JP2020024522A - Information providing apparatus, information providing method and program - Google Patents

Abstract

To improve continuity of dialogue in a system that provides information in a dialogue form.SOLUTION: An information providing apparatus provides information in a conversation form between a virtual personality and a user by allowing the user to select his or her own words. The information providing apparatus includes: a conversation control unit that reads a scenario from a storage device storing the scenario in the conversation form including words of the virtual personality and the words of the user, outputs option of the words of the virtual personality and the words of the user, and accepts the selection of the words from the user; and an insertion control unit inserting a filler which is an expression filling a space between opinions or expression expressing gratitude between the selected words of the user and the words of the virtual personality, on the basis of histories of the virtual personality expressed according to the scenario and the words of the user. The insertion control unit adjusts probability of inserting the filler or the expression expressing gratitude to increase as the number of opinions from when the filler or the expression expressing gratitude was previously inserted increases.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an information providing device, an information providing method, and a program.

  2. Description of the Related Art Conventionally, various chatbots and software agents for providing information, chatting, and the like with users have been proposed.

  Also, in order to realize natural utterance in speech synthesis, a technology that inserts a redundant expression language or the like representing redundant expressions in a conversation into text of spoken words included in text data to be synthesized and outputs the speech. Has been proposed (for example, Patent Document 1). Specifically, it describes that a redundant expression language is inserted based on an insertion rule using probabilities. In addition, as a redundant expression language, fillers such as "er", redundant endings such as "-isu", interjections such as "mah", stagnation of "(silence)", and synonym dictionary data and synonym dictionary data are used. There is a description that the same word or synonym is registered in the dictionary.

JP 2014-48443 A

  Conventionally, chatbots and software agents that interact with text cannot express nuances due to the way of speaking, unlike spoken language, and there has been no example of actively incorporating filler-like expressions.

  On the other hand, the inventor of the present application has found that redundant expressions such as fillers are not always useless in order for users to use chatbots and software agents. An object of the present invention is to improve the continuity of a dialog in a system for providing information to a user in the form of a dialog.

  The information providing apparatus provides information in a form of conversation between the virtual personality and the user by allowing the user to select his or her own dialogue. Further, the information providing apparatus reads a scenario from a storage device that stores a conversational scenario including a virtual personality dialogue and a user's dialogue, and outputs a virtual personality dialogue and a user's dialogue option, A conversation control unit that accepts selection of a dialogue from the user, and a speech between the selected user's dialogue and the virtual personality's dialogue based on the virtual personality spoken according to the scenario and the history of the user's dialogue. And an insertion control unit that inserts a filler or a word of gratitude, which is a word that fills the space between the words. Adjust so as to increase as the number of utterances from when the words to be expressed are inserted before is increased.

  By doing so, fillers and words of thanks can be inserted at preferred positions in a predetermined scenario, and continuity of the dialogue can be improved. In addition, the same scenario can be changed by stochastically determining the insertion of fillers and words of appreciation.

  Further, the insertion control unit analyzes the type of emotion related to the sentence of the dialog using a predetermined feature vector determined in association with the word included in the history of the dialog, and associates the type of the emotion with the type of the emotion in advance. A filler or a word expressing gratitude stored in the storage device may be inserted. In this way, it is possible to select a filler that does not cause any discomfort in the context or a word that expresses gratitude.

  Also, based on the history of dialogue, statistically analyze the relationship between the presence or absence of filler or words expressing appreciation and the number of series of statements in the history, so that the number of series of speech in conversation increases, A learning processing unit that adjusts the probability of inserting a filler or a word expressing appreciation may be further provided. By preferentially selecting words and timings that increase the number of dialogues, it is possible to insert words in a mode with less discomfort and further increase the number of dialogues.

  The contents described in the means for solving the problems can be combined as much as possible without departing from the problems and technical ideas of the present invention. Further, the contents of the means for solving the problems can be provided as an apparatus such as a computer or a system including a plurality of apparatuses, a method executed by a computer, or a program executed by a computer. In addition, you may make it provide the recording medium which holds a program.

  In a system that provides information to a user in an interactive manner, the continuity of the dialog can be improved.

FIG. 1 is a diagram illustrating an example of a configuration of an entire system. It is a block diagram showing an example of composition of a user device and an interactive information providing device. FIG. 5 is a diagram illustrating an example of an image displayed on the user device. FIG. 4 is a diagram visually illustrating an example of a scenario stored in the interactive information providing apparatus. It is a figure showing an example of an output at the time of inserting a filler between dialogues of a scenario. It is a processing flow figure showing an example of conversation offer processing. It is a processing flow figure showing an example of conversation offer processing. FIG. 4 is a diagram illustrating an example of filler or words of appreciation stored in a storage device. It is a processing flow figure showing an example of conversation offer processing concerning a modification. FIG. 14 is a diagram illustrating an example of an image displayed in a modified example. It is a processing flow figure showing an example of conversation offer processing concerning a modification. FIG. 14 is a diagram illustrating an example of an image displayed in a modified example.

  Hereinafter, embodiments for implementing the present invention will be described with reference to the drawings.

<System configuration>
FIG. 1 is a diagram illustrating an example of the entire system according to the embodiment. This system provides a software agent which is a virtual personality whose speech is output by a computer and has a conversation with a user. Further, the content of the conversation proceeds based on information representing a scenario created in advance.

The system includes a user device 1 and an interactive information providing device 2, which are connected via a network 3. The network 3 is a communication network such as the Internet, and computers connected to the network 3 can communicate with each other based on various protocols. The user device 1 includes a smartphone, a tablet, and a PC (Personal Computer).
), And functions as a user interface of the conversation providing system according to the present embodiment by executing a program (also referred to as software or an application) according to the present embodiment. Note that a plurality of user devices 1 may be connected to the network 3. The interactive information providing device 2 is a general computer, and outputs a statement of a virtual personality to the user device 1 via the network 3 by executing the program according to the present embodiment. The process of acquiring the user's response from is repeated. The user's remark is input to the user device 1 when the user selects one of the options output by the interactive information providing device 2, and transmitted to the interactive information providing device 2 via the network 3.

<Apparatus configuration>
FIG. 2 is a block diagram illustrating an example of a configuration of the user device 1 and the interactive information providing device 2.

The user device 1 is a general computer such as a smartphone or a tablet, and includes an input / output interface (I / F) 11, a storage device 12, a communication interface (I / F) 13, a processor 14, a bus 15, It has. The input / output I / F 11 is a user interface such as a touch panel. The input / output I / F 11 displays a statement of a virtual personality to the user, displays options of the user's response, and receives a selection of the user's response based on the operation of the user. The storage device 12 includes a main storage device such as a random access memory (RAM) and a read only memory (ROM) and an auxiliary storage device (secondary storage) such as a hard disk drive (HDD), a solid state drive (SSD), and a flash memory. Device). The main storage device temporarily stores a program read by the processor and data of speech to be received, and secures a work area for the processor. The auxiliary storage device stores a program executed by the processor. The communication I / F 13 is a network module that communicates with, for example, a base station of a mobile phone or an access point of a wireless LAN (Local Area Network), and performs communication based on a predetermined protocol. The processor 14 is an arithmetic processing device such as a CPU (Central Processing Unit), and performs each process according to the present embodiment by executing a program. In the example of FIG. 2, functional blocks are shown in the processor 14. Specifically, the processor 14 functions as the input / output control unit 141. The input / output control unit 141 displays a virtual personality statement on the input / output I / F 11 and receives an input of a user response via the input / output I / F 11. The above components are connected via a bus 15.

The interactive information providing apparatus 2 is a general computer such as a server, and includes an input / output I / F 21, a storage device 22, a communication I / F 23, a processor 24, and a bus 25. The input / output I / F 21 is a user interface such as a keyboard, a mouse, and a display. The storage device 22 is a main storage device such as a RAM and a ROM, and an auxiliary storage device such as an HDD, an SSD, and a flash memory. The main storage device temporarily stores programs read by the processor and data of scenarios stored in the auxiliary storage device, and secures a work area for the processor. The auxiliary storage device stores a program executed by the processor. The communication I / F 23 is, for example, a wired network card, and performs communication based on a predetermined protocol. The processor 24 is an arithmetic processing device such as a CPU, and performs each process according to the present embodiment by executing a program. As illustrated in FIG. 2, the processor 24 functions as a conversation control unit 241, an insertion control unit 242, and a learning processing unit 243. The conversation control unit 241 outputs a virtual personality remark to the user and a choice of the user's response to the remark using a scenario prepared in advance. The insertion control unit 242 determines according to a predetermined rule whether to insert a filler or a word expressing gratitude. The learning processing unit determines whether the number of dialogues with the user has increased by inserting fillers and words expressing appreciation based on the statistical processing, and the insertion control unit 242 converts the words expressing fillers and appreciation into words. Adjust the probability of insertion. The above components are connected via a bus 15.

<User interface>
FIG. 3 is a diagram illustrating an example of an image displayed on the input / output I / F 11 of the user device 1. In the example of FIG. 3, an image 111 representing an agent, an image 112 representing a user, a balloon 113 representing a statement from an agent, and an option 114 of a user response are displayed on an input / output I / F 11 serving as a touch panel. I have. The user can interact with the agent, for example, by tapping and selecting his / her remark from among such options. The number of options is one or more, and is not limited to four.

<Scenario>
FIG. 4 is a diagram visually illustrating an example of a scenario stored in the storage device 22 of the interactive information providing apparatus 2 in advance. In the present embodiment, processing by software such as an agent dialogue or a telephone call represented by a rectangle and options of a user's dialogue represented by a rounded rectangle are arranged in order by arrows. The dialogs of the agents are basically connected in series, and are output to the balloon 113 shown in FIG. 3 in the order of connection by arrows. The user's dialogue options are displayed as options 114 shown in FIG. At this time, the options arranged in parallel are output in parallel, the scenario branches according to the dialogue selected by the user, and the subsequent dialogue is displayed. In addition, the function of the user device may be used to perform the telephone call processing shown in FIG. 4, display of a website (not shown), and the like.

<Inserting filler and words of thanks>
FIG. 5 is a diagram illustrating an example of an output when a filler is inserted between dialogue lines of a scenario. The insertion control unit 242 of the interactive information providing apparatus 2 according to the present embodiment includes “er”, “er”, “un-un”, “hoho”, Insert filler words to fill in the conversation, or words of appreciation such as "thank you". Specifically, the agent may insert these words based on a predetermined probability after the partner's dialogue.

  Further, the inserted word may be stored in advance in association with a type of emotion such as emotion, anger and sorrow. Then, with respect to the words included in the dialogue of the partner, the emotions in the flow of the dialog are analyzed using the feature vector having the emotion type as an element, and words to be inserted are selected according to the emotion type. In addition, the above-described probability may be increased as the number of conversations (the number of dialogues spoken by the user and the agent) since the previous insertion of the filler or the like increases.

  According to the scenario shown in FIG. 4, after the user's dialogue option “what can be done?”, The agent's dialogue “order your product ...” is connected. Are filled with the filler "Um ..." based on the above conditions.

The conversation control unit 241 of the interactive information providing apparatus 2 accumulates a log of the conversation between the user and the agent in the storage device 22, and the learning processing unit 243 inserts fillers or words of thanks to reduce the number of conversations. Machine learning whether or not it has grown. Then, the learning processing unit 243 determines whether or not the words are determined to have increased the number of conversations by inserting them, or at the time when the number of dialogues is determined to be increased by insertion (during a dialogue in a scenario). Feedback is provided to increase the probability of insertion. Whether or not the number of conversations has increased is determined based on whether or not the user has advanced the prepared scenario beyond the number of the inserted dialogues.

<Conversation providing processing>
FIG. 6 is a processing flowchart showing an example of the conversation providing processing executed in the present system. First, the conversation control unit 241 of the interactive information providing apparatus 2 reads a scenario as shown in FIG. 4 (FIG. 6: S1). In this step, based on the order defined in the scenario, an option of an agent's dialog or a user's dialog or an instruction to perform other processing is read. In the example of FIG. 4, the dialogue “Shopping?” Of the agent connected by the arrow is read from “Start”.

  Then, the conversation control unit 241 determines whether the read dialogue or the like is an agent's dialogue (FIG. 6: S2). If it is an agent dialog (S2: YES), the conversation control unit 241 outputs the read agent dialog via the input / output I / F 11 (FIG. 6: S3). For example, as shown in FIG. 3, an agent's dialogue “Is shopping?” Is output. Then, the processing makes a transition to the processing in FIG. 7 via the connector “A”.

On the other hand, when the read dialogue or the like is not the dialog of the agent (S2: NO), the conversation control unit 241 determines whether the read dialogue or the like is an option of the user's dialogue (FIG. 6: S4). If it is not the user's dialogue option (S4: NO), the conversation control unit 241 performs processing according to the scenario (FIG. 6: S5). In this step, for example, processing such as calling a predetermined telephone number and displaying a website of a predetermined URI (Uniform Resource Identifier) is performed. Then, the processing makes a transition to the processing in FIG. 7 via the connector “A”.

  On the other hand, when the read dialogue or the like is a user's dialogue option (S2: YES), the conversation control unit 241 displays the user's dialogue option (FIG. 6: S6). For example, in the case of the scenario shown in FIG. 4, in the next stage of the agent's dialogue “Is shopping?”, The user's dialogue options “I want to see products”, “What can I do?”, “Shop” And the like are read out, and are output as option 118 in FIG. 3 in this step.

  Thereafter, the conversation control unit 241 accepts the user's selection via the input / output I / F 21 (FIG. 6: S7). In this step, one of the dialogue options is selected, for example, according to the position where the user taps the touch panel. Then, the processing makes a transition to the processing in FIG. 7 via the connector “B”.

  FIG. 7 is a process flowchart illustrating an example of the conversation providing process executed in the present system. After S7, the insertion control unit 242 of the interactive information providing apparatus 2 determines whether to insert a filler or a word of thanks (FIG. 7: S8). In this step, it is determined that a filler or a word of thanks is to be inserted based on a predetermined probability. The history of the dialogue of the agent or the user who has spoken according to the scenario may be stored in the storage device 22, and the probability may increase as the number of conversations increases from the previous insertion.

  Further, a word to be inserted may be selected based on the history. For example, it is broken down into words included in the uttered sentence by morphological analysis, and analyzed using a feature vector having an emotion type as an element. The insertion control unit 242 may perform sentiment analysis based on words included in the dialog history of the agent or the user using, for example, Word2Vec or Doc2Vec. Then, the percentages of the emotion types “happy”, “angry”, “sad”, and “easy” are obtained. In addition, the probability that each of the words to be inserted is selected is further adjusted based on the proportion of each emotion.

FIG. 8 is a diagram illustrating an example of fillers or words of appreciation stored in the storage device 22. For example, a word that is stored in the storage device 22 in association with the type of emotion in advance is inserted. The table of FIG. 8 has the type of emotion and the attributes of filler or words of appreciation. In the field of emotion type, “happy”, “angry”, “sad”, and “easy” are registered. In the field of the filler or the words of appreciation, words such as "um", "un-un", "hoho", and "thank you" are registered. Then, as a result of the emotion analysis described above, the probability that the filler or the words of appreciation corresponding to the emotion according to the conversation history are increased. A plurality of tables as shown in FIG. 8 may be prepared according to the characteristics of the agent.

  Further, based on the history of the dialogue of the agent or the user, the learning processing unit 243 sets the filler at the time when it is determined that the number of conversations has been increased by the insertion or when the number of conversations has been determined to be increased by the insertion. Alternatively, further adjustment may be made so as to increase the probability of inserting a word of thanks. The learning process is based on, for example, conditional probabilities obtained from a history when a word is inserted and when the word is not inserted, and the probability that the word is selected is increased as the difference between the number of expected conversations is larger. It may be. Also, based on the conditional probabilities of inserting fillers or words of appreciation at a certain timing and not, based on the conditional probability, the larger the difference between the expected number of dialogues, the higher the probability that words are inserted at that timing. It may be. In addition, as for the combination of the selection of the word and the selection of the timing, the probability that the combination is selected may be increased as the number of conversations expected from the history increases.

  When it is determined that a filler or a word of thanks is to be inserted (S8: YES), the insertion control unit 242 outputs the selected dialog (FIG. 7: S9). In this step, for example, as shown in FIG. 5, fillers or words of thanks that are not included in the scenario are output.

  On the other hand, when it is determined that the filler or the words of thanks are not to be inserted (S8: NO), or after S3 in FIG. 6, the conversation control unit 241 determines whether the scenario read in S1 is completed (FIG. 7: S10). In this step, for example, in the scenario as shown in FIG. 5, when the end of the dialogue or the like connected by the arrow is reached, it is determined that the scenario has ended.

  When the scenario ends (S10: YES), the conversation control unit 241 ends the conversation providing process. On the other hand, when the scenario is not completed (S10: NO), or after S9, the process returns to S1 of FIG. 6 via the connector “C”.

<Effect>
Especially in chat-type software agents, fillers are often considered redundant and useless. However, the present inventors have found that, through a demonstration experiment on a chat-type agent, the number of conversations with the user increases even in the same scenario by inserting fillers or words expressing gratitude. An increase in the number of conversations means an increase in the use time (active rate) of the software, which is considered to increase the user's satisfaction and, in turn, happiness.

  According to the above-described embodiment, fillers and words of appreciation can be inserted at preferred positions in a predetermined scenario, and continuity of dialogue can be improved. In addition, the same scenario can be changed by stochastically determining the insertion of fillers and words of appreciation. Also, by performing learning processing on the conversation history and selecting words and timings that increase the number of conversations, it is possible to insert words with less discomfort, and further increase the number of conversations.

<Modification>
If the number of agents is more than one, fillers or words of appreciation can be similarly inserted. In the present modified example, the role is divided into an agent that speaks a dialogue according to a scenario and an agent that speaks a filler or words of appreciation.

  FIG. 9 is a process flowchart illustrating an example of a conversation providing process according to a modification. This modification is basically the same as the processing in FIG. 6, and corresponding steps are denoted by the same reference numerals. Hereinafter, the description will focus on the differences from FIG.

  In the modification of FIG. 9, it is determined in S2 whether it is the dialog of one of the agents (agent 1), and in S3, the dialog of agent 1 is output.

  FIG. 10 is a diagram illustrating an example of an image displayed in the modification. In the image of FIG. 10, a second agent 115 (agent 2) is displayed in addition to the first agent 111 (agent 1) also shown in FIG.

  After S3, the process transits to S8 in FIG. 11 via the connector "B". FIG. 11 is a process flowchart illustrating an example of a conversation providing process according to a modification. That is, in S8, it is determined whether or not the other agent (agent 2) speaks a filler or a word of appreciation not only after the speech of the user but also after the speech of the agent 1. When it is determined that the filler or the words of thanks are to be inserted (S8: YES), in S9, the agent 2 outputs the selected dialogue.

  FIG. 12 is a diagram illustrating an example of an image displayed in the modification. In the example of FIG. 12, a balloon 116 representing the statement of the agent 2 is displayed. Also, in FIG. 12, after the speech of the agent 1, the dialogue "Hohou" of the agent 2 is inserted. In this way, the agent 2 speaks a filler or a word of thanks after the dialog of the agent 1 or after the dialog of the user. Note that the probability of inserting a word can be determined in the same manner as in the above-described embodiment.

  By doing so, the number of conversations with the user can be increased even in a conversation between the three parties.

  Further, in the above-described modified example, the role is divided into an agent that speaks a dialogue according to a scenario and an agent that speaks a filler or a word of thanks, but the present invention is not limited to such an example. A scenario in which dialogue is assigned to two agents in advance may be used. In this case, for example, one agent issues a filler or thanks to the other agent and the user. Also, the other agent utters a filler or thanks to the one agent and the user. At this time, the probability of inserting a word can be determined in the same manner as in the above embodiment.

<Others>
The configuration of the above-described embodiment is an exemplification, and can be combined or changed as much as possible without departing from the problem and the technical idea of the present invention.

  The present invention includes a computer program that executes the above-described processing. Furthermore, a computer-readable recording medium on which the program is recorded also belongs to the category of the present invention. For the recording medium on which the program is recorded, the above-described processing can be performed by causing a computer to read and execute the program on the recording medium.

Here, the computer-readable recording medium refers to a recording medium in which information such as data and programs is accumulated by electrical, magnetic, optical, mechanical, or chemical action and can be read by a computer. Among such recording media, those removable from a computer include a flexible disk, a magneto-optical disk, an optical disk, a magnetic tape, a memory card, and the like. Further, examples of the recording medium fixed to the computer include a hard disk drive and a ROM.

1: User device 11: Input / output I / F
12: Storage device 13: Communication I / F
14: Processor 141: Input / output control unit 2: Interactive information providing device 21: Input / output I / F
22: Storage device 23: Communication I / F
24: Processor 241: Conversation control unit 242: Insertion control unit 243: Learning processing unit 3: Network

Claims (5)

An information providing device that provides a virtual personality and information in a conversation format with the user by allowing the user to select his or her dialogue,
A scenario is read from a storage device that stores a conversational scenario including the virtual personality dialogue and the user's dialogue, and the virtual personality dialogue and the user's dialogue options are output. A conversation control unit for receiving a selection of dialogue,
Based on the virtual personality and the history of the user's dialogue that are said in accordance with the scenario, a word that fills the space between the selected user's dialogue and the virtual personality's dialogue. An insertion control unit for inserting a filler or words expressing gratitude,
With
The insertion control unit adjusts the probability of inserting the filler or the word expressing gratitude to increase as the number of remarks from when the filler or the word expressing gratitude was previously inserted increases. Information providing device. The insertion control unit analyzes a type of an emotion related to a sentence of the dialog using a predetermined feature vector determined in association with a word included in the history of the dialog, and corresponds to the type of the emotion in advance. The information providing apparatus according to claim 1, wherein the filler or a word expressing appreciation is attached and stored in the storage device. Based on the history of the dialogue, statistically analyze the relationship between the presence or absence of the filler or words expressing appreciation and the number of series of statements in the history, so that the number of series of speech in conversation increases The information providing apparatus according to claim 1, further comprising a learning processing unit configured to adjust a probability of inserting the filler or a word expressing appreciation. An information providing method in which a computer provides information in a form of conversation between a virtual personality and the user by allowing the user to select his or her dialogue,
A scenario is read from a storage device that stores a conversational scenario including the virtual personality dialogue and the user's dialogue, and the virtual personality dialogue and the user's dialogue options are output. Accepts dialogue selection,
Based on the virtual personality and the history of the user's dialogue that are said in accordance with the scenario, a word that fills the space between the selected user's dialogue and the virtual personality's dialogue. Insert filler or words of gratitude,
The insertion of the filler or words of appreciation is the probability of inserting the filler or words of appreciation as the number of remarks from when the filler or appreciation words were previously inserted increases. Information providing method to adjust to increase. By allowing the user to select his or her dialogue, a computer that provides information in the form of a conversation with the virtual personality and the user,
A scenario is read from a storage device that stores a conversational scenario including the virtual personality dialogue and the user's dialogue, and the virtual personality dialogue and the user's dialogue options are output. Accepts dialogue selection,
Based on the virtual personality and the history of the user's dialogue that are said in accordance with the scenario, a word that fills the space between the selected user's dialogue and the virtual personality's dialogue. Insert filler or words of gratitude,
The insertion of the filler or words of gratitude is the probability that the filler or words of gratitude will be inserted as the number of remarks from the previous insertion of the words of filler or gratitude increases. The program that executes the process of adjusting so that it rises.

Images