EATclassifier is a system for identifying expected answer types of the questions.
It has been written as part of an internship in 2009 by Md. Faisal Mahbub Chowdhury.
The EAT is an automatically assigned semantic label (e.g. “PERSON”, “LOCATION”,
“DATE”) which is used by the system to reduce the search space of candidate
answers. In the search of the exact answer to a natural language question, in
fact, only the concepts matching the EAT category assigned to the question
are usually extracted from the retrieved passages, and considered as potential
The system is freely available for usage.
The main class is Processor.java.
Contents of folders other than "/src" -
1. All the test and training files are generated inside "/english" and "/italian" folders.
2. The n-gram statistics file should be kept inside "/ngrams" folder
3. All the training questions in plain text form is written inside "/questions" folder
4. "/svm_multiclass_windows" contains exe for experimenting with the generated test and training files
5. "/Text-NSP-1.09" contains n-gram statistics package.
To use this program, TreeTagger installation is also required which is freely available in web.
##### ----- Please cite the following paper if you use any part of this program:
“Expected Answer Type Identification from Unprocessed Noisy Questions”, by Md. Faisal Mahbub Chowdhury, Matteo Negri, In Proceedings of the 8th International Conference of Flexible Query Answering Systems (FQAS 2009), volume 5822 of Lecture Notes in Computer Science, pages 263-274. Springer, 2009.
Please refer to the above paper for details about the system.
Please write to email@example.com for any question regarding the system or the paper.