JP2001141500A - On-vehicle agent process system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a car driver with the natural language information full of flexibility in a simple manner. SOLUTION: A voice-sound information including voice, as a signal received with a wireless communication part 1 is processed in a syntactic analyzing part for semantic interpretation after voice recognition while an e-mail is processed in the same part as it is, so that the information related to the operation of various on-vehicle units in a car is selected for summarizing with priority, which is outputted from a speaker 5 after voice-sound synthesization. If various on-vehicle units are estimated to require operation, that estimation is automatically judged at an agent process part 4, for voice-guiding which is added with the operation guide information for the on-vehicle unit.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle equipped with a communication terminal for communicating with the outside, which provides information desired by an occupant by voice output based on information transmitted through the communication terminal. It relates to an agent processing device.

[0002]

2. Description of the Related Art With the remarkable development of information communication technology in recent years, various information has been provided to occupants by information communication with the outside in the field of automobiles.

For example, a car navigation system is
In a map previously stored in a predetermined recording medium such as a CD-ROM, the position of the own vehicle is grasped every moment based on a GPS signal transmitted from a satellite and the like, and an optimum route to a travel destination is automatically searched. It is used as a system that provides useful information on road driving.

[0004] An intelligent transportation system (Intell.
igent Transport Systems (hereinafter referred to as "ITS")
Is to provide occupants with various road traffic information such as traffic congestion information and automatic toll collection on toll roads using information and communication technology, etc., to eliminate accidents and traffic congestion, and to save energy and coexist with the environment. It is expected that its development will continue.

[0005] The provision of such information is generally displayed as image information or audio information and shown to the occupant. Information provided from the outside is provided as a regular data string generated according to a predetermined data format. It is specified in advance.

Further, as communication for providing information to occupants, there is a conversation type using a wireless communication system such as a hands-free portable telephone. According to this method, an occupant of a car makes various questions to the outside without grasping the handset by hand, and a clerk (operator) at a remote place responds to the question in a flexible manner. Offers can be made. That is, in this wireless communication system, the real voice of the conversation partner or the like is directly given as audio information.

[0007] In addition, it is also possible to use an in-vehicle computing system such as a so-called in-vehicle personal computer to receive information provided by e-mail from outside through wireless communication using a mobile phone or the like.

[0008]

The information provided by the car navigation system or the ITS in the various fields of automobile communication as described above must be provided in a predetermined format of a specific format. It is necessary to set the data in a specific format in advance and prepare it. For this reason, it is extremely difficult to transmit information that is not prepared as data in a unique format.

On the other hand, when information is provided by a hands-free wireless communication system or e-mail, the occupant can receive the information in a so-called natural language format. Compared to this, it is possible to secure the flexibility of information, but since the information is indefinite, often unnecessary information is given in a mixed manner, and all such information is presented to the occupants in the car Despite the necessity, it is not desirable in terms of information efficiency to transmit much information as it is. Therefore, it is desirable to summarize necessary parts from the information in the natural language and reconstruct the information so that it is easy for the occupant to understand.

[0010] Therefore, an object of the present invention is to convert real voice information or e-mail in a natural language with high flexibility transmitted by wireless communication into a form that can be summarized into a format that can be easily understood by a vehicle occupant, and It is an object of the present invention to provide an in-vehicle agent processing device that allows an occupant to additionally process various necessary information contents and transmit the information by voice output so as to be invisible.

[0011]

Means for Solving the Problems In order to solve the above problems,
According to the first aspect of the present invention, in a vehicle equipped with a communication terminal that communicates with the outside, an in-vehicle that provides information desired by an occupant by voice output based on information transmitted through the communication terminal. An agent processing device, a wireless communication unit that receives a wireless signal provided from the outside, and various on-vehicle units in a vehicle that perform syntax analysis on natural language information in the wireless signal provided from the outside through the wireless communication unit An agent processing unit for selecting and summarizing information regarding the information and reconstructing it into character information;
A voice synthesizing unit for converting the character information reconstructed by the agent processing unit into voice and outputting the voice;

According to the second aspect of the present invention, the agent processing unit includes, in the information on the various on-vehicle units, information on the occupant which is a factor for newly operating the on-vehicle units. If so, the character information is reconstructed by including the operation guide information for instructing the occupant to operate the in-vehicle unit and determining the fact.

According to the third aspect of the present invention, the natural language information includes voice information including a real voice, and further includes voice recognition means for voice-recognizing the voice information and converting it into character information. .

According to the fourth aspect of the present invention, the natural language information is an electronic mail including character information.

According to the invention described in claim 5, the various on-vehicle units include a car navigation system for performing a recommended route search for an arbitrary travel destination, and the car navigation system includes information on the various on-vehicle units as information. When new information on the travel destination of the car is included, a recommended route search is executed based on the new information.

In this specification, natural language information includes both text data, which is character information mixed with kanji and kana, and voice information, such as a conversation with a real voice.

[0017]

DESCRIPTION OF THE PREFERRED EMBODIMENTS <Structure> FIG. 1 is a block diagram showing an on-vehicle agent processing device according to an embodiment of the present invention. This in-vehicle agent processing device extracts the contents of an e-mail in natural language given from the outside by radio, etc. by performing syntax analysis, and extracts the voice data given from the outside by radio to text data by voice recognition. After the content is converted to syntax, the content is subjected to syntax analysis, and various information is additionally processed based on the content, and then output as audio information to the occupant.

More specifically, as shown in FIG. 1, the on-vehicle agent processing device includes a radio communication unit 1 for performing radio communication with the outside on real voice information and various character information as an analog signal, An e-mail transmission / reception device 2 for transmitting / receiving e-mail to / from the outside through the communication unit 1;
Speech recognition / synthesis device 3 that performs voice recognition of externally provided real voice information through wireless communication unit 1 to convert it into text data (character information) and converts text data provided from agent processing unit 4 into voice data, which will be described later.
Then, the contents of the e-mail received by the e-mail transmission / reception device 2 and the text data converted by speech recognition by the speech recognition / synthesis device 3 are subjected to syntax analysis and keyword analysis to extract the contents, and various information is added and processed. And an agent processing unit 4 that outputs the data to the speech recognition / synthesis device 3.
In addition, the wireless communication unit 1, the transmitting / receiving device 2, the voice recognition / synthesis device 3
The agent processing unit 4 is communicatively connected to a bus via a local area network (LAN) in the vehicle, and communicates with the local area network (LAN), like various on-vehicle units such as the car navigation system 7. Among the signals, signals are transmitted and received by an existing multiplex communication method such that a signal is selected on the receiving side so as to permit input of only signals meeting predetermined conditions.

The wireless communication unit 1 uses a general portable telephone that performs wireless communication via the antenna 1a, connects to a general public line through a wireless communication station installed in each region, and connects to a general public line through the general public line. It is possible to receive information from service bureaus and the like.

As the e-mail transmitting / receiving device 2, a so-called general in-vehicle personal computer (personal computing system) is used. It is possible to send and receive e-mails.

The speech recognition and synthesis device 3 is a circuit having both a speech recognition function and a speech synthesis function.

As shown in FIG. 2, the voice recognition function of the voice recognition / synthesizing device 3 is such that the frequency analysis unit 11 analyzes the frequency of a voice signal in a natural language of a real voice received through the wireless communication unit 1, and this is analyzed by the phoneme decomposition unit 12. After the phoneme decomposition, the word recognition unit 13 performs word recognition in order to reduce the error rate, and the sentence recognition unit 14 executes sentence recognition to convert the data into text data similar to an electronic mail. .

The frequency analysis unit 11 divides the audio signal supplied from the wireless communication unit 1 into frames of several ms to several tens of ms, and performs a spectrum analysis on each frame using a fast Fourier transform or the like. Then, after converting the obtained spectrum into a parameter based on the auditory scale, the noise is removed and output to the phoneme decomposition unit 12.

The phoneme decomposing unit 12 decomposes the spectrum data supplied from the frequency analysis unit 11 into phonemes such as consonants and vowels while referring to a phoneme model 15a in a predetermined dictionary file 15. Here, for example, since the feature of the consonant is in the pattern of the time change of the spectrum, the phoneme model 15 in the dictionary file 15 is used.
As a, a voice parameter expressed in time series is used. The phoneme model 15a includes a vector quantization (VQ) for obtaining a representative pattern that minimizes distortion in a category, a learning vector quantization (LVQ) for obtaining a representative pattern so as to preserve boundaries between categories, and a time series. Using a Hidden Markov Model (HMM) in which a pattern is converted into a Markov model, a pattern learned from a large number of data is used. However, since the phoneme decomposition in the phoneme decomposition unit 12 is for an unspecified speaker, the phoneme recognition rate is about 60% to 80%. For example, the probability that a word composed of 8 phonemes is correct is 41% ( Since it is considered to be less than (80% of the fourth power), the result of the phoneme decomposition here requires further word recognition in the word recognition unit 13 and sentence recognition in the sentence recognition unit 14.

The word recognizing unit 13 checks the word model 15b in the dictionary file 15 for each phoneme decomposed by the phoneme decomposing unit 12, calculates the degree of coincidence between the two, and selects a word having a high degree of coincidence. To recognize words.
Here, as the word model 15b in the dictionary file 15, a model that takes into account the deformation of phonemes such as vowelization, vowelization, nasalization, and palatalization of consonants in a word is prepared.

The sentence recognizing unit 14 selects a word sequence that matches a language model (model of word-word connection or grammar) 15c in the dictionary file 15 from the result of the word recognition by the word recognizing unit 13. Has become.

On the other hand, the speech synthesis function of the speech recognition / synthesis device 3 adopts a text speech synthesis system for synthesizing speech from text data which is character information mixed with kanji or kana, and as shown in FIG. It comprises a language processing unit 21 for converting character information containing mixed Chinese characters and kana into a phonetic character string, and an acoustic processing unit 22 for performing regular speech synthesis to convert the phonetic character string into a speech waveform.

The language processing unit 21 searches the word dictionary 15d in the dictionary file 15 to obtain the reading and accent of the word, and determines the reading and accent of each word in the text data based on the search result. Further, the intonation of the sentence is determined from the dependency relation of each phrase, and a phonogram string storing information on the phonogram string, word reading, accent and intonation of the sentence is generated.

The sound processing section 22 reads out the waveform of each phoneme from the waveform dictionary 15e in the dictionary file 15 based on the phonetic character string provided from the language processing section 21, and assembles a speech waveform. . The sound processing unit 2
The audio waveform assembled in 2 is output as audio through the speaker 5 shown in FIG.

The agent processing unit 4 grasps the semantic contents of the given information while reflecting the modularity of knowledge necessary for natural language processing using, for example, a hierarchical model method, and then converts various information into text data. As shown in FIG. 4, the syntax analysis unit 31 analyzes the structure of the text data of the e-mail given from the e-mail transmission / reception device 2 and the text data recognized by the speech recognition / synthesis unit 3 by speech. After the information determined and analyzed here is interpreted by the semantic analysis unit 32, various information is added and edited by the information addition and editing unit 33, and after the utterance content is generated by the utterance content generation unit 34, The sentence generation unit 35 forms the sentence and outputs it as text data.

The syntax analyzer 31 determines whether the text data to be analyzed conforms to a predetermined syntactic rule,
Based on this, the structure of the sentence (main part / predicate, modification / modification, etc.) is extracted, and the extracted structure is output to the semantic analysis unit 32 as syntax tree data obtained by expanding the tree.
Here, the syntax rules handled by the syntax analysis unit 31 include:
Various types such as a regular grammar, a context-free grammar, and a context-dependent grammar are conceivable. However, in consideration of handling a natural language of a free real voice externally given through the wireless communication unit 1, in this embodiment, It is desirable to use a context-free grammar. As a method of analyzing this context-free grammar,
A top-down search in which rewriting rules are applied starting from a start symbol to an input word string, a bottom-up search in which words are first dictionary-derived and a partial structure is assembled, or an LR table is prepared in advance from a grammar. There is a deterministic analysis or the like in which a table is created and analyzed, and any method may be adopted.

The semantic analysis unit 32 automatically narrows down each word by category based on the analysis result of the syntax analysis unit 31 and automatically recognizes the meaning of the words.
First, a predicate (verb) portion is extracted from the syntax tree, which is the analysis result of the syntax analysis unit 31, and a skeleton of a case frame is created. Check that the given constraint written in the matching slot is satisfied, and if so, unify it into that slot, and if it doesn't, try to incorporate it into another syntax category slot. By doing so, words in the parse tree are narrowed down by category. In particular, the semantic analysis unit 32 determines whether or not there is an information transmission source (a requester) when the information is received through the wireless communication unit 1, who is the requester, and whether the requester is a car occupant. If there is information about the requested action requested, and the target object of the requested action (for example, a new destination in the case of a request to change the traveling destination), if these are the most important information, The information is transmitted to the information adding / editing unit 33 as a word in the syntax tree by priority.

The information adding / editing unit 33 selects and receives, from the semantic contents analyzed by the semantic analyzing unit 32, information suitable for the behavior of the occupant, such as the car navigation system shown in FIG. The information of the sender (requester) of the real voice given by e-mail or wireless, the basic character string prepared in advance, and the like are added and output to the utterance content generation unit 34 in the form of a syntax tree. It has become. As the information content of the syntax tree output here, in particular, information related to the driving of the car is selected from the received e-mail or real voice information given wirelessly, and the summary content is included. The operation-related contents of various on-vehicle units of the occupant such as a navigation system include information on operation guidance.

The utterance content generation unit 34 includes the information addition editing unit 3
Based on the information in the form of the syntax tree given from No. 3, what to represent as its internal representation is determined according to a predetermined grammatical constraint.

The sentence generation unit 35 constructs the internal expression determined by the utterance content generation unit 34 according to a predetermined grammar constraint,
The information relating to the traveling of the car and the information of the operation guide are converted into text as kanji-kana mixed text data, and this is output to the speech recognition synthesis device 3.
The text data converted into a sentence by the sentence generation unit 35 is converted into a voice waveform by the voice synthesis function of the voice recognition / synthesis device 3 and is output as voice from the speaker 5.

The voice recognition / synthesis device 3 and the agent processing section 4 read out a software program stored in a predetermined storage device into a memory, and read a software program stored in a predetermined CPU.
May be realized as a functional element when the function is operated, or hardware having a circuit configuration on a substrate may be used.

As the dictionary file 15, it is possible to use a unique one as an in-vehicle agent processing device.
May be used as the internal dictionary used in the above.

Incidentally, reference numeral 6 in FIG.
1 shows a microphone device for inputting an occupant's voice to the user.

<Operation> An example of the operation of the vehicle-mounted agent processing device having the above configuration will be described with reference to the flowcharts of FIGS. Here, a case will be described as an example where a target route is changed through external wireless communication when an optimum route search of the car navigation system 7 is performed in a car.

First, in step S1 in FIG. 5, the occupant of the automobile turns on the power of the on-vehicle agent processing apparatus and starts its operation (agent processing). In this case, the in-vehicle agent processing device holds the standby state until the wireless communication unit 1 receives an external signal.

When any wireless signal is given to the wireless communication unit 1 from the outside, the signal is output to the local area network in the automobile.

If the signal output from the wireless communication unit 1 is an electronic mail, the electronic mail
On the side of the electronic mail transmitting / receiving device 2, it is determined that the signal is an electronic mail by a code or the like attached to the head portion of the signal flowing in the local area network. After converting the data into text data in a format, the text data is transmitted to the agent processing unit 4 through the local area network.

Alternatively, when the signal output from the wireless communication unit 1 is a conversation with a real mobile phone, the voice recognition / synthesis unit 3 sends the signal to the head of the signal flowing in the local area network. Judgment that the signal is an audio signal based on the assigned code or the like,
This is received in the speech recognition synthesis device 3. Then, by the voice recognition function, the frequency analysis unit 11 analyzes the frequency of the voice signal in the natural language of the real voice received through the wireless communication unit 1, and the phoneme decomposition unit 12 performs the phoneme decomposition, and then reduces the error rate. For this purpose, the word recognition unit 13 performs word recognition, and the sentence recognition unit 14 executes sentence recognition to convert the data into text data similar to that of an electronic mail (FIG. 2). The converted text data is transmitted to the agent processing unit 4 via the local area network.

In the agent processing unit 4, the radio communication unit 1
Of the text data (hereinafter, referred to as “received data”) transmitted from the electronic mail transmitting / receiving device 2 or the voice recognition / synthesizing device 3 through the local area network after being received by the parsing unit 31, Here, the information determined to be analyzed is interpreted by the semantic analysis unit 32 (step S2).

As a processing procedure in the semantic analysis unit 32 at this time, first, as in step S9 in FIG. 6, is there an information transmission source (requester) when the information is received through the wireless communication unit 1? It is determined whether or not the requester can be specified, and the process is terminated (step S15). On the other hand, if the requester can be identified, it is determined who the requester is, and in step S10, the name of the requester (request name) is held in the memory.

Next, in step S11, it is determined whether or not there is a request action requested by the client to the occupant of the automobile.
At 12, it is held in the memory. On the other hand, if there is no request action, the process proceeds to step S15 and ends.

Further, in step S13, it is determined whether or not there is information such as what is the target object of the requested action (for example, a new destination in the case of a request to change the traveling destination). Is the step S1
In 4, the information adding / editing unit 33 combines these pieces of information.
To communicate.

After the processing in the semantic analysis unit 32 is completed, next, in step S3 in FIG. 5, the information addition and editing unit 33 adds and edits various information including operation guidance and the like, and utters. After the utterance content is generated by the content generation unit 34, it is formed into a sentence by the sentence generation unit 35 and output as text data including operation guidance and the like.

More specifically, as shown in FIG. 7, the information adding / editing section 33 first reads a basic character string prepared in advance in step S16, and incorporates the basic character string into a syntax tree. Similarly, the requester's name given via the semantic analysis unit 32 (step S17), the request action requested by the client to the occupant of the car (step S18), and the purpose of the request action Target (for example, a new destination in the case of a request to change the travel destination: Step S
19) is also incorporated in the syntax tree. The information in the form of a syntax tree is output to the utterance content generation unit 34 in a state including operation guidance of various on-vehicle units such as the car navigation system 7.

In step S20, the information addition / editing unit 33
Based on the information in the syntax tree format given by the utterance generation unit 34, the utterance content generation unit 34 determines the expression content according to a predetermined grammatical constraint, and the sentence generation unit 35 constructs the expression according to the predetermined grammatical constraint, and is used for driving a car As related information and operation guide information, text is formed as text data mixed with kanji and kana, and this is output to the speech recognition / synthesis device 3.

Then, as in step S 4 in FIG. 5, the text data that has been sentenced by the sentence generation unit 35 is converted into a voice waveform by the voice synthesis function of the voice recognition / synthesis device 3, and is output as voice from the speaker 5. You.

At this time, if the name of the requester is “Mr. XX” and the new destination of the travel destination change request is “で”, for example, “Mr. Asked me to come to △△. Do you want to start searching for routes? "

Here, the occupant (eg, driver) in the car makes a decision on his / her action in response to the voice-guided operation guide (step S5). If the car navigation system 7 is not operated, it is left as it is. The process ends (Step S6). On the other hand, when it is desired to start the route search in accordance with the operation guidance, the vehicle is once stopped, and a predetermined button operation or the like of the car navigation system 7 is performed to instruct a route calculation (step S7), and the result is obtained. At this point, the process ends (step S
8).

By omitting the occupant decision-making process in step S5 and proceeding to the route calculation process in step S4 from step S4, the route calculation in the destination change in the car navigation system 7 is fully automatic. It is convenient to be able to start.

As described above, the real voice information or e-mail in the natural language with high flexibility transmitted from the outside through the wireless communication unit 1 is summarized and converted into a format that can be easily understood by the occupant of the car. In addition, since the occupant can add and process various necessary information contents and transmit the information by voice output so as to be invisible, unnecessary information can be simply presented to the occupant in the vehicle, and the predetermined information can be determined in advance. You do n’t need to prepare data in the format,
It is possible to greatly reduce the labor on the transmission side.

In the above-described embodiment, the operation has been described by taking the change of the traveling destination of the car navigation system 7 as an example. If the information includes a factor that should be used to operate the various on-vehicle units, it is determined that the information is included, and operation guidance information for inviting the occupant to perform operations on the various on-vehicle units is included. It is needless to say that any operation may be performed as long as the character information is reconstructed.

[0057]

According to the first aspect of the present invention, a radio signal received from the outside is received by the radio communication unit, and syntax analysis is performed on the natural language information in the radio signal by the agent processing unit. Since information relating to various on-vehicle units is selected, summarized, and reconstructed into character information, and the reconstructed character information is converted into speech by a speech synthesizing unit and output, for example, Such voice information relating to a real voice or information in a natural language having high flexibility such as an e-mail as claimed in claim 4 is converted into a format that is easy to understand for a vehicle occupant and converted. Various necessary information contents can be additionally processed and transmitted by voice output so that they can be invisible, so unnecessary information can be simply presented to the occupants in the car. And in that the need to prepare the data for pre-determined format is eliminated, it is possible to save significantly the time of the transmission side.

Further, as in the second aspect of the present invention, when the agent processing unit determines that the operation of various on-vehicle units should be performed, the agent processing unit reconstructs the character information by including the operation guide information. Since the operation is automatically called to the occupant, there is an effect that the most necessary information for the occupant can be simply presented.

According to the fifth aspect of the present invention, when new information on a travel destination is included as information on various on-vehicle units, a recommended route is provided in the car navigation system based on the new information. Since the search is executed, it is possible to start the recommended route search at the time of destination change or the like automatically without performing any operation, which is convenient.

[Brief description of the drawings]

FIG. 1 is an overall block diagram showing an on-vehicle agent processing device according to an embodiment of the present invention.

FIG. 2 is a functional block diagram showing an internal configuration of a part of a voice recognition / synthesis device that controls a voice recognition function.

FIG. 3 is a functional block diagram showing an internal configuration of a part that controls a speech synthesis function of the speech recognition and synthesis device.

FIG. 4 is a functional block diagram showing an internal configuration of an agent processing unit.

FIG. 5 is a flowchart showing an operation of the vehicle-mounted agent processing device according to one embodiment of the present invention;

FIG. 6 is a flowchart showing an operation of the vehicle-mounted agent processing device according to one embodiment of the present invention;

FIG. 7 is a flowchart showing an operation of the vehicle-mounted agent processing device according to one embodiment of the present invention;

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 wireless communication unit 2 electronic mail transmitting and receiving device 3 voice recognition and synthesis device 4 agent processing unit 5 speaker 7 car navigation system 1a antenna 11 frequency analysis unit 12 phoneme decomposition unit 13 word recognition unit 14 sentence recognition unit 15 dictionary file 21 language processing unit 22 Sound processing unit 31 Syntax analysis unit 32 Semantic analysis unit 33 Information addition and editing unit 34 Utterance content generation unit 35 Sentence generation unit

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G10L 15/28 G10L 3/00 551Q (72) Inventor Jun Okada 1-7 Kikuzumi, Minami-ku, Nagoya City, Aichi Prefecture No. 10 F-term in Harness Research Institute, Inc. (Reference) 2F029 AA02 AB07 AC02 AC18 5D015 HH11 KK02 KK04 LL01 5H180 AA01 BB05 BB13 FF05 FF25 FF27 9A001 BB04 CC05 DD11 EE02 HH17 HH18 HH33 JJ14 JJ77 KK56

Claims (5)

[Claims] 1. An in-vehicle agent processing device for providing information desired by an occupant by voice output based on information transmitted through the communication terminal in an automobile provided with a communication terminal that communicates with the outside. A wireless communication unit that receives a wireless signal given from the outside; a syntax analysis is performed on natural language information in the wireless signal given from the outside through the wireless communication unit, and information on various on-vehicle units in the vehicle is selected. Summarizing,
An in-vehicle agent processing apparatus comprising: an agent processing unit configured to reconstruct character information; and a speech synthesis unit configured to convert the character information reconstructed by the agent processing unit into voice and output the voice. 2. The in-vehicle agent processing device according to claim 1, wherein the agent processing unit includes a factor in which the occupant newly operates the various in-vehicle units in information on the various in-vehicle units. When the information included in the character information is included, operation guide information indicating that the occupant is operated to operate the on-vehicle unit is included, and the character information is reconstructed. Characteristic in-vehicle agent processing device. 3. The on-vehicle agent processing device according to claim 1, wherein the natural language information includes voice information including a real voice, and converts the voice information into character information by voice recognition. An in-vehicle agent processing device further comprising a voice recognition unit. 4. The on-vehicle agent processing device according to claim 1, wherein the natural language information is an electronic mail including character information. 5. The on-vehicle agent processing device according to claim 1, wherein the various on-vehicle units include a car navigation system that performs a recommended route search for an arbitrary traveling destination. When the car navigation system includes new information about the travel destination of the vehicle as information on the various on-vehicle units, the car navigation system performs a recommended route search based on the new information. Characteristic in-vehicle agent processing device.