Natural Language Processing

  • GOING Chinese IME

  • "GOING Chinese IME (also called Natural Chinese Input)" was originated from "Going Input" which was developed in 1990 by Dr. Wen-Lian Hsu of Acadamia Sinica of Taiwan, and is a Chinese input software built-in with artificial intelligence and rich Chinese semantic and lexicon database. After more than 10 years improvement, GOING Chinese IME has become a very popular and powerful Chinese input software in Taiwan and adopted by about 1.5 million users. It has always been the most popular commercial Chinese input product.
    Having large amount of Chinese semantic and lexicon databases, GOING Chinese IME has gradually become a very important platform for learning and teaching Chinese. Students, teachers, the parents of children, compatriots living abroad, even foreigners choose GOING Chinese IME to learning Traditional Chinese. Allowing users to type Simplified Chinese in Traditional Chinese environment and Traditional Chinese in Simplified one, this input software is common used by alien workers, international companies, and various online surfers who often need to communicate with net friends from China. For the convenience of the old and disadvantaged people, "Magnified Input Window", "Real-time Voice Reading", and "Braille Output", "Chinese Input Method Editor with Auto Correction" functions, Chinese text input is much more efficient and output text is much more readable and the most important contribution is this software successfully assist visually impaired individuals, individuals with learning disabilities, ALS patients and any other handicapped individuals to overcome the obstacles on Chinese text input. The latest version is version 8 which released on April 28, 2005 and this project is ongoing.

    webiconDemo Site URL:

  • Template Generation

  • We propose a template generation approach, called alignment-based surface pattern (ABSP), which integrates semantic information into syntactic patterns for question answering (QA). ABSP uses surface patterns to extract important terms from questions, and constructs the terms’ relations from sentences in the corpus. Surface patterns are syntactic patterns that connect answers and question keywords. A number of online QA systems use patterns to deal with users’ questions and extract answers. ABSPs are generated from question-answer pairs regardless of the question type. The surface patterns, which are automatically generated and selected from training data for any kind of question type, can capture important relations between a question’s terms and the correct answer. In situations involving multiple question keywords and multiple passages, several ABSPs are used together to calculate a score for an answer. They can be used in cross-language QA systems. The proposed method is based on sequence alignment. It incorporates local alignment algorithms and uses dynamic programming to align sentence pairs. Then, we extract similar (syntactic and semantic) parts of the sentences as surface patterns.