WO2017119014A1 - Information processing apparatus, information processing method and computer-readable medium - Google Patents

Information processing apparatus, information processing method and computer-readable medium Download PDF

Info

Publication number
WO2017119014A1
WO2017119014A1 PCT/JP2016/000090 JP2016000090W WO2017119014A1 WO 2017119014 A1 WO2017119014 A1 WO 2017119014A1 JP 2016000090 W JP2016000090 W JP 2016000090W WO 2017119014 A1 WO2017119014 A1 WO 2017119014A1
Authority
WO
WIPO (PCT)
Prior art keywords
question
answer
observation
suggestion
information processing
Prior art date
Application number
PCT/JP2016/000090
Other languages
French (fr)
Inventor
Ramkumar Rajendran
Original Assignee
Nec Corporation
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 Nec Corporation filed Critical Nec Corporation
Priority to PCT/JP2016/000090 priority Critical patent/WO2017119014A1/en
Publication of WO2017119014A1 publication Critical patent/WO2017119014A1/en

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
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G06F16/3329—Natural language query formulation or dialogue systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • G06N5/022—Knowledge engineering; Knowledge acquisition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • G06N5/041—Abduction

Definitions

  • the present invention pertains to an information processing technology, in particular, to a technology of extracting knowledge base from data.
  • a user with a problem will seek the assistance from the expert to solve the problem.
  • the user usually describes the observation that the user observed and states the problem.
  • the expert may understand the problem and provide suggestion to solve the problem.
  • the expert may ask more details to the user to understand the problem and then provide suggestion to solve the problem.
  • the user implements the solution, and may ask for further questions or report a function after implementing the solution.
  • Similar conversations are in online discussion forums (hereinafter, referred to as "ODF"), plant maintenance logs, medical conversation data, etc. Due to availability of vast amount of data, for a new problem, identifying the solution is becoming difficult for users and experts.
  • ODF online discussion forums
  • PTL 1 discloses an example of technology of extracting knowledge from text data that is unstructured data by using a text corpus.
  • the text corpus considered in PTL 1 includes documents from a vehicle service reporting system.
  • Each vehicle service report generated in the vehicle service reporting system of PTL 1 includes an identified problem, symptoms associated with the problem and suggestion to solve the problem, written by the experts.
  • the technology according to PTL 1 can extract knowledge from text data explicitly providing symptoms associated with problem, a problem to be solved, and suggestion to solve the problem.
  • PTL 2 discloses an example of technology of extracting knowledge from online discussion forums.
  • the technology according to PTL 2 includes creating ⁇ question, answer> pair by selecting and ranking suitable answer for the questions from the threads of online forums.
  • PTL 2 discloses a method to rank the answers from plural of answers for a question in online discussion forums.
  • the approach used in PTL 2 can be applied to a thread that has all the details of the problem given by the user in the first post and an answer to the problem is given in a single post of the thread in such a way that the answer can be extracted directly from the single post as a response of a question concerning the problem.
  • a thread in online discussion forums and conversation data of a technical support center often have the root post not including explanation of the symptoms and details of a problem.
  • the root post is, for example, the first post of a thread or the first utterances including a question by the user at a technical support center.
  • an expert may ask the user for more details to provide correct suggestion and the information of a question (referred to as an additional question) asked by the experts and an answer to the additional question may be in the discussions within the thread, the answer to the additional question may be missed while ranking the answers.
  • One of the objects of the present invention is to provide an information processing apparatus capable of capturing scattered information included in verbal interaction.
  • An information processing apparatus includes: analysis means for generating an observation based on a question and an answer to the question, the observation expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and generation means for generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
  • An information processing method includes: generating an observation based on a question and an answer to the question, the observation expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
  • a computer-readable medium stores a program causing a computer to execute: analysis processing of generating an observation based on a question and an answer to the question, the observation expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and generation processing of generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
  • the present invention may be achieved by the program stored in the computer-readable medium.
  • Fig. 1 is a block diagram illustrating an example structure of an information processing apparatus according to first and second exemplary embodiments of the present invention.
  • Fig. 2 is a block diagram illustrating an example structure of the knowledge base according to any of exemplary embodiments of the present invention.
  • Fig. 3 is a block diagram illustrating an example structure of the knowledge base according to any of exemplary embodiments of the present invention.
  • Fig. 4 is a flowchart illustrating an example of operation of the information processing apparatus according to the first and second exemplary embodiments of the present invention.
  • Fig. 5 is a flowchart illustrating an example of operation in the preprocessing of the information processing apparatus according to the first and second exemplary embodiments of the present invention.
  • Fig. 1 is a block diagram illustrating an example structure of an information processing apparatus according to first and second exemplary embodiments of the present invention.
  • Fig. 2 is a block diagram illustrating an example structure of the knowledge base according to any of exemplary embodiments of the present invention.
  • Fig. 3
  • FIG. 6 is a flowchart illustrating example operation of discussion processing of the information processing apparatus according to the first and second exemplary embodiments of the present invention.
  • Fig. 7 is a flowchart illustrating example operation of RFD processing of the information processing apparatus according to the first and second exemplary embodiments of the present invention.
  • Fig. 8 is a flowchart illustrating example operation of AQ processing of the information processing apparatus according to the first exemplary embodiment of the present invention.
  • Fig. 9 is a flowchart illustrating example operation of the AQ processing of the information processing apparatus according to the second exemplary embodiment of the present invention.
  • Fig. 10 is a flowchart illustrating example operation of the creation processing of the information processing apparatus 100 of the second exemplary embodiment of the present invention.
  • FIG. 11 is a block diagram illustrating a schematic example of whole structure of an information processing system according to the first and second exemplary embodiments of the present invention.
  • Fig. 12 is a block diagram illustrating another schematic example of whole structure of an information processing system according to the first and second exemplary embodiments of the present invention.
  • Fig. 13 is a block diagram illustrating an information processing apparatus according to a third exemplary embodiment of the present invention.
  • Fig. 14 is a flowchart illustrating example operation of the information processing apparatus according to the third exemplary embodiment of the present invention.
  • Fig. 15 is a block diagram illustrating a computer capable of achieving the information processing apparatus according to any one of the exemplary embodiments of the present invention.
  • Fig. 12 is a block diagram illustrating another schematic example of whole structure of an information processing system according to the first and second exemplary embodiments of the present invention.
  • Fig. 13 is a block diagram illustrating an information processing apparatus according to a third exemplary embodiment of the present invention.
  • Fig. 14 is
  • FIG. 16 is a block diagram illustrating an example structure, achieved by circuits, of the information processing apparatus 100, achieved by dedicated circuits, according to the first and the second exemplary embodiments of the present invention.
  • Fig. 17 is a block diagram illustrating an example structure of the information processing apparatus, achieved by dedicated circuits, according to the third exemplary embodiment of the present invention.
  • Fig. 1 is a block diagram illustrating an example structure of an information processing apparatus 100 according to the first exemplary embodiment of the present invention.
  • the information processing apparatus 100 illustrated in Fig. 1 includes a raw data reception unit 101, a preprocessing unit 102, a classification unit 103, an analysis unit 104, and a generation unit 105.
  • the information processing apparatus 100 may further include a dictionary storage unit 108.
  • the information processing apparatus according any one of exemplary embodiment of the present invention may be referred to as a "generation apparatus" or a "knowledge base generation apparatus". Note that directions of arrows drawn in Fig. 1 and other figures do not limit directions of data transmission.
  • the raw data reception unit 101 receives raw data.
  • the raw data is, for example, text data of a thread in an online discussion forum.
  • An interaction between a user and an expert to solve a problem is termed as a "thread".
  • the thread may be posts (i.e. items posted by posters) each of which is related with at least one of the other posts in the thread.
  • a post in the posts includes text data and information (i.e. identification such as e-mail address) of a poster who posts the post.
  • the posters such as, a user (e.g. a customer) and an expert are referred to as "participants”.
  • the raw data may be conversation data that is data of conversation at a support center.
  • the support center may be a technical support center of an industrial product provider such as, a personal computer provider, or a car manufacturer.
  • the support center may be a control room supporting plant maintenance.
  • the support center may be a clinic, a hospital, or a medical center.
  • the conversation data may be data in text format.
  • the conversation data may be data in audio format and may be generated by recording and digitizing conversation. Text data converted from the conversation between a user and an expert at a support center is also referred to as the "thread".
  • the user and the expert at the support center are also referred to as "participants”.
  • the preprocessing unit 102 performs preprocessing including splitting the thread into sentences.
  • the preprocessing unit 102 correlates, on the basis of recorded identification included in the thread, the sentences with their respective posters.
  • the preprocessing unit 102 may correlate a sentence in the sentences with a user asking an expert for solution to a problem or the expert providing the solution to the problem.
  • the preprocessing unit 102 may not distinguish between the experts.
  • the preprocessing unit 102 may assume the experts to be one expert.
  • the preprocessing unit 102 converts the raw data in audio format into text data by using existing speech recognition technology, and then performs the preprocessing on the text data to generate the sentences.
  • the preprocessing unit 102 may distinguish speakers of the sentences by using a voice authentication technology.
  • the classification unit 103 classifies the sentences into categories including three main categories that are observation or symptoms of a problem, the problem, and suggestion to solve the problem.
  • a sentence representing observation or symptoms of a problem is referred to as an "Observation”.
  • a category of Observation is referred to as an "Observation category”.
  • a sentence representing a problem is referred to as a "Problem”.
  • a category of Problem is referred to as a "Problem category”.
  • a sentence indicating suggestion to solve a problem is referred to as a "Suggestion”.
  • a category of Suggestion is referred to as a "Suggestion category”.
  • a thread in an online discussion forum may include a short discussion such as a set of a question and an answer.
  • the set of a question and an answer is referred to as a " ⁇ question, answer> pair".
  • the ⁇ question, answer> pair may be classified into either of two types based on who raised the question as described below.
  • a purpose of the question may be to request for more details, more specifically, a) to get more details of the problem, b) to understand the environment of the problem, or c) to get feedback of the solution suggested.
  • RFD a question to request for more details
  • the ⁇ question, answer> pair including a question classified as an RFD and an answer to the RFD is referred to as an " ⁇ RFD, Answer to RFD>" or an " ⁇ RFD, Answer to RFD>” pair.
  • a category of an RFD is referred to as an "RFD category”.
  • the original poster is referred to as an "OP”.
  • the question is an additional question, for example, a) to understand the solution suggested, b) to ask a question about a new problem raised by carrying out the suggestion suggested by the expert, or c) to ask a new question.
  • a question classified as the additional question is referred to as an "AQ”.
  • An answer, by an expert, to an AQ is referred to as an "Answer to AQ”.
  • the ⁇ question, answer> pair including a question classified as an AQ is referred to as an " ⁇ AQ, Answer to AQ>" or an " ⁇ AQ, Answer to AQ>” pair.
  • a category of an AQ is referred to as an "AQ category”.
  • a category of an Answer to AQ is referred to as an "Answer to AQ category”.
  • the categories other than the three main categories described above are the RFD category, the Answer to RFD category, the AQ category, and the Answer to AQ category.
  • a sentence classified into any of the above-described categories is considered to be important to understand the problem or to implement the suggestion.
  • a sentence which is not able to be classified into any of the above-described categories is labeled as supplementary.
  • a sentence labeled as supplementary is referred to as "Supplementary”.
  • a category of a Supplementary is referred to as a "Supplementary category”.
  • a sentence labeled as supplementary is not important to understand the problem or to implement the suggestion.
  • a sentence generated by splitting a thread is classified as: (a) Observation - information needed to solve a problem, and environmental information to understand the problem to suggest solution; (b) Problem - a main question asked by an original poster, and the first problem reported by the original poster; (c) Suggestion - the solution suggested by an expert to solve the problem; (d) RFD - question asked by an expert to get more details about the problem or to know the environment, and any question asked by the expert relevant to the problem; (e) Answer to RFD - an answer provided by the original poster to answer RFD; (f) Additional Questions (AQ) - an additional questions asked by the original poster, for example, after conversation with the expert; (g) Answer to AQ - an answer provided by the expert for AQ; or (h) Supplementary - a sentence which is not important to understand the problem or to implement the suggestion.
  • the classification unit 103 classifies a sentence generated from the thread by preprocessing unit 102 into one of the categories.
  • the classification unit 103 may operate as a machine classifier classifying a sentence into one of the above-described categories.
  • the classification unit 103 may classify the sentence according to a result of machine learning such as supervised learning.
  • the result of machine learning is, for example, a dictionary generated in advance by machine learning to classify a sentence of a thread into one of the categories described above.
  • the method of machine learning may be selected by an operator.
  • the format of the dictionary may be defined so that the classification unit 103 is capable of reading using the dictionary.
  • the dictionary is stored in the dictionary storage unit 108.
  • the classification unit 103 correlates an RFD with an Answer to RFD which is an answer to the RFD.
  • the RFD and the Answer to RFD which is correlated with the RFD is the above-mentioned ⁇ RFD, Answer to RFD > pair.
  • the classification unit 103 correlates an AQ with an Answer to AQ which is an answer to the AQ.
  • the AQ and the Answer to AQ which is correlated with the AQ is the above-mentioned ⁇ AQ, Answer to AQ> pair.
  • the classification unit 103 may first detect, on the basis of the machine learning, a non-related sentence, such as a monologue, greetings, or an acknowledgement, in the sentences generated by splitting the thread.
  • the classification unit 103 may remove the detected non-related sentences from the sentences.
  • the classification unit 103 may classifies, into the categories, the sentence in the sentences from which the non-related sentence is removed.
  • the classification unit 103 may classify the non-related sentence as Supplementary.
  • the dictionary unit 108 stores the dictionary to classify a sentence as one of sentences including an Observation, a Problem, a Suggestion, an RFD, an Answer to RFD, an AQ, an Answer to AQ, and a Supplementary.
  • the dictionary storage unit 108 stores the dictionary to classify a sentence into one of the categories including the Observation category, the Problem category, the Suggestion category, the RFD category, the Answer to RFD category, the AQ category, the Answer to AQ category, and the Supplementary category.
  • the analysis unit 104 analyzes a ⁇ RFD, Answer to RFD> pair and extracts, from the ⁇ RFD, Answer to RFD> pair, information to be integrated into a knowledge base described below.
  • the analysis unit 104 also analyzes a ⁇ AQ, Answer to AQ> pair and extracts, from the ⁇ AQ, Answer to AQ> pair, information to be integrated into the knowledge base.
  • the analysis unit 104 merges an RFD and an Answer to RFD included in an ⁇ RFD, Answer to RFD> pair into an Observation according to the Answer to RFD as follows.
  • the analysis unit 104 confirms whether the Answer to RFD is affirmative or negative. More specifically, the analysis unit 104 confirms whether the Answer to RFD is a yes-type answer (e.g. an answer including a word "yes") or a no-type answer (e.g. an answer including a word "no"). When the Answer to RFD includes a word "yes”, the analysis unit 104 may estimate the Answer to RFD to be an affirmative answer. When the Answer to RFD includes a word "no”, the analysis unit 104 may estimate the Answer to RFD to be a negative answer.
  • a yes-type answer e.g. an answer including a word "yes”
  • a no-type answer e.g. an answer including a word "no”
  • the Answer to RFD is not a yes-type answer or a no-type answer
  • the analysis unit estimates whether the Answer to RFD is affirmative or negative by using natural language processing (NLP) technology, such as a sentiment analysis predicting sentiment of the Answer to RFD.
  • NLP natural language processing
  • the analysis unit 104 estimates the Answer to RFD is a yes-type answer.
  • the analysis unit 104 estimates the Answer to RFD to be a no-type answer.
  • the analysis unit 104 converts the RFD into a declarative sentence according to the Answer to RFD.
  • the analysis unit 104 converts the RFD into an affirmative sentence.
  • the analysis unit 104 convers the RFD into a negative sentence.
  • the analysis unit 104 classifies (e.g. labels) the sentence converted from the RFD as an Observation.
  • the analysis unit 104 When the Answer to RFD is not able to be estimated to be an affirmative answer or a negative answer, the analysis unit 104 skips and ignores the ⁇ RFD, Answer to RFD> pair.
  • the analysis unit 104 may label the RFD and the Answer to RFD included in the ⁇ RFD, Answer to RFD> pair into the Supplementary category.
  • the analysis unit 104 confirms whether the AQ is similar to the Problem asked by OP.
  • the analysis unit 104 may estimate, by using NLP technology described above, whether the AQ is similar to the Problem asked by OP.
  • the analysis unit 104 skips and ignores the ⁇ AQ, Answer to AQ> pair.
  • the analysis unit 104 may classify the AQ and the Answer to AQ included in the ⁇ AQ, Answer to AQ> pair into Supplementary category.
  • the analysis unit 104 classifies, by using the dictionary stored in the dictionary storage unit 108, the Answer to AQ as a Suggestion or an Observation in the same way as the classification unit 103 classifies the sentence generated by splitting the thread.
  • the analysis unit 104 may send the Answer to AQ to the classification unit 103.
  • the classification unit 103 may classify the Answer to AQ as a Suggestion or an Observation.
  • the classification unit 103 may send a result of classification to the analysis unit 104.
  • the analysis unit 104 may skip and ignore the ⁇ AQ, Answer to AQ> pair.
  • the analysis unit 104 may label the AQ and the Answer to AQ into the Supplementary category.
  • the analysis unit 104 convert the AQ included in the ⁇ AQ, Answer to AQ> pair into an Observation by changing, for example, the AQ written as an interrogative sentence to a declarative sentence.
  • the AQ is a question to ask the reason of environmental information such as a situation or an event arising in the environment of OP
  • the analysis unit 104 converts the AQ into an Observation representing the environmental information.
  • the processing of converting the AP to an Observation is not limited to the example described above.
  • the analysis unit 104 classifies (e.g. labels) the sentence converted from the AQ as an Observation.
  • the generation unit 105 generates a knowledge base based on the Problem, the Suggestion, and the Observation generated from the thread.
  • the generation unit 105 stores the knowledge base in the result storage unit 106.
  • the generation unit 105 generates the knowledge base based on the Problem, the Suggestion, and two or more of the Observations when the number of sentences classified as Observations is two or more.
  • the generation unit 105 may correlate the Problem and the Suggestion.
  • the generation unit 105 may correlate the Problem and the Observations.
  • the generation unit 105 may store the Problem, the Suggestion correlated with the Problem, and the Observation correlated with the Problem in the result storage unit 106 as the knowledge base.
  • a knowledge base generated from an online discussion forum can assist a new user to identify the relevant solution for a problem occurring to the new user.
  • the online discussion forum may be provided by an industrial product provider in order to support customers of the industrial product provider, and may be maintained by an expert belonging to the industrial product provider.
  • the knowledge base generated from the unstructured data (i.e. the raw data and the input data) is used to respond to a new problem asked by users in future.
  • the knowledge base generated by the information processing apparatus 100 of the present exemplary embodiment is able to help experts to make correct decision in short time, and to provide relevant information to users. Hence, the time spent by an expert for each problem is reduced and user satisfaction improves.
  • the information processing apparatus 100 may generate a knowledge base for plant maintenance.
  • the knowledge base for plant maintenance By using the knowledge base for plant maintenance, the problem that will arise in the future is quickly detected and suggestion is provided immediately. Hence, the down time in the industry is reduced.
  • the information processing apparatus 100 may generate a knowledge base used for identifying the disease when the patient provides the symptoms associated with the disease.
  • the knowledge base can assist the medical professional to make decision about the disease and to treat the patient immediately.
  • the patient can directly interact with the system which uses knowledge base to understand the reasons for symptoms and to do further actions.
  • Fig. 11 is a block diagram illustrating a schematic example of whole structure of an information processing system 1 according to the present exemplary embodiment of the present invention.
  • the information processing system 1 in Fig. 11 includes the information processing apparatus 100, a knowledge base storage system 200, a suggestion apparatus 300, a training apparatus 400, an acquisition apparatus 500, an ODF system 700, and a terminal apparatus 800.
  • the information processing apparatus 100 may include any of the knowledge base storage system 200, the suggestion apparatus 300, the training apparatus 400 and the acquisition apparatus 500.
  • the information processing system 1 may include an information processing apparatus according to one of exemplary embodiments other than the first exemplary embodiments instead of the information processing apparatus 100 according to the first exemplary embodiment.
  • the acquisition apparatus 500, the ODF system 700 and the terminal apparatus 800 are communicably connected via a network 600 that is a communication network. All or a part of the apparatus included in the information processing system 1 may be communicably connected with the network 600.
  • the ODF system 700 provides an online discussion forum.
  • the acquisition apparatus 500 acquires a thread of the online discussion forum provided by the ODF system 700.
  • the training apparatus 400 generates the dictionary stored in the dictionary storage unit 108 by using machine learning technology. Training data for machine learning to generate the dictionary is provided to the training apparatus 400 by, for example, a user terminal (not illustrated) communicably connected with the training apparatus 400.
  • the knowledge base storage system 200 stores the knowledge base generated by the generation unit 105 and output by the output unit 107 of the information processing apparatus 100.
  • the knowledge base system 200 provides the knowledge base to the suggestion apparatus 300.
  • the suggestion apparatus 300 may receive inquiry indicating, for example, one or more Observations from the terminal apparatus 800, may search for the Problem correlated with the Observation, and may send, for example, the Problem and the Suggestion correlated with the Problem to the terminal apparatus 800.
  • the terminal apparatus 800 is, for example, a computer used by an expert.
  • the expert reads a Problem and an Observation concerning the Problem in the online discussion forum provided by the ODF system 700.
  • the expert searches for the similar Problem as the Problem in the online discussion forum by sending the inquiry indicating the Observation concerning the Problem in the online discussion forum. If the similar Problem is found, the expert may post a Suggestion to the online discussion as a response to the Problem in the online discussion forum with reference to the Suggestion to the similar Problem.
  • Fig. 12 is a block diagram illustrating another schematic example of whole structure of an information processing system 1A according to the present exemplary embodiment of the present invention.
  • the information processing system 1A in Fig. 12 includes the information processing apparatus 100, the knowledge base storage system 200, the suggestion apparatus 300, the training apparatus 400, the terminal apparatus 800, and a recording apparatus 900.
  • the recording apparatus 900 may be achieved by a computer connected with a microphone.
  • the recording apparatus 900 may be achieved by an IC (Integrated Circuit) recorder.
  • the information processing system 1A may include an information processing apparatus according to one of exemplary embodiments other than the first exemplary embodiments instead of the information processing apparatus 100 according to the first exemplary embodiment.
  • the recording apparatus 900 records conversation between an expert and a user (e.g. a customer) as conversation data in an audio format.
  • the recording apparatus 900 transmits the conversation data to the acquisition apparatus 500.
  • the expert and the user may make face to face conversation.
  • the expert and the user may make conversation over the telephone.
  • the acquisition apparatus 500 receives the conversation data, and inputs the conversation data into the information processing apparatus 100 as the input data.
  • the information processing apparatus 100, the knowledge base storage system 200, the suggestion apparatus 300, the training apparatus 400 and the terminal apparatus 800 are the same as those of the information processing system 1 illustrated in Fig. 11.
  • the expert recognizes a Problem and an Observation concerning the Problem in the conversation.
  • the expert searches for the similar Problem as the Problem recognized in the conversation by sending the inquiry indicating the Observation concerning the recognized Problem by using the terminal 800. If the similar Problem is found, the expert may provide the user with a Suggestion with reference to the Suggestion to the similar Problem.
  • Fig. 2 is a block diagram illustrating an example structure of the knowledge base according to the present exemplary embodiment.
  • Fig. 2 illustrates a schematic structure of the knowledge base based on the Problem, the Suggestion, and the Observations generated from a single thread.
  • the generation unit 105 may generate the knowledge base based on Problems, Suggestions, and Observations generated from different threads.
  • the generation unit 105 may correlate the Problem and the Suggestion generated from the same thread.
  • the generation unit 105 may correlate the Observations and the Problem generated from the same thread.
  • the generation unit 105 may integrate the same Observations into an integrated Observation and correlate the integrated Observation with the different threads.
  • the generation unit 105 may store the Problems, the Suggestions each correlated with the Problems, and the Observations each of which is correlated with at least one of the Problems in the result storage unit 106 as the knowledge base.
  • Fig. 3 is a block diagram illustrating an example structure of the knowledge base according to the present exemplary embodiment.
  • Fig. 3 illustrates a schematic structure of the knowledge base based on the Problems, the Suggestions, and the Observations generated from a plurality of threads.
  • the result storage unit 106 stores the knowledge base generated by the generation unit 106.
  • the output unit 107 outputs knowledge base data representing the knowledge base to an apparatus storing the knowledge base.
  • the apparatus may be connected with terminal devices for experts in a support center and may provide the experts with information for suggestion to users.
  • Fig. 4 is a flowchart illustrating an example of operation of the information processing apparatus 100 according to the present exemplary embodiment of the present invention.
  • the dictionary storage unit 108 stores the above-described dictionary of classifying a sentence into the categories described above.
  • a problem asked by a user and the complete interaction between an expert and the user, to solve the problem is referred to as a thread.
  • Input data representing the thread is inputted to the information processing apparatus 100 by an apparatus, such as an apparatus collecting a thread from an online discussion forum, or an apparatus recording conversation between a user and an expert at a support center.
  • an apparatus such as an apparatus collecting a thread from an online discussion forum, or an apparatus recording conversation between a user and an expert at a support center.
  • the raw data reception unit 101 receives input data (Step S101).
  • the input data may be data in audio or data in text format.
  • the preprocessing unit 102 performs preprocessing on the input data (Step S102).
  • Fig. 5 is a flowchart illustrating an example of operation in the preprocessing of the information processing apparatus 100 according to the present exemplary embodiment.
  • the preprocessing unit 102 converts the input data into text data (Step S202).
  • the input data may be data in an audio format (i.e. audio data) of conversation at a support center.
  • the preprocessing unit 102 may extract text from the input data by using speech recognition technology.
  • the input data may be data, other than audio data or text data, which can be converted into text data by the preprocessing unit 102.
  • the text data includes the thread in text format.
  • the preprocessing unit 102 splits the text data generated by converting the audio data into sentences (Step S203).
  • the preprocessing unit 102 may further specify respective speakers of the sentences by, for example, voice recognition.
  • the preprocessing unit 102 may estimate which of a user and an expert is the speaker of a sentence.
  • the preprocessing unit 102 may classify the speakers of the sentences into three or more individual participants.
  • the preprocessing unit 102 may correlate, as a poster of a sentence, the speaker of the sentence with the sentence.
  • the preprocessing unit 102 splits text data, that is the input data in text format, into sentences (Step S203).
  • the input data represents a thread.
  • the preprocessing unit 102 splits the thread into sentences (Step S203).
  • the preprocessing unit 102 specifies posters of the sentences.
  • the thread in an online discussion forum may be a set of posts each of which is related with at least a post in the set.
  • the post is a message, such as a question, an answer or a comment, posted to the online discussion forum.
  • Each of the post includes poster information that is information of a poster who posts a post.
  • the preprocessing device may specify a poster of a sentence on the basis of the poster information of a post including the sentence.
  • the preprocessing unit 102 correlates the sentences with their respective posters.
  • the preprocessing unit 102 may arrange the sentences in order of the sentences in the thread.
  • the processing unit 102 may correlate the sentences with order of the sentences in the thread.
  • the information processing apparatus 100 performs the processing in Step S103 next.
  • the outcome of the preprocessing processing in Step S102 is the sentences and information of respective posters (i.e. participants) of the sentences.
  • the participants are classified into two types, a type of user who asks a question, called as an original poster, and a type of expert who answers the question, called an expert.
  • the expert is referred to as an "expert".
  • the sentences may be arranged according to order of the sentences arisen in conversation, represented by the input data, between a user and an expert.
  • the classification unit 103 removes a non-related data (Step S103).
  • the non-related data is a sentence not related with a question concerning a problem, an answer to the question, and an observation concerning the problem.
  • the non-related data is also referred to as a "non-related sentence".
  • the classification unit 103 may extract, from the sentences, a non-related sentence, such as utterance monologue, greetings, or an acknowledgement.
  • the classification unit 103 removes the non-related sentence from the sentences on the basis of, for example, machine learning.
  • the classification unit 103 classifies related sentences into the categories described above (Step S104).
  • the related sentences are sentences not removed as the non-related sentence.
  • the classification unit 103 labels the related sentences.
  • Step S105 the analysis unit 104 performs the discussion processing.
  • the Observation, the Problem and the Suggestion are used to create the knowledge base as shown in Fig 2.
  • the RFD, the Answer to RFD, the AQ, and the Answer to AQ each are not sufficient to be used alone to create the knowledge base.
  • the ⁇ RFD, Answer to RFD> pair and the ⁇ AQ, Answer to AQ> pair may include information to capture the details and environment information required to understand the problem and to suggest correct suggestion.
  • the analysis unit 104 analyzes the ⁇ RFD, Answer to RFD> pair and the ⁇ AQ, Answer to AQ> pair and derives information to be integrated to the knowledge base.
  • Fig. 6 is a flowchart illustrating example operation of discussion processing of the information processing apparatus 100 according to the present exemplary embodiment.
  • the analysis unit 104 performs RFD processing (Step S301), and performs AQ processing (Step S302).
  • Step S301 the analysis unit 104 merges the RFD and the Answer to RFD that is the answer to the RFD to generate a declarative sentence, and classifies the generated sentence as an Observation.
  • Step S302 the analysis unit 104 generates a declarative sentence on the basis of the AQ and the Answer to AQ, and classifies the generated sentence as an Observation.
  • Fig. 7 is a flowchart illustrating example operation of RFD processing of the information processing apparatus 100 according to the present exemplary embodiment.
  • the analysis unit 104 repeats a loop A illustrated in Fig 7. That is, the analysis unit 104 repeats processing from Step S311 to Step S313 for each RFD in the sentences generated by splitting the thread of the input data.
  • the analysis unit 104 classifies the Answer to RFD as yes or no (Step S311). To merge RFD and Answer to RFD, the analysis unit 104 first confirms whether the Answer to RFD is a yes-type answer or a no-type answer. If the Answer to RFD is not a simple yes-type answer or a simple no-type answer, then using natural language processing (NLP) technology, such as sentiment analysis, the analysis unit 104 predicts sentiment of the Answer to RFD, and classifies the Answer to RFD as affirmative or non-affirmative. When the Answer to RFD is classified as affirmative, the analysis unit 104 estimates the Answer to RFD to be yes. When the Answer to RFD is classified as negative, the analysis unit 104 estimates the Answer to RFD to be no.
  • NLP natural language processing
  • Step S312 When classification does not succeed (NO in Step S312), that is, when the analysis unit 104 does not successfully classify the Answer to RFD as yes or no, the analysis unit 104 skips and ignores the ⁇ RFD, Answer to RFD> pair.
  • the Answer to RFD is "I don't know” or the like, the Answer to RFD may not be classified as yes or no.
  • Step S312 When classification succeeds (YES in Step S312), that is, when the analysis unit 104 successfully classifies the Answer to RFD as yes or no, the analysis unit 104 convert the RFD on the basis of the Answer to RFD into a declarative sentence, and classifies and labels the declarative sentence as an Observation(Step S313).
  • the analysis unit 104 converts the RFD to a declarative sentence based on the type of the Answer to RFD.
  • the following example is a part of conversation, in a PC (Personal Computer) technical support center, when a user reports a problem that the user's PC is not connected to a network. Then if an expert asks for more information, such as whether the user have a firewall program installed in his PC or not, then a question to ask for more information, given below, is categorized as an RFD.
  • PC Personal Computer
  • the analysis unit 104 converts the RFD into "a firewall program is installed in the PC.” If Answer to RFD is "No”, then the analysis unit 104 converts the RFD into "A firewall is not installed in the PC.”
  • the analysis unit 104 may convert an interrogative sentence representing the question of RFD to an affirmative sentence or a negative sentence by using NLP technology.
  • the analysis unit 104 may classify the converted sentence "Firewall installed in the PC.” or "Firewall not installed in the PC.” as an Observation.
  • the analysis unit 104 performs the processing from Step S311 to Step S313 for each of the sentences classified into the RFD category, and the analysis unit 104 ends the operation illustrated in Fig. 7.
  • Step S301 The step next to the Step S301 whose details are illustrated in Fig. 7 is Step S302 whose detail is illustrated in Fig. 8.
  • Fig. 8 is a flowchart illustrating example operation of the AQ processing of the information processing apparatus 100 according to the present exemplary embodiment.
  • the analysis unit 104 repeats a loop B illustrated in Fig 8. That is, the analysis unit 104 repeats processing from Step S321 to Step S325 for each AQ in the sentences generated by splitting the thread of the input data.
  • the analysis unit 104 confirms whether an AQ in the sentences is similar to the Problem asked by the original poster (Step S321). When the AQ is similar to the Problem asked by the user (YES in Step S322), the analysis unit 104 ignores the AQ. That is, the analysis unit 104 skips further processing for the AQ.
  • the analysis unit 104 may classify and label the AQ and the Answer to AQ that is an answer to the AQ as Supplementary.
  • the analysis unit 104 confirms whether the Answer to AQ that is an answer to the AQ is classified as an Observation or a Suggestion (Step S323).
  • the analysis unit 104 may classify the Answer to AQ as a Suggestion or an Observation by using classification technology, such as NLP technology, similar to that used by the classification unit 103. In this case, the analysis unit 104 classifies the Answer to AQ as a Suggestion or an Observation.
  • classification technology such as NLP technology
  • the classification unit 103 may further classify the Answer to AQ as a Suggestion or an Observation.
  • the analysis unit 104 may send identification of the Answer to AQ stored in a storage unit (not illustrated) to the classification unit 103.
  • the analysis unit 104 may send the Answer to AQ to the classification unit 103.
  • the classification unit 103 may send a result of classifying the Answer to AQ to the analysis unit 104.
  • the analysis unit 104 converts the AQ that is the question whose answer is the Answer to AQ to a declarative sentence.
  • the analysis unit 104 classifies and labels the declarative sentence generated by converting from the AQ as an Observation.
  • the analysis unit 104 ignores the AQ and Answer to AQ.
  • the analysis unit 104 may label the AQ as Supplementary.
  • the analysis unit 104 may label the Answer to AQ as Supplementary.
  • the analysis unit 104 performs the processing from Step S321 to Step S325 for each of the sentences classified into the AQ category, and the analysis unit 104 ends the operation illustrated in Fig. 8. Then the operation illustrated in Fig. 6 ends also.
  • Step S302 whose details are illustrated in Fig. 8 is Step S106 illustrated in Fig. 4.
  • the generation unit 105 updates a knowledge base on the basis of the result of processing of Step S104 and Step S105 (Step S106).
  • the knowledge base is stored in the result storage unit 106.
  • the result of processing of Step S104 and Step S105 is the Problem, the Suggestion correlated with the Problem, and Observation correlated with the Problem, which are included in the sentences generated by splitting the thread of the input data.
  • the generation unit 105 When no knowledge base is stored in the result storage unit 106, the generation unit 105 generates a new knowledge base including the Problem, the Suggestion correlated with the Problem, and Observation correlated with the Problem. The generation unit 105 stores the new knowledge base in the result storage unit 106, as illustrated in Fig. 2.
  • the knowledge base illustrated in Fig. 2 is generated for one thread, or one problem asked by an original poster and suggestion to solve the problem.
  • the generation unit 105 merges the Problem, the Suggestion correlated with the Problem, and Observation correlated with the Problem into the knowledge base.
  • the generation unit 105 may perform clustering on the knowledge base so that similar Problems from the same domain are included in a cluster.
  • An example of a knowledge base for a domain is shown in Fig. 2.
  • the generation unit 105 may perform clustering and creating the final knowledge base by using machine learning technology.
  • the Problem, the Suggestion and the Observation included in the knowledge base stored in the result storage unit 106 is referred to as an "existing Problem", an "existing Suggestion”, an “existing Observation”, respectively.
  • the Problem, the Suggestion and the Observation included in the sentences generated by splitting the thread of the input date received in Step S101 is referred to as a "new Problem", a " new Suggestion”, a “ new Observation”, respectively.
  • the generation unit 105 may perform merging the new Problem, the new Suggestion and the new Observation into a knowledge base, for example, as follows.
  • the generation unit stores the new Problem and the new Suggestion in the result storage unit 106.
  • the generation unit 105 detects, in the result storage unit 106, an existing Observation that is the same as the new Observation.
  • the generation unit 105 correlates the existing Observation as the same Observation as the new Observation with the new Problem. In this case, the generation unit 105 may not store the new Observation that is the same as the detected existing Observation.
  • the generation unit 105 stores the new Observation in the result storage unit 106.
  • the generation unit 105 may detect the same existing Problem and existing Suggestion that are the same as the new Problem and new Suggestion. When the generation unit 105 detects the same existing Problem and existing Suggestion that are the same as the new Problem and new Suggestion, the generation unit 105 may not store the new Problem and new Suggestion in the result storage unit. In this case, the generation unit 105 confirms whether the result storage unit 106 stores the existing Observation that is the same as the new Observation correlated with the new Problem. When the generation unit 105 detects no existing Observation that is the same as the new Observation correlated with the new Problem, the generation unit 105 stores the new Observation in the result storage unit 106, and correlates the new Observation with the detected existing Problem.
  • the generation unit 105 may correlate the detected existing Observation with the detected existing Problem.
  • the output unit 107 outputs the knowledge base stored in the result storage unit 106 to an apparatus that utilize the knowledge base (Step S107).
  • One of the effects of the present exemplary embodiment of the present invention is that it is possible to capture scattered information included in verbal interaction such as conversation in a technical support center or a thread in an online discussion forum.
  • the analysis unit 104 converts discussion, such as the ⁇ RFD, Answer to RFD> pair and the ⁇ AQ, Answer to AQ> pair, in the verbal interaction into Observation.
  • the RFD, the Answer to RFD, the AQ and the Answer to AQ each do not always singly give information to understand and solve the problem explicitly.
  • An Observation, into which the discussion is converted, is able to give information to understand and solve the problem explicitly.
  • Fig. 1 is a block diagram illustrating an example structure of the information processing apparatus 100 according to the present exemplary embodiment.
  • the information processing apparatus 100 of the present exemplary embodiment has the same structure as that of the information processing apparatus 100 of the first exemplary embodiment.
  • Elements of the information processing apparatus 100 of the present exemplary embodiment are the same as the elements assigned the same codes in the information processing apparatus 100 of the first exemplary embodiment.
  • the information processing apparatus 100 of the present exemplary embodiment is the same as the information processing apparatus 100 of the first exemplary embodiment except the following differences. Therefore, detailed description of the information processing apparatus 100 of the present exemplary embodiment is omitted except the difference.
  • the analysis unit 104 of the present exemplary embodiment converts the AQ the answer of which is the Answer to AQ into a Problem, and converts the Answer to AQ into a Suggestion.
  • the analysis unit 104 identifies the Observations related to the Answer to AQ.
  • the analysis unit 104 identifies an Observation related to the Answer to AQ by similarity confirmation using NLP technology.
  • the analysis unit 104 correlates the Problem converted from the AQ with the Observation identified as an Observation related with the Problem converted from the AQ.
  • the analysis unit correlates the Problem converted from the AQ with the Suggestion converted from the Answer to AQ.
  • the generation unit 105 generates a new knowledge base on the basis of the Problem converted from the AQ, the Suggestion converted from the Answer to AQ and the Observation identified as an Observation related with the Problem converted from the AQ.
  • the generation unit 105 may merge the new knowledge base into an existing knowledge base that is a knowledge base stored in the result storage unit 106.
  • Fig. 4 is a flowchart illustrating an example of operation of the information processing apparatus 100 according to the present exemplary embodiment of the present invention.
  • the operation, illustrated in Fig. 4, of the information processing apparatus 100 of the present exemplary embodiment is the same as the operation of the information processing apparatus 100 of the first exemplary embodiment 100 except the processing in Step S105 and Step S106.
  • the detailed description of the operation illustrated in Fig. 4 of the information processing apparatus 100 is omitted except the following description of the differences.
  • step S105 when the Answer to AQ is classified as a Suggestion, the analysis unit 104 converts the AQ as a Problem, converts the Answer to AQ as a Suggestion, and identifies an Observation related to the Answer to AQ.
  • the generation unit 105 further generates a knowledge base on the basis of the Problem converted from the AQ, the Suggestion converted from the Answer to AQ and the Observation identified as an Observation related to the Answer to AQ.
  • Fig. 5 is a flowchart illustrating an example of operation in the preprocessing in the present exemplary embodiment.
  • the operation in the preprocessing of the present exemplary embodiment is the same as that of the first exemplary embodiment. Therefore, the detailed description of the operation illustrated in Fig. 5 of the information processing apparatus 100 is omitted.
  • Fig. 6 is a flowchart illustrating example operation of discussion processing of the information processing apparatus 100 according to the present invention.
  • the operation in the discussion processing of the present exemplary embodiment is the same as that of the first exemplary embodiment except the processing in Step S302.
  • the differences in the processing in Step S302 are described later.
  • Fig. 7 is a flowchart illustrating example operation of RFD processing of the information processing apparatus 100 according to the present invention.
  • the RFD processing of the present invention is the same as that of the first exemplary embodiment. Therefore, the detailed description of the RFD processing of the present invention is omitted.
  • Fig. 9 is a flowchart illustrating example operation of the AQ processing of the information processing apparatus 100 according to the present exemplary embodiment.
  • the processing from Step S321 to Step S325 of the present exemplary embodiment is the same as that of the first exemplary embodiment.
  • the following description represents the differences in the AQ processing between the present exemplary embodiment and the first exemplary embodiment.
  • the analysis unit 104 of the information processing apparatus 100 of the first exemplary embodiment ignores the AQ and the Answer to AQ when the Answer to AQ is classified as an Observation (NO in Step S324).
  • the analysis unit 104 of the information processing apparatus 100 of the present exemplary embodiment performs creation processing (Step S326).
  • Fig. 10 is a flowchart illustrating example operation of the creation processing of the information processing apparatus 100 of the present exemplary embodiment.
  • the analysis unit 104 identifies an Observation related to the Answer to AQ, in the sentences generated by splitting the thread of the input data (Step S401).
  • the analysis unit 104 converts the AQ as a Problem (Step S402).
  • the analysis unit 104 converts the Answer to AQ as a Suggestion (Step S403).
  • the analysis unit 104 correlates the Problem converted from the AQ with the Suggestion converted from the Answer to AQ and the Observation identified as an Observation related with the Answer to AQ that is converted to the Suggestion (Step S404).
  • the present invention has the same effect as that of the first exemplary embodiment.
  • the reason is the same as that of the first exemplary embodiment.
  • Another effect of the present exemplary embodiment is that it is possible to capture a common question asked by a poster. Because, when the Answer to AQ is classified as an Observation, the analysis unit 104 converts the AQ to a Problem, converts the Answer to AQ to a Suggestion and identifies the Observation related to the Answer to AQ. The generation unit 105 generates a knowledge base based on the Problem converted from the AQ, the Suggestion converted from the Answer to AQ, and the Observation related with the Answer to AQ.
  • Fig. 13 is a block diagram illustrating an information processing apparatus 100A according to the third exemplary embodiment of the present invention.
  • the information processing apparatus 100A includes an analysis unit 104 and a generation unit 105.
  • the analysis unit 104 generates an Observation based on a question and an answer to the question.
  • the Observation is an expression expressing a situation.
  • the expression may be a sentence.
  • the question and the answer are extracted from verbal interaction between a first person and a second person.
  • the first person has a problem.
  • the second person finds and provides a Suggestion to solve the problem.
  • the generation unit 105 generates a knowledge base based on the problem, the suggestion and the Observation converted from the question.
  • Fig. 14 is a flowchart illustrating example operation of the information processing apparatus 100A according to the present exemplary embodiment.
  • the analysis unit 104 generates an Observation based on a question and an answer to the question (Step S501).
  • the analysis unit 104 of the present exemplary embodiment may convert an RFD into an Observation on the basis of the Answer to RFD that is an answer to the question of the RFD.
  • the analysis unit 104 of the present exemplary embodiment may covert an AQ into an Observation when the Answer to AQ that is an answer to the AQ is an Observation.
  • the generation unit 105 generates a knowledge base based on the Observation generated by the analysis unit 104 (Step S502).
  • the present exemplary embodiment has the same effect as that of the first exemplary embodiment.
  • the reason for the effect is the same as that of the first exemplary embodiment.
  • the information processing apparatus is achieved by circuitry.
  • the circuitry may be one or more computers including at least a processor and a memory storing a program controlling the processor.
  • the computers are communicably connected with one another.
  • the circuitry may be one or more circuits communicably connected with one another.
  • the circuitry may be achieved by one or more computers and one or more circuits.
  • Fig. 15 is a block diagram illustrating a computer capable of achieving the information processing apparatus according to any one of the exemplary embodiments of the present invention.
  • the computer 1000 illustrated in Fig 15 includes a processor 1001, a memory 1002, a storage device 1003 and I/O (Input/Output) interface 1004 which are connected with one another via a bus.
  • a storage medium 1005 is connected with the bus of the computer 1000 so that the processor 1001 is able to access the storage medium 1005.
  • the memory 1002 is, for example, a DRAM (Dynamic Random Access Memory) or the like.
  • the storage device 1003 is, for example, an HDD (Hard Disk Drive), SSD (Solid State Drive) or the like.
  • the storage medium 1005 is a removable storage medium, such as a CD-ROM (Compact Disc Read Only Memory), a USB (Universal Serial Bus) memory or the like.
  • the I/O interface is an interface by which the processor is able to communicate with a device, such as, a keyboard or a mouse.
  • the I/O interface may be communicably connected with a communication network.
  • the processor may be communicate with an apparatus connected with the communication network.
  • the storage medium 1005 stores a program causing the computer 1000 operates as the information processing apparatus according to any one of the exemplary embodiments of the present invention.
  • the processor 1001 reads out the program from the storage medium 1005, and loads the program in the memory 1002.
  • the storage device 1003 may work as the storage medium 1005. By executing a program loaded in the memory 1002, the processor 1000 operates as the information processing apparatus according to any one of the exemplary embodiments of the present invention.
  • the raw data reception unit 101, the preprocessing unit 102, the classification unit 103, the analysis unit 104, the generation unit 105, and the output unit 107 are able to be achieved by the memory 1002 and the processor 1001 executing the program loaded in the memory 1002.
  • the result storage unit 106 and the dictionary storage unit 108 are able to be achieved by the memory 1002 or the storage device 1003.
  • All or a part of the raw data reception unit 101, the preprocessing unit 102, the classification unit 103, the analysis unit 104, the generation unit 105, the result storage unit 106, the output unit 107, and the dictionary storage unit 108 is able to be achieved by a dedicated circuits having functions of respective units.
  • Fig. 16 is a block diagram illustrating an example structure of the information processing apparatus 100, achieved by dedicated circuits, according to the first and the second exemplary embodiments of the present invention.
  • the information processing apparatus 100 illustrated in Fig. 16 includes a raw data reception circuit 1101, a preprocessing circuit 1102, a classification circuit 1103, an analysis circuit 1104, a generation circuit 1105, a result storage device 1106, an output circuit 1107 and a dictionary storage device 1108.
  • Fig. 17 is a block diagram illustrating an example structure of the information processing apparatus 100A, achieved by dedicated circuits, according to the third exemplary embodiment of the present invention.
  • the information processing device 100A illustrated in Fig. 17 includes the analysis circuit 1104 and the generation circuit 1105.
  • the raw data reception circuit 1101 operates as the raw data reception unit 101.
  • the raw data reception unit 101 may be achieved by the raw data reception circuit 1101.
  • the preprocessing circuit 1102 operates as the preprocessing unit 102.
  • the preprocessing unit 102 may be achieved by the preprocessing circuit 1102.
  • the classification circuit 1103 operates as the classification unit 103.
  • the classification unit 103 may be achieved by the classification circuit.
  • the analysis circuit 1104 operates as the analysis unit 104.
  • the analysis unit 104 may be achieved by the analysis circuit 1104.
  • the generation circuit 1105 operates as the generation unit 105.
  • the generation unit 105 may be achieved by the generation circuit 1105.
  • the result storage device 1106 operates as the result storage unit 106.
  • the result storage unit 106 may be achieved by the result storage device 1106.
  • the output circuit 1107 operates as the output unit 107.
  • the output unit 107 may be achieved by the output circuit 1107.
  • the dictionary storage device 1108 operates as the dictionary storage unit 108.
  • An information processing apparatus comprising: analysis means for generating an observation based on a question and an answer to the question, the observation being an expression expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and generation means for generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
  • Supplementary Note 2 The information processing apparatus according to Supplementary Note 1, wherein the analysis means generates the observation by converting a request-for-details question into the observation based on a result of determining whether the answer to the request-for-details question is affirmative, the request-for-details question being the question requesting information by the second user.
  • Supplementary Note 3 The information processing apparatus according to Supplementary Note 1 or 2, wherein the analysis means confirms a similarity between an additional question and the problem, and generates the observation by converting an additional question to which an observation is answered as an answer into the observation based on the similarity, the additional question being the question asked by the first user in addition to the problem in the verbal interaction.
  • An information processing method comprising: generating an observation based on a question and an answer to the question, the observation being an expression expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
  • Supplementary Note 8 The information processing method according to Supplementary Note 6 or 7, further comprising: confirming a similarity between an additional question and the problem, wherein the generating the observation includes converting an additional question to which an observation is answered as an answer into the observation based on the similarity, the additional question being the question asked by the first user in addition to the problem in the verbal interaction.
  • Supplementary Note 9 The information processing method according to Supplementary Note 8, further comprising: generating a new problem and a new suggestion by converting the additional question to which a suggestion is answered as an answer, and by converting the answer to the additional question into the new suggestion; and generating the knowledge base further based on the new problem and the new suggestion.
  • Supplementary Note 10 The information processing method according to Supplementary Note 9, further comprising: extracting an observation related to the new problem from the verbal interaction; and generating the knowledge base further based on the observation related to the new problem.
  • a computer-readable medium storing a program causing a computer to execute: analysis processing of generating an observation based on a question and an answer to the question, the observation being an expression expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and generation processing of generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
  • Supplementary Note 13 The computer-readable medium according to Supplementary Note 11 or 12, wherein the analysis processing confirms a similarity between an additional question and the problem, and generates the observation by converting an additional question to which an observation is answered as an answer into the observation based on the similarity, the additional question being the question asked by the first user in addition to the problem in the verbal interaction.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Human Computer Interaction (AREA)
  • Databases & Information Systems (AREA)
  • Machine Translation (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

One of the objects of the present invention is to provide an information processing apparatus capable of capturing scattered information included in verbal interaction. An information processing apparatus according to an exemplary aspect of the present invention includes: analysis means for generating an observation based on a question and an answer to the question, the observation expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and generation means for generating a knowledge base based on the problem, the suggestion and the observation converted from the question.

Description

INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD AND COMPUTER-READABLE MEDIUM
The present invention pertains to an information processing technology, in particular, to a technology of extracting knowledge base from data.
In technical support center conversations, a user with a problem will seek the assistance from the expert to solve the problem. The user usually describes the observation that the user observed and states the problem. The expert may understand the problem and provide suggestion to solve the problem. The expert may ask more details to the user to understand the problem and then provide suggestion to solve the problem. The user implements the solution, and may ask for further questions or report a function after implementing the solution. These discussions between the user and the expert happen as conversation. Similar conversations are in online discussion forums (hereinafter, referred to as "ODF"), plant maintenance logs, medical conversation data, etc. Due to availability of vast amount of data, for a new problem, identifying the solution is becoming difficult for users and experts.
PTL 1 discloses an example of technology of extracting knowledge from text data that is unstructured data by using a text corpus. The text corpus considered in PTL 1 includes documents from a vehicle service reporting system. Each vehicle service report generated in the vehicle service reporting system of PTL 1 includes an identified problem, symptoms associated with the problem and suggestion to solve the problem, written by the experts. In other words, the technology according to PTL 1 can extract knowledge from text data explicitly providing symptoms associated with problem, a problem to be solved, and suggestion to solve the problem.
PTL 2 discloses an example of technology of extracting knowledge from online discussion forums. The technology according to PTL 2 includes creating <question, answer> pair by selecting and ranking suitable answer for the questions from the threads of online forums. PTL 2 discloses a method to rank the answers from plural of answers for a question in online discussion forums. The approach used in PTL 2 can be applied to a thread that has all the details of the problem given by the user in the first post and an answer to the problem is given in a single post of the thread in such a way that the answer can be extracted directly from the single post as a response of a question concerning the problem.
United States Patent Specification No. 8489601 United States Patent Specification No. 7814048
In the conversational data like technical support center data, or in online discussion forums, symptoms associated with a problem, the problem to be solved, and solution (e.g. suggestion) to solve the problem are not explicitly provided. Discussions within the conversational data may implicitly provide information to understand the problem and/or solution of the problem. The technology according to PTL 1 is not able to extract knowledge base from unstructured data providing symptoms associated with problem, a problem to be solved, and solution to solve the problem not explicitly.
A thread in online discussion forums and conversation data of a technical support center often have the root post not including explanation of the symptoms and details of a problem. The root post is, for example, the first post of a thread or the first utterances including a question by the user at a technical support center. Though an expert may ask the user for more details to provide correct suggestion and the information of a question (referred to as an additional question) asked by the experts and an answer to the additional question may be in the discussions within the thread, the answer to the additional question may be missed while ranking the answers.
By the technology of PTL 1 or PTL 2, it is not possible to acquire scattered information, such as information that can be acquired from one sentence, included in verbal interaction.
One of the objects of the present invention is to provide an information processing apparatus capable of capturing scattered information included in verbal interaction.
An information processing apparatus according to an exemplary aspect of the present invention includes: analysis means for generating an observation based on a question and an answer to the question, the observation expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and generation means for generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
An information processing method according to an exemplary aspect of the present invention includes: generating an observation based on a question and an answer to the question, the observation expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
A computer-readable medium according to an exemplary aspect of the present invention stores a program causing a computer to execute: analysis processing of generating an observation based on a question and an answer to the question, the observation expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and generation processing of generating a knowledge base based on the problem, the suggestion and the observation converted from the question. The present invention may be achieved by the program stored in the computer-readable medium.
According to the present invention, it is possible to capture scattered information included in verbal interaction.
Fig. 1 is a block diagram illustrating an example structure of an information processing apparatus according to first and second exemplary embodiments of the present invention. Fig. 2 is a block diagram illustrating an example structure of the knowledge base according to any of exemplary embodiments of the present invention. Fig. 3 is a block diagram illustrating an example structure of the knowledge base according to any of exemplary embodiments of the present invention. Fig. 4 is a flowchart illustrating an example of operation of the information processing apparatus according to the first and second exemplary embodiments of the present invention. Fig. 5 is a flowchart illustrating an example of operation in the preprocessing of the information processing apparatus according to the first and second exemplary embodiments of the present invention. Fig. 6 is a flowchart illustrating example operation of discussion processing of the information processing apparatus according to the first and second exemplary embodiments of the present invention. Fig. 7 is a flowchart illustrating example operation of RFD processing of the information processing apparatus according to the first and second exemplary embodiments of the present invention. Fig. 8 is a flowchart illustrating example operation of AQ processing of the information processing apparatus according to the first exemplary embodiment of the present invention. Fig. 9 is a flowchart illustrating example operation of the AQ processing of the information processing apparatus according to the second exemplary embodiment of the present invention. Fig. 10 is a flowchart illustrating example operation of the creation processing of the information processing apparatus 100 of the second exemplary embodiment of the present invention. Fig. 11 is a block diagram illustrating a schematic example of whole structure of an information processing system according to the first and second exemplary embodiments of the present invention. Fig. 12 is a block diagram illustrating another schematic example of whole structure of an information processing system according to the first and second exemplary embodiments of the present invention. Fig. 13 is a block diagram illustrating an information processing apparatus according to a third exemplary embodiment of the present invention. Fig. 14 is a flowchart illustrating example operation of the information processing apparatus according to the third exemplary embodiment of the present invention. Fig. 15 is a block diagram illustrating a computer capable of achieving the information processing apparatus according to any one of the exemplary embodiments of the present invention. Fig. 16 is a block diagram illustrating an example structure, achieved by circuits, of the information processing apparatus 100, achieved by dedicated circuits, according to the first and the second exemplary embodiments of the present invention. Fig. 17 is a block diagram illustrating an example structure of the information processing apparatus, achieved by dedicated circuits, according to the third exemplary embodiment of the present invention.
Exemplary embodiments of the present invention are described below with reference to drawings.
<First Exemplary Embodiment>
A first exemplary embodiment of the present invention is described with reference to drawings.
Fig. 1 is a block diagram illustrating an example structure of an information processing apparatus 100 according to the first exemplary embodiment of the present invention. The information processing apparatus 100 illustrated in Fig. 1 includes a raw data reception unit 101, a preprocessing unit 102, a classification unit 103, an analysis unit 104, and a generation unit 105. The information processing apparatus 100 may further include a dictionary storage unit 108. The information processing apparatus according any one of exemplary embodiment of the present invention may be referred to as a "generation apparatus" or a "knowledge base generation apparatus". Note that directions of arrows drawn in Fig. 1 and other figures do not limit directions of data transmission.
The raw data reception unit 101 receives raw data. The raw data is, for example, text data of a thread in an online discussion forum. An interaction between a user and an expert to solve a problem is termed as a "thread". The thread may be posts (i.e. items posted by posters) each of which is related with at least one of the other posts in the thread. A post in the posts includes text data and information (i.e. identification such as e-mail address) of a poster who posts the post. The posters, such as, a user (e.g. a customer) and an expert are referred to as "participants".
The raw data may be conversation data that is data of conversation at a support center. The support center may be a technical support center of an industrial product provider such as, a personal computer provider, or a car manufacturer. The support center may be a control room supporting plant maintenance. The support center may be a clinic, a hospital, or a medical center. The conversation data may be data in text format. The conversation data may be data in audio format and may be generated by recording and digitizing conversation. Text data converted from the conversation between a user and an expert at a support center is also referred to as the "thread". The user and the expert at the support center are also referred to as "participants".
The preprocessing unit 102 performs preprocessing including splitting the thread into sentences. The preprocessing unit 102 correlates, on the basis of recorded identification included in the thread, the sentences with their respective posters. The preprocessing unit 102 may correlate a sentence in the sentences with a user asking an expert for solution to a problem or the expert providing the solution to the problem. When two or more experts participate in a communication of the thread, the preprocessing unit 102 may not distinguish between the experts. The preprocessing unit 102 may assume the experts to be one expert.
When the raw data is in audio format, the preprocessing unit 102 converts the raw data in audio format into text data by using existing speech recognition technology, and then performs the preprocessing on the text data to generate the sentences. The preprocessing unit 102 may distinguish speakers of the sentences by using a voice authentication technology.
The classification unit 103 classifies the sentences into categories including three main categories that are observation or symptoms of a problem, the problem, and suggestion to solve the problem. In the description of the present exemplary embodiment of the present invention, a sentence representing observation or symptoms of a problem is referred to as an "Observation". A category of Observation is referred to as an "Observation category". A sentence representing a problem is referred to as a "Problem". A category of Problem is referred to as a "Problem category". A sentence indicating suggestion to solve a problem is referred to as a "Suggestion". A category of Suggestion is referred to as a "Suggestion category".
A thread in an online discussion forum may include a short discussion such as a set of a question and an answer. The set of a question and an answer is referred to as a "<question, answer> pair". The <question, answer> pair may be classified into either of two types based on who raised the question as described below.
If the expert asks the question, a purpose of the question may be to request for more details, more specifically, a) to get more details of the problem, b) to understand the environment of the problem, or c) to get feedback of the solution suggested. Hereinafter, a question to request for more details is referred to as an "RFD". An answer, by an original poster (i.e. the user who started the thread), to the RFD (i.e. a response by the original poster to the question classified as an RFD) is referred to as an "Answer to RFD". The <question, answer> pair including a question classified as an RFD and an answer to the RFD is referred to as an "<RFD, Answer to RFD>" or an "<RFD, Answer to RFD>" pair. A category of an RFD is referred to as an "RFD category". The original poster is referred to as an "OP".
If an original poster asks the question, the question is an additional question, for example, a) to understand the solution suggested, b) to ask a question about a new problem raised by carrying out the suggestion suggested by the expert, or c) to ask a new question. Hereinafter, a question classified as the additional question is referred to as an "AQ". An answer, by an expert, to an AQ is referred to as an "Answer to AQ". The <question, answer> pair including a question classified as an AQ is referred to as an "<AQ, Answer to AQ>" or an "<AQ, Answer to AQ>" pair. A category of an AQ is referred to as an "AQ category". A category of an Answer to AQ is referred to as an "Answer to AQ category".
The categories other than the three main categories described above are the RFD category, the Answer to RFD category, the AQ category, and the Answer to AQ category.
A sentence classified into any of the above-described categories is considered to be important to understand the problem or to implement the suggestion. A sentence which is not able to be classified into any of the above-described categories is labeled as supplementary. Hereinafter, a sentence labeled as supplementary is referred to as "Supplementary". A category of a Supplementary is referred to as a "Supplementary category". A sentence labeled as supplementary is not important to understand the problem or to implement the suggestion.
In summary, a sentence generated by splitting a thread is classified as:
(a) Observation - information needed to solve a problem, and environmental information to understand the problem to suggest solution;
(b) Problem - a main question asked by an original poster, and the first problem reported by the original poster;
(c) Suggestion - the solution suggested by an expert to solve the problem;
(d) RFD - question asked by an expert to get more details about the problem or to know the environment, and any question asked by the expert relevant to the problem;
(e) Answer to RFD - an answer provided by the original poster to answer RFD;
(f) Additional Questions (AQ) - an additional questions asked by the original poster, for example, after conversation with the expert;
(g) Answer to AQ - an answer provided by the expert for AQ; or
(h) Supplementary - a sentence which is not important to understand the problem or to implement the suggestion.
The classification unit 103 classifies a sentence generated from the thread by preprocessing unit 102 into one of the categories. The classification unit 103 may operate as a machine classifier classifying a sentence into one of the above-described categories. In other words, the classification unit 103 may classify the sentence according to a result of machine learning such as supervised learning. The result of machine learning is, for example, a dictionary generated in advance by machine learning to classify a sentence of a thread into one of the categories described above. The method of machine learning may be selected by an operator. The format of the dictionary may be defined so that the classification unit 103 is capable of reading using the dictionary. The dictionary is stored in the dictionary storage unit 108.
The classification unit 103 correlates an RFD with an Answer to RFD which is an answer to the RFD. The RFD and the Answer to RFD which is correlated with the RFD is the above-mentioned < RFD, Answer to RFD > pair.
The classification unit 103 correlates an AQ with an Answer to AQ which is an answer to the AQ. The AQ and the Answer to AQ which is correlated with the AQ is the above-mentioned <AQ, Answer to AQ> pair.
The classification unit 103 may first detect, on the basis of the machine learning, a non-related sentence, such as a monologue, greetings, or an acknowledgement, in the sentences generated by splitting the thread. The classification unit 103 may remove the detected non-related sentences from the sentences. The classification unit 103 may classifies, into the categories, the sentence in the sentences from which the non-related sentence is removed. When classifying the sentences into the categories, the classification unit 103 may classify the non-related sentence as Supplementary.
The dictionary unit 108 stores the dictionary to classify a sentence as one of sentences including an Observation, a Problem, a Suggestion, an RFD, an Answer to RFD, an AQ, an Answer to AQ, and a Supplementary. In other words, the dictionary storage unit 108 stores the dictionary to classify a sentence into one of the categories including the Observation category, the Problem category, the Suggestion category, the RFD category, the Answer to RFD category, the AQ category, the Answer to AQ category, and the Supplementary category.
To capture the details and environment information required to understand the problem and suggest correct suggestion, the analysis unit 104 analyzes a <RFD, Answer to RFD> pair and extracts, from the <RFD, Answer to RFD> pair, information to be integrated into a knowledge base described below. The analysis unit 104 also analyzes a <AQ, Answer to AQ> pair and extracts, from the <AQ, Answer to AQ> pair, information to be integrated into the knowledge base.
In other words, the analysis unit 104 merges an RFD and an Answer to RFD included in an <RFD, Answer to RFD> pair into an Observation according to the Answer to RFD as follows.
The analysis unit 104 confirms whether the Answer to RFD is affirmative or negative. More specifically, the analysis unit 104 confirms whether the Answer to RFD is a yes-type answer (e.g. an answer including a word "yes") or a no-type answer (e.g. an answer including a word "no"). When the Answer to RFD includes a word "yes", the analysis unit 104 may estimate the Answer to RFD to be an affirmative answer. When the Answer to RFD includes a word "no", the analysis unit 104 may estimate the Answer to RFD to be a negative answer. The Answer to RFD is not a yes-type answer or a no-type answer, the analysis unit estimates whether the Answer to RFD is affirmative or negative by using natural language processing (NLP) technology, such as a sentiment analysis predicting sentiment of the Answer to RFD. When the Answer to RFD is estimated to be affirmative, the analysis unit 104 estimates the Answer to RFD is a yes-type answer. When the Answer to RFD is estimated to be negative, the analysis unit 104 estimates the Answer to RFD to be a no-type answer.
Next, the analysis unit 104 converts the RFD into a declarative sentence according to the Answer to RFD. When the Answer to RFD is estimated to be an affirmative answer, the analysis unit 104 converts the RFD into an affirmative sentence. When the Answer to RFD is estimated to be a negative answer, the analysis unit 104 convers the RFD into a negative sentence. The analysis unit 104 classifies (e.g. labels) the sentence converted from the RFD as an Observation.
When the Answer to RFD is not able to be estimated to be an affirmative answer or a negative answer, the analysis unit 104 skips and ignores the <RFD, Answer to RFD> pair. The analysis unit 104 may label the RFD and the Answer to RFD included in the <RFD, Answer to RFD> pair into the Supplementary category.
The analysis unit 104 confirms whether the AQ is similar to the Problem asked by OP. The analysis unit 104 may estimate, by using NLP technology described above, whether the AQ is similar to the Problem asked by OP. When the AQ is estimated to be similar to the Problem asked by OP, the analysis unit 104 skips and ignores the <AQ, Answer to AQ> pair. The analysis unit 104 may classify the AQ and the Answer to AQ included in the <AQ, Answer to AQ> pair into Supplementary category.
When the AQ is estimated to be not similar to the Problem asked by OP, the analysis unit 104 classifies, by using the dictionary stored in the dictionary storage unit 108, the Answer to AQ as a Suggestion or an Observation in the same way as the classification unit 103 classifies the sentence generated by splitting the thread. The analysis unit 104 may send the Answer to AQ to the classification unit 103. The classification unit 103 may classify the Answer to AQ as a Suggestion or an Observation. The classification unit 103 may send a result of classification to the analysis unit 104.
When the Answer to AQ is classified as a Suggestion, the analysis unit 104 may skip and ignore the <AQ, Answer to AQ> pair. The analysis unit 104 may label the AQ and the Answer to AQ into the Supplementary category.
When the Answer to AQ is classified as an Observation, the analysis unit 104 convert the AQ included in the <AQ, Answer to AQ> pair into an Observation by changing, for example, the AQ written as an interrogative sentence to a declarative sentence. For example, the AQ is a question to ask the reason of environmental information such as a situation or an event arising in the environment of OP, the analysis unit 104 converts the AQ into an Observation representing the environmental information. The processing of converting the AP to an Observation is not limited to the example described above. The analysis unit 104 classifies (e.g. labels) the sentence converted from the AQ as an Observation.
The generation unit 105 generates a knowledge base based on the Problem, the Suggestion, and the Observation generated from the thread. The generation unit 105 stores the knowledge base in the result storage unit 106. The generation unit 105 generates the knowledge base based on the Problem, the Suggestion, and two or more of the Observations when the number of sentences classified as Observations is two or more.
The generation unit 105 may correlate the Problem and the Suggestion. The generation unit 105 may correlate the Problem and the Observations. The generation unit 105 may store the Problem, the Suggestion correlated with the Problem, and the Observation correlated with the Problem in the result storage unit 106 as the knowledge base.
A knowledge base generated from an online discussion forum can assist a new user to identify the relevant solution for a problem occurring to the new user. The online discussion forum may be provided by an industrial product provider in order to support customers of the industrial product provider, and may be maintained by an expert belonging to the industrial product provider.
The knowledge base generated from the unstructured data (i.e. the raw data and the input data) is used to respond to a new problem asked by users in future. The knowledge base generated by the information processing apparatus 100 of the present exemplary embodiment is able to help experts to make correct decision in short time, and to provide relevant information to users. Hence, the time spent by an expert for each problem is reduced and user satisfaction improves.
When the unstructured data is data of conversation at a control room supporting plant maintenance, the information processing apparatus 100 may generate a knowledge base for plant maintenance. By using the knowledge base for plant maintenance, the problem that will arise in the future is quickly detected and suggestion is provided immediately. Hence, the down time in the industry is reduced.
When the unstructured data is data of medical conversation to identify (a) the symptoms of the disease, (b) disease, and (c) remedies for the disease, the information processing apparatus 100 may generate a knowledge base used for identifying the disease when the patient provides the symptoms associated with the disease. The knowledge base can assist the medical professional to make decision about the disease and to treat the patient immediately. The patient can directly interact with the system which uses knowledge base to understand the reasons for symptoms and to do further actions.
Fig. 11 is a block diagram illustrating a schematic example of whole structure of an information processing system 1 according to the present exemplary embodiment of the present invention.
The information processing system 1 in Fig. 11 includes the information processing apparatus 100, a knowledge base storage system 200, a suggestion apparatus 300, a training apparatus 400, an acquisition apparatus 500, an ODF system 700, and a terminal apparatus 800. The information processing apparatus 100 may include any of the knowledge base storage system 200, the suggestion apparatus 300, the training apparatus 400 and the acquisition apparatus 500. The information processing system 1 may include an information processing apparatus according to one of exemplary embodiments other than the first exemplary embodiments instead of the information processing apparatus 100 according to the first exemplary embodiment.
The acquisition apparatus 500, the ODF system 700 and the terminal apparatus 800 are communicably connected via a network 600 that is a communication network. All or a part of the apparatus included in the information processing system 1 may be communicably connected with the network 600.
The ODF system 700 provides an online discussion forum. The acquisition apparatus 500 acquires a thread of the online discussion forum provided by the ODF system 700.
The training apparatus 400 generates the dictionary stored in the dictionary storage unit 108 by using machine learning technology. Training data for machine learning to generate the dictionary is provided to the training apparatus 400 by, for example, a user terminal (not illustrated) communicably connected with the training apparatus 400.
The knowledge base storage system 200 stores the knowledge base generated by the generation unit 105 and output by the output unit 107 of the information processing apparatus 100. The knowledge base system 200 provides the knowledge base to the suggestion apparatus 300.
The suggestion apparatus 300 may receive inquiry indicating, for example, one or more Observations from the terminal apparatus 800, may search for the Problem correlated with the Observation, and may send, for example, the Problem and the Suggestion correlated with the Problem to the terminal apparatus 800.
The terminal apparatus 800 is, for example, a computer used by an expert. The expert reads a Problem and an Observation concerning the Problem in the online discussion forum provided by the ODF system 700. The expert searches for the similar Problem as the Problem in the online discussion forum by sending the inquiry indicating the Observation concerning the Problem in the online discussion forum. If the similar Problem is found, the expert may post a Suggestion to the online discussion as a response to the Problem in the online discussion forum with reference to the Suggestion to the similar Problem.
Fig. 12 is a block diagram illustrating another schematic example of whole structure of an information processing system 1A according to the present exemplary embodiment of the present invention.
The information processing system 1A in Fig. 12 includes the information processing apparatus 100, the knowledge base storage system 200, the suggestion apparatus 300, the training apparatus 400, the terminal apparatus 800, and a recording apparatus 900. The recording apparatus 900 may be achieved by a computer connected with a microphone. The recording apparatus 900 may be achieved by an IC (Integrated Circuit) recorder. The information processing system 1A may include an information processing apparatus according to one of exemplary embodiments other than the first exemplary embodiments instead of the information processing apparatus 100 according to the first exemplary embodiment.
The recording apparatus 900 records conversation between an expert and a user (e.g. a customer) as conversation data in an audio format. The recording apparatus 900 transmits the conversation data to the acquisition apparatus 500. The expert and the user may make face to face conversation. The expert and the user may make conversation over the telephone.
The acquisition apparatus 500 receives the conversation data, and inputs the conversation data into the information processing apparatus 100 as the input data.
The information processing apparatus 100, the knowledge base storage system 200, the suggestion apparatus 300, the training apparatus 400 and the terminal apparatus 800 are the same as those of the information processing system 1 illustrated in Fig. 11.
The expert recognizes a Problem and an Observation concerning the Problem in the conversation. The expert searches for the similar Problem as the Problem recognized in the conversation by sending the inquiry indicating the Observation concerning the recognized Problem by using the terminal 800. If the similar Problem is found, the expert may provide the user with a Suggestion with reference to the Suggestion to the similar Problem.
Fig. 2 is a block diagram illustrating an example structure of the knowledge base according to the present exemplary embodiment. Fig. 2 illustrates a schematic structure of the knowledge base based on the Problem, the Suggestion, and the Observations generated from a single thread.
The generation unit 105 may generate the knowledge base based on Problems, Suggestions, and Observations generated from different threads.
The generation unit 105 may correlate the Problem and the Suggestion generated from the same thread. The generation unit 105 may correlate the Observations and the Problem generated from the same thread. When the same Observations are generated from different threads, the generation unit 105 may integrate the same Observations into an integrated Observation and correlate the integrated Observation with the different threads.
The generation unit 105 may store the Problems, the Suggestions each correlated with the Problems, and the Observations each of which is correlated with at least one of the Problems in the result storage unit 106 as the knowledge base.
Fig. 3 is a block diagram illustrating an example structure of the knowledge base according to the present exemplary embodiment. Fig. 3 illustrates a schematic structure of the knowledge base based on the Problems, the Suggestions, and the Observations generated from a plurality of threads.
The result storage unit 106 stores the knowledge base generated by the generation unit 106.
The output unit 107 outputs knowledge base data representing the knowledge base to an apparatus storing the knowledge base. The apparatus may be connected with terminal devices for experts in a support center and may provide the experts with information for suggestion to users.
Next, an example of operation of the information processing apparatus 100 is described with reference to drawings.
Fig. 4 is a flowchart illustrating an example of operation of the information processing apparatus 100 according to the present exemplary embodiment of the present invention.
At the start of the operation illustrated in Fig. 4, the dictionary storage unit 108 stores the above-described dictionary of classifying a sentence into the categories described above.
As described above, a problem asked by a user and the complete interaction between an expert and the user, to solve the problem, is referred to as a thread.
Input data representing the thread is inputted to the information processing apparatus 100 by an apparatus, such as an apparatus collecting a thread from an online discussion forum, or an apparatus recording conversation between a user and an expert at a support center.
First, the raw data reception unit 101 receives input data (Step S101). The input data may be data in audio or data in text format.
Next, the preprocessing unit 102 performs preprocessing on the input data (Step S102).
Fig. 5 is a flowchart illustrating an example of operation in the preprocessing of the information processing apparatus 100 according to the present exemplary embodiment.
If the input data is not text data (NO in Step S201), the preprocessing unit 102 converts the input data into text data (Step S202). As described above, the input data may be data in an audio format (i.e. audio data) of conversation at a support center. The preprocessing unit 102 may extract text from the input data by using speech recognition technology. The input data may be data, other than audio data or text data, which can be converted into text data by the preprocessing unit 102.
The text data includes the thread in text format. The preprocessing unit 102 splits the text data generated by converting the audio data into sentences (Step S203). The preprocessing unit 102 may further specify respective speakers of the sentences by, for example, voice recognition. The preprocessing unit 102 may estimate which of a user and an expert is the speaker of a sentence. The preprocessing unit 102 may classify the speakers of the sentences into three or more individual participants. The preprocessing unit 102 may correlate, as a poster of a sentence, the speaker of the sentence with the sentence.
If the input data is not audio data (i.e. if the input data is text data) (YES in Step S201), the preprocessing unit 102 splits text data, that is the input data in text format, into sentences (Step S203). The input data represents a thread. In Step S203, the preprocessing unit 102 splits the thread into sentences (Step S203).
The preprocessing unit 102 specifies posters of the sentences. The thread in an online discussion forum may be a set of posts each of which is related with at least a post in the set. The post is a message, such as a question, an answer or a comment, posted to the online discussion forum. Each of the post includes poster information that is information of a poster who posts a post. The preprocessing device may specify a poster of a sentence on the basis of the poster information of a post including the sentence. The preprocessing unit 102 correlates the sentences with their respective posters.
The preprocessing unit 102 may arrange the sentences in order of the sentences in the thread. The processing unit 102 may correlate the sentences with order of the sentences in the thread.
Then the operation illustrated in Fig. 5 ends. The information processing apparatus 100 performs the processing in Step S103 next.
The outcome of the preprocessing processing in Step S102 is the sentences and information of respective posters (i.e. participants) of the sentences. The participants are classified into two types, a type of user who asks a question, called as an original poster, and a type of expert who answers the question, called an expert. In the description of the present exemplary embodiment, the expert is referred to as an "expert". The sentences may be arranged according to order of the sentences arisen in conversation, represented by the input data, between a user and an expert.
Next, the classification unit 103 removes a non-related data (Step S103). The non-related data is a sentence not related with a question concerning a problem, an answer to the question, and an observation concerning the problem. The non-related data is also referred to as a "non-related sentence". The classification unit 103 may extract, from the sentences, a non-related sentence, such as utterance monologue, greetings, or an acknowledgement. The classification unit 103 removes the non-related sentence from the sentences on the basis of, for example, machine learning.
The classification unit 103 classifies related sentences into the categories described above (Step S104). The related sentences are sentences not removed as the non-related sentence. The classification unit 103 labels the related sentences.
Next, the analysis unit 104 performs the discussion processing (Step S105).
The Observation, the Problem and the Suggestion are used to create the knowledge base as shown in Fig 2. The RFD, the Answer to RFD, the AQ, and the Answer to AQ each are not sufficient to be used alone to create the knowledge base. The <RFD, Answer to RFD> pair and the <AQ, Answer to AQ> pair may include information to capture the details and environment information required to understand the problem and to suggest correct suggestion. In step S105, the analysis unit 104 analyzes the <RFD, Answer to RFD> pair and the <AQ, Answer to AQ> pair and derives information to be integrated to the knowledge base.
Fig. 6 is a flowchart illustrating example operation of discussion processing of the information processing apparatus 100 according to the present exemplary embodiment. According to operation in Fig. 6, the analysis unit 104 performs RFD processing (Step S301), and performs AQ processing (Step S302). In Step S301, the analysis unit 104 merges the RFD and the Answer to RFD that is the answer to the RFD to generate a declarative sentence, and classifies the generated sentence as an Observation. In Step S302, the analysis unit 104 generates a declarative sentence on the basis of the AQ and the Answer to AQ, and classifies the generated sentence as an Observation.
Fig. 7 is a flowchart illustrating example operation of RFD processing of the information processing apparatus 100 according to the present exemplary embodiment.
The analysis unit 104 repeats a loop A illustrated in Fig 7. That is, the analysis unit 104 repeats processing from Step S311 to Step S313 for each RFD in the sentences generated by splitting the thread of the input data.
The analysis unit 104 classifies the Answer to RFD as yes or no (Step S311).
To merge RFD and Answer to RFD, the analysis unit 104 first confirms whether the Answer to RFD is a yes-type answer or a no-type answer. If the Answer to RFD is not a simple yes-type answer or a simple no-type answer, then using natural language processing (NLP) technology, such as sentiment analysis, the analysis unit 104 predicts sentiment of the Answer to RFD, and classifies the Answer to RFD as affirmative or non-affirmative. When the Answer to RFD is classified as affirmative, the analysis unit 104 estimates the Answer to RFD to be yes. When the Answer to RFD is classified as negative, the analysis unit 104 estimates the Answer to RFD to be no.
When classification does not succeed (NO in Step S312), that is, when the analysis unit 104 does not successfully classify the Answer to RFD as yes or no, the analysis unit 104 skips and ignores the <RFD, Answer to RFD> pair. When the Answer to RFD is "I don't know" or the like, the Answer to RFD may not be classified as yes or no.
When classification succeeds (YES in Step S312), that is, when the analysis unit 104 successfully classifies the Answer to RFD as yes or no, the analysis unit 104 convert the RFD on the basis of the Answer to RFD into a declarative sentence, and classifies and labels the declarative sentence as an Observation(Step S313). The analysis unit 104 converts the RFD to a declarative sentence based on the type of the Answer to RFD. The following example is a part of conversation, in a PC (Personal Computer) technical support center, when a user reports a problem that the user's PC is not connected to a network. Then if an expert asks for more information, such as whether the user have a firewall program installed in his PC or not, then a question to ask for more information, given below, is categorized as an RFD.
RFD: "Do you have a firewall program installed in your PC?"
If the Answer to RFD is "Yes", then the analysis unit 104 converts the RFD into "a firewall program is installed in the PC." If Answer to RFD is "No", then the analysis unit 104 converts the RFD into "A firewall is not installed in the PC."
The analysis unit 104 may convert an interrogative sentence representing the question of RFD to an affirmative sentence or a negative sentence by using NLP technology. The analysis unit 104 may classify the converted sentence "Firewall installed in the PC." or "Firewall not installed in the PC." as an Observation.
The analysis unit 104 performs the processing from Step S311 to Step S313 for each of the sentences classified into the RFD category, and the analysis unit 104 ends the operation illustrated in Fig. 7.
The step next to the Step S301 whose details are illustrated in Fig. 7 is Step S302 whose detail is illustrated in Fig. 8.
Fig. 8 is a flowchart illustrating example operation of the AQ processing of the information processing apparatus 100 according to the present exemplary embodiment.
The analysis unit 104 repeats a loop B illustrated in Fig 8. That is, the analysis unit 104 repeats processing from Step S321 to Step S325 for each AQ in the sentences generated by splitting the thread of the input data.
The analysis unit 104 confirms whether an AQ in the sentences is similar to the Problem asked by the original poster (Step S321). When the AQ is similar to the Problem asked by the user (YES in Step S322), the analysis unit 104 ignores the AQ. That is, the analysis unit 104 skips further processing for the AQ. The analysis unit 104 may classify and label the AQ and the Answer to AQ that is an answer to the AQ as Supplementary.
When the AQ is not similar to the Problem (NO in Step S322), the analysis unit 104 confirms whether the Answer to AQ that is an answer to the AQ is classified as an Observation or a Suggestion (Step S323).
The analysis unit 104 may classify the Answer to AQ as a Suggestion or an Observation by using classification technology, such as NLP technology, similar to that used by the classification unit 103. In this case, the analysis unit 104 classifies the Answer to AQ as a Suggestion or an Observation.
The classification unit 103 may further classify the Answer to AQ as a Suggestion or an Observation. In this case, the analysis unit 104 may send identification of the Answer to AQ stored in a storage unit (not illustrated) to the classification unit 103. The analysis unit 104 may send the Answer to AQ to the classification unit 103. The classification unit 103 may send a result of classifying the Answer to AQ to the analysis unit 104.
When the Answer to AQ is classified as an Observation (YES in Step S324), the analysis unit 104 converts the AQ that is the question whose answer is the Answer to AQ to a declarative sentence. The analysis unit 104 classifies and labels the declarative sentence generated by converting from the AQ as an Observation. When the Answer to AQ is classified as a Suggestion (NO in Step S324), the analysis unit 104 ignores the AQ and Answer to AQ. The analysis unit 104 may label the AQ as Supplementary. The analysis unit 104 may label the Answer to AQ as Supplementary.
The analysis unit 104 performs the processing from Step S321 to Step S325 for each of the sentences classified into the AQ category, and the analysis unit 104 ends the operation illustrated in Fig. 8. Then the operation illustrated in Fig. 6 ends also.
The step next to the Step S302 whose details are illustrated in Fig. 8 is Step S106 illustrated in Fig. 4.
The generation unit 105 updates a knowledge base on the basis of the result of processing of Step S104 and Step S105 (Step S106). The knowledge base is stored in the result storage unit 106. The result of processing of Step S104 and Step S105 is the Problem, the Suggestion correlated with the Problem, and Observation correlated with the Problem, which are included in the sentences generated by splitting the thread of the input data.
When no knowledge base is stored in the result storage unit 106, the generation unit 105 generates a new knowledge base including the Problem, the Suggestion correlated with the Problem, and Observation correlated with the Problem. The generation unit 105 stores the new knowledge base in the result storage unit 106, as illustrated in Fig. 2.
As described above, the knowledge base illustrated in Fig. 2 is generated for one thread, or one problem asked by an original poster and suggestion to solve the problem.
When a knowledge base is stored in the result storage unit 106, the generation unit 105 merges the Problem, the Suggestion correlated with the Problem, and Observation correlated with the Problem into the knowledge base.
The generation unit 105 may perform clustering on the knowledge base so that similar Problems from the same domain are included in a cluster. An example of a knowledge base for a domain is shown in Fig. 2. The generation unit 105 may perform clustering and creating the final knowledge base by using machine learning technology.
Hereinafter, the Problem, the Suggestion and the Observation included in the knowledge base stored in the result storage unit 106 is referred to as an "existing Problem", an "existing Suggestion", an "existing Observation", respectively. the Problem, the Suggestion and the Observation included in the sentences generated by splitting the thread of the input date received in Step S101 is referred to as a "new Problem", a " new Suggestion", a " new Observation", respectively.
The generation unit 105 may perform merging the new Problem, the new Suggestion and the new Observation into a knowledge base, for example, as follows. The generation unit stores the new Problem and the new Suggestion in the result storage unit 106. The generation unit 105 detects, in the result storage unit 106, an existing Observation that is the same as the new Observation. When the existing Observation that is the same as the new Observation is detected in the result storage unit 106, the generation unit 105 correlates the existing Observation as the same Observation as the new Observation with the new Problem. In this case, the generation unit 105 may not store the new Observation that is the same as the detected existing Observation. When no existing Observation that is the same as the new Observation is detected in the result storage unit 106, the generation unit 105 stores the new Observation in the result storage unit 106.
The generation unit 105 may detect the same existing Problem and existing Suggestion that are the same as the new Problem and new Suggestion. When the generation unit 105 detects the same existing Problem and existing Suggestion that are the same as the new Problem and new Suggestion, the generation unit 105 may not store the new Problem and new Suggestion in the result storage unit. In this case, the generation unit 105 confirms whether the result storage unit 106 stores the existing Observation that is the same as the new Observation correlated with the new Problem. When the generation unit 105 detects no existing Observation that is the same as the new Observation correlated with the new Problem, the generation unit 105 stores the new Observation in the result storage unit 106, and correlates the new Observation with the detected existing Problem. When the generation unit 105 detects an existing Observation that is the same as the new Observation correlated with the new Problem, and the detected existing Observation is not correlated with the detected existing Problem, the generation unit 105 may correlate the detected existing Observation with the detected existing Problem.
The output unit 107 outputs the knowledge base stored in the result storage unit 106 to an apparatus that utilize the knowledge base (Step S107).
One of the effects of the present exemplary embodiment of the present invention is that it is possible to capture scattered information included in verbal interaction such as conversation in a technical support center or a thread in an online discussion forum. Because the analysis unit 104 converts discussion, such as the <RFD, Answer to RFD> pair and the <AQ, Answer to AQ> pair, in the verbal interaction into Observation. The RFD, the Answer to RFD, the AQ and the Answer to AQ each do not always singly give information to understand and solve the problem explicitly. An Observation, into which the discussion is converted, is able to give information to understand and solve the problem explicitly.
<Second Exemplary Embodiment>
Next, a second exemplary embodiment of the present invention is described with reference to drawings.
Fig. 1 is a block diagram illustrating an example structure of the information processing apparatus 100 according to the present exemplary embodiment. The information processing apparatus 100 of the present exemplary embodiment has the same structure as that of the information processing apparatus 100 of the first exemplary embodiment.
Elements of the information processing apparatus 100 of the present exemplary embodiment are the same as the elements assigned the same codes in the information processing apparatus 100 of the first exemplary embodiment. The information processing apparatus 100 of the present exemplary embodiment is the same as the information processing apparatus 100 of the first exemplary embodiment except the following differences. Therefore, detailed description of the information processing apparatus 100 of the present exemplary embodiment is omitted except the difference.
When the Answer to AQ is classified as a Suggestion, the analysis unit 104 of the present exemplary embodiment converts the AQ the answer of which is the Answer to AQ into a Problem, and converts the Answer to AQ into a Suggestion. The analysis unit 104 identifies the Observations related to the Answer to AQ. The analysis unit 104 identifies an Observation related to the Answer to AQ by similarity confirmation using NLP technology. The analysis unit 104 correlates the Problem converted from the AQ with the Observation identified as an Observation related with the Problem converted from the AQ. The analysis unit correlates the Problem converted from the AQ with the Suggestion converted from the Answer to AQ.
The generation unit 105 generates a new knowledge base on the basis of the Problem converted from the AQ, the Suggestion converted from the Answer to AQ and the Observation identified as an Observation related with the Problem converted from the AQ. The generation unit 105 may merge the new knowledge base into an existing knowledge base that is a knowledge base stored in the result storage unit 106.
Next, an operation of the information processing apparatus 100 of the present exemplary embodiment of the present invention is described with reference to drawings.
Fig. 4 is a flowchart illustrating an example of operation of the information processing apparatus 100 according to the present exemplary embodiment of the present invention. The operation, illustrated in Fig. 4, of the information processing apparatus 100 of the present exemplary embodiment is the same as the operation of the information processing apparatus 100 of the first exemplary embodiment 100 except the processing in Step S105 and Step S106. The detailed description of the operation illustrated in Fig. 4 of the information processing apparatus 100 is omitted except the following description of the differences.
In step S105, when the Answer to AQ is classified as a Suggestion, the analysis unit 104 converts the AQ as a Problem, converts the Answer to AQ as a Suggestion, and identifies an Observation related to the Answer to AQ. In Step S106, the generation unit 105 further generates a knowledge base on the basis of the Problem converted from the AQ, the Suggestion converted from the Answer to AQ and the Observation identified as an Observation related to the Answer to AQ.
Fig. 5 is a flowchart illustrating an example of operation in the preprocessing in the present exemplary embodiment. The operation in the preprocessing of the present exemplary embodiment is the same as that of the first exemplary embodiment. Therefore, the detailed description of the operation illustrated in Fig. 5 of the information processing apparatus 100 is omitted.
Fig. 6 is a flowchart illustrating example operation of discussion processing of the information processing apparatus 100 according to the present invention. The operation in the discussion processing of the present exemplary embodiment is the same as that of the first exemplary embodiment except the processing in Step S302. The differences in the processing in Step S302 are described later.
Fig. 7 is a flowchart illustrating example operation of RFD processing of the information processing apparatus 100 according to the present invention. The RFD processing of the present invention is the same as that of the first exemplary embodiment. Therefore, the detailed description of the RFD processing of the present invention is omitted.
Fig. 9 is a flowchart illustrating example operation of the AQ processing of the information processing apparatus 100 according to the present exemplary embodiment. The processing from Step S321 to Step S325 of the present exemplary embodiment is the same as that of the first exemplary embodiment. The following description represents the differences in the AQ processing between the present exemplary embodiment and the first exemplary embodiment.
The analysis unit 104 of the information processing apparatus 100 of the first exemplary embodiment ignores the AQ and the Answer to AQ when the Answer to AQ is classified as an Observation (NO in Step S324). When the Answer to AQ is classified as an Observation (NO in Step S324), the analysis unit 104 of the information processing apparatus 100 of the present exemplary embodiment performs creation processing (Step S326).
Fig. 10 is a flowchart illustrating example operation of the creation processing of the information processing apparatus 100 of the present exemplary embodiment.
The analysis unit 104 identifies an Observation related to the Answer to AQ, in the sentences generated by splitting the thread of the input data (Step S401). The analysis unit 104 converts the AQ as a Problem (Step S402). The analysis unit 104 converts the Answer to AQ as a Suggestion (Step S403). The analysis unit 104 correlates the Problem converted from the AQ with the Suggestion converted from the Answer to AQ and the Observation identified as an Observation related with the Answer to AQ that is converted to the Suggestion (Step S404).
The present invention has the same effect as that of the first exemplary embodiment. The reason is the same as that of the first exemplary embodiment.
Another effect of the present exemplary embodiment is that it is possible to capture a common question asked by a poster. Because, when the Answer to AQ is classified as an Observation, the analysis unit 104 converts the AQ to a Problem, converts the Answer to AQ to a Suggestion and identifies the Observation related to the Answer to AQ. The generation unit 105 generates a knowledge base based on the Problem converted from the AQ, the Suggestion converted from the Answer to AQ, and the Observation related with the Answer to AQ.
This may improve a performance of decision suggested by a system utilizing the knowledge base based on the Problem converted from the AQ, the Suggestion converted from the Answer to AQ, and the Observation related with the Answer to AQ. It is possible to provide the information that is more relevant to the user.
<Third Exemplary Embodiment>
Next, an third exemplary embodiment of the present invention is described in detail with reference to drawings.
Fig. 13 is a block diagram illustrating an information processing apparatus 100A according to the third exemplary embodiment of the present invention.
The information processing apparatus 100A includes an analysis unit 104 and a generation unit 105. The analysis unit 104 generates an Observation based on a question and an answer to the question. The Observation is an expression expressing a situation. The expression may be a sentence. The question and the answer are extracted from verbal interaction between a first person and a second person. The first person has a problem. The second person finds and provides a Suggestion to solve the problem. The generation unit 105 generates a knowledge base based on the problem, the suggestion and the Observation converted from the question.
Next, operation of the information processing apparatus 100A is described.
Fig. 14 is a flowchart illustrating example operation of the information processing apparatus 100A according to the present exemplary embodiment.
The analysis unit 104 generates an Observation based on a question and an answer to the question (Step S501). In Step S501, The analysis unit 104 of the present exemplary embodiment may convert an RFD into an Observation on the basis of the Answer to RFD that is an answer to the question of the RFD. The analysis unit 104 of the present exemplary embodiment may covert an AQ into an Observation when the Answer to AQ that is an answer to the AQ is an Observation. The generation unit 105 generates a knowledge base based on the Observation generated by the analysis unit 104 (Step S502).
The present exemplary embodiment has the same effect as that of the first exemplary embodiment. The reason for the effect is the same as that of the first exemplary embodiment.
<Other Exemplary Embodiment>
The information processing apparatus according to any one of the exemplary embodiments of the present invention is achieved by circuitry. The circuitry may be one or more computers including at least a processor and a memory storing a program controlling the processor. The computers are communicably connected with one another. The circuitry may be one or more circuits communicably connected with one another. The circuitry may be achieved by one or more computers and one or more circuits.
Fig. 15 is a block diagram illustrating a computer capable of achieving the information processing apparatus according to any one of the exemplary embodiments of the present invention.
The computer 1000 illustrated in Fig 15 includes a processor 1001, a memory 1002, a storage device 1003 and I/O (Input/Output) interface 1004 which are connected with one another via a bus. A storage medium 1005 is connected with the bus of the computer 1000 so that the processor 1001 is able to access the storage medium 1005.
The memory 1002 is, for example, a DRAM (Dynamic Random Access Memory) or the like. The storage device 1003 is, for example, an HDD (Hard Disk Drive), SSD (Solid State Drive) or the like. The storage medium 1005 is a removable storage medium, such as a CD-ROM (Compact Disc Read Only Memory), a USB (Universal Serial Bus) memory or the like. The I/O interface is an interface by which the processor is able to communicate with a device, such as, a keyboard or a mouse. The I/O interface may be communicably connected with a communication network. The processor may be communicate with an apparatus connected with the communication network.
The storage medium 1005 stores a program causing the computer 1000 operates as the information processing apparatus according to any one of the exemplary embodiments of the present invention. The processor 1001 reads out the program from the storage medium 1005, and loads the program in the memory 1002. The storage device 1003 may work as the storage medium 1005. By executing a program loaded in the memory 1002, the processor 1000 operates as the information processing apparatus according to any one of the exemplary embodiments of the present invention.
The raw data reception unit 101, the preprocessing unit 102, the classification unit 103, the analysis unit 104, the generation unit 105, and the output unit 107 are able to be achieved by the memory 1002 and the processor 1001 executing the program loaded in the memory 1002.
The result storage unit 106 and the dictionary storage unit 108 are able to be achieved by the memory 1002 or the storage device 1003.
All or a part of the raw data reception unit 101, the preprocessing unit 102, the classification unit 103, the analysis unit 104, the generation unit 105, the result storage unit 106, the output unit 107, and the dictionary storage unit 108 is able to be achieved by a dedicated circuits having functions of respective units.
Fig. 16 is a block diagram illustrating an example structure of the information processing apparatus 100, achieved by dedicated circuits, according to the first and the second exemplary embodiments of the present invention.
The information processing apparatus 100 illustrated in Fig. 16 includes a raw data reception circuit 1101, a preprocessing circuit 1102, a classification circuit 1103, an analysis circuit 1104, a generation circuit 1105, a result storage device 1106, an output circuit 1107 and a dictionary storage device 1108.
Fig. 17 is a block diagram illustrating an example structure of the information processing apparatus 100A, achieved by dedicated circuits, according to the third exemplary embodiment of the present invention.
The information processing device 100A illustrated in Fig. 17 includes the analysis circuit 1104 and the generation circuit 1105.
The raw data reception circuit 1101 operates as the raw data reception unit 101. The raw data reception unit 101 may be achieved by the raw data reception circuit 1101. The preprocessing circuit 1102 operates as the preprocessing unit 102. The preprocessing unit 102 may be achieved by the preprocessing circuit 1102. The classification circuit 1103 operates as the classification unit 103. The classification unit 103 may be achieved by the classification circuit. The analysis circuit 1104 operates as the analysis unit 104. The analysis unit 104 may be achieved by the analysis circuit 1104. The generation circuit 1105 operates as the generation unit 105. The generation unit 105 may be achieved by the generation circuit 1105. The result storage device 1106 operates as the result storage unit 106. The result storage unit 106 may be achieved by the result storage device 1106. The output circuit 1107 operates as the output unit 107. The output unit 107 may be achieved by the output circuit 1107. The dictionary storage device 1108 operates as the dictionary storage unit 108. The dictionary storage unit 108 may be achieved by the dictionary storage device 1108.
A part or all of the above-described exemplary embodiments may be described as the following supplementary notes, but are not limited by the following.
(Supplementary Note 1)
An information processing apparatus comprising:
analysis means for generating an observation based on a question and an answer to the question, the observation being an expression expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and
generation means for generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
(Supplementary Note 2)
The information processing apparatus according to Supplementary Note 1, wherein
the analysis means generates the observation by converting a request-for-details question into the observation based on a result of determining whether the answer to the request-for-details question is affirmative, the request-for-details question being the question requesting information by the second user.
(Supplementary Note 3)
The information processing apparatus according to Supplementary Note 1 or 2, wherein
the analysis means confirms a similarity between an additional question and the problem, and generates the observation by converting an additional question to which an observation is answered as an answer into the observation based on the similarity, the additional question being the question asked by the first user in addition to the problem in the verbal interaction.
(Supplementary Note 4)
The information processing apparatus according to Supplementary Note 3, wherein
the analysis means generates a new problem and a new suggestion by converting the additional question to which a suggestion is answered as an answer, and by converting the answer to the additional question into the new suggestion, and
the generation means generates the knowledge base further based on the new problem and the new suggestion.
(Supplementary Note 5)
The information processing apparatus according to Supplementary Note 4, wherein
the analysis means extracts an observation related to the new problem from the verbal interaction, and
the generation means generates the knowledge base further based on the observation related to the new problem.
(Supplementary Note 6)
An information processing method comprising:
generating an observation based on a question and an answer to the question, the observation being an expression expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and
generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
(Supplementary Note 7)
The information processing method according to Supplementary Note 6, wherein
the generating the observation includes converting a request-for-details question into the observation based on a result of determining whether the answer to the request-for-details question is affirmative, the request-for-details question being the question requesting information by the second user.
(Supplementary Note 8)
The information processing method according to Supplementary Note 6 or 7, further comprising:
confirming a similarity between an additional question and the problem, wherein
the generating the observation includes converting an additional question to which an observation is answered as an answer into the observation based on the similarity, the additional question being the question asked by the first user in addition to the problem in the verbal interaction.
(Supplementary Note 9)
The information processing method according to Supplementary Note 8, further comprising:
generating a new problem and a new suggestion by converting the additional question to which a suggestion is answered as an answer, and by converting the answer to the additional question into the new suggestion; and
generating the knowledge base further based on the new problem and the new suggestion.
(Supplementary Note 10)
The information processing method according to Supplementary Note 9, further comprising:
extracting an observation related to the new problem from the verbal interaction; and
generating the knowledge base further based on the observation related to the new problem.
(Supplementary Note 11)
A computer-readable medium storing a program causing a computer to execute:
analysis processing of generating an observation based on a question and an answer to the question, the observation being an expression expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and
generation processing of generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
(Supplementary Note 12)
The computer-readable medium according to Supplementary Note 11, wherein
the analysis processing generates the observation by converting a request-for-details question into the observation based on a result of determining whether the answer to the request-for-details question is affirmative, the request-for-details question being the question requesting information by the second user.
(Supplementary Note 13)
The computer-readable medium according to Supplementary Note 11 or 12, wherein
the analysis processing confirms a similarity between an additional question and the problem, and generates the observation by converting an additional question to which an observation is answered as an answer into the observation based on the similarity, the additional question being the question asked by the first user in addition to the problem in the verbal interaction.
(Supplementary Note 14)
The computer-readable medium according to Supplementary Note 13, wherein
the analysis processing generates a new problem and a new suggestion by converting the additional question to which a suggestion is answered as an answer, and by converting the answer to the additional question into the new suggestion, and
the generation processing generates the knowledge base further based on the new problem and the new suggestion.
(Supplementary Note 15)
The computer-readable medium according to Supplementary Note 14, wherein
the analysis processing extracts an observation related to the new problem from the verbal interaction, and
the generation processing generates the knowledge base further based on the observation related to the new problem.
While the invention has been particularly shown and described with reference to exemplary embodiments thereof, the invention is not limited to these embodiments. It will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention as defined by the claims.
Reference sings List
1 Information processing system
1A Information processing system
100 Information processing apparatus
100A Information processing apparatus
101 Raw data reception unit
102 Preprocessing unit
103 Classification unit
104 Analysis unit
105 Generation unit
106 Result storage unit
107 Output unit
108 Dictionary storage unit
200 Knowledge base storage unit
300 Suggestion apparatus
400 Training apparatus
500 Acquisition apparatus
600 Network
700 ODF system
800 Terminal apparatus
900 Recording apparatus
1000 Computer
1001 Processor
1002 Memory
1003 Storage device
1004 I/O interface
1005 Storage medium
1101 Raw data reception circuit
1102 Preprocessing circuit
1103 Classification circuit
1104 Analysis circuit
1105 Generation circuit
1106 Result storage device
1107 Output circuit
1108 Dictionary storage device

Claims (11)

  1. An information processing apparatus comprising:
    analysis means for generating an observation based on a question and an answer to the question, the observation being an expression expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and
    generation means for generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
  2. The information processing apparatus according to claim 1, wherein
    the analysis means generates the observation by converting a request-for-details question into the observation based on a result of determining whether the answer to the request-for-details question is affirmative, the request-for-details question being the question requesting information by the second user.
  3. The information processing apparatus according to claim 1 or 2, wherein
    the analysis means confirms a similarity between an additional question and the problem, and generates the observation by converting an additional question to which an observation is answered as an answer into the observation based on the similarity, the additional question being the question asked by the first user in addition to the problem in the verbal interaction.
  4. The information processing apparatus according to claim 3, wherein
    the analysis means generates a new problem and a new suggestion by converting the additional question to which a suggestion is answered as an answer, and by converting the answer to the additional question into the new suggestion, and
    the generation means generates the knowledge base further based on the new problem and the new suggestion.
  5. The information processing apparatus according to claim 4, wherein
    the analysis means extracts an observation related to the new problem from the verbal interaction, and
    the generation means generates the knowledge base further based on the observation related to the new problem.
  6. An information processing method comprising:
    generating an observation based on a question and an answer to the question, the observation being an expression expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and
    generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
  7. The information processing method according to claim 6, wherein
    the generating the observation includes converting a request-for-details question into the observation based on a result of determining whether the answer to the request-for-details question is affirmative, the request-for-details question being the question requesting information by the second user.
  8. The information processing method according to claim 6 or 7, further comprising:
    confirming a similarity between an additional question and the problem, wherein
    the generating the observation includes converting an additional question to which an observation is answered as an answer into the observation based on the similarity, the additional question being the question asked by the first user in addition to the problem in the verbal interaction.
  9. A computer-readable medium storing a program causing a computer to execute:
    analysis processing of generating an observation based on a question and an answer to the question, the observation being an expression expressing a situation, the question and the answer being extracted from verbal interaction between a first person and a second person, the first person having a problem, the second person finding and providing a suggestion to solve the problem; and
    generation processing of generating a knowledge base based on the problem, the suggestion and the observation converted from the question.
  10. The computer-readable medium according to claim 9, wherein
    the analysis processing generates the observation by converting a request-for-details question into the observation based on a result of determining whether the answer to the request-for-details question is affirmative, the request-for-details question being the question requesting information by the second user.
  11. The computer-readable medium according to claim 9 or 10, wherein
    the analysis processing confirms a similarity between an additional question and the problem, and generates the observation by converting an additional question to which an observation is answered as an answer into the observation based on the similarity, the additional question being the question asked by the first user in addition to the problem in the verbal interaction.
PCT/JP2016/000090 2016-01-08 2016-01-08 Information processing apparatus, information processing method and computer-readable medium WO2017119014A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/JP2016/000090 WO2017119014A1 (en) 2016-01-08 2016-01-08 Information processing apparatus, information processing method and computer-readable medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2016/000090 WO2017119014A1 (en) 2016-01-08 2016-01-08 Information processing apparatus, information processing method and computer-readable medium

Publications (1)

Publication Number Publication Date
WO2017119014A1 true WO2017119014A1 (en) 2017-07-13

Family

ID=55272539

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2016/000090 WO2017119014A1 (en) 2016-01-08 2016-01-08 Information processing apparatus, information processing method and computer-readable medium

Country Status (1)

Country Link
WO (1) WO2017119014A1 (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7814048B2 (en) 2006-08-14 2010-10-12 Microsoft Corporation Knowledge extraction from online discussion forums
US8489601B2 (en) 2010-07-08 2013-07-16 GM Global Technology Operations LLC Knowledge extraction methodology for unstructured data using ontology-based text mining
US20150206443A1 (en) * 2013-05-03 2015-07-23 Samsung Electronics Co., Ltd. Computing system with learning platform mechanism and method of operation thereof

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7814048B2 (en) 2006-08-14 2010-10-12 Microsoft Corporation Knowledge extraction from online discussion forums
US8489601B2 (en) 2010-07-08 2013-07-16 GM Global Technology Operations LLC Knowledge extraction methodology for unstructured data using ontology-based text mining
US20150206443A1 (en) * 2013-05-03 2015-07-23 Samsung Electronics Co., Ltd. Computing system with learning platform mechanism and method of operation thereof

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
KIM ET AL: "Incorporating tutoring principles into interactive knowledge acquisition", INTERNATIONAL JOURNAL OF HUMAN-COMPUTER STUDIES, ACADEMIC PRESS, NEW YORK, NY, US, vol. 65, no. 10, 16 August 2007 (2007-08-16), pages 852 - 872, XP022202606, ISSN: 1071-5819, DOI: 10.1016/J.IJHCS.2007.06.003 *
NIEDERMAIR G T ET AL: "Linguistic processing related to speech understanding in spicos II", SPEECH COMMUNICATION, ELSEVIER SCIENCE PUBLISHERS, AMSTERDAM, NL, vol. 9, no. 5-6, 1 December 1990 (1990-12-01), pages 565 - 585, XP024228788, ISSN: 0167-6393, [retrieved on 19901201], DOI: 10.1016/0167-6393(90)90031-4 *

Similar Documents

Publication Publication Date Title
JP6544131B2 (en) INFORMATION PROCESSING APPARATUS AND INFORMATION PROCESSING PROGRAM
JP6857581B2 (en) Growth interactive device
Mirheidari et al. Detecting Signs of Dementia Using Word Vector Representations.
US10923115B2 (en) Dynamically generated dialog
JP6791825B2 (en) Information processing device, dialogue processing method and dialogue system
JP2017016566A (en) Information processing device, information processing method and program
CN109192194A (en) Voice data mask method, device, computer equipment and storage medium
KR102285142B1 (en) Apparatus and method for recommending learning data for chatbots
WO2019156103A1 (en) Faq consolidation assistance device, faq consolidation assistance method, and program
Dudzik et al. Context in human emotion perception for automatic affect detection: A survey of audiovisual databases
US10825558B2 (en) Method for improving healthcare
JP2020077159A (en) Interactive system, interactive device, interactive method, and program
CN111145903A (en) Method and device for acquiring vertigo inquiry text, electronic equipment and inquiry system
Yadav et al. A novel automated depression detection technique using text transcript
Prabhakaran et al. Detecting institutional dialog acts in police traffic stops
Elmitwally et al. Personality Detection Using Context Based Emotions in Cognitive Agents.
CN113095204B (en) Double-recording data quality inspection method, device and system
WO2017119014A1 (en) Information processing apparatus, information processing method and computer-readable medium
Pevy et al. Predicting the cause of seizures using features extracted from interactions with a virtual agent
JP5109571B2 (en) Information processing system and information processing program
Saeki et al. Confusion Detection for Adaptive Conversational Strategies of An Oral Proficiency Assessment Interview Agent.
Pucci et al. Speech emotion recognition with artificial intelligence for contact tracing in the COVID‐19 pandemic
US20200327889A1 (en) Nurse operation assistance terminal, nurse operation assistance system, nurse operation assistance method, and nurse operation assistance program recording medium
JPWO2019103007A1 (en) Information processing device and information processing method
US11715554B1 (en) System and method for determining a mismatch between a user sentiment and a polarity of a situation using an AI chatbot

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: 16702216

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: 16702216

Country of ref document: EP

Kind code of ref document: A1