KR19980078071A - Scheduling from Natural Memo Sentences - Google Patents

Abstract

The present invention relates to a method for managing schedules from natural language memo sentences that can be managed by recognizing natural language memo sentences input by a user.

The schedule management method from the natural language memo sentence of the present invention is a natural language memo sentence input by the user in the agent system consisting of an interface agent responsible for input and output to the user, a plurality of application agents responsible for the application and the corresponding application programs. A method of managing a schedule by processing the information, comprising: extracting information from a natural language sentence input by a user in the natural language processing portion of the interface agent to construct a generic template, which is a general-purpose information structure, and selecting the application agent in the dispatcher portion Delivering a generic template to the agent; A schedule agent, one of the application agents, extracts schedule related information from the generic template, constructs a schedule template, and analyzes the time and keywords in the schedule template to semanticly register and search the schedule program. Acting on behalf of the client.

According to the present invention, the schedule management agent system receives the natural language schedule memo familiar to the user, automatically organizes it, records it, and searches it efficiently, thereby maximizing user convenience.

Description

Scheduling from Natural Memo Sentences

1 is a diagram showing the structure of a multi-agent system for applying the present invention.

2 is a flowchart for explaining a processing procedure of an input sentence in the NTA of FIG.

3 is a diagram illustrating an example of reconstruction of sentences and combining of units in the preprocessing step of FIG.

4 is a diagram showing an example of generic template configuration in the generic template configuration step of FIG.

5 is a diagram showing the detailed structure of the schedule management agent of FIG.

6 is a diagram showing an example of a schedule template configured in FIG.

* Explanation of symbols for the main parts of the drawings

2: NTA (Note Taking Agent), 4: Household Account Agent, 6: Thesaurus, 8: Schedule Agent, 10: Household Program, 12: Schedule Program

The present invention relates to a schedule management agent (Agent) system for managing a schedule input by a user, and more particularly to a schedule management method from a natural language memo sentence that can manage the schedule by recognizing a natural language memo sentence input by the user will be.

Today, there is a need for a software agent, a computer program that responds independently to a given situation and responds independently, without user intervention, based on knowledge and reasoning skills. Agent system composed of such software agents can handle a variety of applications, such as email management, calendaring, cashier, reservation management.

However, in the conventional schedule management program, the schedules to be recorded by the user are manually classified into anniversary, task management, and general schedule, and then the schedule is managed by recording the place, companion, time and information according to each category. do. These calendars need to record items in different ways according to the type of schedule from the user's point of view, and even when searching for schedules, the schedules recorded by the user must be searched through the interface supported by the schedule management program. Therefore, the user can input a schedule through natural language, and a new color schedule management is required to maximize the user's convenience by using natural language sentences in search.

In addition, a natural sentence used by a user frequently has a form of a brief memo that is difficult to process using a general natural language processing technique. In general, a memo is often omitted from the main components such as surveys and endings, and often includes a symbol and a symbol having a specific meaning and a date, time, amount, telephone number, and e-mail address. Therefore, there is a need for a memo processing technique for correctly analyzing a memo.

Accordingly, an object of the present invention is to provide a schedule management method from a natural language memo sentence that manages a schedule using an input of a natural language sentence, which is devised by the above circumstances.

Another object of the present invention is to provide a schedule management method from a natural memo sentence that can efficiently analyze the memo sentence input by the user.

It is still another object of the present invention to provide a schedule management method from natural language memo sentences in which a schedule can be efficiently searched through a natural language query.

In order to achieve the above object, the schedule management method from the natural language memo sentence according to the present invention is an interface agent responsible for input and output to the user, the agent system consisting of a plurality of application agents and the corresponding application programs for the application A method of managing a schedule by processing a natural language memo sentence input by a user, comprising: extracting information from a natural language sentence input by a user in the natural language processing portion of the interface agent to construct a generic template which is a general information structure, and in the dispatcher portion Selecting a corresponding application agent and delivering a generic template to the selected application agent; A schedule agent, one of the application agents, extracts schedule-related information from the generic template, constructs a schedule template, and analyzes the time and keywords in the schedule template to mean registration and retrieval to a schedule management program. Characterized in that it comprises a step.

Other objects and advantages of the present invention other than the above objects will become apparent from the description of the embodiments of the present invention with reference to the accompanying drawings.

Hereinafter, exemplary embodiments of the present invention will be described with reference to FIGS. 1 to 6.

FIG. 1 illustrates the structure of an agent system for applying a schedule management method from a natural language memo sentence according to the present invention. The agent system of FIG. 1 is an NTA (Note Taking Agent) which is an interface agent responsible for input / output with a user. Agent, 2) and a plurality of application agents (4, 8) responsible for the application is composed of the respective application programs (10, 12).

NTA (2), which is in charge of processing natural memo sentences, refers to a program that allows users to easily access and use various types of applications with the same interface without having to learn how to use the applications. In other words, the NTA (2) accepts natural language sentences, organizes them, presents them in a desired form when the user wants them, and also tries to maximize the user's convenience by acquiring the user's personal taste or information. The NTA 2 is composed of a natural language processing portion for processing natural language input and a dispatcher portion for selecting an application agent suitable for user input.

Sentences input from the user to NTA (2) generally have a short memo format. The method of processing the memo format input statement in this NTA 2 is as shown in FIG.

First, when an input sentence as shown in FIG. 3 is input from the user to the NTA 2, that is, a memo sentence such as attending the HCI conference (Lotte Hotel) from February 18 to 20, the preprocessing step of the second step is performed. Reorganize sentences and combine numbers. In detail, the reconstruction of the sentence is to be described later by processing the date and time portion first so that the memo sentence can be analyzed, and separating and storing the portion that can be regarded as the amount portion and the annotation, that is, the (Lotte Hotel) portion. Only the parts that can be parsed are delivered in the parsing step (stage 6). In addition, combining numbers means combining numbers, rhetorical and crown adjectives and unity dependent nouns, which can be semantically clearly determined, before the next morphological analysis step. Accordingly, the input text is converted into 'date month: 2 date: 18 to date: 20 attending HCI conference' as shown in FIG. 3 through the preprocessing step as described above. Here, the part represented by '...' due to the combination of numbers is regarded as one token, which is the basic processing unit in parsing, and the efficiency of sentence analysis process is improved.

In the fourth stage, the morphological analysis is performed on the characters applied from the second stage, and then, in the sixth stage, a complete sentence is formed and parsed. In this parsing step, although the input sentence has been converted to facilitate the analysis in the preprocessing step, the main sentence components such as the ending or the survey may still be omitted, so the operability such as 'attendance' in the example sentence of FIG. Only vocabulary classified as nouns is analyzed as a part of verbs such as 'to attend'.

The post-processing step of the eighth step selects an appropriate sentence according to the analysis result of the memo of the sixth step. If the analysis result of the memo has ambiguity, one analysis result is selected, and if the analysis fails, it is responsible for deriving one analysis result from the partial analysis result.

Subsequently, in the tenth step, a generic template, which is a general-purpose information structure, may be transmitted to the memo analysis result and various additional information to various application agents 4 and 8. An example of the configuration of the generic template according to the above input statement is as shown in FIG.

In step 12, the dispatcher portion of the NTA 2 selects an application agent 4 or 8 capable of semanticly analyzing the contents of the generic template, and delivers the generic template to the selected agent.

Accordingly, the application agent 4, 8 extracts the information required for the application from the generic template received from the NTA 2, processes each memo 10 and 12 on behalf of the user, and processes the memo. Return the result in (2). At this time, the NTA 2 and the application agents 4 and 8 use a thesaurus 6 containing the semantic hierarchy of words to grasp the meaning of the memo. In addition, KQML, which is a representative language and protocol, is used for communication between the NTA 2 and the application agents 4 and 8. It defines the forms of performance required for dialogue between agents, such as Tell, Ask, Reply, Sorry, and so on.

Here, the schedule management agent 8 in charge of the scheduling application has a structure as shown in FIG. 5, and receives a general template and execution forms such as Tell and Ask in a message form from the NTA 2 and processes it. That is, the schedule agent 8 analyzes the message received from the NTA 2 and retrieves the schedule that has automatically recorded or registered the schedule in the schedule program 12 on behalf of the user.

Specifically, when the NTA 2 requests the calendar agent 8 to record a memo entered by the user as a schedule, or to retrieve a schedule corresponding to the query entered by the user, the calendar agent 8 may request the NTA. The schedule template is constructed by extracting information important for schedule management from the generic template delivered from (2) through the NTA interface. An example of the configuration of this schedule template is as shown in FIG.

The schedule template includes slots related to the time describing the schedule, such as general repetition flags, recurrence conditions, start dates, and slots related to the contents of the schedule, such as place, companion, and action. Time-related slots are constructed using normalized time expressions described in the temporal portion of the generic template, and content-related slots are used to analyze the parsing results of generic templates using the thesaurus (6) and stopword-list. It is constructed using one result.

When the NTA 2 requests the schedule registration, the scheduling agent 8 classifies the previously configured schedule template by using the time type slot's value and behavior, and the attributes of the words in the keyword slot. Drive the schedule management program 12 to store the schedule.

In addition, when the NTA 2 requests a schedule query, the schedule agent 8 compares the schedule template configured from the user's query with the schedule template previously read and configured the schedule template corresponding to the user's query. Search for. At this time, the comparison between the schedule templates is largely divided into the comparison of the time portion and the comparison of the keyword portion. The comparison of the time parts is done by converting the contents of the time slots in the two schedule templates to be compared to a set of time points and time periods, each representing absolute time, and then comparing the two sets. The comparison of the content parts is made by comparing the meanings of the words in the accompanying, place, action, keyword slots of the two templates using the thesaurus 6. For example, if a word in that slot of a calendar template constructed from a query and a word in that slot of a calendar template constructed from a saved query are the same word, synonym, or parent-child, the calendar agent may I think it has a connection.

In this way, the NTA (2) performs pre-processing on the natural language input from the user before morphological analysis, enabling analysis of memos that cannot be analyzed by a general sentence analyzer, and combining the morpheme analysis and parsing by the number unit. By reducing the target, it is effective to increase the efficiency of sentence analysis. In addition, the scheduling agent not only enables the application program through a simple natural language memo by running an application program on behalf of a user, but also effectively searches a schedule by analyzing the memo.

As described above, according to the schedule management method from the natural language memo sentence according to the present invention, the schedule management agent system receives the natural language schedule memo familiar to the user to automatically organize and record it and to search efficiently Has the effect of maximizing. In addition, the present invention can be applied to a variety of applications such as e-mail management, cashier, information retrieval, so that the user can easily use the natural language without learning how to use the application program processing each application field have.

Those skilled in the art will appreciate that various changes and modifications can be made without departing from the technical spirit of the present invention. Therefore, the technical scope of the present invention should not be limited to the contents described in the detailed description of the specification but should be defined by the claims.

Claims (3)

In a method for managing a schedule by processing a natural language memo sentence input by the user in an agent system consisting of an interface agent responsible for input and output to the user, and a plurality of application agents and applications according to the application, Extracting information from the natural language sentence input by the user in the natural language processing portion of the interface agent, constructing a generic template that is a general-purpose information structure, selecting the application agent in the dispatcher portion, and delivering the generic template to the selected application agent; A schedule agent, one of the application agents, extracts schedule-related information from the generic template, constructs a schedule template, and analyzes the time and keywords in the schedule template to mean registration and retrieval to a schedule management program. A schedule management method from a natural language memo sentence, characterized in that it comprises a step. The method of claim 1, And the interface agent and the schedule management agent use a thesaurus containing semantic hierarchies of words to grasp the meaning of the memo sentence. The method of claim 1, The natural language processing portion of the interface agent A preprocessing step of performing sentence reconstruction and unity combining on the natural language sentence input from the user; A morpheme analysis step of analyzing a morpheme of a sentence applied from the preprocessing step; A parsing step of parsing the syntax so as to form a complete sentence to be analyzed; And a post-processing step of selecting an appropriate sentence in accordance with the syntax analysis result.