Attached are three gifs representing screen dumps from our Keizai demo. You may want to find a place for them in the workshop report.
In addition, here are a few notes/comments on each gif.
1) Keizai is a multi-lingual information retrieval system. In this screen the user has selected a Korean newspaper data base as the search space (Donga on-line news 1999) and typed in an English query term "machine translation."
After selecting "Translate query", the results show up in the bottom frame. Based on a bilingual dictionary and the data base corpus itself, many possibilities for Korean search terms are provided, along with an English translation. The user may select any number of terms to be included in the search query, based on the user's knowledge of Korean or on the basis of the English translation of the query term and context.
2) Once a query has been submitted, a set of ranked documents is returned. In addition to highlighting the query term that produced the document, proper names and places occurring in the document are listed and translated, to help determine the relevancy of the document.
The document currently under scrutiny is marked in the left column and a summarization (in Korean) appears in the right frame.
3) The whole document may be requested and translated, or merely the summarization. Shown here is the translation of the summarization. Please remember that the translation system is a simple word-for-word translation, with no syntactic or word-choice component.