JPH07192001A - Natural language processor - Google Patents

Abstract

PURPOSE:To provide a natural language processing unit with a high efficiency using a server client model. CONSTITUTION:The natural language processing unit consisting of a client 200 making a request of natural language processing, a server 207 executing the natural language processing and a network section 213 connecting them electronically is provided with a means sending nondecisive data generated during the natural language processing altogether simultaneously from the server 207 to the client 200.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a natural language processing apparatus in which a plurality of terminals are connected to a network and natural language processing can be efficiently performed in these terminals.

[0002]

2. Description of the Related Art As is well known, with the development of computer networks, many computers of various types are connected to the same network. In addition, the communication speed has been improved, and it has become possible to easily send and receive large amounts of data between computers.

On the other hand, not all computers connected to the network have high capabilities, so that a server / client model in which a high-performance computer and a low-performance computer are connected and jobs are shared becomes popular. ing. In this model, for example, when using natural language processing such as machine translation, speech recognition, summarization, kana-kanji conversion, and document proofreading support, a highly capable server machine takes charge of jobs that require computer power, and network Natural language processing is performed by issuing a processing request from a client machine with low capability connected via the above.

A server for machine translation as natural language processing
As an example realized with the client model, Japanese Patent Laid-Open No. 4-
The "inter-network communication processing device" disclosed in Japanese Patent No. 127748 and the "machine translation network system" disclosed in Japanese Patent Laid-Open No. 5-128144 can be mentioned.

However, in these systems, only the deterministic data can be transferred, and the translation result is uniquely determined and transferred to the client terminal which issued the processing request. Therefore, there is a problem in that the client terminal that makes the translation request cannot obtain other translation result candidates of the translated sentence, for example.

As an example of realizing Kana-Kanji conversion as natural language processing by a server / client model, Wn
n (Jserver) (KABA's "Wnn + GMW Introduction", Iwanami Shoten, 1990). Even in Wnn, basically, only deterministic data is transferred, the server grasps the status of the client, and when other data (next candidate for kana-kanji conversion) is requested by the client, other data is responded accordingly. I am trying to transfer. For this reason, frequent communication is performed between the server and the client, and when there is a discrepancy between the actual situation of the client and the situation of the client known to the server due to, for example, communication failure, malfunction occurs. There was a risk of causing. Further, this method has a problem that although the amount of communication per time is small, the number of inquiries to the server increases, and the number of processes in the server increases.

[0007]

As described above, in the conventional server-client model for natural language processing, only deterministic data can be transferred. Therefore, for example, in machine translation or kana-kanji conversion, other translation candidate or There is a problem that a candidate or the like cannot be obtained on the client side, or a correct candidate cannot be obtained due to a difference in situation recognition between the server and the client. There is also a problem that the number of inquiries between the server and the client increases, which causes an increase in load on the network.

Therefore, the present invention aims to provide a natural language processing apparatus capable of efficiently performing natural language processing by using a server / client model by enabling transfer of data including non-deterministic data. I am trying.

[0009]

In order to solve the above-mentioned problems, the present invention provides at least one processing request terminal section for making a natural language processing request and at least a natural language processing in response to the natural language processing request. In a natural language processing device comprising a single natural language processing section and a network section that electronically connects the processing request terminal section and the natural language processing section, in the natural language processing section, the processing request terminal section is provided. When transmitting the processed data to the non-deterministic data portion that cannot be uniquely determined during the natural language processing, the transmitting means also transmits the candidate data.

The transmitting means is a means for changing the amount of non-deterministic data to be transmitted according to the amount of data that can be accepted by the processing request terminal unit, and an amount of non-deterministic data to be transmitted according to the situation of the network unit. Is preferably provided, and a means for transmitting after completion of all natural language processing such as word utilization expansion is provided.

[0011]

According to the present invention, data including non-deterministic data can be transferred.
It can be done efficiently by using the client model.

[0012]

Embodiments will be described below with reference to the drawings. FIG. 1 shows a schematic configuration of a natural language processing apparatus, which is a machine translation processing apparatus according to the first embodiment of the present invention. In the figure, reference numeral 1 indicates a processing request terminal unit. The processing request terminal unit 1 is composed of a control unit 2 and a communication unit 3. Reference numeral 4 denotes a natural language processing unit. The natural language processing unit 4 is composed of a communication unit 5 and a natural language processing engine 6. Reference numeral 7 denotes a network unit connecting these.

FIG. 2 shows a concrete structure of each part of the apparatus of the embodiment. In this device, an electronic computer is used for the processing request terminal unit and the natural language processing unit. In the figure,
Reference numeral 200 denotes a processing request terminal unit. The processing request terminal unit 200 includes an input unit 201. Input unit 201
Is for inputting a language for requesting processing, a processing request command, and input for correcting processing results. Generally, a pointing device using a keyboard, a mouse, a touch panel, a pen, or the like. It is composed of.

The input data from the input unit 201 is the control unit 2
Sent to 05. The control unit 205 includes a display unit 202,
It is connected to the recording unit 203, the editing unit 204, the communication unit 206 and the like, and controls these and transfers data.
The display unit 202 displays a command or a character string input from the input unit 201 or a translation result, and is generally composed of a CRT display or a bitmap display. The display unit 202 may be a goggle type one or one that is projected on a wall surface.

The recording unit 203 is used to record the data input from the input unit 201, the translation data obtained via the communication unit 206, or the data edited by the editing unit 204. The editing unit 204 is used for editing the data input from the input unit 201, the translation data obtained via the communication unit 206, and the like.

The communication unit 206 transmits the data input from the input unit 201 or the data recorded in the recording unit 203 to the natural language processing unit 207 described later, or receives the transmission data from the natural language processing unit 207. It is for doing. The communication unit 206 can receive the data transmitted from the natural language processing unit 207 even when the data is non-deterministic data. A specific example of this will be described later.

On the other hand, the natural language processing section 207 is constructed as follows. That is, in the drawing, 208 indicates a communication unit. The communication unit 208 has a role of receiving the translation request sentence or command transmitted from the communication unit 206 and transmitting the translated data. It should be noted that if the translation result has ambiguity,
It has a function of transmitting it as non-deterministic data to the processing request terminal unit 200. A specific example of this will be described later.

In the figure, reference numeral 209 indicates a control unit. The control unit 209 is connected to the recording unit 210, the translation unit 211, etc., and controls these and transfers data. The recording unit 210 temporarily records the data received by the communication unit 208 and the data translated by the translation unit 211. The translation unit 211 refers to the dictionary of the translation dictionary unit 212 and translates the data received via the communication unit 208 or the data recorded in the recording unit 210.

The processing request terminal unit 200 and the natural language processing unit 207 are connected via the communication path 213. Next, the operation of the apparatus configured as described above will be described. In the following description, the processing request terminal unit 200 will be called a client, and the natural language processing unit 207 will be called a server.

FIG. 3 shows an example of a processing procedure in the processing request terminal unit 200, that is, the client in the apparatus of this embodiment. When there is an input through the input unit 201, the client first determines whether or not it is the original sentence for which a translation request is made (step 301). If it is an original sentence, the data is transferred to the recording unit 203, and is recorded here. This is designated as data A (step 302).

When it is determined in step 301 that the input data is not the original sentence for the translation request, it is determined that it is a command to the client, and it is determined whether or not it is a translation start request (step 303). ). If it is not the translation start request, the processing according to the input command is performed (step 304). If the request is a translation start request, the data A recorded in the recording unit 203 is transmitted to the natural language processing unit 207, that is, the server (step 305).

Since the server translates this document, the client waits until the translation is completed (step 306). The translation process in the server will be described separately. During this time, the client may be waiting for input of another command or a translation request sentence. When the translation in the server is completed and the translation result data is transferred, it is judged whether or not it is non-deterministic data including ambiguity (step 307), and if it is non-deterministic data. Expands it for easy reference (step 308). After that, the translation result is presented to the operator by displaying the data on the display unit 202 (step 309).

FIG. 4 shows an example of the processing procedure in the natural language processing unit 207, that is, the server in the apparatus of this embodiment. The server waits to receive the translation data C from the client (step 401).
When the translation data C is received, the data C is translated by the translation unit 2
It is transferred to 11 and translated (step 402). Translation department 2
When the translation is completed in 11, the translation unit 211 causes the translation result D and the data including the ambiguity (nondeterministic data) E to be obtained.
And all are received (step 403), and these are transferred to the client (step 404).

The specific operation of the apparatus of this embodiment will be described below with reference to FIGS. FIG. 5 shows a display example on the display unit 202 in the client. In FIG. 5, the upper part of the screen is the original text display portion P, which is used as a screen for inputting and editing a translation request sentence. In FIG. 5, the lower part of the screen is a translated text display portion Q, which is used to display the translation result and the information accompanying it. FIG. 5 shows a situation in which a sentence for which a translation request is made is input and the result is displayed in the original sentence display area P.

Here, an example of the situation in which a translation start request is input from the input unit 201 and a translation result is obtained by the process shown in FIGS. 3 and 4 and which is shown in the translated sentence display unit Q of the display unit 202 is shown below. It is shown in FIG. Although the translation result displayed in the translated sentence display portion Q only shows the first solution, for example, in this example, "you" is ambiguous among the words in the translated sentence. That is, there are other translations. In the apparatus of this embodiment, since the "other translations" are also sent from the server to the client, it is possible to display the other translations by designating "you" on the screen of the client, and presented in the first solution. Can be replaced with a phrase.

FIG. 7 shows an example of this state.
In this figure, as another candidate for "you", "you",
You can see that there are "You" and "You". At this time,
The apparatus of this embodiment does not make an inquiry to the server.
Therefore, the number of communications can be reduced, and the server does not need to store various information including the translated word information used for translating this sentence, so that the computer resources can be effectively used.

Non-deterministic data in the translation result is not limited to the translated word. For example, the non-deterministic data may include the word-to-word modifier relations in the original sentence. Figure 8
Shows an example of translated sentence "I saw a girl with a telescope." In this case, the phrase with leads
There is a case where it is interpreted as “a girl with a telescope” depending on girl, and a case where it is interpreted as “seen with a telescope” due to saw and cannot be uniquely determined. In this case, too, ambiguous data from the server (non-deterministic data)
Since all of the above have been transferred, the above non-deterministic data can be presented to the user without querying the server.

FIG. 9 and FIG. 10 show an example of data transferred between the client and server to perform these processes. FIG. 9 shows an example of data sent from the server to the client to generate the situation of FIG. As shown in this figure, one word of the translated sentence is expressed as one line. The first column in each row is the first
It is solution data. The second column is a list of other candidates for the data in the first column. The third column is the corresponding original sentence word, and the fourth column is the part of speech. The client collects the data in the first column to prepare the display data shown in FIG. 6 and collects the data in the second column to prepare the display data shown in FIG.

FIG. 10 shows an example of data transmitted from the server to the client to generate the situation of FIG. As shown in this figure, the data structure is the same as the data structure in FIG. 9, but the range of non-deterministic data exceeds the range of words and is expressed as one line in units of phrases. The first column of each line shows the phrase number, the second column shows the translated sentence by phrase, the third column shows the corresponding original sentence word, and the fourth column shows the part of speech. There is also auxiliary data indicating how these data are constructed to generate a translated sentence. In this way, the client collects the transmitted data in the second column according to the auxiliary data to create the display data shown in FIG.

FIG. 11 shows a schematic configuration of a natural language processing apparatus according to the second embodiment of the present invention, which is also a machine translation processing apparatus. The apparatus of this embodiment has a function of changing the amount of transmission data depending on the condition of the communication path. In the figure, reference numeral 1101 indicates a processing request terminal unit. The processing request terminal unit 1101 is configured similarly to the processing request terminal unit 200 in FIG.

Reference numeral 1100 denotes a natural language processing section.
1103 is a communication unit. Here, the processing request terminal unit 1
The translation request sentence or command transmitted from the communication unit 101 is received, or the translated data is transmitted. Then, when the translation result has ambiguity, the communication unit 1103 treats the processing request terminal unit 1 as non-deterministic data.
It has a function of transmitting to 101. A specific example of this will be described later.

The recording unit 1104 temporarily records the data received by the communication unit 1103 and the data translated by the translation unit 1106. A control unit 1105 includes a recording unit 110.
4, connected to the translation unit 1106, the non-deterministic data creation unit 1108, etc., and controls these and transfers data. The translation unit 1106 refers to the dictionary of the translation dictionary unit 1107 and translates the data received via the communication unit 1103 and the data recorded in the recording unit 1104.

The nondeterministic data creating unit 1108 is a feature of the apparatus of this embodiment, and refers to the nondeterministic data determination rule unit 1110 based on the status of the communication path 1102 obtained by the communication path monitor 1109 to make a request. Processing terminal unit 11
The data to be transferred to 01 is determined as described later. Next, the operation of the apparatus configured as described above will be described. In the following description, the processing request terminal unit 1101 is a client,
The natural language processing unit 1100 will be called a server.

FIG. 12 shows the non-deterministic data creating unit 1108.
The processing procedure in is shown. First, in order to detect the congestion degree of the communication channel, the communication amount per unit time of the communication channel is measured by the communication channel monitor 1109, and whether or not it is equal to or more than the value N defined by the nondeterministic data determination rule unit 1110. Is determined (step 1201). If it is N or more, the communication amount is reduced in order to avoid the trouble of the communication path, and the communication amount is restricted according to the rule recorded in the rule recording unit (step 1205). If it is equal to or less than N, the speed of data that can be transferred by the communication path is measured, and it is determined whether or not it is equal to or less than the value M defined by the nondeterministic data determination rule unit 1110 (step 1202). If it is M or less, it takes a lot of time to send a large amount of data, so the communication amount is reduced, and the communication amount is restricted according to the rule recorded in the rule recording unit of the nondeterministic data determination rule unit 1110. If it is M or more, the cost per unit communication amount of the communication path is obtained, and it is determined whether or not this is the value L or more defined by the nondeterministic data determination rule unit 1110 (step 1203).

If L or more, a large amount of data is very costly to transmit, so the communication amount is reduced, and communication is performed according to the rule recorded in the rule recording unit of the nondeterministic data determination rule unit 1110. Limit the amount. If it is equal to or less than L, the communication amount is not limited (step 1204). FIG. 13 and FIG. 14 show an example of the communication amount restriction performed by the device of this embodiment. In addition, all data without data restriction is shown in FIG.
As shown in.

In the data restriction example shown in FIG. 13, of the non-deterministic data, up to the second candidate of the translated word is transmitted, and the data of the third candidate and thereafter are not transmitted. Also, FIG.
In the data restriction example shown in FIG. 4, only the non-deterministic data of the dialect is sent, and the non-deterministic data of other parts of speech are not sent. FIG. 15 shows a schematic configuration of a natural language processing apparatus according to the third embodiment of the present invention, which is also a machine translation processing apparatus.

In the apparatus of this embodiment, the amount of transmission data is changed according to the capability of the processing request terminal unit. In FIG. 15, reference numeral 1501 denotes a processing request terminal unit. Hereinafter, this will be referred to as a client.
The client includes an input unit 1502 and a display unit 1503.
And a condition determining unit 1504, a control unit 1505, a communication unit 1506, an editing unit 1507, and a recording unit 1508.

The input section 1502, the display section 1503, the editing section 1507, and the recording section 1508 have the same functions as those included in the request processing terminal section of the apparatus shown in FIG. The condition determination unit 1504 has a size of the display unit 1503,
This is for determining the amount of non-deterministic data that can be received at one time from the capacity that can be recorded in the recording unit 1508.
In addition, the communication unit 1506 sets the above restrictions to the natural language processing unit 1.
509 (hereinafter, referred to as a server), and has the same function as the communication unit 206 in FIG. The control unit 1505 also has the same function as the control unit 205 in FIG. 2 except that it also controls the condition determination unit 1504.

On the other hand, the natural language processing unit 1509, that is, the server, includes a communication unit 1510, a recording unit 1511, a control unit 1512, a translation unit 1513, and a translation dictionary unit 151.
4, a nondeterministic data creation unit 1515, and a nondeterministic data determination rule unit 1516. Communication unit 1
510 has the same function as the communication unit 1103 in FIG. 11 except that it can receive the limiting condition from the client. The recording unit 1511, the translation unit 1513, the translation dictionary unit 1514, and the control unit 1512 also have the same functions as the respective units in FIG.

The non-deterministic data creating section 1515 refers to the non-deterministic data decision rule section 1516, etc.
Although it functions in the same manner as the non-deterministic data creating unit in FIG. 11, in the embodiment shown in FIG. 11, the limiting data for creation is obtained from the communication channel monitor, whereas in this embodiment, it is obtained from the data received from the client. The difference is that you are trying to get. Then, these client and server are connected via a communication path 1517.

That is, in the apparatus of this embodiment, the amount of non-deterministic data to be transmitted is determined according to the amount of data that can be accepted by the client. The present invention is not limited to the above embodiments.
That is, in each of the above-described embodiments, the processing for all data transmitted from the client to the server for the processing request is transmitted to the client after the server completes the processing.
It may be transmitted at the time when a certain unit of processing is completed. For example, a client may ask the server, "I have a pen.You.
If a request `` have a pencil. '' is issued, the translation of `` I have a pen.
The data may be transmitted to the client when the translation of “” is completed.

The above embodiments are examples of a machine translation device as a natural language processing device, but the invention is not limited to this, and other natural language processes such as voice recognition, kana-kanji conversion, and document calibration support. It can be applied when ambiguity data is output as a processing result such as character recognition. For example, taking the Kana-Kanji conversion as an example, the nondeterministic data is a homophone of the conversion result, and the nondeterministic data shown in FIG. 9 is as shown in FIG. 16, for example. When applied to kana-kanji conversion, the first candidate, other conversion candidates, reading, part-of-speech, etc. are sent from the server to the client for each word resulting from the conversion. Taking character recognition as an example,
The non-deterministic data is another character candidate in the recognition result, and the non-deterministic data shown in FIG. 9 is as shown in FIG. 17, for example. When applied to character recognition, other recognition result candidates are also sent from the server to the client for each word that is a recognition result.

As shown in these examples, the natural language data to be transferred is transferred after the utilization expansion is completed. As a result, the client does not need to have any function related to natural language processing, and the resources of the client can be effectively used. However, some processing may be undertaken by the client. Besides, the present invention can be applied to processes such as voice recognition, spell checking, proofreading, and search. Further, although the embodiment is an example of English-Japanese translation, the same applies to other language pairs. It can also be applied to natural language processing for languages other than Japanese.

[0044]

As described above, according to the present invention,
By enabling transfer of data including non-deterministic data, natural language processing can be efficiently performed using the server / client model.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a natural language processing apparatus according to a first embodiment of the present invention.

FIG. 2 is a specific block configuration diagram showing each part of the device.

FIG. 3 is a diagram showing a flow of processing in a processing request terminal unit in the same apparatus.

FIG. 4 is a diagram showing a flow of processing in a natural language processing unit in the same apparatus.

FIG. 5 is a diagram showing a screen display example of a processing request terminal unit in the same apparatus.

FIG. 6 is a diagram showing a screen display example of a processing request terminal unit in the same apparatus.

FIG. 7 is a diagram showing a screen display example of a processing request terminal unit in the same apparatus.

FIG. 8 is a diagram showing a screen display example of a processing request terminal unit in the same apparatus.

FIG. 9 is a diagram showing an example of transfer data between a processing request terminal unit and a natural language processing unit in the same device.

FIG. 10 is a diagram showing an example of transfer data between a processing request terminal unit and a natural language processing unit in the same device.

FIG. 11 is a block configuration diagram of a natural language processing device according to a second embodiment of the present invention.

FIG. 12 is a diagram showing a flow of processing for changing the amount of transmission data according to the state of the communication path in the same device.

FIG. 13 is a diagram showing an example of transmission data in which the transmission data amount is limited in the device.

FIG. 14 is a diagram showing an example of transmission data in which the transmission data amount is limited in the apparatus.

FIG. 15 is a block configuration diagram of a natural language processing device according to a second embodiment of the present invention.

FIG. 16 is a diagram showing an example of transfer data between a processing request terminal unit and a natural language processing unit when the present invention is applied to a Kana-Kanji conversion process.

FIG. 17 is a diagram showing an example of transfer data between a processing request terminal unit and a natural language processing unit when the present invention is applied to character recognition processing.

[Explanation of symbols]

1,200,1101,1501 ... Processing request terminal unit (client) 4,207,1100,1509 ... Natural language processing unit (server)

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location G06K 9/03 Z 9289-5L

Claims (5)

[Claims] 1. At least one processing request terminal unit for making a natural language processing request, at least one natural language processing unit for carrying out natural language processing in response to a natural language processing request from this terminal unit, and said processing. In a natural language processing device including a requesting terminal unit and a network unit that electronically connects the natural language processing unit, a natural language is used when processing data is transmitted from the natural language processing unit to the processing requesting terminal unit. A natural language processing apparatus comprising: a transmission unit that also transmits non-deterministic data that cannot be uniquely determined during processing. 2. The transmission means comprises means for changing the amount of non-deterministic data to be transmitted in accordance with the amount of data that can be accepted by the processing request terminal unit. Natural language processor. 3. The natural language processing apparatus according to claim 1, wherein said transmitting means comprises means for changing the amount of non-deterministic data to be transmitted according to the situation of said network unit. 4. The natural language processing apparatus according to claim 1, wherein said transmitting means comprises means for transmitting data after completing all natural language processing such as word development. 5. The transmitting means comprises means for simultaneously transmitting data necessary for word development and the like so that a part of the natural language processing is performed by the processing request terminal unit. The natural language processing device according to claim 1.