The Just-IN-TIME INFORMATION RETRIEVAL (JITIR) agent is a user-related information based on the user's current context information, providing information related to the user's current work context, and does not require a clear instruction. It can be seen as a search engine that does not need to provide a query.
Context-related environments may be a current working environment on a computer, such as email, webpage, and document, or a natural environment in which people are located, by camera, microphone, GPS, etc., and pass to the computer.
When Jitir obtains information related to the user context, as much as possible is provided to the user in the form of non-dispersible user attention, but this information is also accessible. After the current context environment of the user changes, the information provided by Jitir will change.
The three important features of Jitir: the initiative, the non-mandatory and accessibility of information providing, and the information provided are always related to the current context of the user.
Active: People from search engines, such as Yahoo! Before getting information, there will be some questions in the mind and have some keywords. However, Jitir has more active, and does not need to have the idea of inquiry, and even don't even know about the information on the work environment. In addition, the information obtained by the user's query by the search engine is limited to the user's perception of the current working environment.
Non-mandatory but visible line: When a phone rang, it provides a message, some people call you, even from the electric music, I can know the identity of the phone, but it forcibly enters our life: no matter what we How important is the current work, can't be disturbed, it is scattered our attention. If the phone is closed, all calls are forwarded to the message box. Such a silent phone is impossible to affect our work, but its accessibility is greatly reduced. The information provided by Jitir is between the two extremes. The desktop alarm system is a similar example. JITIR provides information on the user as much as possible without dispersing the user's attention, and can even be ignored, but if there is information requested by the user, it can be accessed by the user. Regardless of the information provided is useful or useless, it is determined by the user to browse it.
Context Relationship: Announcement System, such as news services, alarm signals, all have initiative, but this information is independent of the user context. For example: a warning may sound, when receiving an email or stock price or has a news that is interested in users, interest-related news. The information provided by these announcement systems should attract the user's attention and pull the user from the current work. The system of this class is to provide new information from a changing information source to the user. JITIR is based on the constant changing context environment, providing information from a certain source of information to the user. It does not affect the current work of use, and more useful information related to user work is provided.
Auto Help System, like Microsoft Office Assistant, in line with Jitir definitions. However, the automatic help system is limited to a particular area, which only provides help information from a particular domain from a particular database.
Remembrance Agent (RA) is a JITIR running on Emacs, which is designed to suit the user's current work. It continuously displays a list of content related to the document currently being read or written at the bottom of the EMACS window and is arranged according to the correlation of the content. This list will change when the user is typed in new content, or changed the current environment.
The approach works as follows:
First, the user works, reads Email or WebPage, or writes the papers in the EMACS environment. RA will extract some of the current content list from a database based on the current work content of the user to the user at the bottom of the Emacs window. After 5 seconds, the current work content of the user has changed, and the list of information provided by the RA update. If a certain moment, the user looked at the information provided by RA. If you don't intend to this information, return directly to the current job. If it is interested in the information provided by the RA, the corresponding operation is made, and the RA will provide more detailed information based on the user's operation to provide more detailed information. If the user is right-click on an information of interest, the RA will pop up a small window above the message and display this information with the keyword related to the current work content. If the user wants to see a full text of a certain message, he can use the keyboard Ctrl and the number of this information (Arabic, with # instead) or click on this information number directly on this information number. The full text of the relevant information replaces the current work content displayed on Emacs. RA does not make the information provided to its own recursive calls. The user can also evaluate the information provided by RA, with 1 to 5 levels, which will be recorded in a log file, which is convenient for the accuracy of the information. If the user wants yourself to query, you can query information related to this information by clicking or selecting this information on providing information and typing Ctrl and Q, you will query information related to this information, or type Ctrl and F and enter you want Query keyword. Replacing the database can be used with a combination key Ctrl D. Margin Notes is a web-based agent that automatically adds annotations for this web page when a web is loaded into the browser.
Jimminy is also called wearable RA. It provides information for users based on the user's natural environment: where is the user, he is talking, and now. It supplies information to the user through a display device attached to the head on a wearable computer.
All the above three Jitir have a same information retrieval system as a background, SAVANT. He contains a document indexer and a retrieval engine. Document Index utilizes a template structure to index document according to the file type. For example: it can identify archive mail files and focus on source, topics, date, and mail text. The retrieval engine receives the selected text from the user's current working environment (read or written email, or a web page, etc.) and returns a list of documents that may have the most active on the current environment.
Savant contains two programs: RA-RETRIEVE retrieves based on query, RA-INDEX makes an index to retrieve the faster run.
Template match: Identify document or query type, word, index.
Data Fusion:
filter
CONTINUE ...
