Method and device for recognising a phonetic sound sequence or character sequence

• US 20040199389A1
• Filed: 02/12/2004
• Published: 10/07/2004
• Est. Priority Date: 08/13/2001
• Status: Active Grant

• Current Assignee: Keep GmbH
• Original Assignee: Keep GmbH
• Inventors: Geiger, Hans
• Granted Patent: US 7,966,177 B2
• Time in Patent Office: Days
• US Class Current: 704/254
• CPC Class Codes: G10L 15/16 using artificial neural networks
• CPC Class Codes: G10L 15/18 using natural language modelling
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units

Abstract

The invention relates to a method for recognizing a phonetic sound sequence or a character sequence, e.g. according to the ASCII standards, comprising the following steps: a) the sequence is fed to a neuronal network, b) in said neuronal network, a sequence of characteristics is formed from the phonetic sequence or character sequence, by taking into consideration phonetic and/or lexical stored information and/or based on a character string sequence (blank characters), c) the character sequence thus formed is compared with a characteristic combination that has a defined statement content, said combination being formed from stored lexical and semantic information, based on the characteristic sequence, d) step c is repeated using new character combinations until, by the reduction of contradictions, a character combination is found that at least largely corresponds with the character sequence, e) the statement content of the character combination with the least number of contradictions is output as the result and/or an action assigned to the statement content is carried out.

Claims

1-12. -12. (Canceled)

13. A method for recognizing a phonetic sequence of sounds or a sequence of characters, for example based on the ASCII standard, with the following steps in the method:

a) the sequence is sent to a neural network

b) in the neural network a sequence of features is formed from the phonetic sequence or the sequence of characters, taking into account phonetic and/or lexical stored information and/or based on a string of characters (spaces),

c) the sequence of features that is generated is compared with a combination of features (H) that has a defined meaningful content and that is formed from saved lexical and semantical information on the basis of the sequence of features,

d) step c) is repeated with new combinations of features until, by minimizing contradictions, a combination of features that at least to a great extent agrees with the sequence of features is found,

e) the meaningful content of the combination of features having the least number of contradictions is then output as a result, and/or an action that is based on the meaningful content of said combination of features is performed if said result is plausible, otherwise a new defined sequence of features is formed by the neural network in accordance with b), and steps c) to e) are repeated.

14. The method of claim 13, wherein the meaningful content of an agreeing combination of features is not output as the result unless said meaningful content is contained in a stored area of permissible meaningful contents.

15. The method of claim 13, wherein agreeing combinations of features found in previous recognition steps are stored in a temporary memory, and the features and/or combinations of features temporarily stored in the temporary memory are preferably used for a comparison of features.

16. The method of claim 15, wherein the information stored in the temporary memory is assigned to a specific statement area or topic area on the basis of stored lexical information, and primary features from said statement area or topic area are used to form the combinations of features.

17. The method of claim 13, wherein, if no agreeing combination of features is found, the non-agreeing part of the sequence is changed on the basis of a combination of features having as many feature agreements as possible, and the changed sequence is sent to the recognition method in step c).

18. The method of claim 17, wherein the non-agreeing part of the sequence of features is changed using a fuzzy logic, and the change of the non-agreeing part of the sequence of features is made taking into account the similarity and sequence of the features with the sequence of the meaningful content defined in feature combinations (H), provided said similarity exceeds a defined value or value range.

19. The method of claim 13, wherein, if no agreeing combination of features is found, the sequence of features in step b) is re-defined, and sent to the recognition method in step c).

20. The method of claim 13, wherein, in generating the sequences of features, vectors are formed that are characteristic of the sequence and/or properties of phonetic or character sequences, and the comparison of features is performed on the basis of the comparison of said vectors.

21. A device for detecting a phonetic sequence of sounds or a sequence of characters for example based on the ASCII standard, having the following components:

an input device for inputting speech and/or inputting characters, for example a keyboard,

a neural network that accepts the phonetic sequence or the sequence of characters input in the input device,

a memory containing lexical, semantic and/or phonetic information,

a first level of a neural network that forms a sequence of features from the phonetic sequence or sequence of characters, taking into account the information in the memory,

a memory area having permissible meaningful contents or command statements,

a second level of the neural network that, using the information from the memory, forms combinations of features having a defined meaningful content,

a comparison device for comparing the sequences of features with the combinations of features,

an evaluation device for evaluating contradictions of posited combinations of features with the sequence of features and/or the lexical/semantic information stored in the memory, and optionally the command statements in the memory area,

a device that, depending on the results, causes the first level of the neural network to form a new sequence of features,

an output device to display the meaningful contents of the least contradictory combination of features and/or a device to perform an action that is associated with the meaningful content of said combination of features.

22. The device of claim 21, characterized by a fuzzy logic to change at least a part of a sequence of features, taking into account the similarity and sequence of features with combinations of features stored in the memory, said fuzzy logic being connected on the output side to the second level.

23. A device of claim 21, characterized by a temporary memory that stores recognized combinations of features from the previous recognition methods, said temporary memory being connected on the output side to the second level of the neural network.

24. A device of claim 21, characterized by a device that, depending on the result, causes the first level of the neural network to form a new sequence of features.