JP2010020531A - Conversation apparatus and conversation method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a conversation apparatus and a conversation method which produce a conversation sentence by substituting a variable part in a succeeding conversation sentence with a substitution candidate substituted by a variable part set previously in a previous conversation sentence or a keyword predicted to emerge in a conversation sentence succeeding the previous conversation sentence based on the substitution candidate. <P>SOLUTION: The conversation apparatus includes a conversation sentence storage section for storing a plurality of conversation sentences for each scene, and a variable part information storage section for setting a variable part in the conversation sentence and storing a substitution candidate for substituting by the variable part. The conversation apparatus stores substitution candidates substituted by the variable part in the order of conversation, retrieves a conversation sentence stored in the conversation sentence storage section using the substitution candidate as a key, then determines whether or not the variable part is in the retrieved conversation sentence, and if the variable part therein is determined, substitutes the variable part with the substitution candidate used as the key to produce a next conversation sentence. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a replacement candidate that has been replaced with a preset variable section in a previous conversation sentence, or a keyword that is predicted to appear in a conversation sentence after the previous conversation sentence based on the replacement candidate. The present invention relates to a conversation apparatus and a conversation method for generating a conversation sentence by replacing a variable part in a sentence. The conversation device and conversation method of the present invention are used in a conversation device and conversation method for translating between two languages.

  For example, Patent Document 1 discloses a conversation control device. This Patent Document 1 is a conversation control device that sequentially outputs a series of conversations and topics prepared in advance to a user while responding to the user's conversation. The conversation control device stores a plurality of plans each having an answer sentence and next candidate designation information for designating a next candidate answer sentence that is scheduled to be output in the next order of the answer sentences in the plan storage unit. Then, the plan is selected according to the first utterance of the user, the answer is output, and then the next candidate is selected based on the next candidate designation information of the plan according to the second utterance of the user. By outputting the answer sentences, a plurality of answer sentences are output in a predetermined order according to the user's utterance.

  Further, Patent Document 2 is a speech understanding method that makes it possible to interpret an understanding of an input utterance in which the flow of dialogue is appropriate using a prediction effect obtained by storing the utterance context. This speech understanding method obtains the utterance type of the utterance proposition content by searching the utterance type table, and corresponds to the proposition pattern included in the key slot of the fully instantiated plan instance in the plan instance buffer. The expectation level of the proposition content is set low, and the expectation level of the proposition content corresponding to the proposition pattern included in the key slot of the plan instance or the non-instantiated subgoal that is not fully embodied is set high. Thus, the speech understanding control unit obtains a final score of each recognition result using the utterance type and the proposition content expectation.

Further, Patent Document 3 relates to a multilingual interpreting apparatus and method that can communicate by conversation in a form that repeats a scene according to a preset script, and a fixed sentence is associated with a preset relationship between terminal apparatuses. For a conversation in a format that is exchanged according to, the language type of one standard sentence is different, and multiple expressions that are in parallel with each other are prepared and displayed in different languages corresponding to the screen. is there.
JP 2007-115142 A JP-A-4-307664 JP-A-2002-82984

The conversation control device of Patent Document 1 is a conversation control device that outputs a next candidate answer sentence in response to the utterances of the first and second users, but does not reduce the user's input effort.
Moreover, although patent document 2 calculates | requires the speech recognition result of the next input stochastically based on the previous input, the effort of the next input cannot be reduced.
Moreover, although patent document 3 displays a predetermined sentence, the sentence displayed next cannot be changed flexibly.
The present invention solves the above-described problem, and replaces a replacement candidate with a preset variable section in a previous conversation sentence, or a conversation sentence after the previous conversation sentence based on the replacement candidate. An object of the present invention is to provide a conversation apparatus and a conversation method for generating a conversation sentence by replacing a keyword predicted to appear with a variable part in a subsequent conversation sentence.

In order to solve the above-described problem, a conversation apparatus according to the present invention stores a conversation sentence storage unit that stores a plurality of conversation sentences composed of either a variable unit or a fixed unit, and a replacement candidate to be replaced with the variable unit. A variable phrase information storage unit, a search unit for searching for the first conversation sentence stored in the conversation sentence storage unit, and a replacement phrase from among the replacement candidates of the variable part of the first conversation sentence searched by the search unit A replacement candidate determination unit for determining the replacement phrase, and an output unit for replacing the replacement phrase determined by the replacement candidate determination unit with the variable part of the first conversation sentence, and outputting the replacement phrase. Is used as a key to search for a second conversation sentence from the conversation stored sentence, and the output unit includes the second conversation sentence in the second conversation sentence that has the same variable part as the first conversation sentence. The variable part is replaced with the replacement word and output.
In particular, the conversation sentences stored in the conversation sentence storage unit are classified according to scenes, and step numbers classified for each step of the conversation are stored in association with the conversation sentences.
The conversation device of the present invention is classified into scenes such as restaurants, hotels, airports, shopping and hospitals, and further classifies each scene into a type, class, age of a user, sex, etc. The conversation sentence is stored in the conversation sentence storage unit. The conversation sentence is composed of either a variable part or a fixed part, and replacement candidates to be replaced with the variable part are stored in the variable part information storage part. The conversation sentence storage unit and the variable part information storage unit are provided, and the conversation sentence stored in the conversation sentence storage unit is searched and searched with the replacement candidate replaced with the variable part set in advance in the previous conversation sentence as a key. If there is a variable part in the received conversation sentence, the replacement candidate used as the key is replaced with the variable part, and the subsequent conversation sentence is generated, so that the conversation sentence to be output next is generated by reducing the user's trouble. be able to.

Further, in the embodiment, the conversation device according to the present invention stores the variable part replacement candidate and the keyword that is expected to be included in the conversation sentence of the next step of the conversation sentence having the replacement candidate in association with each other. An input prediction storage unit is further included, and the search unit searches the replacement phrase from the replacement candidate words in the next input prediction storage unit, and obtains a keyword corresponding to the searched replacement word from the next input prediction storage unit The second conversation sentence stored in the conversation sentence storage unit is searched using the acquired keyword as a key.
This makes it possible to search for a conversation sentence including a keyword that is expected to appear in a conversation sentence after the previous conversation sentence, using the replacement candidate replaced with the previous conversation sentence as a key. Predicted conversation sentences can be generated.

In the embodiment, the conversation device according to the present invention further includes an input unit for inputting characters and voices, and the search unit uses the sentence input from the input unit or the input keyword as a key. The first conversation sentence or the second conversation sentence stored in the storage unit is searched.
Thereby, the conversation sentence which a user needs can be output.

In the embodiment, in the conversation device according to the present invention, when the conversation sentence storage unit includes a plurality of variable parts in one stored conversation sentence, the conversation unit corresponds to a predetermined variable part replacement candidate and the replacement candidate. It is characterized in that it is stored with a co-occurrence relationship with other variable parts to be changed or replacement candidates with fixed parts.
Thereby, a predetermined relationship can be given to a plurality of replacement candidates.

In the embodiment of the conversation device according to the present invention, the output unit outputs the first conversation sentence or the second conversation sentence generated by the replacement when instructed by the input unit. .
Thereby, it is possible to output a conversation sentence with good timing.

In the conversation device of the present invention, in the embodiment, the conversation sentence storage unit stores the stored conversation sentence, time information for proceeding to the conversation sentence of the next step of the conversation sentence, and the next step number of the next step. They are stored in association with each other.
As a result, the sentence to be displayed after the designated time or the designated time has elapsed can be associated.

In the embodiment, the conversation device of the present invention further includes a measuring unit that measures a time from when the first conversation sentence is output to the output unit, and includes time information of the first conversation sentence retrieved by the retrieval unit. When it is determined that the time information and the time measured by the measuring means have exceeded the time information, the search unit uses the next step number stored in association with the first conversation sentence as a key. A second conversation sentence having a next step number is searched from the conversation storage sentence.
Thereby, it is possible to output a conversation sentence with good timing.

In the embodiment, the conversation device according to the present invention further includes a translation conversation sentence storage unit that stores a second language translation sentence corresponding to the first language conversation sentence stored in the conversation sentence storage unit, and the variable unit. A translation variable section information storage section for storing a translation of the second language corresponding to the replacement candidate of the first language stored in the information storage section, and the output section outputs the conversation sentence generated by the replacement In this case, a translated sentence in the second language is output.
This makes it possible to talk using different languages.

In another aspect, the present invention provides a conversation method, a conversation sentence storage unit that stores a plurality of conversation sentences composed of either a variable part or a fixed part, and a replacement candidate that replaces the variable part. And a variable step information storage unit for storing a search step for searching for the first conversation sentence stored in the conversation sentence storage unit, and replacement from among the replacement candidates for the variable part of the searched first conversation sentence A replacement candidate determining step for determining a phrase; and an output step for replacing the determined replacement phrase with a variable part of the first conversation sentence and outputting the variable, and the search step uses the replacement phrase as a key. A second conversation sentence is retrieved from the conversation stored sentence, and the output step includes a variable part of the second conversation sentence when the second conversation sentence is a variable part that is the same as the first conversation sentence. Replace with the replacement word and output.
The conversation method of the present invention searches for a conversation sentence stored in a conversation sentence storage unit using a replacement candidate replaced with a preset variable part in a previous conversation sentence as a key, and a variable part in the retrieved conversation sentence. If there is, the replacement candidate used as the key is replaced with the variable part, and the subsequent conversation sentence is generated, so that it is possible to generate the conversation sentence to be output next while reducing the user's trouble.

  According to the conversation device and the conversation method of the present invention, it is predicted that a replacement candidate replaced with a preset variable part in the previous conversation sentence, or a conversation sentence after the previous conversation sentence based on the replacement candidate is predicted. By substituting the keyword to be replaced with the variable part in the subsequent conversation sentence, it is possible to reduce the user's trouble and generate the next conversation sentence to be output based on the keyword of the previous conversation sentence. .

  The conversation apparatus of the present invention is provided as one function of a mobile terminal, a mobile phone, a personal computer, or the like, or is configured as a single conversation apparatus. The conversation device of the present invention is equipped with the same hardware as that of a mobile terminal, a mobile phone, a personal computer and the like. That is, the conversation device includes an arithmetic processing unit (CPU), a main memory (RAM), a read only memory (ROM), an input device, an output device, an external storage device such as a hard disk, a magnetic disk driver, an optical disk driver, a CD driver, A DVD driver and a wired or wireless communication interface for communicating with an external network are provided. The ROM, the external storage device, the magnetic disk, the optical disc, the CD, or the DVD stores a program for causing the conversation device of the present invention to function or a program for causing a computer to execute the conversation method of the present invention. These programs, or programs acquired from the network via the communication interface, are read by the CPU into the main memory and executed, whereby the conversation device or the conversation method of the present invention is realized.

The conversation device of the present invention includes an input unit 1, a storage unit 2, a control unit 3, an output unit 4, and a clock 8. Further, a communication interface (not shown) is provided as necessary.
The input unit 1 is operated by a user operating a keyboard, a mouse, a pointer, and the like, and inputs sentences, words, instructions, and the like into the conversation device of the present invention. For the user's instruction, a STEP move button is prepared, and when this STEP move button is operated, the conversation apparatus performs processing to advance the processing flow from the current STEP to the next STEP. Alternatively, a button for proceeding with the process is prepared, and the process proceeds when the user operates this button. In addition to these, a microphone and a camera may be provided to input a conversation. In order to input a conversation sentence or a word from a microphone, it is necessary to provide a voice recognition device. Moreover, in order to input a conversation sentence or a word from a camera, it is necessary to provide a character recognition device.
The storage section 2 includes a conversation sentence storage section 21, a variable section information storage section 22, a conversation history storage section 23, a next input prediction storage section 24, a translation conversation sentence storage section 25, a translation variable section information storage section 26, and a next input prediction keyword. A list 27, a display list 28, and a program are stored. The conversation sentence storage unit 21, the variable part information storage unit 22, the conversation history storage unit 23, the next input prediction storage unit 24, the translated conversation sentence storage unit 25, and the translation variable part information storage unit 26 are hereinafter referred to as the conversation sentence database 21, variable. It may be called a part information database 22, a conversation history database 23, a next input prediction database 24, a translation conversation sentence database 25, and a translation variable part information database 26.
The control unit 3 includes a search unit 31, a determination unit 32, and a replacement unit (or replacement candidate determination unit) 33.
The above-described portions provided in the storage unit 2 and the control unit 3 are portions where the ROM 5 and the RAM 6 function when the CPU 7 reads a program and sequentially executes processing, and are not portions that actually exist. However, each of the above parts can be realized by hardware such as a logic circuit.
The output unit 4 may be a display device such as a liquid crystal display device, a plasma display device or an organic EL display device, a printing device such as a printer, or a signal output device such as a communication interface. Further, when the conversation device outputs a voice, a speaker is provided.
The clock 8 is a clock signal generator of the conversation device, measures the elapsed time by counting the clock of the clock 8, and outputs the second conversation sentence after the elapsed time set in the first conversation sentence. Alternatively, the second conversation sentence is output when the time set in the first conversation sentence is reached.

The conversational sentence storage unit 21 stores a plurality of conversational sentences composed of either a variable part or a fixed part, and the conversational sentence stores a plurality of conversational sentences for each scene in the order of conversation steps. In this case, a plurality of conversation sentences may be stored in each step.
The variable part information storage unit 22 sets a variable part in the conversation sentence, and stores replacement candidates to be replaced with the variable part. When a plurality of variable parts are set in one conversation sentence, replacement candidates to be replaced with the variable parts may have a co-occurrence relationship. Here, the co-occurrence relationship is said to be a co-occurrence relationship between parts when a part is changed and the other part is not changed and the sentence becomes inaccurate. The co-occurrence relationship describes how it is changed in the related part. Specifically, when the noun phrase of the variable part is changed from the singular to the plural, the a or an It is stored in the conversation sentence that the article is deleted or the drink counting method is changed from “a glass of” to “a cup of” when a hot drink is changed from a cold drink. Further, the co-occurrence relationship may be provided not only between the variable parts but also between the variable part and the fixed part.
The conversation history storage unit 23 stores the replacement candidates replaced by the variable unit in the order of conversation.
The next input prediction storage unit 24 stores keywords that are expected to appear in conversation sentences after the previous conversation sentence, based on the replacement candidate replaced with the variable part in the previous conversation sentence.
The search unit 31 searches the conversation sentence stored in the conversation sentence storage unit 21 using the keyword in the sentence input from the input unit or the input keyword as a key. The next input prediction keyword predicted to appear in the next conversation sentence based on the replacement candidate stored in the conversation history storage unit 23 or the replacement candidate stored in the next input prediction storage unit 24 is used as a key. A conversation sentence in the conversation sentence storage unit 21 is searched.
The determination unit 32 determines whether or not the variable unit is included in the searched conversation sentence.
The replacement unit (or replacement candidate determination unit) 33, when the determination unit 32 determines that the variable unit is included in the conversation sentence searched by the search unit 31, the replacement candidate used as the key or A replacement word / phrase is determined so that the next input prediction keyword is replaced with the variable part, and the replacement word / phrase is replaced with the variable part.
The translated conversation storage unit 25 stores a second language translation corresponding to the first language conversation stored in the conversation storage unit 21.
The translation variable part information storage unit 26 stores the second language translation corresponding to the first language replacement candidates stored in the variable part information storage unit 22.
The next input prediction keyword list 27 stores the next input prediction keyword obtained by searching the next input prediction storage unit 24.
The display list 28 stores conversation sentences obtained as search results.

The operation of the conversation apparatus of the present invention will be schematically described with reference to the sentence shown in FIG.
FIG. 2 (a) shows conversation sentences stored in the conversation step order by scene in the conversation apparatus of the present invention. FIG. 2B shows conversation sentences sequentially output by the conversation apparatus of the present invention based on the user's instruction.
(STEP 1) Now assume that the user selects the scene “Restaurant”. The conversation device outputs the conversation sentence of STEP 1 of the scene “Restaurant”. In the example of FIG. 2A, the scene “Restaurant” is selected, and there are a plurality of conversation sentences of STEP 1 such as “Please show me the menu” and “Please give me a cup of coffee”. In response to this output, the user selects “Please have a cup of coffee.”
(STEP 2) Then, the conversation device searches the database to obtain the next conversation sentence based on the keywords “coffee” and “one cup” of the selected sentence, and changes to STEP 2 in FIG. As shown in the figure, “Please give me a hot note” and “The ordered item has not come yet” is obtained and output. At the time of this output, the keyword in the conversation sentence selected in STEP 1 is reflected, and the “thing” in the sentence is replaced with “coffee” and outputted. That is, as shown in STEP 2 of FIG. 2B, “ordered coffee has not come yet” is output.
(STEP 3) After that, when the user instructs to input next as the last conversation sentence of the scene of this restaurant, as shown in STEP 3 of FIG. "I'll treat you." When outputting this, the conversation device reflects that the sentence containing “one cup” was selected in the conversation sentence of STEP1 prior to STEP3, “I will treat you”, “ "Let's check" is unnecessary, leave "Please give me an account" and output this.

STEP2 in FIG. 2 (a) is a case where the user has instructed the next input. By setting the designated time in advance, as shown in STEP2 in FIG. STEP2
Can be output. Although the designated time is set in FIG. 2C, the time may be set. In the case of STEP2 in FIG. 2 (c), the keyword in the conversation sentence selected in STEP1 is reflected and “Ordered coffee has not come yet” is output.

The conversation device according to the present invention operates as described above, and the details thereof will be described together with the operations of the respective portions of the present invention with reference to FIG. Since the conversation apparatus of FIG. 3 has a translation function, the case where a translation is simultaneously output will be described.
(Example 1)
The conversation sentence database 21 stores one or more conversation sentences for each scene and in the order of conversation, that is, in the order of STEP1 to STEPn. FIG. 3 shows a scene “Restaurant”. As shown in FIG. 3A, two sentence examples are stored in STEP1, two sentence examples are stored in STEP2, and three sentence examples are stored in STEP3. Scenes may be classified into scenes such as hotels, airports, shopping, and hospitals in addition to restaurants. Furthermore, each scene may be classified by type, class, user age, sex, and the like.
As shown in FIG. 3A, the conversation sentence database 21 has columns of sentence number, STEP, sentence example, translated sentence, keyword, non-display keyword, timer designated time [sec], and display sentence number after the designated time elapses. It is a list with the horizontal axis and sentence number on the vertical axis. This list is shown for explaining the invention, and may be stored in any form in the memory.

In the conversation apparatus of the present invention, the variable part and the fixed part are set in advance in the sentence example. The sentence number 0002 in FIG. 3A sets two variable parts. That is, <E0001> and <E0002>. The part other than the variable part corresponds to the fixed part. Replacement candidates for replacement with the variable part are registered in the variable part information database 22 in advance. As shown in FIG. 3B, the variable portion <E0001> is “coffee, tea,...”. The variable portion <E0002> is “one cup, two cups...”. When there are a plurality of variable parts in one sentence, a plurality of replacement candidates have a co-occurrence relationship.
The sentence number 0010 sets the variable part <E0003>. The replacement candidate to be replaced with the variable part is “ice, hot” as shown in the variable part database 22 of FIG.
The variable information database 22 also stores translations of replacement candidates. That is, the translation of the sentence number 0002 is “Can I have <T0002><T0001>.”, The translation of the variable part T0002 is “a cup, two cup…”, and the variable part <T0001> is “coffee, tea … ”.

  If the conversation text storage unit 21 has the list shown in FIG. 3A, it is assumed that the user has input “Please have a cup of coffee.” When the conversation sentence database 21 is searched using “coffee” and “one cup” in the sentence inputted by the user as keywords, the variable part <E0001> of the sentence number 0002 “<E0001> should be <E0002> please” is replaced with the candidate “ When “coffee” is selected and the variable portion <E0002> is “1 cup” as a replacement candidate, they match. That is, when the sentence number 0002 is replaced with “Please have a cup of coffee.”, It matches the sentence entered by the user. At the same time, a translation corresponding to this sentence number is acquired. Since the translation of the sentence number 0002 is “Can I have <T0002> <T0001>.”, The variable part <T0002> is “a cup of” and the variable part <T0001> is “coffee”. Therefore, the translation of sentence number 0002 is “Can I have a cup of coffee?”. As a result, as shown in FIG. 4, “Please have a cup of coffee” and “Can I have a cup of coffee?” Are displayed on the display unit.

The above is a case where the conversation sentence database 21 is searched based on the sentence input by the user, but another search method will be described.
First, when the user selects a scene, the conversation apparatus automatically outputs a STEP1 sentence example of the scene. When the scene “Restaurant” is selected, STEP 1 has sentence numbers 0001 and 0002, as shown in FIG. That is, “Please show me the menu.” And “Please <E0001> to <E0002>.” These two sentence examples are output and selected by the user. The sentence number 0002 is “<E0001> should be <E0002>”. If this display is difficult for the user to understand, the first replacement candidate is displayed. The fact that the first replacement candidate can be replaced may be displayed by flash, display color, character type, and font. That is, it is displayed that the first replacement candidate “coffee” of <E0001> and the first replacement candidate “1 cup” of <E0002> can be replaced.
If the user sees these two sentence examples and the user selects sentence number 0002, for example, the default is displayed in the variable part <E0001>, and the replacement candidate “coffee, tea...” Of the variable part <E0001> is displayed. Is done. Similarly, a default is displayed in the variable part <E0002>, and replacement candidates “one cup, two cups...” Of the variable part <E0002> are displayed. If there is only one conversation sentence in STEP1, there is no need to select a sentence.
When the user selects the replacement candidate “coffee” for the variable part <E0001> and the replacement candidate “1 cup” for the variable part <E0002>, the sentence number 0002 becomes “Please have a cup of coffee.” This is displayed. . At the same time, the translation of sentence number 0002 “Can I have a cup of coffee?” Is also displayed. (Fig. 4)

  As described above, the sentence example searched or selected by any procedure is stored in the conversation history database 23. FIG. 3C shows a storage example of the conversation history database 23. The horizontal axis includes a scene name, a STEP, a sentence number, a keyword, a timer designated time, and a display sentence column after the designated time has elapsed. The retrieved or selected sentence examples are sequentially stored on the vertical axis. Since FIG. 3C has one history, the scene name is “Restaurant” and the STEP is “1” as shown by the arrow (1) from FIG. 3A to FIG. 3C. , “0002” as the sentence number, “E0001 = coffee, E0002 = 1 cup” as the keyword, “600” as the timer designated time, and “0011” as the displayed sentence after the designated time has elapsed. In this way, the replacement candidate replaced in each variable part is stored as a keyword. In this case, the keywords are E0001 = coffee and E0002 = 1.

Next, a conversation sentence to be output when an input instruction is received from the user is predicted in advance.
That is, the next input prediction database 24 is searched using all the keyword information stored in the conversation history database 23, in this case, “E0001 = coffee, E0002 = 1 cup” as a key. The next input prediction database 24 is a database prepared in advance in the conversation device of the present invention. As shown in FIG. 3D, when coffee is a keyword, the next conversation sentence is based on this coffee. Keywords that are expected to appear in are registered in advance. That is, for coffee, ice and hot are predicted to be included in the next conversation. For tea, ice and hot are predicted to be included in the next conversation. If the keyword is full, one person is predicted to be included in the next conversation sentence. Similarly, keywords predicted to be included in the next conversation sentence are collected and registered with respect to the keywords of the previous conversation sentence.
In this example, since “coffee” and “one cup” are keywords, when the next input prediction database is searched using this as a key, keywords predicted to be included in the next conversation sentence are “ice” and “hot”. , "One person" is obtained. This is indicated by two arrows (2) from FIG. 3 (c) to FIG. 3 (d). These keywords are stored in the next input prediction keyword list 27. The next input prediction keyword list 27 is shown in FIG. This is indicated by the arrow (3) from FIG. 3 (d) to FIG. 3 (e).

Next, the conversation sentence database 21 is searched using each of the keywords “ice”, “hot”, and “one person” stored in the next input predicted keyword list 27 as keys. The search searches the keyword column of the conversation sentence database 21. The search object here is an example of a STEP sentence after the sentence “Please have a cup of coffee” that was input last time (that is, STEP 1), and the variable part is a conversation sentence in which the next input prediction keyword list 27 is expanded. That is, a search is made assuming that “ice”, “hot”, or “one person” is included in the variable part of STEP2, <E0003>.
The search is now for “ice”, “hot”, “one person” and hits E0003 = ice, hot ... ”, so sentence number 0010“ <E003> having this E0003 as a keyword. "Please give me a request." And the translated sentence "<T0003><T0001>,please." With the sentence number 0010 is obtained as a search result. This is indicated by the arrow (4) from FIG. 3 (e) to FIG. 3 (a).

Next, the above search results (ie, “Please do with <E0003>” and the translated text “<T0003><T0001>, please.”) And the previous conversation “Please give me a cup of coffee.” And “Can I have a cup of coffee?” Is displayed.
If there is a variable part in the sentence, a list box is displayed near the variable part so that the user can select a replacement candidate from the list box.
As a result of the search, when there are a plurality of sentences, they may be displayed side by side on the screen. At this time, the search result may be displayed at the top of the screen, or the color may be changed or blinked.
In the above description, the conversation device predicts the next conversation sentence to be output in advance, and when the user gives an instruction, the conversation sentence predicted in advance is output. You may start.

Moreover, the conversation apparatus of this invention is good to memorize | store the style of the translation sentence output previously, and to change the style of subsequent translated sentences according to the style.
That is, if the style of the translated text that was output before is a normal style such as “Can / Will you…?” Or “May I…?”, The translation after this translation will be “Can / Will you” …? "Or" May I…? "
However, if the style of the translation that was output before is a polite style like “Could / Would you…?”, The translation after this translation will change to “Could / Would you…?” Let
In addition, if the style of the translated text that was output before was a regular style, such as “I want to…” “I going to…” or “I go to…”, the translation after this translation will be “I want to…” Want to ... "" I going to ... "or" I go to ... "
However, if the style of the translation that was output before was a colloquial style such as “I wanna…” “I'm gonna…”, the translation after this translation would be “I wanna…” “I'm Use a colloquial style such as “gonna…” or “I gotta…”.

(Example 2)
FIG. 5 is a diagram for explaining the second example. FIG. 5 shows that there are three STEP1 sentence examples, the variable part of the sentence number 0011 of STEP2 is <E0005>, and the replacement candidates for the variable part name E0005 in the variable part database 22 are E0001 and E0003. Unlike 3, the others are the same. In this example 2, the sentence number 0011 has the variable part <E0005> in the sentence expanded based on the variable part information database, and “ordered <E0005> has not yet arrived” is “ordered <E0001>, <E0003> has not come yet. " That is, the variable portion <E0005> can be replaced by either of the replacement candidates <E0001> and <E0003>.
For example, assume that there are three sentence examples of STEP1 and the user selects sentence number 0002. As a result, <E0001>, <E0002> are stored as keywords in the conversation history database 23. Since the variable part <E0005> of the sentence number 0011 is replaced with the keyword <E0001>, “the ordered coffee has not yet come”.
In this way, the replacement candidate selected by the user can be reflected in the sentence number 0011 of STEP2. This example 2 explains that the variable part of the STEP2 sentence example is replaced by the selection of STEP1, and is stored in the conversation history database 23 without using the next input prediction database 24 as in the example 1. A case where STEP2 sentence examples were searched using the above keywords was explained.
In addition, the explanation of the above example 2 is the case when one sentence is selected in STEP1, but after "Please give me a cup of coffee." The keyword E0001 in the conversation history database stores “coffee” and “tea”. Therefore, in this case, there are two replacement candidates for the variable part <E0005> of the sentence number 0011, “coffee” and “tea”, and the user can select a plurality of candidates.

(Example 3)
FIG. 6 is a diagram for explaining the third example. FIG. 6 is substantially the same as FIG. 3, but FIG. 6 illustrates a case where the timer designated time set in the keyword is activated.
That is, when the sentence number 0002 is selected or searched in STEP 1 and stored in the conversation history database 23, the timer set for the keyword starts. Here, 600 sec is set. Then, the clock of the clock 8 is counted, and the designated sentence number 0011 is displayed after the designated time has elapsed. That is, after 600 seconds, sentence number 0011 “Ordered <E0001> has not yet arrived” is obtained, and the variable part <E0001> is replaced with the replacement candidate “Coffee”, and “Ordered coffee has not yet arrived. "And the translation"I'm still waiting for the <T0001>. "

(Example 4)
FIG. 7 is a diagram for explaining the fourth example. FIG. 7 displays a conversation sentence designated after a designated time has elapsed in accordance with time information in the selected or searched sentence.
That is, as shown in FIG. 7A, it is assumed that the sentence number 0101 is selected or searched in STEP1. The sentence number 0101 is “We are waiting at <E0011> after <E0010> minutes”. The keywords of this sentence number are “E0011” and “waiting”. The timer designated time is “E0010 × 60”.
This sentence example is stored in the variable part information database 22 as shown in FIG. That is, the replacement candidate of the variable part name E0010 is an arbitrary integer. Replacement candidates for the variable part name E0011 are Osaka Station, Kyoto Station,.
Then, it is stored in the conversation history database 23 as shown in FIG. In other words. Keyword E0011 = Osaka Station is waiting for you. The timer specified time is E0010 × 60 = 1800. Therefore, as sentence number 0101, “waiting at Osaka Station after 30 minutes” is entered. The display sentence numbers after the designated time elapse are 0110 and 0111. When saved in the conversation history database 23, the clock of the clock 8 starts counting, and when the specified time elapses, display sentences 0110 and 0111 are displayed. In other words, there are two sentences, “Would you like to get to Osaka Station yet?” And the translation “Have we arrive at Osaka yet?”, “Why are you late?” And the translation “What's the reason of delay?” Is displayed.

(Example 5)
FIG. 8 is a diagram for explaining the fifth example. FIG. 8 displays a sentence specified after a specified time according to time information in the selected or searched sentence.
That is, as shown in FIG. 8A, it is assumed that the sentence number 0201 is selected or searched in STEP1. Sentence number 0201 is “Waiting for train <E0030> hour <E0031> minutes”. The keyword of this sentence number is “E0030 time and E0031 minutes, train”. The timer designated time is “E0030 time E0031 minutes”.
This sentence example is stored in the variable part information database 22 as shown in FIG. That is, the replacement candidate for the variable part name E0030 is an integer from 0 to 24. The replacement candidates for the variable part name E0031 are integers from 0 to 59.
And it is preserve | saved in the conversation history database 23 as shown in FIG.8 (c). In other words. The keywords E0011 = 10 and E0031 = 15 are trains. Therefore, 0201 is input “Waiting for a train departing at 10:15”. The timer designated time is E0030 hours and E0031 minutes, and the display sentence numbers after the designated time elapse are 0210 and 0211. When saved in the conversation history database, the time of the clock 8 is watched, and display sentences 0210 and 0211 are displayed when the designated time has elapsed. In other words, at 10:15, “<E0030> Hours <E0031> Can the train leave yet?” And the translation “Has the <T0030>: <T0031> train come yet?” Two sentences are displayed: "Are you late?" And the translation "What's the reason of delay?"

Examples 1 to 5 described above are processed and implemented as in the following flowchart.
9 to 13 are flowcharts for executing the processes of the first to second examples.
When the conversation apparatus of the present invention starts operation, first, in step S1, it is determined whether or not there is a scene selection by the user. If there is no scene selection by the user, the conversation device ends its operation. If there is a selection, the process proceeds to step S2 to display a STEP1 sentence example of the selected scene. If there are multiple STEP1 sentence examples, all of the multiple sentence examples are displayed. Next, in step S3, it is determined whether or not there is a desired conversation sentence in the conversation sentence displayed to the user. If it is determined that there is no sentence example desired by the user, the process proceeds to step S4, where an input from a button or the like for changing the STEP displayed on the screen is received, and it is determined whether or not the user has instructed to change the STEP displayed. If the user gives an instruction to change the displayed STEP, the current display content is cleared in step S5. Then, in step S6, a STEP sentence example designated by the user is displayed. Then, it returns to step S3.

  If it is determined in step S4 that there is no instruction to change the displayed STEP, the process proceeds to step S7 to determine whether or not the user input method is a keyword. That is, if a keyword has been input, the process proceeds to step S8 to accept the keyword input. Next, in step S9, a sentence example search using keywords is performed. This sentence example search will be described later with reference to the sentence example search flowchart shown in FIG. If a sentence has been input, the process proceeds to step S10 and accepts an input of a free sentence. Next, in step S11, a sentence example search using a sentence is performed. This sentence example search will be described later with reference to the sentence example search flowchart shown in FIG. The search result of step S9 or step S11 is displayed in step S12. Then, it returns to step S3.

When the user determines that there is a desired sentence example in step S3, the process proceeds to step S13, and the user selects the desired sentence example. If there is only one sentence in STEP1, step S13 may be skipped. If there is a variable part in the sentence, the variable part is displayed as a list box and the replacement candidate is displayed. Then, the user is made to select a replacement candidate. Next, in step S14, the translation of the selected sentence is displayed. In step S15, the conversation history is updated. This process will be described later with reference to FIG.
Next, it is determined whether or not there is a next input instruction from the user. If there is no instruction, the process returns to step S16. When the user gives a next input instruction, in step S17, it is determined whether or not the selected sentence is a sentence of the final STEP of this scene. If it is the final STEP sentence, the process returns to step S1. If it is not the final STEP sentence, the process proceeds to step S18 to predict the next input. This step S18 will be described later with reference to FIG. In step S19, the prediction result is displayed, and the process returns to step S3. When displaying a prediction result in step S19, it is good to memorize | store the style of the translation sentence output previously, and to change the style of subsequent translation sentences according to the style. That is, it stores whether the style of the translation sentence output previously is a normal style, a polite style, or a colloquial style, and the subsequent output translation is styled so as to match the previous style.
In the flowchart of FIG. 9, in step S <b> 16, the process after step S <b> 16 instructed by the user is performed and the prediction result is displayed, but the process of step S <b> 16 is skipped and as described in Example 1 above, The next conversation sentence may be predicted in advance before the user's instruction. Or in step S18, a next input prediction process may be implemented and a prediction result may be displayed by a user's instruction | indication.
Furthermore, although the above description is a case where sequential processing is performed, a move button of the processing flow may be prepared so that an instruction to return from the current processing STEP to the previous processing STEP may be given.

  Details of the sentence example search in step S9 are as shown in FIG. That is, the keyword column in the conversation sentence database 21 is searched using the input keyword as a key. If there is a variable part at the time of this search, the variable part information database 22 is referred to and the variable part is expanded and searched. When there are a plurality of input keywords, either an AND search including all the input keywords or an OR search including any of the input keywords is performed.

  Details of the sentence example search in step S11 are as shown in FIG. That is, the sentence example column in the conversation sentence database 21 is searched using the inputted sentence as a key. If there is a variable part at the time of this search, the variable part information database 23 is referred to and the variable part is expanded and searched. In the case of this search, it is basic to extract sentences that exactly match the input sentence, but morphological analysis is performed from the input sentence so that most of the sentences that are not exact matches can also be extracted. Then, keywords may be cut out, and sentences containing these keywords in the sentence may be searched.

  Details of step S15 are shown in FIG. That is, the scene name to which the selected sentence belongs, the STEP number, the sentence number, and the keyword in the sentence are stored in the conversation history database 23. If there is a variable part in the sentence, each variable part in the sentence and a replacement candidate replaced with each variable part are also stored in the conversation history database 23. By storing the conversation history in this way, the words assigned to each variable part in the sentence are also recorded as a history in order to use the previously input contents for the next input prediction.

  Details of the next input prediction process in step S18 are shown in FIG. That is, in step S18-1, the next input prediction database 24 is searched using each keyword in the conversation history database 24 as a key. In this search, how far the information stored in the conversation history database 23 is used for prediction of the next input is (1) use all the history remaining in the conversation history database 23, (2) specify in advance Use the history up to the number of times before (X = 1, 2,...), (3) Change how many times the history is used depending on the number of hits, (4) Set the threshold for the number of hits in advance If the number of hits in the search is small, increase the number of times the history is hit. If there are many hits in the search, reduce the number of history. There is a method in which the number of hits is changed in this way so that the number of hits becomes a preset threshold value, and which method is used is determined in advance. When the search result is obtained, the next input predicted keyword list 27 stores the next input predicted keyword.

Next, in step S18-2, using each word in the next input prediction keyword list 27 as a key, the keyword column is searched from the conversation sentence database 21 in which the STEP number +1 of the sentence selected last time is hit. Store the sentence in the display sentence list. The sentence hit by the keyword stored in the non-display keyword column in the conversation sentence database 21 is not displayed, and is not stored or added to the display sentence list.
For example, in the restaurant scene of FIG. 3, if it is assumed that the user inputs “Please give me a cup of coffee.”, The sentence number 0002 “<E0001> should be <E0002>.” Is hit, and E0001 = coffee, E0002. = 1 cup.
According to the prediction of the next input, STEP2 “I need your help at <E0003>” and “I have not ordered <E0001> yet” respectively. Or it is displayed on the screen as “The ordered coffee has not yet come.”
Next, assuming that “Please do it hot”, it is displayed on the screen together with the translation of the selected sentence, and information on the selected sentence is stored in the conversation history storage unit 23.
In the next input prediction step S18-2, the next input prediction storage unit 24 is searched using the keywords (in this example, “coffee”, “1 cup”, “request”) stored in the conversation history storage unit 23 as keys. The hit “ice”, “hot”, and “one person” are stored in the next input prediction keyword list 27. The conversation sentence storage unit 21 is searched using “ice”, “hot”, and “one person” stored in the next input prediction keyword list 27 as keys. At this time, “one person” is displayed in the non-display keyword column. Sentence number 0021 “Let's make a split” and sentence number 0022 “I'm going to pay” are hits, but these two sentences are not displayed on the screen because they are hits with hidden keywords.

Next, in step S18-3, the conversation sentence database 21 is searched for a sentence belonging to the STEP having the STEP number +1 of the previously selected sentence, and the hit sentence is stored in the display sentence list 28. The sentence hit by the keyword stored in the non-display keyword column in the conversation sentence database 21 is not displayed, and is not stored or added to the display sentence list.
For example, assume the conversation sentence storage unit 21 in FIG. (FIG. 16 is based on FIG. 5 and adds sentence number 0000 and changes the STEP of each sentence.)
If the variable part E0000 = 1 of sentence number 0000 “<E0000> people but do you have seats?” Is selected and “Is there one person but do you have seats?” Is selected and displayed on the screen together with the translation, select The sentence information is stored in the conversation history storage unit 23.
Next, in STEP2 sentence number 0002 “Please set <E0001> to <E0002>.” E0001 = Coffee, E0002 = 1 cup, “Please give me a cup of coffee.” STEP3 sentence number 0010 “ Suppose that you have selected E0004 = Hot, “Please give me hot”. By the input so far, the conversation history storage unit 23 becomes as shown in FIG.
Consider performing next input prediction from this state. In step S18-3, when a sentence belonging to the previous input STEP + 1, that is, a sentence belonging to STEP4, is searched, sentence number 0020 “Please give me an account”, sentence number 0021 “Let's make a split account”, sentence number 0022 “I The sentence “0021” is hit. At this time, since the keyword “1 person” exists in the conversation history storage unit 23, the sentence number 0021 and the sentence number having “1 person” as a non-display keyword are present. 0022 is not displayed on the screen.
Note that steps S18-1 and S18-2 may be skipped if the search is difficult in Example 1. Similarly, Steps S18-1 and S18-2 may be skipped in the case of Example 3 to Example 5.

  Next, in step S18-4, it is determined whether the sentence in the display sentence list 28 has a variable part. If there is no variable part, the next step S18-5 is skipped. If there is a variable part, in step S18-5, the conversation history database 23 and the variable part information database 22 are referred to and the related variable part is replaced with a word assigned in advance. Variable parts that are not assigned in advance are replaced with default words preset in each variable part.

Next, a flowchart for executing the processes of Examples 3 to 5 will be described with reference to FIGS.
The flowchart of FIG. 14 is different in that there are steps S21 to S23, and the other steps are the same as the flowchart of FIG. That is, there are steps S21 and S22 between steps S15 and S16. If the determination in step S17 is Y, the process of step S23 is performed, and then the process returns to step S1.

A step S21 decides whether or not the timer designated time of the selected sentence is greater than zero. If the timer designated time is 0, step S22 is skipped and the process proceeds to step S16. If it is greater than 0, the process of step S22 is performed. The process of step S22 will be described later with reference to FIG.
Step S23 is a process for resetting the timer. After the reset process, the process returns to step S1.

FIG. 15 is a flowchart illustrating step S22. The processing in step S22 is performed in parallel with the main flow shown in FIG. That is, it is started as a thread. Therefore, when the timer reset is instructed in the main flow, the process of step S22 is ended.
The first step S22-1 of the timer process is a timer count start. The timer count starts when a sentence is stored in the conversation history database 23. Next, in step S22-2, it is half-determined whether the timer designated time has elapsed. If not, the process proceeds to step S22-3, and it is determined whether there is a timer reset instruction. If there is no reset instruction, the process returns to step S22-2 to continue the timer count. If there is a timer reset instruction, the process skips to step S22-6 and executes a timer reset process.
If the timer elapsed time has elapsed in step S22-2, the process proceeds to step S22-4, the conversation sentence database 21 is referred to, and the sentence to be displayed after the specified time has elapsed is stored in the display list. Next, in step S22-5, the minutes in the display sentence list 28 are output to the display unit. In step S22-6, a timer reset process is performed.

As shown in FIGS. 5 and 6 in Examples 1 and 2 above, the statement designation numbers 0002 and 0003 have timer designation times set, but the flowcharts in Examples 1 and 2 (FIG. 9) show the timer designations. There is no processing time step S21. Therefore, timer processing is not performed in Examples 1 and 2.
Predict the next conversation sentence in advance as in Examples 1 and 2 above, as in Examples 3 to 5, in addition to predicting the next conversation sentence when the time is set, it is equipped with a temperature sensor, The next conversation sentence may be predicted according to the output of the temperature sensor. That is, in the restaurant, when it is detected that the temperature is above a certain level, it is preferable to display to order a cold drink, and when it is detected that the temperature is below a certain level, it may be displayed to order a warm drink.

It is a block diagram of the conversation apparatus of this invention. It is a figure which illustrates roughly the flow of the process which the conversation apparatus of this invention implements. It is a figure explaining the process of Example 1 with the conversation apparatus of this invention. It is a figure which shows the example of a display of the conversation apparatus of this invention. It is a figure explaining the process of Example 2 by the conversation apparatus of this invention. It is a figure explaining the process of Example 3 by the conversation apparatus of this invention. It is a figure explaining the process of Example 4 with the conversation apparatus of this invention. It is a figure explaining the process of Example 5 with the conversation apparatus of this invention. It is a 1st flowchart figure of the conversation apparatus of this invention. It is a flowchart figure which shows the process of step S10 in the flowchart of FIG. It is a flowchart figure which shows the process of step S11 in the flowchart of FIG. It is a flowchart figure which shows the process of step S15 in the flowchart of FIG. It is a flowchart figure which shows the process of step S18 in the flowchart of FIG. It is a 2nd flowchart figure of the conversation apparatus of this invention. It is a flowchart figure which shows the process of step S22 in the flowchart of FIG. FIG. 14 is a flowchart for explaining a case where a keyword stored in a non-display keyword column is hit in step S13-3 of FIG.

Explanation of symbols

2 Storage unit 3 Control unit 8 Clock 21 Conversation sentence storage unit (database)
22 Variable part information storage part (database)
23 Conversation history storage (database)
24th order input prediction storage (database)
25 Japanese translation memory section (database)
26 Translation variable part information storage part (database)
27 next input prediction keyword list 28 display list 31 search unit 32 determination unit 33 replacement unit

Claims (10)

A conversational sentence storage unit for storing a plurality of conversational sentences composed of either a variable part or a fixed part;
A variable part information storage unit for storing replacement candidates to be replaced with the variable part;
A search unit for searching for a first conversation sentence stored in the conversation sentence storage unit;
A replacement candidate determining unit that determines a replacement phrase from among the replacement candidates of the variable part of the first conversation sentence searched by the search unit;
An output unit that replaces the replacement phrase determined by the replacement candidate determination unit with the variable part of the first conversation sentence and outputs the replacement phrase;
The search unit searches for a second conversation sentence from the conversation storage sentence using the replacement phrase as a key,
The conversation device, wherein the output unit replaces the variable part of the second conversation sentence with the replacement phrase when the searched second conversation sentence includes the same variable part as the first conversation sentence.   The conversation apparatus according to claim 1, wherein the conversation sentences stored in the conversation sentence storage unit are classified according to scenes, and further, step numbers classified for each conversation step are stored in association with the conversation sentences. Further comprising a next input prediction storage unit that associates and stores the replacement candidate of the variable part and a keyword that is expected to be included in the conversation sentence of the next step of the conversation sentence having the replacement candidate,
The search unit searches the replacement word from the replacement candidate word in the next input prediction storage unit, acquires a keyword corresponding to the searched replacement word from the next input prediction storage unit, and uses the acquired keyword as a key The conversation device according to claim 1, wherein the second conversation sentence stored in the conversation sentence storage unit is searched. It is further equipped with an input part for inputting text and voice,
4. The search unit according to claim 1, wherein the search unit searches the first conversation sentence or the second conversation sentence stored in the conversation sentence storage unit using the sentence input from the input unit or the input keyword as a key. 2. The conversation device according to item 1.   In the case where there is a plurality of variable parts in one stored conversation sentence, the conversation sentence storage unit is replaced with a predetermined variable part replacement candidate and another variable part or fixed part that is changed corresponding to the replacement candidate. The conversation apparatus according to claim 1, wherein a co-occurrence relationship is stored with a replacement candidate for the item.   The conversation device according to claim 1, wherein the output unit outputs the first conversation sentence or the second conversation sentence generated by the replacement when instructed by the input unit.   The said conversational sentence memory | storage part has matched and memorize | stored the memorize | stored conversational sentence, the time information which advances to the conversational sentence of the next step of this conversation sentence, and the next step number of the next step. 2. The conversation device according to 2. A measuring means for measuring a time from when the first conversation sentence is output to the output unit;
When the time information of the first conversation sentence searched by the search unit is acquired and it is determined that the time information and the time measured by the measuring unit exceed the time information, the search unit 8. The conversation device according to claim 7, wherein a second conversation sentence having a next step number is searched from the conversation stored sentence using a next step number stored in association with a sentence as a key. Furthermore, a translated conversation sentence storage unit that stores a translation sentence of a second language corresponding to the conversation sentence of the first language stored in the conversation sentence storage unit, and a first language stored in the variable part information storage unit A translation variable section information storage section for storing a translation of the second language corresponding to the replacement candidate;
The conversation device according to claim 1, wherein the output unit outputs a translated sentence in a second language when outputting the conversation sentence generated by the replacement. A conversational sentence storage unit that stores a plurality of conversational sentences composed of either a variable part or a fixed part, and a variable part information storage part that stores replacement candidates to be replaced with the variable part,
A search step for searching for the first conversation sentence stored in the conversation sentence storage unit; and a replacement candidate determination step for determining a replacement phrase from among the replacement candidates of the variable part of the searched first conversation sentence;
An output step of replacing the determined replacement phrase with the variable part of the first conversation sentence and outputting the variable part;
The search step searches for a second conversation sentence from the conversation storage sentence using the replacement phrase as a key,
The said output step WHEREIN: When there exists a variable part same as the said 1st conversation sentence in the searched 2nd conversation sentence, the conversation method which replaces the variable part of the said 2nd conversation sentence with the said replacement phrase, and outputs it.

Images