DATA PROCESSING METHOD, PRESENTATION METHOD, AND CORRESPONDING APPARATUSES

• US 20140019133A1
• Filed: 06/24/2013
• Published: 01/16/2014
• Est. Priority Date: 07/12/2012
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Bao, Sheng Hua, Chen, Jian, Liu, Wen, Qin, Yong, Shi, Qin, Su, Zhong, Zhang, Shi Lei
• Granted Patent: US 9,158,752 B2
• Time in Patent Office: Days
• US Class Current: 704/257
• CPC Class Codes: G06F 40/20 Natural language analysis semantic analysis of natural language G06F40/30
• CPC Class Codes: G06F 40/279 Recognition of textual entities
• CPC Class Codes: G10L 15/18 using natural language modelling
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A data processing method includes obtaining text information corresponding to a presented content, the presented content comprising a plurality of areas; performing text analysis on the text information to obtain a first keyword sequence, the first keyword sequence including area keywords associated with at least one area of the plurality of areas; obtaining speech information related to the presented content, the speech information at least comprising a current speech segment; and using a first model network to perform analysis on the current speech segment to determine the area corresponding to the current speech segment, wherein the first model network comprises the first keyword sequence.

Claims

1. A data processing method, comprising:

obtaining text information corresponding to a presented content, the presented content comprising a plurality of areas;

performing text analysis on the text information to obtain a first keyword sequence, the first keyword sequence including area keywords associated with at least one area of the plurality of areas;

obtaining speech information related to the presented content, the speech information at least comprising a current speech segment;

using a first model network to perform analysis on the current speech segment to determine the area corresponding to the current speech segment, wherein the first model network comprises the first keyword sequence.

2. The method according to claim 1, wherein the using a first model network to perform analysis on the current speech segment to determine the area corresponding to the current speech segment comprises:

obtaining a confidence degree of at least one area keyword in the first keyword sequence, wherein the higher similarity an area keyword has with respect to the current speech segment, the higher confidence degree is obtained for the area keyword;

determining, if a first condition is met, that the area corresponding to the current speech segment is an area associated with an area keyword involved in the first condition;

wherein the first condition comprises at least one of the following;

that a confidence degree of an area keyword reaches a first threshold;

that all confidence degrees of multiple area keywords associated with a same area reach a second threshold;

that a sum of confidence degrees of multiple area keywords associated with a same area reaches a third threshold.

3. The method according to claim 2,

wherein the first model network further comprises a first phoneme sequence;

wherein the using a first model network to perform analysis on the current speech segment to determine the area corresponding to the current speech segment further comprises;

obtaining at least one phoneme adjacent to the current speech segment based on the first phoneme sequence;

determining a pronunciation similarity between the at least one phoneme and corresponding text information of the at least one area keyword, the corresponding text information of the at least one area keyword comprising a context of the at least one area keyword in the text information;

if the pronunciation similarity between the at least one phoneme and the corresponding text information of the at least one area keyword reaches a fourth threshold, raising the confidence degree of the area keyword with the pronunciation similarity reaching the fourth threshold.

4. The method according to claim 2, wherein the method further comprises:

obtaining a plurality of second keyword sequences, at least one of the second keyword sequences corresponding to at least one area of the plurality of areas and at least one of the second keyword sequence comprising at least one keyword;

obtaining a confidence degree of at least one keyword in at least one of the second keyword sequences, wherein a keyword having a higher similarity with the current speech segment has a higher confidence degree;

the determining, if a first condition is met, that the area corresponding to the current speech segment is an area associated with an area keyword involved in the first condition comprises;

if the first condition is met and a second condition also is met, determining that the area corresponding to the current speech segment is the area associated with the area keyword involved in the first condition, and taking the area corresponding to the current speech segment as the current area;

wherein the second condition comprises that a confidence degree of a second keyword sequence corresponding to the current area is less than a fifth threshold, the confidence degree of the second keyword sequence corresponding to the current area being obtained based on a confidence degree of a keyword included in the second keyword sequence corresponding to the current area.

5. The method according to claim 2, wherein the method further comprises:

obtaining a plurality of second keyword sequences, at least one of the second keyword sequences corresponding to at least one area of the plurality of areas and at least one of the second keyword sequences comprising at least one keyword;

using a second model network to perform analysis on the speech segment, the second model network comprising the second keyword sequences and a second phoneme sequence;

the determining, if a first condition is met, that the area corresponding to the current speech segment is an area associated with an area keyword involved in the first condition comprises;

if the first condition is met and a third condition also is met, determining that the area corresponding to the current speech segment is the area associated with the area keyword involved in the first condition;

wherein the third condition comprises that when using the second model network to perform the analysis on the current speech segment, the times of obtaining outputs based on the second phoneme sequence reaches a sixth threshold.

6. The method according to claim 1, further comprising:

obtaining a plurality of second keyword sequences, at least one of the second keyword sequences corresponding to at least one area of the plurality of areas and at least one of the second keyword sequences comprising at least one keyword;

using a second model network to perform analysis on the current speech segment to determine a keyword corresponding to the current speech segment, the second model network comprising the second keyword sequences.

7. The method according to claim 6, wherein the using a second model network to perform analysis on the current speech segment to determine a keyword corresponding to the current speech segment comprises:

obtaining a confidence degree of at least one keyword in the second keyword sequences, wherein the higher similarity degree a keyword has with respect to the current speech segment, the higher confidence degree is obtained for the keyword;

determining, if the confidence degree of the at least one keyword reaches a seventh threshold, that the keyword corresponding to the current speech segment is the keyword with the confidence degree reaching the seven threshold.

8. The method according to claim 7, wherein the second model network further comprises a second phoneme sequence;

wherein the using a second model network to perform analysis on the current speech segment to determine a keyword corresponding to the current speech segment comprises;

obtaining at least one phoneme adjacent to the current speech segment based on the second phoneme sequence;

determining a pronunciation similarity between the at least one phoneme and corresponding text information of the at least one keyword, the corresponding text information of the at least one keyword comprising a context of the at least one keyword in the text information;

if the pronunciation similarity between the at least one phoneme and the corresponding text information of the at least one keyword reaches an eighth threshold, increasing the confidence degree of the keyword with the pronunciation similarity reaching the eighth threshold.

9. The method according to claim 7, wherein the second model network further comprises a second phoneme sequence;

wherein the determining, if the confidence degree of the at least one keyword reaches a seventh threshold, that the keyword corresponding to the current speech segment is the keyword with the confidence degree reaching the seven threshold comprises;

obtaining at least one phoneme adjacent to the current speech segment based on the second phoneme sequence;

if the confidence degree of the at least one keyword reaches the seventh threshold, determining the keyword with the confidence degree reaching the seventh threshold to be a candidate keyword;

determining a pronunciation similarity between the at least one phoneme and corresponding text information of the candidate keyword, the corresponding text information of the candidate keyword comprising a context of the candidate keyword in the text information;

if the pronunciation similarity between the at least one phone and one piece of the corresponding text information of the candidate keyword reaches a ninth threshold, determining that the keyword corresponding to the current speech segment is the keyword of which the context is the text information with the pronunciation similarity reaching the ninth threshold.

10. A presentation method, the method comprising:

obtaining text information corresponding to a presented content, the presented content comprising a plurality of areas;

performing text analysis on the text information to obtain a plurality of second keyword sequences, at least one of the second keyword sequences corresponding to at least one area of the plurality of areas and at least one of the second keyword sequences comprising at least one keyword;

obtaining speech information related to the presented content, the speech information at least comprising a current speech segment;

obtaining a confidence degree of at least one keyword in at least one of the keyword sequences, wherein the higher similarity a keyword has with the current speech segment, the higher confidence degree is obtained for the keyword;

obtaining the confidence degree of the second keyword sequence corresponding to the current area of the plurality of areas based on the confidence degree of the keyword;

jumping away from the current area in response to the confidence degree of the second keyword sequence corresponding to the current area being less than a tenth threshold.

11. The method according to claim 10, further comprising:

obtaining a first keyword sequence, the first keyword sequence comprising an area keyword associated with at least one area of the plurality of areas;

obtaining a confidence degree of at least one area keyword in the first keyword sequence based on the first keyword sequence, wherein the higher similarity an area keyword has with respect to the current speech segment, the higher confidence degree is obtained for the area keyword;

wherein the jumping away from the current area comprises, if a fourth condition is met, jumping to an area associated with an area keyword involved in the fourth condition, and taking the area associated with the area keyword involved in the fourth condition as the current area;

wherein the fourth condition comprises at least one of the following;

that a confidence degree of an area keyword reaches an eleventh threshold;

that all confidence degrees of multiple area keywords associated with a same area reach a twelfth threshold;

that a sum of confidence degrees of multiple area keywords associated with a same area reaches a thirteenth threshold.

12. The method according to claim 10, the jumping away from the current area comprises:

if a fifth condition is met, jumping to an area corresponding to a second keyword sequence involved in the fifth condition, and taking the area associated with the area keyword involved in the fifth condition as the current area, wherein the fifth condition comprises;

that a confidence degree of at least one second keyword sequence reaches a fourteenth threshold.

13. The method according to claim 10, further comprising:

if the confidence degree of the at least one keyword reaches a fifteenth threshold, determining that the keyword corresponding to the current speech segment is the keyword with the confidence degree reaching the fifteenth threshold;

marking in the presented content the keyword with the confidence degree reaching the fifteenth threshold.

14. The method according to claim 10, further comprising:

obtaining at least one phoneme adjacent to the current speech segment based on a second phoneme sequence;

determining a pronunciation similarity between the at least one phoneme and corresponding text information of the at least one keyword, the corresponding text information of the at least one keyword comprising a context of the at least one keyword in the text information;

if the pronunciation similarity between the at least one phoneme and the corresponding text information of the at least one keyword reaches a sixteenth threshold, increasing the confidence degree of the keyword with the pronunciation similarity reaching the sixteenth threshold.

15-19. -19. (canceled)