JP2002132803A - User support system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve the problem that it is difficult to design an agent which accurately interprets various words that a user speaks and other requests and answer in a short time. SOLUTION: A collection of words that the user is assumed to speak is placed on an origin server 20, where a user's spoken word is specified. According to the result, a user terminal 12 is connected to an expert server where an expert agent which should answer to the user is present. A chat agent as the expert agent is installed on a chat server 24 and has an agent action collection for chatting with the user. Decentralized processing by the separation of the user's spoken word collection and agent action collection is carried out and index retrieval is carried out to specify the user's spoken words.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] The present invention relates to user support technology. This utterance particularly relates to a system that supports user operations, search, and other processes using an agent.

[0002]

2. Description of the Related Art With the spread of the Internet and mobile phones, more and more people have come into contact with computers and various information terminals. Over the past few years, the sales volume of personal computers and other equipment has grown significantly,
So-called beginners are increasing rapidly. With a dramatic improvement in CPU power and memory capacity, a GUI (graphical graphics
The user interface is also being improved. For beginners, make full use of the advanced functions of the device,
Accessing the required information correctly is difficult without a GUI or other help.

As one of such user support technologies, use of an agent is known. The agent is mainly a personified character, appears on the computer screen, acquires the utterance of the user, and transmits this to the computer. The result of the processing in the computer is shown to the user as an utterance of the agent. The presence of the agent frees the user from being stubborn against the machine of the computer, and makes the computer feel more familiar.

[0004]

On the other hand, on the side of designing an agent, its proper operation and refinement are extremely laborious. The user is a living person, and the content of the utterance may be almost infinite. It is virtually impossible to handle all such utterances, but the user expects the agent to know exactly what utterances and respond immediately. Even if a considerable number of user utterances are assumed, it is very difficult to search for and specify words actually spoken by the user with a reaction time close to real time. Also, as the expected number of utterances increases, the amount of data to be handled by the server that controls the agent increases, and the load increases.

The present invention has been made in view of such a situation, and an object of the present invention is to provide a technique for realizing an agent that responds to a user's various utterances and other requests in the shortest possible time. Another object of the present invention is to provide a technique capable of continuously and efficiently increasing the accuracy of a response by an agent.

[0006]

A user support apparatus according to the present invention has an electronic user utterance collection created on the assumption of a user's utterance, and specifies the contents of the utterance when the utterance is input from the user. And a second block having an electronic agent action book created assuming the action of the agent for the purpose of responding to the invention, and causing the agent to respond to the utterance of the user. including. Further, the first and second blocks are configured separately, and the user utterance collection and the agent action collection are held in a separated state. According to this configuration, it is easy to separately access the user utterance collection and the agent action collection, so that the process of specifying the user's utterance and the process of determining the response by the agent can be performed in parallel. it can. As described above, the “agent” is not always visible as an anthropomorphic character, and may refer to a program or other processing function that assists a user in an invisible place, such as a system back end. The “action” of the agent indicates an utterance, an image, an action, or any other processing element or processing operation related to the processing for support.

[0007] The user support system may include a plurality of the second blocks, and each of the blocks may provide a specialized service to the user. For example, a part or all of the first block and the plurality of second blocks are configured as different network nodes,
It may communicate with the user via the network. In this case, the user corresponds to the client, and the function corresponding to the first and second blocks is the server, and a server-client system is constructed.

The first block is an entrance or portal server for specifying the utterance of the user, and an appropriate one may be selected from a plurality of second blocks according to the content of the utterance specified there. The second block may be provided for each content of service to the user, for example, news, fortune-telling, travel, cooking, business,
It may be provided for each topic such as health. In this case, the refinement of the agent is relatively easy because each of the second blocks has a clear theme. Also, since processing can be performed on different nodes for each topic,
The load can be reduced by the distributed processing.

The first block includes an utterance search unit for searching for the utterance of the user in the user utterance collection, and when the user's utterance is not found in the user utterance collection, A notification unit for notifying a system administrator (hereinafter, simply referred to as an administrator). In that case, the administrator can separately deal with the user or revise the user utterance collection and the agent action collection.

[0010] The system may further include a recording unit for acquiring a record of access to the system by the user. In this case, the second block may select an utterance of the agent for the response from a plurality of candidates according to a situation of access by the user.
For example, when a user such as "the Continent" for the agent A, if the initial visits by the user, agent A may If you answer "Hello", but if a second time, "Hey, meet again I'm sorry. " In this case, the most suitable one is selected based on the access history of the user, and a more natural or more friendly conversation is realized.

[0011] The first block may include an index storage unit for holding an index for the content of the user utterance collection. In this case, when the utterance of the user is input, the search unit can perform an initial search, that is, a search for narrowing down, on the index storage unit, and the search can be speeded up. Realize.

It is to be noted that any combination of the above-described components and any conversion of the expression of the present invention between a method, an apparatus, a system, a recording medium, and the like are also effective as embodiments of the present invention.

[0013]

FIG. 1 shows an overall configuration of a network system 10 including a user support system 16 according to an embodiment. Here, the user terminal 12 and the user support system 16 are connected via the Internet 14. The user terminal 12 is a PC (personal computer), a PDA, a mobile phone connectable to the Internet 14, or any other hardware.

The user support system 16 has a starting server 20 as a first processing block, a chat server 24 and a menu server 26 as second processing blocks, respectively, which are connected to the Internet 14 respectively. . Therefore, the origin server 20, the talking server 24, and the menu server 26 are configured as different network nodes, and the processing of the user's utterance and the agent's utterance can be distributed. In addition, the agent can be another node according to the field in which the agent is in charge, thereby facilitating maintenance for each agent. The names of the chat server 24 and the menu server 26 are also given according to their field of specialization or specialty. Hereinafter, the chat server 24, the menu server 26, and the like are collectively referred to as “specialized servers”, and the agents placed on those servers are referred to as “specialized agents”. The user support system 16 may be integrally configured, for example, may be configured in one portal site. However, in the following description, an example in which the user support system 16 is configured as a separate node and the origin server 20 acts as a portal server for the user terminal 12 will be described. Think.

Although details will be described later, the outline of the processing in FIG. 1 is as follows. When the user activates the user terminal 12, a local agent mounted inside the user terminal 12 appears. The local agent waits for the user's first utterance (hereinafter, this utterance is referred to as “processing start utterance”). The processing start utterance is transmitted to the origin server 20 via the Internet 14. At this time, the user terminal 12 is displaying the page of the origin server 20 by the WWW browser.

A collection of user utterances, that is, a collection of supposed utterances that the user will make, is mounted on the origin server 20, and the processing start utterance is collated with the utterance and the utterance content is specified. As a result, a specialized agent to respond to the processing start utterance is specified, and the URL of the specialized server (U in FIG.
RLa, URLb) are sent to the browser of the user terminal 12. The page of the specialized server is displayed on the screen of the user terminal 12 that has acquired the URL, and the specialized agent appears. Each specialized server has a collection of agent actions for specialized agents, and responds to the processing start utterance and subsequent user utterances (hereinafter referred to as “normal utterances”). In the following, utterance is mainly used as an example of the agent's action, but the present invention is not limited to this, and the user may use gestures and other actions, changes in colors and textures on images, search operations by the agent, and other program processing. May be supported.

When the user makes a new utterance to the specialized agent, that is, a normal utterance, the utterance is acquired and sent to the origin server 20 (indicated as URLs in the figure). The origin server 20 again specifies the specialized agent that should respond to the utterance, and transmits the URL of the specialized server to the user terminal 12 again. Hereinafter, similarly, 1. 1. Identification of user utterance by origin server 20 2. Identification of a specialized server responsible for the identified utterance 3. Response to the user by the specialized agent placed on the specialized server. A series of processing of requesting or promoting the user's normal utterance by the specialized agent is repeated. Since the processing always returns to the origin server 20 and is newly started from there,
There is a name of the origin server 20.

FIG. 2 shows the internal configuration of the origin server 20.
In the figure, "H" is utterance data, "I" is an index search of utterance, "F" is the name of a file holding the URL of the page of the specialized server to respond to the specified user utterance, and "X" is Each utterance that could not be identified (hereinafter also referred to as an unknown utterance) is shown. 2 can be realized by a CPU, a memory, or other LSI of an arbitrary computer in terms of hardware, and is realized by a program loaded in the memory in terms of software. Draws functional blocks realized by. Therefore, it will be understood by those skilled in the art that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof. Hereinafter, similar considerations are valid throughout the present specification.

The communication unit 30 communicates with the specialized server and the user terminal 12 via the Internet 14. The utterance acquisition unit 32 acquires the utterance of the user from the user terminal 12 and sends the utterance to the utterance search unit 34. The utterance search unit 34 checks the first character of the utterance against the index file 36, first narrows down by index, and then
Identify the content of the utterance. At this time, a phrase search is performed for the entire utterance of the user. The phrase search is a process of finding a match including the order of words. If no corresponding phrase is found by the phrase search, as described later, the utterance is divided into morphemes, and a keyword or word is searched for.

The index file 36 is generated by arranging the supposed utterances of the user stored in the user utterance collection 38 in the order of the Japanese syllabary.
Even if is large, the user's utterance can be searched very quickly. As will be described later, since the user utterance collection can be easily enhanced in this embodiment, the capacity of the user utterance collection 38 is expected to be very large in a short period of time, and the merit of speeding up by index search is great.

When an utterance is specified in the index file 36, a file in which information such as a URL of a specialized server to respond to the utterance is described is stored in the index file 36.
Then, the file body contained in the user utterance collection 38 is opened, and the URL is found. In the user utterance collection 38, one file is formed for one utterance. The found URL is set in the browser of the user terminal 12 via the communication unit 30, and the access destination of the user terminal 12 is the specialized server. Strictly speaking,
The URL does not refer to the home page of the specialized server, but to an individual page for responding to the user's utterance, and one or more pages are assigned to each utterance. . Multiple cases will be described later.

There is no problem when a message that completely matches the user's utterance is stored in the user's utterance collection 38, but this does not often occur particularly in the process of enriching the user's utterance collection 38. In this case, the utterance search unit 34 decomposes the utterance of the user into morphemes by a known method, and performs a search again by taking the logical product of the nouns among them, thereby obtaining the most probable utterance from the user utterance collection 38. Find out. The utterance targeted for the re-search and the utterance not found through the re-search are both recorded as unknown utterances in the unknown utterance file 40, and this is sent to the administrator of the origin server 20 via the notification unit 42 by e-mail or the like. Notified by The administrator newly registers the utterance and the URL of the page of the specialized server to be responded to in the user utterance collection 38, and stores the index of the utterance in the index file 3.
6 and a process including the utterance of the specialized agent on the page is designed. According to this maintenance, the unknown utterances can be added to the user utterance collection 38 without any difficult process, so that the user utterance collection 38 is very easily enriched.

The access recording unit 44 records the status of access to the specialized server in the access information file 46 for each user. This makes it possible to make different responses to the same user utterance. For example, a user for the first time visited the chat server 24 is "Hello"
In such a case, the specialized agent of the chat server 24 (hereinafter, also referred to as the chat agent) is referred to as "Hello!" On the other hand, when revisit to the chat server 24 by the user, chat agent can be referred to as "Hello. What about from there.", Fine-grained corresponding to realize. The access recording unit 44 informs the utterance search unit 34 of the access status. When a plurality of pages of the specialized server for responding to the user's utterance are described in the user utterance collection 38 as in the present example, the utterance search unit 34 finds an appropriate one based on the access situation. Select and set the URL in the browser of the user terminal 12.

FIG. 3 shows the internal structure of the index file 36, and FIG. 4 shows the internal structure of the user utterance collection 38. The index file 36 has a Japanese syllabary column 100, a user utterance column 102, and a file name column 104. The utterances of the user are arranged in the order of the Japanese syllabary. If the first character is "A", the utterances are classified under "A" in the Japanese syllabary column 100, and so on.

The user utterance collection 38 includes a file name column 104,
It has a user's utterance field 102 and a page field 120 for a specialized server to respond to. For example, the page of the specialized server responding to the user utterance “Haya” is “URLa43”,
The combination of the utterance “Haya” and “URLa43” constitutes the file f044. The user's utterances are grouped for each specialized server, and, for example, a user utterance 110 to be in charge of the talking server 24 and a user utterance 112 to be in charge of the menu server 26 are described as a group. The former relates to general greetings, and the latter relates to cooking and recipes. Index file 36 and user utterance collection 3
8 are linked by the file name, for example, in the utterance of "Hello", in the index file 36 file f044 is described, this is the user utterance collection 3
8 file f044.

[0026] As shown in FIG. 4, page of the response to the "Hello" There are two URLa1 and URLa2.
Here, it is assumed that URLLa1 is transmitted to the user who first comes to the chatting server 24, and URLa2 is transmitted in the case of returning.

FIG. 5 shows an example of an internal description of the access information file 46. In the figure, a user “user1” has visited specialized servers “talking”, “menu”, and “auction” in the past. On the other hand, the user “user2” visits a specialized server “travel” and “PC”. Therefore, when user2 visits the chat server 24, as described above, the chat agent utters a speech for the first visitor, and when user1 visits the chat server 24, a greeting for the return visitor is uttered.

FIG. 6 shows an internal configuration of the talking server 24 as an example of a specialized server. The communication unit 60 communicates with the user terminal 12, the origin server 20, and the like via the Internet 14. URLa1 or URL when the URL specified by the utterance search unit 34 of the originating server 20, that is, the user's utterance as in Figure 4 of the "Hello"
a2 is given to the agent action book 62 via the communication unit 60. The agent action collection 62 includes, in addition to the utterances of the specialized agents, agent data 72 describing images and actions thereof. In addition, a page is provided one-to-one with the URL specified by the utterance search unit 34. Specifically, the page 64 of the URLa1 and the page 6 of the URLa2 are provided.
6. It is configured as a URL 68 page. Each page is a so-called Web page, not only the speech by the talking agent, but also the display of images and actions, the services that the agent should perform,
For example, it also searches for information. In other words, one Web page is provided for each utterance of the user, and the flexibility and enhancement of the response are intended.

Since each page has substantially the same configuration, only the page 64 of URLa1 is described in detail in FIG. UR
The La1 page 64 has an agent output unit 70, a user utterance acquisition unit 74, and a specific process execution unit 76. These are the CGIs that work behind this page (Common Gateway
Interface). Agent output unit 7
0 responds to the user's utterance by the talking agent based on the agent data 72. As described above, the specific process execution unit 76 performs any process other than the response by the utterance, such as information search and execution of various programs. As an example, the utterance of the user who arrived at this page is "I want to know today's news"
If so, the talking agent is Internet 1
4 to retrieve news and show it to the user. The user utterance acquisition unit 74 acquires the subsequent normal utterance by the user and notifies the origin server 20 of this. As a result, a new specialized server is specified by the origin server 20.

FIG. 7 shows the internal configuration of the user terminal 12. The communication unit 130 communicates with the origin server 20, the talking server 24, the menu server 26, and the like via the Internet 14. The user interface 138 is a general term for a general configuration for prompting and inputting a user's decision and includes a keyboard, a mouse, a display device, and various data interfaces. The local agent output unit 132 reads out the local agent data 134 and provides it to the user through the user interface 138. The user's processing start utterance and normal utterance are acquired by the user utterance input unit 136 via the user interface 138, and transmitted to the origin server 20 via the communication unit 130 and the Internet 14. The general configuration of the embodiment has been described above, and the processing will be described below with a specific example.

FIG. 8 shows a screen 150 displayed when the user activates the user terminal 12. Here the local agent 152 appears and says, "Welcome! Let's talk to me." The user inputs a "Hello" in the input area 154, and press the send button. The input area 154 may be configured to appear when the user clicks on the local agent 152, and the local agent 152 may continue to speak random words while no click is made, or may ask the user a question and prompt the user to speak. You may. In any case, was entered "Hello" is sent to the origin server 20 as the start of processing speech, it has been identified chat server 24 as a specialized server from its contents, the access destination of the user terminal 12 is transferred to the page.

FIG. 9 shows a screen 150 displayed at that time. Here, the chat agent 156 is displayed, but in this embodiment, the local agent 15 is displayed.
By using the same agent image as in No. 2, it is shown that seamless conversation continues. The chat agent 156 speaks "Hello! I am Pico of chat agent. ...".
When the user inputs “Tell me a recipe” in the input area 154 and sends it, this is acquired by the origin server 20 and one page of the menu server 26 is newly specified. The URL of the specified page is sent to the user terminal 12, and the access destination of the user terminal 12 moves to the page.

FIG. 10 shows a screen 150 appearing at that time. Here, a menu agent 160 has newly appeared,
Say, "Let's leave it to me, the menu agent." Then he asks, "What kind of food do you want to know?" This means that when the utterance of the user is vaguely asking for a recipe, there are too many candidates, and the narrowing-down condition is acquired in advance from the user. The user inputs “China” in the input area 154 and transmits this. This normal utterance is sent to the origin server 20, where the specialized server specified is still the menu server 26, but moves to another page therein.

FIG. 11 shows a screen 150 showing the page to which the user has moved. Here, in order to further narrow down the search conditions, the menu agent 160 asks, "Well, what about the ingredients. Please choose one."
At the lower part of the screen 150, an area 162 for displaying a food candidate with a radio button appears.
"Fish" and the like. The user here is "cow"
Is selected and the send button is pressed.

FIG. 12 shows a screen 150 appearing at that time. The screen 150 in FIG. 12 is a page different from the screen 150 in FIG. 11, but this time, the access destination did not change based on the utterance, and the user jumped to the page linked to the above-mentioned “cow” radio button. The result is changing. In this way, by incorporating the options prepared in advance into the utterance by the specialized agent, the user's intention can be confirmed more easily and reliably as a general theory.

In FIG. 12, the search on the Internet 14 by the specific processing execution unit 76 has already been completed under the search conditions of the previous user, that is, “China” and “cow” and “recipe”. The menu agent 160 speaks, "What about this?", And the title 170 of "Fried beef and pepper" and other recipes is displayed. These titles 1
A link is affixed to 70 so that a detailed explanation page can be accessed. In the search result column 172, a site for introducing a Chinese recipe is displayed, for the convenience of the user. In any case, starting from this screen 150, the user can know the details of the recipe. In the same figure, as the normal utterance of the further user, “I want to see the timetable of the airplane” is displayed in the input area 15.
4 has been entered. When this is transmitted, the travel agent of the travel server (not shown) is specified in the origin server 20, and necessary processing is started.

On the other hand, although not shown, if the user inputs, for example, "I want to know Egyptian royal cuisine," this may be treated as an unknown utterance. In that case,
This utterance is notified as it is to the system administrator, and the user utterance collection 38 and the index file 36 are updated. In addition, a new page is provided in the menu server to respond to, and thereafter, information regarding the Egyptian royal cuisine is correctly returned for this utterance.

The present invention has been described based on several embodiments. These embodiments are exemplifications, and it is understood by those skilled in the art that various modifications can be made to the combination of each component and each processing process, and such modifications are also within the scope of the present invention. By the way. Here are some examples.

In FIG. 8 and the like, the user's utterance is performed on a text basis, but this may be performed by voice recognition. In that case, the user is in a form that is more interacting with the specialized agent. Specialized agents may also speak by voice.

The unknown utterance was considered to be an utterance that could not be specified in the user utterance collection 38, but could be specified in the user utterance collection 38, but the response by the specialized agent was actually incomplete or could not be realized at all. In that case, the utterance may be called an unknown utterance. For example, when the specific process execution unit 76 performs a search based on the user's utterance “I want to know the recipe”, if the number of candidates is too large to satisfy the user's wishes, the utterance to the administrator is also unknown. Notifications may be made to improve the response by specialized agents.

In the embodiment, the utterance by the specialized agent is appropriately selected in consideration of the access situation of the user to the specialized server. Alternatively, the utterance may be selected based on the attribute information of the user. For example, if the user is a woman, the professional agent may select a relatively soft expression, and if the user is an elderly user, a more precise expression may be selected.

In the embodiment, the local agent 1
Although the image of the chat agent 52 and the image of the chat agent 156 are matched, it is not necessary to limit to this. Further, the local agent 152 may not be mounted on the user terminal 12, but may be mounted on the origin server 20 such as a “processing start agent” appearing at the time of startup.

[0043]

According to the present invention, an agent that responds to a user's various utterances and other requests in a relatively short time is realized. Further, the accuracy of the response by the agent can be continuously and efficiently increased.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram of a network system including a user support system according to an embodiment.

FIG. 2 is a configuration diagram of an origin server included in the user support system.

FIG. 3 is a configuration diagram of an index file included in an origin server.

FIG. 4 is a configuration diagram of a user utterance collection included in an origin server.

FIG. 5 is a configuration diagram of an access information file included in an origin server.

FIG. 6 is a configuration diagram of a talking server included in the user support system.

FIG. 7 is a configuration diagram of a user terminal that uses the user support system.

FIG. 8 is a diagram showing a local agent appearing when a user terminal is started.

FIG. 9 is a diagram showing a chatting agent appearing when the user speaks.

FIG. 10 is a diagram showing menu agents appearing when a user asks for a recipe.

FIG. 11 is a diagram showing a manner in which a menu agent confirms a search narrowing condition to a user.

FIG. 12 is a diagram showing a scene where a menu agent presents a search result to a user.

[Explanation of symbols]

12 user terminals, 16 user support systems, 2
0 Origin server, 24 Talking server, 26
Menu server, 32 utterance acquisition unit, 34 utterance search unit, 36 index file, 38 user utterance collection, 42 notification unit, 44 access recording unit, 6
2 agent action book, 104 file name, 1
20 Page address of specialized server to respond to, 13
2 local agent output section, 152 local agent, 156 chatting agent,
160 Menu agent.

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G06F 3/16 320 G06F 3/16 320H 340 340Z (72) Inventor Tsugufumi Matsuoka 2 Keihanhondori, Moriguchi-shi, Osaka 5-5-5 Sanyo Electric Co., Ltd. F-term (reference) 5B075 ND20 NK10 NK46 PQ04 5E501 AA02 AB15 AC15 BA05 CA08 CB15 DA01 EA21 FA14 FA32 FA43

Claims (7)

[Claims] An electronic user utterance collection created assuming an utterance of a user, a first block for specifying the content of the utterance when the utterance is input from the user, and a response to the utterance. A second block that has an electronic agent action book created on the assumption of a target agent action, and causes the agent to respond to the utterance of the user, wherein the first and second blocks are A user support system, wherein the user utterance collection and the agent action collection are separated by being configured separately. 2. The system according to claim 1, wherein the system is connected to a terminal of the user via a network, and the first block and the second block are configured as different network nodes. A user support system as described. 3. The system according to claim 1, wherein the system includes a plurality of the second blocks, each of which provides a specialized service to the user. User support system. 4. The utterance search unit that searches for the utterance of the user in the utterance collection of the user, when the utterance of the user is not found in the utterance collection of the user. The user support system according to any one of claims 1 to 3, further comprising: a notification unit that notifies a system administrator of the fact. 5. The recording apparatus according to claim 1, further comprising a recording unit configured to acquire a record of access to the system by the user, wherein the second block is configured to determine a plurality of actions of the agent for the response according to a situation of access by the user. 2. The method according to claim 1, wherein the selection is made from candidates.
5. The user support system according to any one of items 1 to 4. 6. The method according to claim 1, wherein the second block selects an action of the agent for the response from a plurality of candidates according to attribute information of the user.
6. The user support system according to any one of claims 1 to 5, 7. The first block has an index storage unit that holds an index for the content of the user utterance collection, and the search unit stores the index in the index storage unit when the user's utterance is input. 7. The user support system according to claim 1, wherein an initial search is performed on the user support system.