1 Introduction

There are many studies and applications which manage a schedule to support personal actions. However there are not enough systems which support our action by using personal Action history.

Almost of all our actions in daily life are similar to the past actions. It is not smart to make such a plan on each occasion. If we are able to reassess and arrange past acitons statistically, we would make better plans.

In this research, we propose the system "AcTrec" for the purpose of supporting the personal action. It becomes aware of the user's situation and records personal actions. By using such action data, it can evaluates and corrects user's schedule. AcTrec consists of three elements; "personal situation awareness and recording", "retrieval of the action history" and "cooperation with a schedule".

2 Personal Situation Awareness and Recording

Some researches on action awareness are done. They install various sensors in real space, and are aware of action by collecting the information acquired from there. However, the sensor embedded in environment is capable of limited application. So these systems are not realistic.

In AcTrec, context information, such as contents of operation, a place, and time, is acquired with the Personal Digital Assistant (PDA) customized for individuals.User selects a action in some groups categorized at the start of the action, then enters information about the place and the description of its action at the end of the action. Thereby, the system gets the context information. In the future, we aims at acquiring context information in half-automatic by guessing a personal situation from the the relevance of a place, time and a action. We also considers adopting GPS to get information.

The context information acquired in the PDA is sent to the database server. At that time, it is transformed into XML data in consideration of the convenience of future operations.

3 Retrieval of the Action History

In this system, user can retrieve the personal action history accumulated at the database from a web browser. Retrieval system consists of servlet, relational database (RDB) and XML database (XMLDB).We adopt PostgreSQL for RDB and Xindice for XMLDB. The rough flow of the retrieval is as follows.

1. Enter the search condition from a web browser.

2. Servlet which receives the search conditon retrieve RDB, and takes relevant records.

3. Servlet takes XML data corresponding to the records from XMLDB.

4. Servlet transforms XML into HTML by Style Seet, and displays a result list on a browser.

Simultaneously with the display of a result list, statistics data is also displayed such as the average action pattern and the time distribution of a searched action within the specified period.Statistics data is obtained by extracting and processing the group and time zone of action from a retrieval result.In this system, 24 hours is divided at intervals of 1 minute, the frequency of appearance according to action group is measured for every per minute, and action with the highest frequency of appearance is used as the average action pattern in the time zone in each time zone (Figure.). Graph is categorized by color according to action group, and what has high frequency is displayed more deeply. Therefore, we are able to recognize the average action pattern in the specified period visually.Similarly, a graph of the time distribution of the searched action within the specified period is also displayed. These processings and the result of those are used also in the cooperation with a schedule.

Fugure1: (a)Average action pattern from Monday to Friday, (b)Average action pattern on Saturday and Sunday

4 Cooperation with a Schedule

By statistical work to the context information which is collected by personal situation awareness and recording, it is possible to make our schedule better. In other words, we can generate the optimal plan in alignment with the past action, and evaluate and correctio a schedule . AcTrec generates and corrects schedule along with the following flows.

1. A user selects kind of actions.

2. A system displays the average action pattern, the time distribution of the searched action and the similar action record in past which are obtained by carrying out statistics processing of the past action data.

3. A user inputs a schedule based on the shown statistics data.

4. A system saves the inputted data in the database for schedule accumulation, and a user downloads a schedule to PDA.

5. When there is a gap between a schedule and actual action, a system presents a correction proposal.

6. A user corrects a schedule to reference for the shown data.

A user generates and corrects schedule from a web browser. Xindice which is XML database is used for the database for schedule accumulation.

By displaying statistics data, a user becomes possible examining how much time should be assigned about that schedule.

When the scheduled action inputted is not done, a system suggests the correction of schedule. The procedure of its correction is the same as that of the generation almost.

5 Summary

We propose the system "AcTrec" which becomes aware of the user's situation and records personal actions for the purpose of supporting the personal action. From now on, we are going to advance mounting about evaluation and correction of the schedule, and to examine validity.