Speech recognition with score calculation

A speech recognition method for calculating a score indicating a likelihood that one or more speech recognition results of input speech including an unknown word is a correct speech recognition of the unknown word, and for performing continuous speech recognition on said speech based on the score, said speech recognition method comprising the steps of:

• extracting features of the input speech;
• storing a dictionary database that is divided into a standard dictionary area and an unknown word dictionary area;
• calculating said score of a word sequence for each candidate for the unknown word in the word sequence by using the extracted features on the basis of the unknown word dictionary area in which unknown-word-forming elements are stored, said elements forming a speech recognition result corresponding to an unknown word, and for classifying the speech recognition result by an attribute thereof;
• and
• selecting a word sequence of speech recognition results, which represent the original input speech on said score,
• wherein the unknown word dictionary includes a general purpose branch and an attribute branch connected in sequence, a loop branch is connected to the general purpose branch, wherein unknown words are repeatedly processed through the general purpose branch a predetermined number of times after going through the loop branch.

A computer-readable medium storing a computer program that when executed on a processor performs a method for calculating a score indicating a likelihood that one or more speech recognition results of input speech that includes an unknown word is a correct speech recognition result of the unknown word and for performing continuous speech recognition on the input speech based on the score, said method comprising the steps of:

• extracting features of the input speech;
• storing a dictionary database that is divided into a standard dictionary area and an unknown word dictionary area;
• calculating said score of a word sequence for each candidate for the unknown word in the word sequence by using the extracted features on the basis of the unknown word dictionary area in which unknown-word-forming elements are stored, said elements forming a speech recognition result corresponding to an unknown word, and for classifying the speech recognition result by an attribute thereof;
• and
• selecting a word sequence of speech recognition results, which represent the original input speech on said score,
• wherein the unknown word dictionary includes a general purpose branch and an attribute branch connected in sequence, a loop branch is connected to the general purpose branch, wherein unknown words are repeatedly processed through the general purpose branch a predetermined number of times after going through the loop branch.