The Reader's Helper: A Personalized Document Reading Environment

Summaries

The annotation agent automatically generates a summary of the document based on the concept hits found in the text. For instance, given the annotated document in figure 2, a user might want to read only the sentences that have hits in them for the concept Design. This document is generated automatically by the RH after it is done analyzing the document. Figure 3 shows a summary of the document annotated in figure 2. The toolbar button invokes the summary function to present the reader with a list of available summaries. In figure 3 each sentence containing a keyword phrase for the concept Design is listed in the order of appearance in the document . This interface provides a way for the reader to view all relevant sentences at once. Associated with each sentence is a hypertext link back into the same sentence in the annotated document. Another summary generated by the RH lists all key phrases for each concept and the number of hits each phrase has in the document. Each hit is represented as a hypertext link which points back to the location of the hit in the annotated document. The value of this summary is to show how broad the coverage is with respect to the concept's key phrases.

The use of RH outside the target document

The RH also provides access to an archive of previously annotated documents. Each time a document is annotated in the RH, a local copy of the document is stored in the reader's private document archive so that it can be accessed again.

Calendar Interface

Readers can use a calendar interface to the archive for retrieving documents based on their date of annotation. The calendar keeps track of all documents annotated by recording the date, time and the most relevant concept on the day the annotation occurred. There are two ways to view the calendar: as a standard calendar month (days listed in columns from left to right, etc.) with entries on each day an annotation occurred, or as a timeline so that more information about the document can be presented to the user.

Similar Documents Interface

Another way to access the document archive is by viewing documents that are similar in content. This is possible using the similar documents interface. Each time a document is annotated, the RH annotation agent records how each concept in the reader's profile performed (i.e. the concept score is logged with respect to the document's unique ID). When a reader requests to see all documents which are similar to the current document in the browser, the system uses the concept information for the current document to generate a list of documents from the archive that have scored similarly. For instance, in figure 4 we see the current state of the concepts after annotating a document ( & ). The ExpSys concept (Expert Systems) has a score of 80% and the NLP concept (Natural Language Processing) has a score of 63%. Now the system must generate a collection of documents which have scored similarly to the way this document scored. In other words, what documents in my collection or the group's collection have an emphasis on both expert systems and natural language understanding? Because the RH annotation agent records all concept scores for each document annotation, the answer can be easily generated and presented to the reader. Figure 4 shows the title of the document previously annotated which generated the scores in the concept area: Adding Intelligence to the Interface. The similar document subsystem creates a list for each concept of the documents which have scored above a user defined threshold (e.g. 50%). The list is composed of the title of the document which is a hypertext link to the document in the archive. From this list we can see that one document scored as well as the original document did with both concepts: The Expert Surgical Assistant: An Intelligent Virtual Environment with Multimodal Input. This document is emphasized in the area as MOST SIMILAR since it is most like the document we have currently been viewing.