Information processing device, method and computer program product for processing voice recognition data

1. An information processing device comprising:

a memory controller to store in a memory:

a plurality of character strings included in a voice text obtained by performing a voice recognition on voice data;

a node index indicating start position information and end position information of each of the character strings in the voice text;

a voice recognition score obtained by performing the voice recognition; and

a voice index of voice position information indicating a position of each of the character strings in the voice data;

a detector to detect reproduction section information which indicates a reproduced section of the voice data;

an obtainer to obtain reading information, which represents at least a part of a character string representing reading of a phrase in a text written down from the voice data that has been reproduced, and obtain insertion position information, which indicates a character insertion position in the written text;

a searcher to specify, as a target character string for searching, a character string that, from among the plurality of character strings stored in the memory, has the corresponding voice position information included in the reproduction section information, and search the specified character string for a character string that includes reading indicated by the reading information;

a determiner to, when a value of the voice recognition score corresponding to the character string obtained by the searcher is equal to or greater than a display threshold value, determine to display the character string obtained by the searcher;

a display controller to display the character string, which is determined by the determiner, on a display unit;

a history updater to perform storage control to store, in a candidate history memory, candidate history data in which the character string obtained by the searcher, the voice recognition score, and the character insertion position are associated, and to update the candidate history data according to a change in a text;

a selector to, when the character string displayed by the display controller is subjected to a selection operation, select the character string; and

a threshold value updater to decide on the display threshold value, which is used in comparison with a voice recognition score by the determiner, using at least one of a voice recognition score of the candidate history data and a voice recognition score of the character string selected by the selector.

2. The device according to claim 1 , further comprising a follow-on searcher to search for a character string which, as a node index of a start position thereof, has a node index of an end position of the character string selected by the selector, wherein

the display controller displays, on the display unit, the character string which is retrieved by the follow-on searcher.

3. The device according to claim 1 , further comprising a multiple-candidate-presentation instruction receiver to receive a multiple-candidate-presentation instruction, wherein

the searcher searches for a character string corresponding to the character insertion position obtained by the obtainer, and

the display controller calculates a position for displaying an input candidate from the character insertion position obtained by the obtainer, the number of character strings retrieved by the searcher, and a length of each character string, so as to display the character string retrieved by the searcher on the display unit.

4. The device according to claim 3 , wherein

when the searcher cannot retrieve a character string corresponding to the character insertion position, the searcher searches for a character which coincides with a character present before a certain number of characters than current character insertion position, and

the display controller displays the retrieved coincident character as the input candidate on the display unit.

5. The device according to claim 1 , further comprising a converter to convert a lattice structure obtained by the voice recognition result into a trie structure.

6. The device according to claim 5 , wherein the converter includes

a generator to generate, from the voice data, a word lattice having a voice recognition score;

an extractor to, from each arc of the word lattice, extract M number of best arcs in descending order of the voice recognition scores, M being an integer; and

a storage to store therein identification information, which is used in identifying the M number of best arcs that are extracted, in a trie tree.

7. An information processing method comprising:

storing, by a memory controller, in a memory:

a plurality of character strings included in a voice text obtained by performing a voice recognition on voice data;

node index indicating start position information and end position information of each of the character strings in the voice text;

a voice recognition score obtained by performing the voice recognition; and

a voice index of voice position information indicating a position of each of the character strings in the voice data;

detecting, by a detector, reproduction section information which indicates a reproduced section of the voice data;

obtaining, by an obtainer, reading information, which represents at least a part of a character string representing reading of a phrase in a text written down from the voice data that has been reproduced, and obtaining insertion position information, which indicates a character insertion position in the written text;

specifying, by a searcher, as a target character string for searching, a character string that, from among the plurality of character strings stored in the memory, has the corresponding voice position information included in the reproduction section information, and searching the specified character string for a character string that includes reading indicated by the reading information;

determining, by a determiner, when a value of the voice recognition score corresponding to the character string obtained at the searching is equal to or greater than a display threshold value, to display the character string obtained at the searching;

displaying, by a display controller, the character string, which is determined at the determining, on a display unit;

storing, by a history memory control unit, in a candidate history memory, candidate history data in which the character string obtained at the searching, the voice recognition score, and the character insertion position are associated;

updating, by a history updater, the candidate history data, which is stored in the candidate history memory, according to a change in a text;

selecting, by a selector, when the character string displayed at the displaying is subjected to a selection operation, the character string; and

deciding on, by a threshold value updater, the display threshold value, which is used in comparison with a voice recognition score at the determining, using at least one of a voice recognition score of the candidate history data and a voice recognition score of the character string selected at the selecting.

8. A computer program product comprising a non-transitory computer readable medium containing an information processing program, wherein the information processing program, when executed by a computer, causes the computer to function as:

a memory controller to store, in a memory:

a plurality of character strings included in a voice text obtained by performing a voice recognition on voice data;

a node index indicating start position information and end position information of each of the character strings in the voice text;

a voice recognition score obtained by performing the voice recognition; and

a voice index of voice position information indicating a position of each of the character strings in the voice data;

a detector to detect reproduction section information which indicates a reproduced section of the voice data;

an obtainer to obtain reading information, which represents at least a part of a character string representing reading of a phrase in a text written down from the voice data that has been reproduced, and obtain insertion position information, which indicates a character insertion position in the written text;

a searcher to specify, as a target character string for searching, a character string that, from among the plurality of character strings stored in the memory, has the corresponding voice position information included in the reproduction section information, and search the specified character string for a character string that includes reading indicated by the reading information;

a determiner to, when a value of the voice recognition score corresponding to the character string obtained by the searcher is equal to or greater than a display threshold value, determine to display the character string obtained by the searcher;

a display controller to display the character string, which is determined by the determiner, on a display unit;

a history updater to perform storage control to store, in a candidate history memory, candidate history data in which the character string obtained by the searcher, the voice recognition score, and the character insertion position are associated, and to update the candidate history data according to a change in a text;

a selector to, when the character string displayed by the display controller is subjected to a selection operation, select the character string; and

a threshold value updater to decide on the display threshold value, which is used in comparison with a voice recognition score by the determiner, using at least one of a voice recognition score of the candidate history data and a voice recognition score of the character string selected by the selector.