Nomura Laboratory at the Kyushu Institute of Technology has developed a software of automatic translation called TESS (TExt Summarization System). The strategy for summarization consists of the following steps
The tags that are given to sentences during the first step contain linguistic informations (e.g. question, opinion, expression of a desire,judgment etc...) which are mainly determined mainly by the structure of the end of the sentence in Japanese, as well as information about the relationships between sentences (e.g. addition, parallelism, contradiction etc...) determined using various linguistic information (e.g. anaphora, conjunctions etc ...). The importance of a sentence is quantified from the information written on the corresponding tag, and the system keeps only the sentences that are judged to be important. The selected sentences are finally rearranged in order to obtain a summary in correct natural language (pronoun rearrangement, division of complex sentences etc...).