Natural Language Processing Principles for Improving Deaf Writing

这项计划的目标是设计一个电脑系统，以提高听力受损人士的写作能力，这些听力受损人士的写作能力落后于听力正常人士。这种系统的好处取决于系统检测错误并提供有益的纠正建议的能力。这项工作将借鉴第二语言习得方面的工作，对美国手语和英语的语言比较，以及对听力受损个体的写作样本进行分析，以便对预期错误进行分类。预期的错误将作为注释的“错误规则”添加到英语语法中。在解析包含已识别错误之一的句子时，系统将使用注释来生成纠正响应。然后可以对用户进行查询，以获取纠正用户输入句子所必需的信息。最后，将生成一个正确的句子。除了对听力受损人群的潜在好处之外，这项工作将导致真正的智能计算机辅助教学系统，这将需要当前自然语言处理能力的进步。特别是，将在处理格式不良的输入以及为生成句子识别足够的表示方面取得进展。

The objective of the project is to design a computer system aimed at improving the writing skills of the hearing impaired whose writing ability lags behind that of their hearing counterparts. The benefit of such a system hinges on the system's ability to detect errors and to offer beneficial corrective advice. The work will draw on work in second language acquistion and a linguistic comparison of American Sign Language and English as well as analysis of writing samples from hearing impaired individuals, in order to develop a taxonomy of expected errors. The expected errors will be added as annotated "mal-rules" to a grammar of English. Upon parsing a sentence containing one of the identified errors, the system will use the annotations to generate a corrective response. The user may then be queried in order to acquire information necessary to correct the user's input sentence. Finally, a correct sentence will be generated. In addition to the potential benefit to the hearing impaired population, this work will lead toward a truly intelligent computer aided instruction system which will require advances in current natural language processing capabilities. In particualr, advances will be made in dealing will ill-formed input as well as in identifying sufficient representations for sencence generation.