====================================================== THE BIDIRECTIONAL LTAG DEPENDENCY PARSER Libin Shen lshen@bbn.com java -classpath parser -quick For example, the parser is installed at ~/binc/, then you can use the following command to parse a sample file, ~/binc/pardata/text_pos.smp. java -classpath ~/binc/binc_01.jar -Xmx1500m parser -quick ~/binc/pardata/text_pos.smp ~/binc > output It is recommended to run it on a Linux machine with 1.5 GB MEM or more. It is required to run it with JDK 1.5 or JRE 5.0. The input of the parser is in Ratnaparkhi's POS tag format, which is WORD1_POS1 WORD2_POS2 WORD3_POS3 ... Empty categories with POS tag NONE are filtered out in pre-processing. pardata/output.smp showes the output of text_pos.smp Sentence 2 shows an example of adjunction, and sentence 8 shows an example of predicate coordination. The parser runs at the speed of more than 4 sentences a second on a Xeon 3G Hz CPU, but it may take a couple of minutes to load about 4.6M features at the beginning. Please let me know if you have any questions about the parser. Have fun. ======================================================