Document State

We are now ready to define a new model for documents, which will be called the document state model. A document can be seen as an entity present at some point in time. The document as a unit has some state within that instant of time. This document will also have some properties and relations. The properties of the document and its relationship to other documents can also be seen as units of information with state.

  
Figure: A breakdown of document state

Figure  illustrates this model of document state. Some portions of the document state can be derived automatically, others by semantic interpretation of a user when reading the document. We can separate the document state into three constituent parts (based on properties and relations):

• Internal State: Properties that are of this type generally refer to the content or internal representation of the document. For example we can say that a document contains the word ``apple.'' Generally, things that are part of the internal state of the document are properties we can define by looking inside the document. These properties may be non-obvious, and may be derived by some semantic understanding of the document.
• External State: Properties that can be considered meta-data are generally held as part of external state. The author, date of publication, and location are good examples of external state. However, it is important to note that properties are not necessarily exclusive (i.e. a property can be part of more than one state). For example, the author of a document can be ``David'' by virtue of the fact that we know that David wrote it and emailed it to us. This knowledge is part of some external state. By reading the document we can find that it is ``by: David.'' So we have derived an authorship relation by semantic understanding, and so the author being ``David'' is also part of the internal state.
• Relative State: There are a set of fuzzy properties of a document that encapsulate relationships between one document and another. One document can cite another or link to it. These relationships can sometimes be derived automatically. Sometimes, however, the relationships are derived based on some semantic understanding of a user. This semantic undestanding may be obvious to everyone, to a limited group of users, or to the single user alone. For example, one document can be related to another by virtue of the fact that the reader read both while eating a chocolate ice cream cone. As stated before, we will place non-trivial inter-document properties in this category.

The final important idea to consider is that state of a document may change. We will make the claim that documents present state is also a function of its previous sates. For example, while a document (call it myThesis) may have been located on goose.lcs.mit.edu, it may have subsequently moved to haystack.lcs.mit.edu. A user looking for the document myThesis by constraining his search system to the previous location may come up empty handed if the system does not remember that the document used to be on goose.lcs.mit.edu. Any piece of state that has previously existed may be of importance to the user who will later look for the document based on this previous state. Additionally, the more expressive an implemented system can be in maintaining all the state information of a given document, the better it will be able to respond to user queries.