JPH1131149A - Intelligent interface system and document retrieval method using the system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain an intelligent interface (IIF: intelligent interface) system which reduces the load on a user to the operation of a multi-function application as a natural language front end and a document retrieval method that uses it. SOLUTION: This IIF system 1 consists of a module 11 which carries out whole control, a module 12 which is connected to the module 11 that executes whole control and executes sentence normalization, a module 13 which executes a sentence structure analysis, a module 14 that carries out CBR processing and a DM(Discourse Manager) module 15 that manages discourse processing by a natural language. Then a user access the module 11 which carries out whole control through an AM 3.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an intelligent interface system and a document search method using the same, and more particularly, to an intelligent interface system capable of reducing a user's load for operating a multifunctional application. It relates to a document search method using the same.

[0002]

2. Description of the Related Art Currently, applications (software) used in personal computers are becoming more and more multifunctional, and the load of multifunctional applications is increasing when users start using them or try to use them. It tends to be larger.

[0003] In the above-mentioned tendency, various solutions have been devised on the GUI as a solution. For example,
There are function buttons arranged on icons and toolbars, which are intended to reduce the load on the operator by operating them.

[0004]

However, there is a problem that if the number of these functions is large, it is difficult to determine which function is required, and the load on the user for operating a multifunctional application is not necessarily reduced.

[0005] As a countermeasure against the above problem, an interface for making an inquiry in a natural language or giving an instruction is conceivable. This is not necessarily a complete answer, but works well with the application to allow for existing operations (GUI, etc.)
Can help. In particular, it can be expected to have the effect of lowering the threshold for using software for beginners.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems and countermeasures therefor. As a natural language front end as described above, it is possible to reduce a user's load for operating a multifunctional application. Intelligent Interface (IIF: Intellig
ent Interface) system and a document search method using the same.

[0007]

In order to achieve the above object, in the intelligent interface system according to the present invention, a module for executing overall control and a module for executing the overall control are connected. And a module that executes a sentence normalization process, a module that executes the sentence structure analysis process, is connected to a module that executes the overall control, and a module that is connected to the module that executes the overall control, and executes the CBR process When,
A DM (Discourse Mana) connected to a module for executing the overall control and managing a query process in a natural language.
ger) module.

[0008] Further, according to claim 2,
In the interface system, the module for executing the sentence normalization processing includes a morpheme dictionary, a notation level normalization such as morpheme control, morpheme unit analysis (modal information), and a control dictionary using a morpheme dictionary. It performs processing such as vocabulary control processing, vocabulary level normalization such as phrase unit analysis (functional information), and expression normalization using paraphrase rules.

Further, according to the third aspect of the present invention,
In the interface system, the module for executing the sentence structure analysis processing includes attribute processing using an attribute dictionary, that is, recognition of time, person name, and place name, phrase correction for time, digitization, and dependency relation hypothesis. , And the like.

[0010] Further, according to claim 4 of the present invention,
In the interface system, the module for executing the CBR processing includes a script (command / parameter sequence) according to rules, matching with a case, case processing using a case dictionary for analyzing the intention of an input sentence, and rules. , And the like.

[0011] Further, according to claim 5, the intelligent
In the interface system, the DM (Di
If the user input is ambiguous or does not contain enough information, the module of the scourse manager can perform one or more queries in a natural This is to execute the control to perform in the form.

According to a sixth aspect of the present invention, in the document search method using the intelligent interface system, a morphological analysis is performed on a sentence input by a user using a morphological dictionary. A first step, a second step of performing notation level normalization on the morpheme list (prototype) generated in the first step, and a control on the morpheme list (rounding notation addition) generated in the second step A third step of performing vocabulary level normalization using a dictionary, a fourth step of normalizing expressions using paraphrasing rules, and a phrase list generated by the fourth step using an attribute dictionary. A fifth step of performing attribute processing, and a clause list generated in the fifth step,
A sixth step of performing a dependency process using a grammar heuristic, a seventh step of performing a case search using a case dictionary on the hypothesis graph generated in the sixth step, and a sixth step of performing a case search using the case dictionary. An eighth step of obtaining the analysis candidate and executing a process of determining the analysis result; and a script generation process of performing a script generation process on the analysis result generated in the eighth step based on a script generation rule to generate a script. 9 steps.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Referring now to the accompanying drawings, an embodiment of an intelligent interface system (hereinafter abbreviated as IIF system) and a document search method using the same according to the present invention will be described. 1. Outline of IIF system 2. Function of IIF system 3. Limitations of the IIF system 4. Operation of IIF system Configuration of IIF system This will be described in detail in the order of development of the IIF system.

[1. Overview of IIF System] The purpose of the IIF system is to reduce the user's load on the operation of a multifunctional application as much as possible, and the user generally uses the application operation instead of remembering the operation. It allows the application to be operated in natural language expressions.

That is, upon receiving the user's plain language,
It allows you to give instructions in general terms without knowing the application-specific terminology.
For example, in the case of "I want to see the document of IIF made yesterday", command: search Enter

Further, it is possible to save labor for searching for a function. That is, it is possible to execute a function from a natural language input without searching for a necessary function from a menu or the like when the function is increased or the version is upgraded.

Further, a question is returned for an ambiguous input. That is, humans do not always input completely. It is expected that the information will be insufficient input or will be vaguely expressed. In such a case, an input is prompted by a question and answer to supplement the information necessary for the analysis.

For example, a user: “I want to see a document.” ↓ IIF: “What kind of document?” ↓ A user: “I want to see a document of IIF created yesterday.” ↓ Execution.

FIG. 1 is a block diagram showing the relationship and positioning of the IIF with other applications and the like. IIF
Reference numeral 1 denotes an information manager (hereinafter abbreviated as IM) connecting a plurality of applications 2a and 2b.
3 Here, IM3 controls the entire system. Also, IIF1, synonym 4 and the like are called as necessary. Also, the applications (Ichitaro 2a, Sanshiro 2b) operate according to instructions from IM3. The IIF 1 analyzes the natural language sentence received from the IM 3 and executes a script (command-parameter sequence).
The synonym 4 performs an operation of expanding the vocabulary (search keyword) passed from the IM 3 and returning a synonym expanded keyword.

[2. Functions of IIF System] (Basic Algorithm) Functions provided by the IIF system will be described. First, as a basic algorithm, IIF
Analyzes natural language input by "CBR = Case Based Reasoning". here,
A case prepares a natural language sentence that is assumed to be input by a user and a result of analysis thereof, which is called a case. Also,
The CBR is a method that compares an actual input with a case, extracts an analysis result of the closest case, and makes an adjustment to obtain an analysis result for the input sentence.

(Analysis of Natural Language Input) Next, regarding analysis of natural language input, the IIF can issue an instruction to an application by natural language input. Below, the features of natural language analysis in IIF are presented.

First, it absorbs fluctuations in Japanese expressions and accurately recognizes objects. In the example of FIG. 2, the positions of the attribute name and the attribute value are interchanged, but express the same meaning. In both cases, attribute name = sender and attribute value = Hosokawa can be recognized.

Second, the difference in dependency between words is recognized, and the sentence is correctly analyzed. In the example of FIG. 3, the sentence shape is the same, but the dependency relationship differs depending on the vocabulary. By analyzing these, a correct dependency relationship can be extracted.

Third, it is possible to understand the deep intention of the user. As shown in FIG. 4, in the case of a direct expression, it is easy to determine the operation. On the other hand, in the case of an indirect expression, it is necessary to perform an operation taking in its intention.

Fourth, if different vocabulary expresses the same thing, it is recognized. For example, FIG.
Indicates an action, and FIG. 5B indicates a case.
FIG. 5C shows the case of the modified clause.

Fifth, the various sentences are absorbed to understand the meaning of the sentence. For example, examples shown in FIGS. 6A and 6B are given.

Sixth, even if the sentence type is the same, it can be recognized that different actions are performed depending on the characteristics of the vocabulary. For example, there is an example shown in FIG.

(Command Extraction) Commands are extracted based on the characteristics of predicate verbs and vocabulary (for this, refer to (Natural Language Input Analysis)).

(Parameter Extraction) The IIF system can extract parameters in the following cases.

First, an explicit parameter name is presented, for example, "I want to search for a document with importance 1" → parameter name = importance value = 1.

Second, use of attribute information, for example, "I want to search for documents of Hosokawa" = person name attribute → parameter name = creator value = Hosokawa.

Further, use of attribute information, for example, “I want to search for yesterday's mail” = time attribute → parameter name =
Creation date value = yesterday

Third, a characteristic expression, for example, “I want to search for notification mail” → parameter name = auxiliary attribute value = notification

Fourthly, among other things, items that are not recognized as parameters are extracted from the sentence, which can be search keywords. For example, “I want to search for documents in New York” → “New York” is cut out as a parameter.

(Question) Input in a natural language is not always complete, and information necessary for executing a command may be missing or may include an ambiguous expression having a plurality of interpretations. In such a case, the command cannot be executed immediately.

In this case, some interpretation can be given by the IIF, but an interpretation different from the user's intention may be given. In that case, there is a possibility that an irreparable situation may occur (such as deletion). Therefore, in the case described above, it is necessary to ask the user a question to confirm the intention or to input additional information.

As the method, the following forms can be considered. First, the intention is confirmed by a natural language inquiry, that is, a dialogue with the user. Secondly, a question is returned by a dialog, that is, the user is required to additionally set necessary information. These are not mutually exclusive, and it is considered that each of them should be used properly according to a necessary situation. In particular, in consideration of the affinity with the application, the number of situations in which the inquiry by the dialog according to the second is mainly adopted is increased.

[3. Limitation of IIF System] (Acceptable Input) The IIF system has an example of an expected natural language input sentence. By comparing the input sentence and the case, an appropriate analysis result can be obtained. There are several conditions to be able to be analyzed by the IIF system as shown below.

First, as shown in FIGS. 8A and 8B, the predicate verb is clear and at the end of the sentence. In addition, in order to respond to the user's flexible input, it is also possible to analyze "the sentence end is a sa verb".

Second, as shown in FIG. 9, necessary particles,
Include inflected forms.

Third, as shown in FIG. 10, the sentence is a single sentence and a compound sentence (multiple sentences are not analyzed). When a multiple sentence is input, an attempt is made to analyze as incompletely as possible. However, there is no guarantee that the analysis results are valid.

Fourth, as shown in FIG. 11, when an expression is to be treated as a simple character string (without analysis),
The section is surrounded by “″” or “′”.

Fifth, there are other things such as being a standard word and a general wording.

(Extractable Information) In the IIF, necessary information can be extracted by referring to cases. In that case, the sentence must satisfy the above-mentioned condition. In the case, there is a description of how to handle information and how to extract information based on the description.

Referring to FIG. 1, as a result of analyzing a sentence, a script (command-parameter string) is extracted and IM3
Pass to.

[4. Operation of IIF System] (Mounting target) In the IIF system, the following 3
One of the functions can be realized. That is, first, command execution, second, refinement, and third, item generation (action pad). Hereinafter, each operation will be described in detail.

(Command Execution) As the command execution, the target function will be described first. The following two types of functions can be executed as target functions. IIF system is IM
Since only the operations are performed, only what can be executed by the IM is basically targeted. IM functions Application functions that can be called from IM

Next, the target command will be described.
It does not cover all functions that can be executed by IM. Target commands are narrowed down according to the following policy. Limit the scope of coping vocabulary as much as possible. That is,
In order to maintain the consistency of case descriptions, keep them within a range that can be managed by humans. The integrity management of this case will require some automation in the future. Also, the case dictionary has become huge,
Avoid speed and accuracy degradation. An increase in dictionary size when the number of cases increases is inevitable. Focus on commands that are expected to be entered in natural language. The main functions of IM (mail transmission, etc.) are main, and other functions are added.

(Narrowing down) First, with respect to the user input expected in "Narrowing down", in "Narrowing down", parameters are extracted by specifying narrowing conditions. Since free input is allowed, the following input can be considered. Input only the narrowing conditions, for example, "unread mail" Add "refine" as a predicate verb, for example,
"I want to narrow down the emails of yesterday."
For example, "I want to print a Hosokawa document"

Second, regarding the analysis of a sentence, all of the above three input patterns can be accepted as IIF. However, the following operation is internally performed in consideration of the execution of “narrowing down”.

Input of only narrowing conditions, “Narrow down”
Is analyzed by supplementing the above. As a result, it is possible to cope with the same processing as the command execution. For example, analysis is performed as “unread mail” → “want to narrow down unread mail”. Add "refine" as a predicate verb. If analysis is performed as it is, it can be handled by the same processing as command execution. For example, “I want to narrow down the emails of yesterday” → analyze as it is. Add a predicate verb other than "refine".
Analyze as is. Only when a command corresponding to “narrow down” is included as an analysis result, it is returned as a result. For example, "I want to print a Hosokawa document" →
A narrow-down command + print command is obtained by analysis.
As a result, "refine" is returned. Further, for example, a “new arrival check desired” → new arrival check command is obtained by analysis. This is determined as analysis failure.

(Item Generation) First, regarding the input of the user expected in “item generation”, characters are input to the memo item in “item generation”. The input to the memo item is quite free in nature. Unlike the case of “command execution” described above, it is conceivable that the input content is often not satisfied as a sentence. That is,
In the case of command execution, the condition as a sentence to be accepted is satisfied, but in the case of item generation, it is an input as a memo and is often not accepted.

Second, with regard to attribute extraction, it is expected that there are a wide variety of expressions described as memos, even if conditions for sentences accepted by the IIF are satisfied. Describing all such cases is difficult to maintain as a dictionary and is not practical.

However, it is possible to extract the time attribute and the personal name attribute detected in the middle of the analysis. However, in this case, since it is not an analysis based on a case, it is not possible to judge what kind of information is described. In order to extract other information or to understand the user's intention at all, it is necessary to prepare a dedicated mechanism.

As this dedicated mechanism, an attribute extraction mechanism using template matching can be assumed. This is an intermediate mechanism between IIF case search and natural language help keyword search.

Third, IIF in “item generation”
According to the present invention, the time attribute and the personal name attribute are extracted without preparing the above-mentioned dedicated mechanism. The extracted attributes can be passed to the IM as parameters.

It is conceivable to execute a necessary command (for example, sending an e-mail) by creating some cases. For example, a sentence such as “Contact Hosokawa” is changed to →
It is impossible to determine whether the intention is to send the mail → attach the TODO attribute to the memo.

Unlike the case of command execution, selecting a plurality of commands at this point is no different from selecting from a menu. Considering the other way around, even if the case analysis is performed with the possibility of multiple selection remaining, a similar selection screen will be presented in the normal processing (menu) and the IIF processing, and the user will be presented with a selection screen. Can be a source of confusion. From the viewpoint of "reducing the load on the user", it can be determined that the merit of introducing the case analysis here is small.

(“Narrowing down” and synonym expansion) Although the search keyword obtained by the narrowing down is the one inputted by the user,
In some cases, it is not always possible to obtain sufficient information simply by searching with the vocabulary as it is. For example, there is a case where a keyword to be searched is input based on an ambiguous memory, and a case where a document of another person is searched (may not necessarily use an expected vocabulary). To solve this, synonym expansion is performed. The following three methods can be considered as a method of synonym expansion.

First, a synonym dictionary for search (system dictionary) is used. This search synonym dictionary (system dictionary)
Contains synonyms collected widely. Since an unspecified number of users are considered, vocabulary not necessarily used is also included. For example, taking New York as an example, → New York State, New York, NY, N.N.
Y. Empire State, string education, etc.

Second, search synonym dictionary (user dictionary)
Is used. This search synonym dictionary (user dictionary)
Synonyms registered by the user are stored. This makes it possible to use synonyms required according to individual circumstances. For example, if you want to narrow down to the synonyms required for individual users, taking the above New York as an example,
→ New York State, New York.

Further, a unique synonym group can be created using a synonym dictionary for search (user dictionary).
For example, taking the Takato team developing the present invention as an example, → Synonym expansion of natural language help, IIF, summarizer, etc. can be performed.

Third, the case description corresponds. That is, there may be synonyms depending on the situation. In this case, the search synonym dictionary cannot cope with the problem, so the search keyword is described in the case. However, the description in the case cannot be confirmed or registered by the user. Therefore, if you target many vocabulary,
Conversely, the convenience of the user may be impaired.

Therefore, in the present invention, the description in the case is dealt with by narrowing it to the minimum necessary. IM
In, management of documents is considered as one of the important factors. Therefore, the present invention addresses the problem by narrowing the target to “documents”.

By linking the case description and the synonym dictionary,
Enable the user to perform the required deployment. For example, taking "conference materials" as an example, Becomes

[5. Configuration of IIF System] (Module Configuration) FIG. 12 shows a module configuration of the IIF system 1. As shown in FIG. 12, the IIF 1 includes a module 11 for executing overall control, a module 12 for executing sentence normalization connected to the module 11 for executing overall control, and a module 13 for executing sentence structure analysis. And a module 14 for executing CBR processing,
(Discourse Manager) module 15, and the user accesses the module 11 for executing the overall control via the IM 3.

(Role of Each Module and Dictionary Configuration) The module 12 for executing sentence normalization uses a morphological dictionary to normalize notation levels such as morphological analysis, control of numerics, and morphological unit analysis (modal information). Vocabulary control processing using a control dictionary, vocabulary level normalization such as analysis of phrase units (functional information), and normalization of expressions using paraphrase rules.

Module 13 for executing the above sentence structure analysis
Is attribute processing using an attribute dictionary, that is, recognition of time, person name and place name, phrase correction for time, digitization,
Processing such as creation of a dependency relation hypothesis is performed.

Module 14 for executing the above CBR processing
According to the case processing using the case dictionary using the case dictionary for analyzing the intention of the input sentence →
Execute processing such as creation of a script (command / parameter string).

(Processing Flow) The relationship between the processing flow between each module and the dictionary used therein is as shown in FIG. That is, when the user inputs a sentence, the first
Second, sentence normalization processing, second, sentence structure analysis processing, third,
The CBR process is executed respectively.

As a first sentence normalization process, first, a morphological analysis is performed on a sentence input by the user using the morphological dictionary 20. The morpheme list (prototype) obtained by morphological analysis is then subjected to a notation level normalization process. The morpheme list of the notation level normalization processing word performs vocabulary level normalization processing using the control dictionary 21, and then normalizes the expression based on the paraphrase rule 22.

Next, as a second sentence structure analysis process, an attribute process using the attribute dictionary 23 is performed on the phrase list generated as a result of the expression normalization process. For lists, the grammar heuristic 24
Of the dependency by the user.

Next, as a third CBR process, a case search using the case dictionary 25 is performed on the hypothesis graph obtained by the dependency process, an analysis candidate is obtained, and a process of determining the analysis result is executed. The script generation processing is performed on the analysis result after the finalization processing based on the script generation rule 26 to generate a script.

(DM Processing Flow) DM (Discourse Ma
nager) 15 manages a query processing in a natural language. That is, if the user's input is ambiguous or does not include sufficient information, a query is made to supplement it. The DM 15 executes control for performing one or more queries in a natural manner.

If an unclear point is found in the CBR process during this process, a "DM script" describing "what to ask back" is passed. In DM15,
As shown in FIG. 4, a question sentence to the user is created based on the instruction.

As a response to the response to the inquiry of the DM 15, if the inquiry to the user is not completed at once, as shown in FIG. 15, an inquiry sentence is continuously created.

Further, the judgment based on which the DM 15 operates is described in the dictionary.

In IM3, a query using a dialog is mainly used. In some cases, DM15 is used in situations where a natural language inquiry is required.

(Dictionary) The dictionaries used in the present invention include a morphological dictionary 20, a control dictionary 21, a attribute dictionary 23, and a case dictionary 25, fourthly. Hereinafter, each will be described in detail.

The first morphological dictionary 20 is a dictionary referred to when performing morphological analysis. This stores information on basic vocabulary, such as notation, reading, part of speech, and attributes (person name attribute, place name attribute, etc.).

In a natural language, there are a plurality of expression methods for expressing the same thing. The second control dictionary 21 collects them as much as possible, thereby simplifying the case description and reducing the amount of description. For example, create → create search → search.

Conventionally (concepts in case version natural language help), expressions representing various concepts are stored in a concept dictionary. These concept descriptions are, in this version,
Most of them can be described as examples. However, because of many variations of expressions, there are some that are not realistic to describe as examples. For example, time, person name, place name, and the like. These describe the variation of the expression as the third attribute dictionary 23 and reduce the description amount of the case dictionary.

The fourth case dictionary 25 stores cases assuming user input. In the past (case version natural language help), the relationship between each case was so weak that the expression described in one case could not be used from another case . In the present invention, the relationship between cases is strengthened so that descriptions can be shared. Therefore, the description format becomes complicated, but flexible analysis can be performed with a smaller description amount.

It is necessary to prepare a dictionary for each corresponding application. In the present invention, since it is mounted on the IM, the dictionary is compatible with the IM. When applications are different, it is generally considered that dictionaries cannot be used in common. The task is to build a completely new dictionary.

[6. Development of IIF System] The following development functions can be assumed and added to the IIF system according to the present invention.

First, multiple sentences can be handled. Second,
It is possible to cope with an incomplete input sentence, that is, a situation where a simplified notation is input. This includes, for example,
Use template matching. However, if pattern analysis of the expected input is necessary and the intention of the sentence can be completely extracted, it can be applied to all aspects.

Third, a learning function can be added. That is, a study function can be added after examining what kind of learning is effective for the user. Fourth, user customization can be added. Again, it is necessary to consider what customization is effective for the user.

Fifth, enhancement of natural language querying;
That is, an inquiry is mainly performed by a dialog.
However, the dialog is not always effective in all cases. In some situations, natural language questions are appropriate. In addition, it is necessary to strengthen questioning in natural language.

Sixth, the cooperation with the application is strengthened. According to the present invention, a part of the information on the situation at that time is obtained from the application, and the operation is separated based on the information. When considering the strengthening of cooperation with applications, it is necessary to subdivide into more detailed situations. In addition, the number of cases is expected to be enormous, and it is necessary to provide a means for ensuring consistency between cases.

[0090]

As described above, in the intelligent interface system according to the present invention, the intelligent interface (IIF) capable of reducing the load on the user for operating a multifunctional application as a natural language front end. : In
telligent Interface) system can be obtained.

[Brief description of the drawings]

FIG. 1. IIF and Information Manager (I
FIG. 3B is a block diagram showing a relationship between the program M) and related programs.

FIG. 2 is an explanatory diagram showing an example of analyzing a natural language input.

FIG. 3 is an explanatory diagram showing an example of analyzing a natural language input.

FIG. 4 is an explanatory diagram illustrating an example of analysis of a natural language input.

FIG. 5 is an explanatory diagram showing an example of analyzing a natural language input.

FIG. 6 is an explanatory diagram showing an example of analyzing a natural language input.

FIG. 7 is an explanatory diagram showing an example of analyzing a natural language input.

FIG. 8 is an explanatory diagram showing conditions for enabling analysis of IIF.

FIG. 9 is an explanatory diagram showing conditions for enabling analysis of IIF.

FIG. 10 is an explanatory diagram showing conditions for enabling analysis of IIF.

FIG. 11 is an explanatory diagram showing conditions for enabling analysis of IIF.

FIG. 12 is an explanatory diagram showing a module configuration of the IIF system.

FIG. 13 is a flowchart showing the flow of processing of the IIF system.

FIG. 14 is an explanatory diagram showing a query returning process in a DM natural language.

FIG. 15 is an explanatory diagram showing a query returning process in a DM natural language.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 IIF system 2a, 2b application 3 Information manager (IM) 4 Synonyms 11 Module for executing overall control 12 Module for executing sentence normalization 13 Module for executing sentence structure analysis 14 Module for executing CBR processing 15 DM 20 Morphological dictionary 21 Control Dictionary 22 Paraphrase Rule 23 Attribute Dictionary 24 Grammar Heuristic 25 Case Dictionary 26 Script Generation Rule

Claims (6)

[Claims] A module that executes overall control; a module that is connected to the module that executes overall control and executes a sentence normalization process; and a module that is connected to a module that executes overall control and performs sentence structure analysis processing. CBR connected to a module to execute, and a module to execute the overall control.
A DM (Discourse Mana) connected to a module for executing a process and a module for executing the overall control, and managing a query process in a natural language.
ger) module; and an intelligent interface system. 2. A module for executing the sentence normalization processing includes a morphological analysis using a morphological dictionary, a control of a numeral, a normalization of a notation level such as a morphological unit analysis (modal information), and a vocabulary using a controlled dictionary. 2. The intelligent interface system according to claim 1, wherein processing such as control processing, vocabulary level normalization such as phrase unit analysis (functional information), and expression normalization using a paraphrase rule are performed. . 3. A module for executing the sentence structure analysis processing includes attribute processing using an attribute dictionary, that is, recognition of time, person name and place name, phrase correction for time, digitization, creation of a dependency relation hypothesis, 2. The intelligent interface system according to claim 1, wherein the intelligent interface system performs processing such as: 4. A module for executing the CBR processing includes: matching with a case, case processing using a case dictionary for analyzing an intention of an input sentence, creation of a script (command / parameter sequence) according to rules, 2. The intelligent interface system according to claim 1, wherein the intelligent interface system performs processing such as: 5. A module of the DM (Discourse Manager), in which the user input is ambiguous or does not include sufficient information, is performed once or in order to make a query to supplement the input. 2. The intelligent interface system according to claim 1, wherein control for executing a plurality of queries in a natural manner is performed. 6. An intelligent interface
In a document search method using a system, a first step of performing morphological analysis on a sentence input by a user using a morphological dictionary, and a notation level for a morphological list (prototype) generated in the first step A second step of performing normalization; a third step of performing vocabulary level normalization on the morpheme list (addition of rounding notation) generated in the second step by using a controlled dictionary; A fourth step of performing normalization; a fifth step of performing attribute processing on the phrase list generated in the fourth step using an attribute dictionary; and a fourth step of performing attribute processing on the clause list generated in the fifth step. A sixth step of performing a dependency process using a grammar heuristic, and a seventh step of performing a case search using a case dictionary on the hypothesis graph generated in the sixth step, An eighth step of obtaining an analysis candidate generated in the seventh step and executing a process of determining the analysis result; and performing a script generation process on the analysis result generated in the eighth step based on a script generation rule. Do
A ninth step of generating a script, comprising: