Dialogue system and program

Abstract

To provide an interactive system capable of presenting an appropriate answer to a user's question.SOLUTION: The avatar terminal (12) autonomously replies to the user's question by using the large-scale language model (124, 161) (S7). However, when the large-scale language model (124, 161) cannot answer the question, the transmission means (22, S9) transmits the question to the messenger application server (18) in order to ask for external help. Then, when an answer to the question is obtained from the messenger application server (18), the answering means (22, S13) outputs the answer to the user. Since the answer is obtained from the messenger application server (18), even when the avatar cannot autonomously answer, it is possible to output an appropriate answer to the question of the user.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a dialogue system and a program, and more particularly to a dialogue system and a program in which, for example, an avatar (autonomous agent) receives and answers questions from a user. [Background technology]

[0002] An example of this type of dialogue system is disclosed in Patent Document 1. In the technology of Patent Document 1, a first avatar 301a corresponding to a first character and a second avatar 301b corresponding to a second character are displayed on the same screen, and if a script of an answer to a question entered by a user in a conversation with a chatbot is included in a pre-prepared script, avatar 301a is made to speak the answer script, and if the script of an answer to the question is not included in the pre-prepared script, avatar 301b is made to speak the results of a web search. Therefore, even if a script of an answer is not prepared in advance, the system can respond to a user's question. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7334800 Summary of the Invention [Problem to be solved by the invention]

[0004] Although Patent Document 1 can respond to user questions, it does not specify how avatar 301b specifically performs a web search. However, in the invention described in Patent Document 1, the results of a web search performed autonomously by avatar 301b are presented to the user, and the content presented to user 201 (web search results by avatar 301b) may not be appropriate.

[0005] Therefore, a primary object of the present invention is to provide a novel dialogue system and program.

[0006] Another object of the present invention is to provide a dialogue system and program that can present appropriate answers to user questions. [Means for solving the problem]

[0007] In order to solve the above problems, the present invention employs the following configuration: Note that the reference symbols and supplementary explanations in parentheses indicate the correspondence with the embodiments described to facilitate understanding of the present disclosure, and do not limit the present disclosure in any way.

[0008] A first embodiment is an interactive system that includes an avatar terminal in which an avatar interacts with a user, a language model accessible by the avatar terminal, and a messenger application server accessible by the avatar terminal. The avatar terminal includes a determination means for determining whether an answer to a user's question can be provided using the language model, a response means for outputting an answer based on the language model when the determination means determines that an answer is provided, and a transmission means for transmitting the question to the messenger application server when the determination means determines that an answer is not provided. When the response means receives an answer to the question from the messenger application server, it outputs the answer to the user.

[0009] In the first embodiment, a dialogue system (10: reference numerals illustrating corresponding parts in the embodiments, not intended to be limiting; the same applies below) includes an avatar terminal (12) in which an avatar interacts with a user, a language model (124, 161), and a messenger application server (18). A determination means (22, 40e, S5) determines whether a user's question can be answered using the language model. When the determination means determines that a question can be answered, a response means (22, S7) in the avatar terminal (12) autonomously answers the user's question using the language model (124, 161). However, when the determination means (22, 40e, S5) determines that a user's question cannot be answered using the language model, i.e., when a user's question cannot be answered using the language model (124, 161), a transmission means (22, 40g, S9) transmits data of the question to the messenger application server (18). The messenger application server (18) can be accessed by an operator terminal (20) operated by one or more operators, and the operators input answers to questions from the operator terminal to the messenger application server (18). That is, the answers are obtained from the messenger application server (18). Then, when an answer to the question is obtained from the messenger application server (18), an answering means (22, S13) outputs the answer to the user.

[0010] According to the first embodiment, when the avatar terminal cannot provide an answer using the language model, it asks a question to an operator via the messenger application server to obtain an answer, so that an appropriate answer can be presented to the user's question.

[0011] A second embodiment is a dialogue system dependent on the first embodiment, in which the avatar terminal includes a prompt update means for updating a prompt in the language model when a response is received from the messenger application server.

[0012] In the second embodiment, the prompt update means (22, 40h, S15) updates the prompt of the language model (124, 161) by, for example, storing data of a question from a user and data of an answer from the messenger application server (18) as a set.

[0013] According to the second embodiment, the prompt in the language model is updated, so that the next time the avatar terminal receives the same question, it will be able to answer using only the language model, eliminating the need to make an external inquiry.

[0014] The third embodiment is a dialogue system dependent on the first or second embodiment, in which the avatar terminal has an unanswerable list in which questions that cannot be answered by the language model are preset, and the judgment means judges that the user's question is unanswerable when it is on the unanswerable list.

[0015] In the third embodiment, for example, unanswerable questions are set in advance as rules in the prompt storage area (42d). The determination means (22, 40e, S5) refers to the list of unanswerable questions and determines whether the current user question can be answered using the language model (124, 161).

[0016] According to the third embodiment, when the determining means determines that a reply is not possible, an inquiry can be made promptly to an external party, for example, to a messenger application server.

[0017] A fourth example is a program for a dialogue system that includes an avatar terminal in which an avatar interacts with a user, a large-scale language model accessible by the avatar terminal, and a messenger application server accessible by the avatar terminal, in which the avatar terminal functions as a determination means for determining whether an answer to a user's question can be made using the language model, an answering means for outputting an answer using the large-scale language model when the determination means determines that an answer is possible, and a sending means for sending the question to the messenger application server when the determination means determines that an answer is not possible, and the answering means outputs the answer to the user when it receives an answer to the question from the messenger application server.

[0018] The fourth embodiment is also expected to have the same effects as the first embodiment. [Effects of the Invention]

[0019] According to this invention, when an avatar cannot autonomously answer a user's question, it requests an answer from an external source via a messenger application server and outputs the answer from the messenger application server to the user, thereby providing an appropriate answer to the user's question.

[0020] The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description of the preferred embodiments with reference to the drawings. [Brief explanation of the drawings]

[0021] [Figure 1] FIG. 1 is a block diagram showing an example of an interactive system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a block diagram showing an example of an avatar server in the embodiment of FIG. [Figure 3] FIG. 3 is a diagram showing an example of a memory map of the RAM in the avatar server shown in FIG. [Figure 4] FIG. 4 is a flow diagram showing an example of the operation of the avatar server in the embodiment of FIG. DETAILED DESCRIPTION OF THE INVENTION

[0022] Referring to FIG. 1 , a dialogue system 10 of this embodiment includes an avatar terminal 12, which is connected to a network 14 by wire or wirelessly. An avatar as an autonomous agent is displayed on the avatar terminal 12, and the avatar autonomously converses with a user using a local language model 124 such as ChatGPT (registered trademark) and / or a large-scale language model 161 described later. For example, the avatar answers questions uttered by the user. In other words, the avatar is an autonomous agent that converses with the user.

[0023] 1 is a general-purpose PC, but is called an "avatar terminal" because it displays an avatar, which is a CG agent, to interact with the user. Other terminals, such as a smartphone or tablet PC, can also be used as the avatar terminal 12.

[0024] Furthermore, the term "avatar" should be understood to encompass remotely controlled robots and CG agents, including those that interact with users partially or completely autonomously. For convenience, even if the avatar is a robot, the robot will be referred to as an "avatar terminal."

[0025] Avatar terminal 12 is further provided with a camera 121 for taking a picture of the user, a microphone 122 for listening to the user's speech, and a speaker 123 for speaking a response to the user.

[0026] In association with this avatar terminal 12, a local language model 124 that can only be accessed by the avatar terminal 12 is connected to the network 14 via a chat server 125 that is an interface for processing dialogue.

[0027] The local language model 124 is used to generate answers to user questions. Like the large-scale language model 161 described below, the local language model 124 is a natural language processing model trained by machine learning using a large amount of text data, and can be used for various natural language processing such as text classification, sentiment analysis, information extraction, text summarization, text generation, and question answering. In the embodiment, the local language model 124 is a large-scale language model, but it may also be a medium-scale language model or a small-scale language model that is smaller in scale.

[0028] The dialogue system 10 includes a large-scale language model 161 such as ChatGPT that is connected to the network 14 via a chat server 16, and this large-scale language model 161 is provided so as to be accessible from the avatar terminal 12 via the network 14. The avatar terminal 12 may pose a question to this chat server 16 and request an answer.

[0029] A Slack server 18 is also connected to the network 14. The Slack server 18 is a server for a messenger app such as Slack (product name) that enables group chat or team communication. However, it should be noted that other messenger apps can be used, such as "WhatsApp" (registered trademark), "WeChat" (registered trademark), and "Messenger" (product name), but the term "Slack" will be used here to represent them. The Slack server 18 can therefore be called a messenger app server.

[0030] Multiple channels are set up in the Slack server 18, and one or more operators can input opinions and replies to the corresponding channel through operator terminals 201, 202, ... 20n. This channel function allows information to be concentrated by related topic.

[0031] In the embodiment of FIG. 1, one operator may operate the avatar terminal 12 and one operator terminal 20, and one or more other operators may operate the other operator terminals 20.

[0032] Fig. 2 is a block diagram showing an example of the electrical configuration of avatar terminal 12 shown in Fig. 1. As shown in Fig. 2, avatar terminal 12 includes a CPU 22, which is connected to a RAM 26, an input / output interface (I / F) 28, and a communication interface (I / F) 30 via an internal bus 24.

[0033] The CPU 22 is responsible for overall control of the avatar terminal 12. The RAM 26 is a main storage device and is used as a work area or buffer area for the CPU 22. Although not shown, the avatar terminal 12 is provided with an HDD and ROM as auxiliary storage devices. However, a non-volatile memory such as an SSD may be used instead of or in addition to the HDD.

[0034] In addition to the camera 121 and microphone 122, an input device 32 is connected to the input / output I / F 28. The input device 32 may include a keyboard, a computer mouse, and a touch panel. The touch panel is provided on the display surface of a display 34 connected to the input / output I / F 28. The input / output I / F 28 is also connected to the speaker 123 and display 34 mentioned above.

[0035] The display 34 is, for example, a liquid crystal display device. The camera 121 is a CCD camera, the microphone 122 is a sound-collecting microphone, and the speaker 123 is a stereo speaker.

[0036] The communication I / F 30 is a wired or wireless interface for transmitting and receiving data to and from external computers such as the chat server 16 and the slack server 18 via the network 14 under the control of the CPU 22.

[0037] In addition, the input / output I / F 28 outputs operation data (or operation information) input from the input device 32 to the CPU 22, and outputs image data generated by the CPU 22 to the display 34, causing a screen or image corresponding to the image data to be displayed on the display 34.

[0038] In addition, the input / output I / F 28 converts the user's voice detected by the microphone 122 into digital voice data and outputs it to the CPU 22, converts the voice data output by the CPU 22 into an analog voice signal and outputs it from the speaker 123, and outputs video data of the user's video captured (detected) by the camera 121 to the CPU 22.

[0039] The electrical configuration of avatar terminal 12 shown in FIG. 2 is an example and is not necessarily limited to this.

[0040] Furthermore, if the avatar terminal 12 is a smartphone, it is equipped with a call circuit for making calls via a mobile phone network, or via a mobile phone network and a public telephone network; however, in this embodiment, such calls are not made, and therefore are not shown.

[0041] Similarly, if the avatar terminal 12 is a robot, it will have movable parts such as arms, but since these movable parts are not involved in the dialogue system 10 of this embodiment, they are not shown in the figure.

[0042] Fig. 3 is an illustrative view showing an example of a memory map of RAM 26 of avatar terminal 12 shown in Fig. 2. As shown in Fig. 3, RAM 26 includes a program storage area 36 and a data storage area 38. A control program for avatar terminal 12 is stored in program storage area 36.

[0043] The control programs executed on the avatar terminal 12 include a display control program 40a, an operation detection program 40b, a communication program 40c, a voice recognition program 40d, an answer possibility determination program 40e, an answer creation program 40f, an inquiry program 40g, and a prompt update program.

[0044] Display control program 40a is a program for displaying an image on display 34 (FIG. 2) in avatar terminal 12 using image generation data in image generation data storage area 42a (described below) of data storage area 38.

[0045] The operation detection program 40b is a program for detecting operation data input from the input device 30 and storing the data in an operation data storage area 42b (described later) of the data storage area .

[0046] The communication program 40c is a program for communicating with external devices, such as the chat servers 125 and 16 and the slack server 18.

[0047] The voice recognition program 40d is a program for recognizing the user's voice input through the microphone 122. Voice data as the recognition result is stored, as necessary, in a temporary storage area 42c (described later) of the data storage area 38. For example, if the user's voice is a question, the user's voice data is stored as question data in the temporary storage area 42c.

[0048] When avatar terminal 12 is unable to provide an answer by itself and requests an answer from slack server 18, this voice data is transmitted to slack server 18 by communication program 40c in accordance with inquiry program 44g (described later).

[0049] The answerability determination program 40e is a program for determining whether the avatar terminal 12 can answer a question from a user. In other words, the answerability determination program 40e is a program for determining whether the avatar terminal 12 can answer a user's question autonomously using only the local language model 124 and / or the large-scale language model 161. For example, when the local language model 124 and / or the large-scale language model 161 does not have information or knowledge about the question, it determines that the question cannot be answered. Alternatively, when it is not appropriate to present the answer obtained from the local language model 124 and / or the large-scale language model 161 to the user, the answerability determination program 40e determines that the question cannot be answered.

[0050] In this embodiment, a list of questions that cannot be answered in the local language model 124 and / or the large-scale language model 161 is determined according to rules preset in the prompt storage area 42d, which will be described later. That is, the answerability determination program 40e refers to this prompt storage area 42d to determine whether an answer is possible. Questions that cannot be answered are determined to be unanswerable, and other questions are determined to be answerable. However, if the local language model 124 and / or the large-scale language model 161 does not have enough knowledge to answer the question, it is determined to be unanswerable.

[0051] Examples of questions that should not be answered include specific questions about fees, or questions about personal privacy.

[0052] The answer generation program 40f is a program for generating answers based on answers given by the Slack server 18, in addition to generating answers using the local language model 124 and / or the large-scale language model 161.

[0053] Inquiry program 40g is a program for inquiring of an answer from Slack server 18 when avatar terminal 12 determines by the above-described answer possibility determination program 40e that it cannot answer the user's question itself. For example, when answer possibility determination program 40e determines that it cannot answer the question, it posts to Slack server 18 what kind of question is causing the user trouble. For example, it sends data of the question from the user stored in temporary storage area 42c to Slack server 18.

[0054] The prompt update program 40h is a program for updating the prompt of the local language model 124 and / or the large-scale language model 161 so that when a user's question cannot be answered using only the local language model 124 and / or the large-scale language model 161 and an answer is made using an answer obtained from the Slack server 18, the local language model 124 can thereafter answer the question itself. That is, the answer obtained from the Slack server 18 is added to the prompt as an answer example. For example, the prompt of the local language model 124 and / or the large-scale language model 161 is updated by storing, in the prompt storage area 42e described below, the voice data (question data) from the user stored in the temporary storage area 42c and the answer data of the answer obtained from the Slack server 18 and also stored in the temporary storage area 42c.

[0055] Although not shown, the HDD or SSD provided in the avatar terminal 12 stores a browser and other application programs other than the control program of the present disclosure in addition to the operating system and middleware of the avatar terminal 12. However, since the other application programs are programs separate from the control program, i.e., the information processing program, of this embodiment, they are not included in the information processing program.

[0056] The data storage area 38 of the RAM 26 is provided with a prompt data storage area 42d in addition to the image generation data storage area 42a, the operation data storage area 42b, and the temporary storage area 42c described above.

[0057] The image generation data 46a includes data for generating image data of various screens or various objects.

[0058] Operation data storage area 42b temporarily stores data on operations detected by operation detection program 40b for avatar terminal 12. The operation data is used for processing by CPU 22 and is sent to chat servers 125 and 16 and Slack server 18 as needed.

[0059] As described above, the temporary storage area 42c stores question data, answer data, and the like.

[0060] The prompt data storage area 42d stores prompts to be spoken to the user from the avatar terminal 12 or to instruct the generation of answers to questions from the user. The prompts include the user's utterances (the recognition results of the speech recognition program 40d), or commands, roles (autonomous agents in this case), personalities (or personas), and dialogue histories and / or dialogue examples with the user. The prompt data storage area 42d also stores, as rules, questions that cannot be answered using the local language model 124 and / or the large-scale language model 161. For example, questions about fees, rules, and other questions that could cause problems if answered incorrectly are set as a list of questions that must not be answered.

[0061] Although not shown, data storage area 38 stores other data necessary for the execution and operation of the control program of avatar terminal 12, and also includes flags and other counters (timers) necessary for the execution of the control program.

[0062] The operation of the interactive system 10 of this embodiment will be described below with reference to Fig. 4 etc. The flow chart of Fig. 4 is executed repeatedly, for example, at the frame frequency.

[0063] 4, in step S1, CPU 22 of avatar terminal 12 speaks to the user from speaker 123 (FIG. 1) using a prompt for autonomous dialogue set in local language model .

[0064] Then, in step S3, the CPU 22 determines whether the voice of the user input from the microphone 122 (FIG. 1) has been recognized by the voice recognition program 40d. That is, after recognizing the voice, it determines whether the meaning of the question from the user has been understood. The voice data of the question from the user is stored in the temporary storage area 42c as described above.

[0065] Then, in step S5, the CPU 22 determines whether the user's question can be answered by using only the local language model 124 and / or the large-scale language model 161 in accordance with the answerability determination program 40e.

[0066] For example, if the local language model 124 and / or the large-scale language model 161 does not have the knowledge to answer the user's question, or if it is set as inappropriate to use the local language model 124 and / or the large-scale language model 161 to answer (if such an answer is created), the judgment in step S5 is "NO."

[0067] If the answer is "YES" in step S5, that is, if it is determined that the user's question can be answered using only the local language model 124 and / or the large-scale language model 161, then in step S7 the CPU 22 inputs the user's question into the local language model 124 and / or the large-scale language model 161 in accordance with the answering program 40f, and obtains an answer from the local language model 124 and / or the large-scale language model 161. The answer is spoken from the speaker 123, and the process ends.

[0068] If the CPU 22 determines "NO" in step S5, that is, if it determines that the user's question cannot be answered using only the local language model 124 and / or the large-scale language model 161, then in the next step S9, it sends question data of the user's question to the Slack server 16 via the communication interface 36 (FIG. 2). That is, the avatar (CPU 22) sends the user's question data to a Slack channel where an answer to the user's question can be obtained. This question data is sent from the Slack server 18 to the operator terminals 201, 202, ..., 20n.

[0069] One or more operators operating the respective operator terminals 201, 202, . . . , 20n create answers to questions sent from the Slack server 18 and send the answers to the corresponding channels of the Slack server 18.

[0070] If the Slack server 18 receives a response from one operator terminal within a certain period of time, it returns the response to the avatar terminal 12.

[0071] However, when answers are received from multiple operator terminals, it is conceivable that the first answer will be sent to the avatar terminal 12. Alternatively, when answers are received from multiple operator terminals, the avatar terminal 12 may select the answer that has received the most "likes." In other words, one answer may be selected from multiple answers by voting on the Slack server 18.

[0072] Alternatively, when answers are obtained from multiple operator terminals, the avatar terminal 12 may receive all of the answers, and the CPU 22 of the avatar terminal 12 may use the local language model 124 and / or the large-scale language model 161 to select the optimal answer.

[0073] When an answer is obtained from the Slack server 18 in this manner, "YES" is determined in step S11, and in the following step S13, the CPU 22 responds to the user's question by speaking from the speaker 123 in accordance with the answer obtained in step S11.

[0074] When an answer is obtained from the Slack server 18 in the previous step S11, in step S15, if necessary, the prompt data of the local language model 124 and / or the large-scale language model 161 is updated. For example, the prompt of the local language model 124 and / or the large-scale language model 161 can be updated by storing the data of the question from the user stored in the temporary storage area 42c and the data of the answer obtained from the Slack server 18 and also stored in the temporary storage area 42c in the prompt storage area 42e.

[0075] For example, the answer from the Slack server 18 is stored as is as prompt data together with the question data.

[0076] Alternatively, the response from the Slack server 18 is returned to the local language model 124, adjusted so that the response style (phrasing or nuance) is similar to other responses, and then stored as prompt data.

[0077] For example, when a user asks the avatar terminal 12, "What is the phone number of AVITA Corporation?", if the local language model 124 and / or the large-scale language model 161 does not have information that can respond to the question, a "NO" determination is made in step S5, and the question is sent to the Slack server 18 in step S9 described above. Then, when the Slack server 18 responds, "It's 080-xxx-yyy," the prompt in the local language model 124 and / or the large-scale language model 161 is updated so that the next time the same question is asked, the answer can be obtained from the local language model 124 and / or the large-scale language model 161. In other words, when the Slack server 18 responds, "It's 080-xxx-yyy," the data of that dialogue (stored in the temporary storage area 42c) is added as a dialogue example for the prompt in the local language model 124 and / or the large-scale language model 161 of the avatar terminal 12. This allows the avatar terminal 12 to answer the same question next time using only the local language model 124 and / or the large-scale language model 161.

[0078] Alternatively, if the local language model 124 of the avatar terminal 12 is capable of answering questions about the telephone numbers of various companies, only the part of the list that manages those telephone numbers may be rewritten according to the answer "It's 080-xxx-yyy."

[0079] In the dialogue system 10 shown in FIG. 1, an operator can operate the avatar terminal 12 and input prompts corresponding to the user's question into the local language model 124 and / or the large-scale language model 161 to obtain an answer. In this case, if the dialogue cannot be carried out successfully using the local language model 124 and / or the large-scale language model 161, external help is sought. In the above-described embodiment, in such a case, help is sought from an external person via the Slack server 18. In other words, an answer can be obtained from the operator operating the operator terminal 20 connected to the Slack server 18 through the operator terminal 20.

[0080] When getting a response from an operator, the operator at the avatar terminal 12 and the operator at the operator terminal 20 may be the same person, but by combining it with a messenger app such as Slack, it is possible to connect with various operators (persons in charge) at any time, which will result in a faster response when the avatar is in trouble.

[0081] In the above-described embodiment, when the autonomous avatar of the avatar terminal 12 answers a user's question, if the answer cannot be obtained using the local language model 124 alone, the answer to the user's question is obtained by inquiring from an external server such as the chat server 16 or the Slack server 18, thereby enabling the avatar to provide an accurate answer to the user's question.

[0082] In addition, when an answer is received from an external source, the prompt is updated as needed, so that the next time the same question is asked, the system will be able to answer autonomously using its own large-scale language model.

[0083] However, in addition to updating the prompt, there is also a method for modifying the local language model 124 of the avatar terminal 12. The local language model 124 may be fine-tuned by additional learning using the answer obtained from the Slack server 18 and the user's question immediately preceding this answer.

[0084] Furthermore, if the same results are obtained from the steps in the flow charts shown in the above embodiments, the order in which the steps are processed can be changed. [Explanation of symbols]

[0085] 10...Dialogue systems 12...Avatar terminal 122 ...Mike 123 ...Speaker 124 …Local Language Model 125, 16...Chat server 14...Network 161 ...Large-scale language models 18...Slack server 201, 202, ..., 20n ... operator terminals 22...CPU 26...RAM 30...Communication I / F

Claims

1. The system includes an avatar terminal in which an avatar interacts with a user, a language model accessible to the avatar terminal, and a messenger application server accessible to the avatar terminal; The avatar terminal includes: a determining means for determining whether a user's question can be answered using the language model; answering means for outputting an answer based on the language model when the determining means determines that an answer is possible; and a sending means for sending the question to the messenger application server when the determining means determines that the question cannot be answered; The answering means, when receiving an answer to the question from the messenger application server, outputs the answer to the user.

2. The dialogue system according to claim 1 , wherein the avatar terminal includes a prompt update means for updating a prompt of the language model when a response is received from the messenger application server.

3. 3. The dialogue system according to claim 1, wherein the avatar terminal is provided with an unanswerable list that pre-sets a list of questions that cannot be answered by the language model, and the determination means determines that the user's question is unanswerable when the question is on the unanswerable list.

4. A program for a dialogue system including an avatar terminal in which an avatar interacts with a user, a language model accessible to the avatar terminal, and a messenger application server accessible to the avatar terminal, The avatar terminal, a determining means for determining whether a user's question can be answered using the language model; answering means for outputting an answer based on the language model when the determining means determines that an answer is possible; and When the determining means determines that the question cannot be answered, the device functions as a sending means for sending the question to the messenger application server; The answering means is a program of a dialogue system that, when receiving an answer to the question from the messenger application server, outputs the answer to the user.

Images