Errors are the most common phenomena while typing texts. Different kinds of errors may occur while typing texts. To help correct these errors in Bangla text, there have been much research and many works but most of them are not context-sensitive. Bangla in itself is a language with a complex structure, which makes it a very difficult language to work with, but it is also one of the world’s most spoken languages. Around 210 million people all over the world use it as either their first or second language. So, error detection and correction in Bangla typing is a significant field of work. In the present work, the focus is to find the best language model to detect and correct errors in Bangla typing among some language models with a very large corpus to gain the best accuracy by a context-sensitive approach. To be precise, six language models are chosen to work with. All of which are based on the N-gram algorithm. Along with each of those models, the minimum edit distance is calculated. After finding a possible correction word, accuracy and failure rate were calculated to proceed toward the decision on choosing the most optimum one among these six language models.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

An Indicative-Metric-Based Context-Sensitive Approach to Autocorrect Bangla Spelling

  • Rahnuma Islam Meem,
  • Faisal Bin Abul Kasem,
  • Muhammad Ifte Khairul Islam,
  • B. M. Marjan Khan,
  • Md. Tarek Habib

摘要

Errors are the most common phenomena while typing texts. Different kinds of errors may occur while typing texts. To help correct these errors in Bangla text, there have been much research and many works but most of them are not context-sensitive. Bangla in itself is a language with a complex structure, which makes it a very difficult language to work with, but it is also one of the world’s most spoken languages. Around 210 million people all over the world use it as either their first or second language. So, error detection and correction in Bangla typing is a significant field of work. In the present work, the focus is to find the best language model to detect and correct errors in Bangla typing among some language models with a very large corpus to gain the best accuracy by a context-sensitive approach. To be precise, six language models are chosen to work with. All of which are based on the N-gram algorithm. Along with each of those models, the minimum edit distance is calculated. After finding a possible correction word, accuracy and failure rate were calculated to proceed toward the decision on choosing the most optimum one among these six language models.