Introduction
National Diet Library (NDL) has been conducting library cooperative projects such as the National Union Catalog Network, training and exchange programs. In 2002, NDL launched the Collaborative Reference Database Project. This is an experimental project that examines new possibilities of reference service suited to the Internet age. I will outline and introduce the current state of this unique and unprecedented project.
 

1. Outline

1.1 What is the Collaborative Reference Database Project?
Apart from the NDL, there are thousands of public, university and special libraries in Japan. These libraries provide reference service every day. Reference service is an intellectual activity that links library users and information resources. However, records and practical information about the service have not been utilized so far even if they were recognized as valuable information. We believe that a database that accumulates case data, in other words, records of inquiries, answers and know-how on reference, will be valuable and useful to support reference librarians and to enable users themselves to find information they need. 
The Collaborative Reference Database Project collects case data from the NDL and other libraries and promotes library cooperation in reference service that will be conducted focusing on the database.
We called widely for participation regardless of library types such as national, public, university and special, and about 150 libraries joined in the project. This shows how high the libraries' expectation is for this test project. We are trying hard to meet their expectations. 

1.2 Background
The Collaborative Reference Database Project was set up in 2002 when the Kansai-kan of the NDL was established. It will shift to full-scale operation after a three-year experimental stage. The main schedule of the three-year test operation is as follows:
 

1.3 Verification items 
The verification items to be examined through the operation are as follows:

• Effectiveness of reference examples and other data when released on the web
• Improvement in libraries' reference works enabled by the integrated database of reference examples from libraries
• Methods of library cooperation around the database of reference examples
• Standard format for exchanging and sharing reference examples

2.1 Data
Data that will be accumulated in the Collaborative Reference Database System are the following:

• Data of reference examples
• Data of manuals to find resources on specific subjects 

* Above two data are collectively called "case data," and will form the main part of the data. 
• Data of special collections (e.g. private collections held in each library)
• Profile data of member libraries

2.2 Basic construction of the system
The system has been developed around the case data provided by member libraries to have functions such as data registration, data management, search, and support to the member libraries. Member libraries can access the system directly through the Internet to register, revise, delete, and search the data.
When the system is open to the public, general users will also be able to search the data with a web browser.

<DIAGRAM:  System framework>   *Click the image for a larger and clearer one.

We invited the first participating libraries from September to November 2003 and 148 libraries became members (NDL and two of its branch libraries located in the executive and judicial agencies of the government, 78 public libraries, 51 university libraries, and 16 special libraries). 

<Member libraries divided by category of libraries>

<Member prefectural libraries (in color)>

We have found many things to consider before the beginning of the third year of the experiment:

1. Organizing an instruction manual and guidelines on how to make case data

As many libraries do not have the experience of making a database of reference case data, librarians do not share a common idea on how to edit the data. In order to protect users' privacy as well as to maintain the quality of the case data, we need to draw up an instruction manual and guidelines.
2. Securing the number of participating libraries and data (to take advantage of scale)

In order to make the database useful to both member libraries and general users, we need larger number of member libraries and input of numerous high-quality case data.
3. Extending functionality

We assume that there will be requests for improvement in the current system, such as search functions and communication assistance.