到“小孩有两个冰淇淋吃”。
由“如果小孩没有冰淇淋掉地上,小孩就有两个冰淇淋吃”发展到“因为小孩的冰淇淋掉地上了,所以小孩又哭了”。发展的前提和结果本身都是复合判断。
假设对默认语境的改变是对合理的否定。
S108、根据所述的语境知识库中改变的信息改变所述的生成的信息或 改变所述的发展;
因为语境的改变使得信息或发展变得不合理了,所以根据所述的语境知识库中改变的信息改变所述的生成的信息或改变所述的发展。
信息A在假设语境中变得与受众知识库里的信息矛盾时,使信息A变为另一条信息(设为信息C)。信息C在默认语境中是与受众知识库中的信息矛盾的、在假设语境中是与受众知识库中的信息不矛盾的。
具体地,假设使语境知识库中的信息M变成信息N,在假设语境中把信息A改变成信息C。信息C在默认语境的意思和在假设语境中的意思相同,但信息C与受众知识库中的信息由在默认语境中的矛盾变为在假设语境中的不矛盾了。
例如,在默认语境中生成信息A“一个人说是将军的儿子,将军也承认是这人的父亲”,信息A在默认语境里是合理的;假设把语境知识库中的一条信息M“将军是男性”改为信息N“将军是女性”,信息A在假设语境里是不合理的;假设语境中把信息A改变成信息C“一个人说是将军的儿子,但将军不承认是这人的父亲”,信息C在默认语境中是不合理的而在假设语境中是合理的;显示信息C。
具体地,假设使语境知识库中的信息M变成信息N,在假设语境中把信息A改变成信息C。信息C在默认语境的意思和在假设语境中的意思不相同,而且信息C与受众知识库中的信息由在默认语境中的矛盾变为在假设语境中的不矛盾了。
例如,建立知识库,获得在受众知识库里有“王凡导演了中国最多的电影”和“李明在中国导演界生孩子最多”等信息;在默认语境中生成信 息A“王凡是中国最高产的导演”或者“李明不是中国最高产的导演”,信息A在默认语境里是合理的;假设把语境知识库中的一条信息M“这是在讨论电影”改为信息N“这是在讨论生孩子”,信息A在假设语境中就变得不合理了;在假设语境中把信息A改变成信息C“李明是中国最高产的导演”,信息C的意思在默认语境中和在假设语境中是不同的,信息C在默认语境中是不合理的而在假设语境中是合理的;显示信息C。
信息A的意思改变时,使信息A变为另一条信息(设为信息E)。信息E和信息A的形式相同。信息E在假设语境中的意思和在默认语境中的意思不同。
具体地,假设使语境知识库中的信息M变成信息N,在假设语境中把信息A改变成信息E,信息E和信息A的形式相同。信息E在假设语境中的意思和在默认语境中的意思不同。先显示信息E再显示信息N。
例如:生成信息A“如果老板不收回他今天上午对我说的话,我就离开这家公司”,在含有一条信息M“老板和我的地位是平等的”等信息的默认语境中信息A的意思是“强烈地要求老板道歉”;假设把语境知识库中的信息M改为信息N“我被老板解雇了”;把具有“如果老板不收回他今天上午对我说的话,我就离开这家公司”形式和“我强烈地要求老板道歉”意思的信息A变成了具有同样形式但“我无奈地服从”意思的信息E;先显示信息E再显示信息N。
信息A的意思改变或发展方式改变时,使由信息A发展到信息B改变为由信息A发展到另一条信息(设为D)。
具体地,假设使语境知识库中的信息M变成信息N,信息A的意思 变化时,使信息A发展到信息D。信息A的意思改变成为另一个意思时,由这另一个意思发展到信息D。先显示信息A再显示信息D。
例如,生成信息A:“一个员工说:前天我刚一进电梯,里面就突然全黑了,吓死我了”由信息A发展到信息B“经理听了说:我马上安排对电梯安全检查”。语境中的一条信息M是“员工正在谈论安全问题”。在默认的语境中信息A的意思是“员工担忧在电梯里可能会出危险”。假设把信息M改变成信息N“员工在谈论照明问题”,在假设语境里由信息A的意思“员工在抱怨电梯里太黑暗”发展到信息D“给你们每人配一把手电筒”。先显示信息A再显示信息D。
具体地,假设使语境知识库中的信息M变成信息N,信息A发展的方式改变成另一种方式时。信息A由这种改变的发展方式发展到信息D。先显示信息A再显示信息D。
例如,由信息A“今年降水和温度适当”发展到信息B“今年的庄稼可能有好收成”。语境信息中有一条信息M“降水和温度合适是庄稼有好收成的必要条件”。发展是有前提可能有结论的方式。假设改变信息M为信息N“降水和温度合适是庄稼有好收成的充分条件”,发展是有前提必然有结论的方式。由在假设语境中由信息A发展到信息D“今年的庄稼肯定有好收成”。显示信息A发展到信息D。
例如,生成一条语境信息“小孩说:我要吃苹果”。生成信息A“现在是到小孩睡觉而还没到大人睡觉时间”,发展到信息B“妈妈:苹果睡了”。语境中的一条信息M是“苹果和小孩的相同点起决定的作用”,因此类比发展到了信息B。假设把信息M改变为信息N“大苹果和大人的相同点起 决定的作用”,发展采用不同的方式。大苹果类比大人发展。由信息A发展到信息D“小孩:小的可能睡了,大的肯定还没睡”。可以显示语境信息,可以显示信息A发展到信息B,显示信息A发展到信息D。
假设对默认语境的改变是对合理的否定;在假设语境中的改变是对合理的否定之否定,使机器人有幽默感。
S109、显示所述的改变的生成的信息、假设的信息或者改变的发展。
最后显示改变的生成的信息、假设的信息或者改变的发展。具体实施中显示信息C,先显示信息E再显示信息N,或者先显示信息A再显示D等。信息A和信息E的意思不同而形式相同,显示信息A和显示信息E的效果一样。
显示出了机器人的幽默感。
综上所述,本发明特有的理论基础是:幽默是对合理的潜意识地否定之否定。本发明根据机器人信息储存量大,运算速度快的特点对人类的心理活动进行模拟,通过综合技术手段使机器人的幽默感体现在让受众感觉到机器人对合理性的巧妙处理。机器人先有意识地对合理否定之否定,受众潜意识地回溯机器人的这个过程。合理性通过显示的信息与受众所知道并认可的自然、社会和社会等领域的事实和规则等所有的信息的矛盾性来判断,矛盾则不合理,不矛盾则合理。例如根据思维的同一律,确定的表达形式在确定的语境中意思确定才是合理的。合理性针对特定的受众,对一些受众是合理的对另外的受众则不一定。一些人认为某些是事情是真实的,某些规则是应该接受的,另一些人则可能不这样认为,例如多数人认为“地球围绕太阳转”是真实的,极少数人则认为“太阳围绕地球转”是 真实的。
本发明的技术手段分三个层次,第一层技术手段是通过心照不宣地分享显示的内容由矛盾到不矛盾,达到有幽默感的目的,主要涉及心理学和哲学领域。第二层次技术手段是通过受众和机器人拥有共同语境达到心照不宣地分享的目的,通过由默认语境转换到假设语境的达到显示的内容由矛盾到不矛盾的目的,主要涉及逻辑学和语言学领域。第三个层次的技术手段是通过建立语境知识库达到有共同语境的目的,通过改换数据库达到改换语境的目的,主要通过受众知识库建立语境知识库,通过群体知识库建立受众知识库,主要涉及计算机领域。
受众回溯感受机器人生成幽默信息过程,对机器人显示的内容最初感到不合理,感到紧张和困惑,怀疑是自己理解错误或是智能机器人错误,接着判定自己很可能没错同时机器人可能也没错,就试图合理化机器人的表达,后来发现在假设语境中机器人的表达是合理的,是对合理的否定之否定,最后通过共同的背景知识心照不宣地发现和分享这个秘密而感到智力自豪、认同、默契、放松和愉快等而发笑。
本发明从一般性的信息中生成幽默,内容丰富,效果明显,克服了其它生成机器人幽默性格的技术依赖人类现成幽默材料和主要生成文字游戏的局限。
生成机器人幽默信息的效果可测试。一种测试方法是先使被测试的机器人在图灵测试中回答问题,这时机器人骗过人类的成功率设为S;再让幽默感强的一些人和机器人一起幽默地回答问题,这时机器人骗过人类的成功率设为T;则机器人模仿人类的幽默和模仿人类的其它特性的成功比 率是T/S。如果T/S接近或大于1,则表明机器人模仿人类的幽默比较成功。另一种测试方法是让幽默感强的一些人和机器人一起匿名地而且幽默地回答问题,由人类受众判定各个回答者幽默的分数;机器人在所有回答者中的名次,就是机器人幽默感的成绩。如果排名在前百分之五十,说明机器人的幽默感相比人类较强。
如图2所示,为本发明公开的一种基于知识库生成机器人幽默性格信息的系统实施例1的结构示意图,所述系统可以包括:
划分模块201,用于划分群体;
当需要基于知识库生成机器人幽默性格信息时,首先分别以年龄、性别、教育、专业、职业、阶层、语言、民族、种族、兴趣、爱好、信仰、宗教、地区、国家或文化等群体特点作为标准划分群体。划分群体使受众有针对性。
具体地,群体可以根据国籍分为“中国人”、“美国人”和“日本人”等。
具体地,群体可以根据职业可以分为“医生”、“教师”、“农民”等。
第一建立模块202,用于建立以各类群体预知为信息分类标准的群体知识库;
然后,建立以各类群体预知为信息分类标准的群体知识库,而不是以内容里含有某些关键词为标准来组织建立群体知识库。某类群体很可能知道的信息为预知信息。群体预知信息构成群体知识单元,群体知识单元构成群体知识库。群体知识库最简单的可仅由一个群体知识单元构成,复杂 的可由多个群体知识单元构成。仅有一个群体知识单元构成的群体知识库只能面向这一个群体;有许多群体知识单元构成的群体知识库能面向许多群体。群体知识库为受众知识库提供信息来源。
例如,仅有一个“中国人”群体知识单元构成的群体知识库仅能使中国人感受到机器人的幽默。以“中国人”群体预知为标准把“北京是个城市”、“北京在中国的北方”、“中国有悠久历史”、“越南在中国的南面”和“长江比黄河长”等等这些“中国人”群体很可能知道的信息组织在一起建立“中国人”群体知识单元,进而建立群体知识库。
具体地,群体知识库里可以既有“中国人”群体知识单元,又有“美国人”群体知识单元,又有“德国人”群体知识单元;既有“工人”群体知识单元,又有“农民”群体知识单元,又有“律师”群体知识单元,又有“医生”群体知识单元等。
具体地,以“初中生”群体预知为标准把“初中生”群体很可能知道的信息“加减乘除运算规则”等信息组织在一起建立“初中生”群体知识单元。
具体地,以“中国人初中生”预知为标准把“中国人初中生”群体可能知道的信息组织在一起建立“中国人初中生”群体知识单元。“中国人初中生”很可能知道的信息既包括“中国人”群体的整体很可能知道的信息和“初中生”群体整体很可能知道的信息,也包括只有“中国人”群体的“初中生”部分也即“初中生”群体的“中国人”部分才很可能知道的信息,例如关于中国中考的信息。
具体地,以“中国人男性初中生”预知为标准把“中国人男性初中生” 群体预知的信息组织在一起形成“中国男性初中生”群体知识单元。
判断模块203,用于判断受众所属的群体类别;
分别以所属年龄、性别、教育、国家或专业等受众特点为标准分别判断受众所属的群体类别。
例如,判断某受众属于“中国人”群体。
例如,判断某受众属于“男性”、“中国人”、“初中生”群体。
第二建立模块204,用于从群体知识库中建立以受众预知为信息分类标准的受众知识库;
从群体知识库中建立以受众预知为信息分类标准的受众知识库。受众很可能知道的信息为预知信息。
具体地,受众所属群体在群体知识库中有现成的群体知识单元时,受众知识库的信息直接采用受众所属群体的群体知识单元。例如,群体知识库里有“中国人初中生”群体知识单元时,“中国人初中生”的受众知识库直接采用“中国人初中生”群体知识单元。
具体地,受众所属群体在群体知识库中没有现成的群体知识单元时,以多个群体特点作为交集的受众的受众知识库采用具有这些特点的群体知识单元的并集。例如“中国人足球迷”是以“中国人”特点和“足球迷”特点作为交集的受众。群体知识库里有“中国人”群体知识单元和“足球迷”群体知识单元而没有“中国人足球迷”群体知识单元时,“中国人足球迷”受众知识库采用“中国人”群体知识单元和“足球迷”群体知识单元的并集。受众的特点越多,受众知识库中的信息就越多。如果无受众的特点,受众知识库的信息就只有全人类都知道的内容。
具体地,受众所属群体在群体知识库中没有现成的群体知识单元时,以多个群体特点作为并集的受众的受众知识库采用具有这些特点的群体知识单元的交集。例如“有中国人也有足球迷”的受众,是以“中国人”特点和“足球迷”特点作为并集的。群体知识库里有“中国人”群体知识单元和“足球迷”群体知识单元而没有“有中国人也有足球迷”群体知识单元时,“有中国人也有足球迷”受众知识库采用“中国人”群体知识单元和“足球迷”群体知识单元的交集。
受众知识库使受众和机器人的信息交流有共同的语境。
生成模块205,用于生成一条信息或者由这条信息发展到另一条信息;
生成一条信息(设为信息A)或者由这条信息发展到另一条信息(设为信息B)。信息的形式可以是词语、词组、句子等语言或文字,也可以是词语、词组、句子等语言或文字等形式表达的语气、表情、动作、行为、事件等。
在特定语种中,同一个词、词组或句子的形式在不同的语境下可以表示不同的意思。同一个词、词组、句子、语气、表情、动作、行为、事件等在不同的语境下可以表示不同的意思。
具体地,单个的词就能表达信息。例如称呼香港的一位著名艺人为张国荣、国荣、Leslie、荣少、哥哥、张叔或张猴子等就表达不同的意思。
例如,“他突然站了起来”,既是这句子形式表示了腿快速地由弯曲到伸直的动作的浅层意思,也是这动作在不同的语境中表示生气、高兴、激动、冲动、服从、反对、屈服、赞同、支持、示威、离开、接受指示或者采取行动等不同的深层意思。
例如,“今天天气不错”,在天气预报的语境中表达的是天气的物理状态,在社交的语境中表达的主要是友好和善意。
信息A发展到信息B时,信息A既可以是生成的信息,也可以是接收的信息。
例如,“我有点饿了,我想去吃饭了。”由前一句发展到后一句。两句都是新生成的信息。
例如,对话“你的房子经常漏雨吗?”“不,只有下雨时才漏。”也是由前一句信息发展到后一句。前一句可以是别人现成的信息。
第三建立模块206,用于建立语境知识库;
把以构成信息A或信息B句子的词语为关键词从受众知识库里搜索到的信息和受众刚刚接收到的信息作为语境知识库的内容。把搜索到的含有的关键词在内容上、时间上或空间上与将要显示的信息的关系密切的信息排列在前面,或把受众最近时间获得的信息排列在前面,建立语境知识库。语境知识库使显示的信息形式有准确的意思。
例如:如想要建立“李明在草原上骑马”的语境知识库时,把“李明”、“草原”、“骑”和“马”作为关键词在受众知识库中进行搜索;把搜索到的关于马的信息、关于大型哺乳动物的信息、关于哺乳动物的信息、关于动物的信息和关于生物的信息依先后顺序排列;把搜索到的关于受众所在地点的马的信息、关于其它地点的马的信息、关于其它城市的马信息、关于其它地区的马的信息和关于其它国家的马的信息依先后顺序排序;把搜索到的关于最近数分钟内的马的信息、关于最近数小时内的马的信息、关于最近数天内的马的信息、关于最近数月内马的信息、关于最近数年内马 的信息和关于最近数十年内的马的信息依先后顺序排列;把搜索到的受众最近一分钟内获得的马的信息、受众最近数分钟内获得的马的信息、受众最近数十分钟内获得的马的信息、受众最近数小时内获得的马的信息、受众最近数天内获得的马的信息和受众最近数月内获得的马的信息依先后顺序排列;把搜索到的受众最近一分钟内获得的信息、受众最近数分钟内获得的信息、受众最近数十分钟内获得的信息、受众最近数小时内获得的信息依先后顺序排列。
处理模块207,用于假设改变语境知识库中的一条信息使所述的生成的信息变得与受众知识库里的信息矛盾、使所述的生成的信息意思改变或使所述的发展方式改变,根据所述的语境知识库中改变的信息改变所述的生成的信息或改变所述的发展,和显示所述的改变的生成的信息、假设的信息或改变的发展。
假设使语境知识库中的一条信息(设为信息M)变化成另一条信息(设为信息N)使所述的生成的信息A变得与受众知识库里的信息矛盾、使所述的生成的信息A意思改变或使所述的信息A发展到信息B的方式改变。例如:由信息M“他今天晚上在家。”假设改为信息N“他今天晚上不在家。”假设改变语境知识库中的一条信息使默认语境变成了假设语境。
信息M与信息N不相容。
信息M与语境知识库里的所有信息不矛盾。信息N与语境知识库里的除了信息M及由信息M引起的信息以外的所有信息不矛盾。
信息M和受众知识库里的所有信息可能不矛盾。信息N和受众知识库里的所有信息可能不矛盾,前者的可能性一般大于后者的可能性。例如 一般在默认语境知识库中,将军是男性。事实上在受众知识库中将军可能是男性,也可能是女性,两者与受众知识库中的信息可能不矛盾,但前者的可能性一般大于后者。
想幽默被感知的快些时,使信息M在语境知识库排序中靠前些;想幽默被感知的慢些时,使信息M在语境知识库排序中靠后些。
想有些讽刺的效果时,使信息N与受众知识库里的信息矛盾,而且显示真实、合规、赞赏或友善等正面信息;想有些开玩笑的效果时,使信息N与受众知识库里的信息矛盾,而且显示虚假、违规、贬低或攻击等负面信息。例如:假如李明有两辆自行车,就会分给朋友一辆;假如李明有两辆汽车,就会分给朋友一辆;假如李明有两艘游艇,就会分给朋友一艘。假如李明有两艘宇宙飞船,就会分给朋友一艘。假设信息与客观现实矛盾的可能性越来越大,合理性越来越小。对李明由赞扬到讽刺越来越明显。
想有些滑稽的效果时,使信息N与受众知识库里的信息不矛盾而且把生成的信息的意思变成现实呈现在受众面前。当假设信息变为真实存在时,意思变为现实直接呈现在受众面前,受众无经过需思考就可以理解在默认语境中不合理的信息在假设语境中的合理,浅显易懂,显得滑稽。例如信息“老虎和羊和平地待在一起”在默认的语境下是不合理的,但假设语境中是否合理就需要经过受众的思考。如果假设信息“老虎吃很饱时不会去骚扰猎物”真实存在,把老虎喂得很饱,使得“老虎和羊和平地待在一起”直接在展现在受众面前就显得滑稽。
想有些诡辩效果时,把信息N存在时显示的信息与受众知识库里的信息不矛盾当成信息M存在时显示的信息与受众知识库里的信息不矛盾,即 把假设条件下的合理当成默认条件下的合理,也即把在假设语境中的合理当做默认语境中的合理。
想增强幽默的效果时,使信息N的意思或者在信息N存在时显示的信息的意思违背受众不愿意接受的社会规则或不愿认可的事实等。受众因得到虚拟的满足而且难以受到惩罚而产生强烈的智力自豪感。例如对其它类别受众攻击的笑话。
假设使语境知识库中的信息M变成信息N,在假设语境里使信息A变得与受众知识库里的信息矛盾。
例如,生成信息A“一个人说是将军的儿子,将军也承认是这人的父亲”。假设把语境知识库中的一条信息M“将军是男性”改为信息N“将军是女性”,信息A在假设语境里与受众知识库里的信息矛盾。
假设使语境知识库中的信息M变成信息N,在假设语境里使信息A的意思改变。
例如,生成信息A“如果老板不收回他今天上午对我说的话,我就离开这家公司”,假设把语境知识库中的一条信息M“老板和我的地位是平等的”改为信息N“我被老板解雇了”,信息A的意思由“我强烈地要求老板道歉”变为“我无奈地服从”。
假设使语境知识库中的信息M变成信息N,在假设语境里使信息A发展方式改变。
例如,由信息A“今年降水和温度适当”发展到信息B“今年的庄稼可能有好收成”。语境信息中有一条信息M“降水和温度合适是庄稼有好收成的必要条件”。发展是有前提可能有结论的方式。假设改变信息M为 信息N“降水和温度合适是庄稼有好收成的充分条件”,发展是有前提必然有结论的方式。在假设语境中由信息A发展到信息D“今年的庄稼肯定有好收成”。
例如,生成一条语境信息“小孩说:我要吃苹果”。生成信息A“现在是到小孩睡觉而还没到大人睡觉时间”,发展到信息B“妈妈:苹果睡了”。语境中的一条信息M是“苹果和小孩的相同点起决定的作用”,因此类比发展到了信息B。假设把信息M改变为信息N“大苹果和大人的相同点起决定的作用”,发展采用不同的方式。大苹果类比大人发展。由信息A发展到信息D“小孩:小的可能睡了,大的肯定还没睡”。
信息A和信息B既可以是简单判断,也可以是复合判断。
例如,生成A“我的家乡用无线电,所以我的家乡没有电话线”。一条语境信息M是“只要用无线电,就没有电话线”这是以充分条件方式的发展。假设把信息M改变为信息N“只要用无线电,就没有电话线而且只有用无线电,才没有电话线”,这是以前提为充分必要条件的发展。在假设语境中由信息A发展到信息B“我的家乡没电话线,所以我的家乡用无线电”。
语境和发展的改变可以不只一次。
例如,由“小孩冰淇淋掉地上了”发展到“小孩哭了”,进一步发展到“路人替小孩买了一个冰淇淋”,“进一步发展到小孩仍有一个冰淇淋吃”,进一步发展到“小孩不哭了。
有一条语境信息“只有冰淇淋掉地上了小孩才哭”,“小孩冰淇淋掉地上了”以前提为必要条件的方式,发展到“小孩哭了”。另有一条语境信息“只有小孩哭了路人才会替小孩买冰淇淋”,“小孩哭了”以前提为必要条 件方式发展到“路人替小孩买了一个冰淇淋”。
假设这两条语境信息分别改为“没有冰淇淋掉地上小孩也会哭”或者“没有小孩哭路人也会替小孩买冰淇淋”。冰淇淋掉地上不再是小孩哭的必要条件或者小孩哭不再是路人替小孩买冰淇淋的必要条件。由“小孩冰淇淋没掉地上”发展到“小孩哭了”或者发展到“小孩没哭”,由“小孩没哭”进一步发展到“路人替小孩买了一个冰淇淋”,进一步发展到“小孩有两个冰淇淋吃”。
由“如果小孩没有冰淇淋掉地上,小孩就有两个冰淇淋吃”发展到“因为小孩的冰淇淋掉地上了,所以小孩又哭了”。发展的前提和结果本身都是复合判断。
假设对默认语境的改变是对合理的否定。
因为语境的改变使得信息或发展变得不合理了,所以根据所述的语境知识库中改变的信息改变所述的生成的信息或改变所述的发展。
信息A在假设语境中变得与受众知识库里的信息矛盾时,使信息A变为另一条信息(设为信息C)。信息C在默认语境中是与受众知识库中的信息矛盾的、在假设语境中是与受众知识库中的信息不矛盾的。
具体地,假设使语境知识库中的信息M变成信息N,在假设语境中把信息A改变成信息C。信息C在默认语境的意思和在假设语境中的意思相同,但信息C与受众知识库中的信息由在默认语境中的矛盾变为在假设语境中的不矛盾了。
例如,在默认语境中生成信息A“一个人说是将军的儿子,将军也承认是这人的父亲”,信息A在默认语境里是合理的;假设把语境知识库中 的一条信息M“将军是男性”改为信息N“将军是女性”,信息A在假设语境里是不合理的;假设语境中把信息A改变成信息C“一个人说是将军的儿子,但将军不承认是这人的父亲”,信息C在默认语境中是不合理的而在假设语境中是合理的;显示信息C。
具体地,假设使语境知识库中的信息M变成信息N,在假设语境中把信息A改变成信息C。信息C在默认语境的意思和在假设语境中的意思不相同,而且信息C与受众知识库中的信息由在默认语境中的矛盾变为在假设语境中的不矛盾了。
例如,建立知识库,获得在受众知识库里有“王凡导演了中国最多的电影”和“李明在中国导演界生孩子最多”等信息;在默认语境中生成信息A“王凡是中国最高产的导演”或者“李明不是中国最高产的导演”,信息A在默认语境里是合理的;假设把语境知识库中的一条信息M“这是在讨论电影”改为信息N“这是在讨论生孩子”,信息A在假设语境中就变得不合理了;在假设语境中把信息A改变成信息C“李明是中国最高产的导演”,信息C的意思在默认语境中和在假设语境中是不同的,信息C在默认语境中是不合理的而在假设语境中是合理的;显示信息C。
信息A的意思改变时,使信息A变为另一条信息(设为信息E)。信息E和信息A的形式相同。信息E在假设语境中的意思和在默认语境中的意思不同。
具体地,假设使语境知识库中的信息M变成信息N,在假设语境中把信息A改变成信息E,信息E和信息A的形式相同。信息E在假设语境中的意思和在默认语境中的意思不同。先显示信息E再显示信息N。
例如:生成信息A“如果老板不收回他今天上午对我说的话,我就离开这家公司”,在含有一条信息M“老板和我的地位是平等的”等信息的默认语境中信息A的意思是“强烈地要求老板道歉”;假设把语境知识库中的信息M改为信息N“我被老板解雇了”;把具有“如果老板不收回他今天上午对我说的话,我就离开这家公司”形式和“我强烈地要求老板道歉”意思的信息A变成了具有同样形式但“我无奈地服从”意思的信息E;先显示信息E再显示信息N。
信息A的意思改变或发展方式改变时,使由信息A发展到信息B改变为由信息A发展到另一条信息(设为D)。
具体地,假设使语境知识库中的信息M变成信息N,信息A的意思变化时,使信息A发展到信息D。信息A的意思改变成为另一个意思时,由这另一个意思发展到信息D。先显示信息A再显示信息D。
例如,生成信息A:“一个员工说:前天我刚一进电梯,里面就突然全黑了,吓死我了”由信息A发展到信息B“经理听了说:我马上安排对电梯安全检查”。语境中的一条信息M是“员工正在谈论安全问题”。在默认的语境中信息A的意思是“员工担忧在电梯里可能会出危险”。假设把信息M改变成信息N“员工在谈论照明问题”,在假设语境里由信息A的意思“员工在抱怨电梯里太黑暗”发展到信息D“给你们每人配一把手电筒”。先显示信息A再显示信息D。
具体地,假设使语境知识库中的信息M变成信息N,信息A发展的方式改变成另一种方式时。信息A由这种改变的发展方式发展到信息D。先显示信息A再显示信息D。
例如,由信息A“今年降水和温度适当”发展到信息B“今年的庄稼可能有好收成”。语境信息中有一条信息M“降水和温度合适是庄稼有好收成的必要条件”。发展是有前提可能有结论的方式。假设改变信息M为信息N“降水和温度合适是庄稼有好收成的充分条件”,发展是有前提必然有结论的方式。由在假设语境中由信息A发展到信息D“今年的庄稼肯定有好收成”。
例如,生成一条语境信息“小孩说:我要吃苹果”。生成信息A“现在是到小孩睡觉而还没到大人睡觉时间”,发展到信息B“妈妈:苹果睡了”。语境中的一条信息M是“苹果和小孩的相同点起决定的作用”,因此类比发展到了信息B。假设把信息M改变为信息N“大苹果和大人的相同点起决定的作用”,发展采用不同的方式。大苹果类比大人发展。由信息A发展到信息D“小孩:小的可能睡了,大的肯定还没睡”。可以显示语境信息,可以显示信息A发展到信息B,显示信息A发展到信息D。
假设对默认语境的改变是对合理的否定;在假设语境中的改变是对合理的否定之否定,使机器人有幽默感。
最后显示改变的生成的信息、假设的信息或者改变的发展。具体实施中显示信息C,先显示信息E再显示信息N,或者先显示信息A再显示D等。信息A和信息E的意思不同而形式相同,显示信息A和显示信息E的效果一样。
显示出了机器人的幽默感。
综上所述,本发明特有的理论基础是:幽默是对合理的潜意识地否定之否定。本发明根据机器人信息储存量大,运算速度快的特点对人类的心 理活动进行模拟,通过综合技术手段使机器人的幽默感体现在让受众感觉到机器人对合理性的巧妙处理。机器人先有意识地对合理否定之否定,受众潜意识地回溯机器人的这个过程。合理性通过显示的信息与受众所知道并认可的自然、社会和社会等领域的事实和规则等所有的信息的矛盾性来判断,矛盾则不合理,不矛盾则合理。例如根据思维的同一律,确定的表达形式在确定的语境中意思确定才是合理的。合理性针对特定的受众,对一些受众是合理的对另外的受众则不一定。一些人认为某些是事情是真实的,某些规则是应该接受的,另一些人则可能不这样认为,例如多数人认为“地球围绕太阳转”是真实的,极少数人则认为“太阳围绕地球转”是真实的。
本发明的技术手段分三个层次,第一层技术手段是通过心照不宣地分享显示的内容由矛盾到不矛盾,达到有幽默感的目的,主要涉及心理学和哲学领域。第二层次技术手段是通过受众和机器人拥有共同语境达到心照不宣地分享的目的,通过由默认语境转换到假设语境的达到显示的内容由矛盾到不矛盾的目的,主要涉及逻辑学和语言学领域。第三个层次的技术手段是通过建立语境知识库达到有共同语境的目的,通过改换数据库达到改换语境的目的,主要通过受众知识库建立语境知识库,通过群体知识库建立受众知识库,主要涉及计算机领域。
受众回溯感受机器人生成幽默信息过程,对机器人显示的内容最初感到不合理,感到紧张和困惑,怀疑是自己理解错误或是智能机器人错误,接着判定自己很可能没错同时机器人可能也没错,就试图合理化机器人的表达,后来发现在假设语境中机器人的表达是合理的,是对合理的否定之 否定,最后通过共同的背景知识心照不宣地发现和分享这个秘密而感到智力自豪、认同、默契、放松和愉快等而发笑。
本发明从一般性的信息中生成幽默,内容丰富,效果明显,克服了其它生成机器人幽默性格的技术依赖人类现成幽默材料和主要生成文字游戏的局限。
生成机器人幽默信息的效果可测试。一种测试方法是先使被测试的机器人在图灵测试中回答问题,这时机器人骗过人类的成功率设为S;再让幽默感强的一些人和机器人一起幽默地回答问题,这时机器人骗过人类的成功率设为T;则机器人模仿人类的幽默和模仿人类的其它特性的成功比率是T/S。如果T/S接近或大于1,则表明机器人模仿人类的幽默比较成功。另一种测试方法是让幽默感强的一些人和机器人一起匿名地而且幽默地回答问题,由人类受众判定各个回答者幽默的分数;机器人在所有回答者中的名次,就是机器人幽默感的成绩。如果排名在前百分之五十,说明机器人的幽默感相比人类较强。
本说明书中各个实施例采用递进的方式描述,每个实施例重点说明的都是与其他实施例的不同之处,各个实施例之间相同相似部分互相参见即可。对于实施例公开的装置而言,由于其与实施例公开的方法相对应,所以描述的比较简单,相关之处参见方法部分说明即可。
专业人员还可以进一步意识到,结合本文中所公开的实施例描述的各示例的单元及算法步骤,能够以电子硬件、计算机软件或者二者的结合来实现,为了清楚地说明硬件和软件的可互换性,在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方 式来执行,取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能,但是这种实现不应认为超出本发明的范围。
结合本文中所公开的实施例描述的方法或算法的步骤可以直接用硬件、处理器执行的软件模块,或者二者的结合来实施。软件模块可以置于随机存储器(RAM)、内存、只读存储器(ROM)、电可编程ROM、电可擦除可编程ROM、寄存器、硬盘、可移动磁盘、CD-ROM、或技术领域内所公知的任意其它形式的存储介质中。
对所公开的实施例的上述说明,使本领域专业技术人员能够实现或使用本发明。对这些实施例的多种修改对本领域的专业技术人员来说将是显而易见的,本文中所定义的一般原理可以在不脱离本发明的精神或范围的情况下,在其它实施例中实现。因此,本发明将不会被限制于本文所示的这些实施例,而是要符合与本文所公开的原理和新颖特点相一致的最宽的范围。

Claims (2)

  1. 一种基于知识库生成机器人幽默性格信息的方法,其特征在于,所述方法包括:
    划分群体;
    建立以各类群体预知为信息分类标准的群体知识库;
    判断受众所属的群体类别;
    从所述群体知识库中建立以受众预知为信息分类标准的受众知识库;
    生成一条信息或者由这条信息发展到另一条信息;
    建立语境知识库;
    假设改变语境知识库中的一条信息使所述的生成的信息变得与受众知识库里的信息矛盾、使所述的生成的信息意思改变或使所述的发展方式改变;
    根据所述的语境知识库中改变的信息改变所述的生成的信息或改变所述的发展;
    显示所述的改变的生成的信息、假设的信息或改变的发展。
  2. 一种基于知识库生成机器人幽默性格信息的系统,其特征在于,包括:
    划分模块,用于划分群体;
    第一建立模块,用于建立以各类群体预知为信息分类标准的群体知识库;
    判断模块,用于判断受众所属的群体类别;
    第二建立模块,用于从所述群体知识库中建立以受众预知为信息分类 标准的受众知识库;
    生成模块,用于生成一条信息或者由这条信息发展到另一条信息;
    第三建立模块,用于建立语境知识库;
    处理模块,用于假设改变语境知识库中的一条信息使所述的生成的信息变得与受众知识库里的信息矛盾、使所述的生成的信息意思改变或使所述的发展方式改变,根据所述的语境知识库中改变的信息改变所述的生成的信息或改变所述的发展,和显示所述的改变的生成的信息、假设的信息或改变的发展。
PCT/CN2019/080348 2018-07-27 2019-03-29 一种基于知识库生成机器人幽默性格信息的方法及系统 WO2020019745A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP2020573335A JP7079530B2 (ja) 2018-07-27 2019-03-29 知識ベースに基づき、ロボットのユーモラスな性格情報を生成する方法及びシステム
US16/965,308 US20210117813A1 (en) 2018-07-27 2019-03-29 Method and system for generating humorous personality information for robot by using knowledge base

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810843762.6A CN109033375B (zh) 2018-07-27 2018-07-27 一种基于知识库生成机器人幽默性格信息的方法及系统
CN201810843762.6 2018-07-27

Publications (1)

Publication Number Publication Date
WO2020019745A1 true WO2020019745A1 (zh) 2020-01-30

Family

ID=64646110

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/080348 WO2020019745A1 (zh) 2018-07-27 2019-03-29 一种基于知识库生成机器人幽默性格信息的方法及系统

Country Status (4)

Country Link
US (1) US20210117813A1 (zh)
JP (1) JP7079530B2 (zh)
CN (1) CN109033375B (zh)
WO (1) WO2020019745A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109033375B (zh) * 2018-07-27 2020-02-14 张建军 一种基于知识库生成机器人幽默性格信息的方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2014115952A1 (ko) * 2013-01-25 2014-07-31 포항공과대학교 산학협력단 유머 발화를 이용하는 음성 대화 시스템 및 그 방법
CN105975531A (zh) * 2016-04-29 2016-09-28 华南师范大学 基于对话知识库的机器人对话控制方法和系统
CN107291960A (zh) * 2017-08-07 2017-10-24 大国创新智能科技(东莞)有限公司 基于大数据与小数据相关分析实现幽默的方法和系统
CN107564542A (zh) * 2017-09-04 2018-01-09 大国创新智能科技(东莞)有限公司 基于幽默识别的情感交互方法和机器人系统
CN109033375A (zh) * 2018-07-27 2018-12-18 张建军 一种基于知识库生成机器人幽默性格信息的方法及系统

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2008001350A2 (en) * 2006-06-29 2008-01-03 Nathan Bajrach Method and system of providing a personalized performance
JP5436152B2 (ja) 2009-11-10 2014-03-05 日本電信電話株式会社 質問応答装置、質問応答方法、質問応答プログラム
US10467916B2 (en) * 2010-06-15 2019-11-05 Jonathan Edward Bishop Assisting human interaction
CN103400054A (zh) * 2013-08-27 2013-11-20 哈尔滨工业大学 计算机辅助心理咨询自动问答机器人系统
JP6087333B2 (ja) 2014-11-26 2017-03-01 株式会社ユニバーサルエンターテインメント 情報提供システム、情報提供方法、及びプログラム
WO2018045081A1 (en) * 2016-08-31 2018-03-08 Taechyon Robotics Corporation Robots for interactive comedy and companionship
CN107368572A (zh) * 2017-07-13 2017-11-21 张勇 多功能智能人机交互方法及系统
CN107885837A (zh) * 2017-11-09 2018-04-06 北京光年无限科技有限公司 一种用于智能机器人的交互输出方法以及智能机器人
US10642939B2 (en) * 2018-02-24 2020-05-05 Twenty Lane Media, LLC Systems and methods for generating jokes

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2014115952A1 (ko) * 2013-01-25 2014-07-31 포항공과대학교 산학협력단 유머 발화를 이용하는 음성 대화 시스템 및 그 방법
CN105975531A (zh) * 2016-04-29 2016-09-28 华南师范大学 基于对话知识库的机器人对话控制方法和系统
CN107291960A (zh) * 2017-08-07 2017-10-24 大国创新智能科技(东莞)有限公司 基于大数据与小数据相关分析实现幽默的方法和系统
CN107564542A (zh) * 2017-09-04 2018-01-09 大国创新智能科技(东莞)有限公司 基于幽默识别的情感交互方法和机器人系统
CN109033375A (zh) * 2018-07-27 2018-12-18 张建军 一种基于知识库生成机器人幽默性格信息的方法及系统

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
LIN, HONGFEI ET AL: "Computational humor researches and applications", JOURNAL OF SHANDONG UNIVERSITY (NATURAL SCIENCE), vol. 51, no. 7, 31 July 2016 (2016-07-31), pages 1 - 10 *
YANG, YIXIAN ET AL: "Primary Exploration on Automatic Generation of Microblog Joke", JOURNAL OF BEIJING UNIVERSITY OF POST AND TELECOMMUNICATIONS (SOCIAL SCIENCES EDITION), vol. 14, no. 1, 28 February 2012 (2012-02-28), pages 25 - 29 *

Also Published As

Publication number Publication date
CN109033375B (zh) 2020-02-14
CN109033375A (zh) 2018-12-18
JP2021530055A (ja) 2021-11-04
JP7079530B2 (ja) 2022-06-02
US20210117813A1 (en) 2021-04-22

Similar Documents

Publication Publication Date Title
Glucksberg et al. Experimental Psycholinguistics (PLE: Psycholinguistics): An Introduction
Gelman et al. Mother-child conversations about gender: Understanding the acquisition of essentialist beliefs
Clark Adoptive Youth Ministry (Youth, Family, and Culture): Integrating Emerging Generations into the Family of Faith
Topping et al. Hindrances for parents in enhancing child language
US20060179022A1 (en) Counterpart artificial intelligence software program
Barrett Anthropomorphism, intentional agents, and conceptualizing God
Demo Hacking agency: Apps, autism, and neurodiversity
Adu-Gyamfi Childhood construction and its implications for children’s participation in Ghana
Martin “Girls don't talk about garages!”: Perceptions of conversation in same‐and cross‐sex friendships
WO2020019745A1 (zh) 一种基于知识库生成机器人幽默性格信息的方法及系统
Irwin ‘Hanging out with mates’: Friendship quality and its effect on academic endeavours and social behaviours
Holland et al. Catching on, wising up and learning from your mistakes: young people's accounts of moral development
ZHANG et al. Negativity bias in emergent online events: Occurrence and manifestation
Brown Informed is best: how to spot fake news about your pregnancy, birth and baby
Dominick et al. Young and oppressed
Xu Learning morality with siblings: The untold tale of a mid-twentieth century Taiwanese family
DiTecco New technology, same old stigma: An analysis of feminist discourses and sex work stigma in sex robot media
Cheng The art of logic in an illogical world
ÜSTÜNDAĞ et al. A content analysis of TV series/shows watched by the secondary and high school students in Turkey
Frederick Partying with a purpose: Finding meaning in an online" party'n'play" subculture
Schumacher Raising Awareness About Gender Biases and Stereotypes in Voice Assistants
Frith et al. A Feminist Companion to Research Methods in Psychology
Zhu et al. Measuring Information in Text Explanations
Ake That mystery category “fourthness” and its relationship to the work of CS Peirce
Budzynska et al. Structure of persuasive communication and elaboration likelihood model

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: 19841748

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2020573335

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19841748

Country of ref document: EP

Kind code of ref document: A1