Data reliability is very important in storage systems. To increase data reliability there are techniques based on error-correcting codes(ECCs). These techniques introduce redundant bits into the data stored in the storage system to be able to do error correction. The error correcting codes based error correction has been studied extensively in the past. In this thesis, a new error correction technique based on the redundant information present within the data is studied. To increase the data reliability, an error correction technique based on the characteristics of the text that the data represents is discussed. The focus is on correcting the data that represents English text using the parts-of-speech property associated with the English text. Three approaches, pure count based approach, two-step HMM based approach and bit error minimization based approach, have been implemented. The approach based on two-step HMM has outperformed the other two approaches. Using this newly proposed technique with the existing error correcting techniques would further increase the data reliability and would complement the performance of existing error correcting techniques.
