Dictum - выражение, слово, изречение, острота, обещание, предписание, приказание (лат.)
 
Citation extraction system Dictum®
Man is all made up of questions, while life and
the world around - of answers to these questions.
B.Akunin. The Diamond Chariot.


Natural language is the most widespread knowledge representation language. Text is one of the forms of transmitting knowledge (and thoughts) through time and space; it can be read centuries later and thousands miles away from its origin place. There are huge amounts of electronic knowledge bases in the form of natural languages texts. Modern searching systems allow to find documents, which contain the necessary information. However users must extract the required information from those documents by themselves. "Required information" is understood to mean an answer to a question, which has made one to use the searching system. If it is not a large text then answer extraction is not difficult. Answer extraction becomes a hard-to-solve problem as text grows in size.

This problem gets harder with redundant wordings characteristic of special literature, particularly legal literature. What is meant is redundancy to the user's question. For example, an article from Fiscal Code may take up several printed pages, whereas syntactically it is a single Russian sentence. In most cases a user is interested not in the entire article, but in a little part, which is related to his/her question.

For exaplre, the articlce from Traffic Rules, which regulates stopping of a vehicle, has volume about half-pages whereas the citation with a key word "tunnel" consists only of five words, namely "Stopping is prohibited in tunnels". In this example length of citation relative to the lehgth of original sentence makes 2%.
Thus the citaion is an extract having a quantity indicator, which characterizes percentage of the relevant information in the found sentence.

Computer program DictumR developed by DictumR Software Lab is designed for automatic citation extraction from natural language texts. A user's question is formulated in the usual form as a set of key words. Compactness of extracted citations reduces intellectual expenditures on the analysis of the received information and alluws setting up access to text information with the help of a mobile device with the small display size, in particular cellular phone.

You may become acquainted with work of the system based on the current Russian Traffic Rules in online-mode. Access to a demo version for mobile phone is organized from the address wap.dictum.ru.

Computer adviser
© Dictum Ltd., 2003 - 2009