Method, apparatus, and product for automatic generation of lexical features for speech recognition systems

An apparatus for performing a teaching process associated with a lexicon database containing words and associated lexical properties, comprising:

• a grammar database comprised of lexical property tests having rules and related constraints for combining words in acceptable word combinations for speech recognition, wherein certain constraints are switchable constraints, and switchable constraints are enabled during speech recognition and disabled during the teaching process;
• a receiver configured to receive exemplary clauses containing permissible word combinations for speech recognition;
• a parser configured to parse the exemplary clauses using applicable lexical property tests;
• and
• a modifier configured to update the lexicon database with additional lexical properties for selected words in the lexicon database, corresponding to words in the received exemplary clauses.

A computer program product comprising:

• a computer usable medium having computer readable code embodied therein for performing a teaching process associated with a lexicon database containing words and associated lexical properties, the computer usable medium comprising:
• a grammar database comprised of lexical property tests having rules and related constraints for combining words in acceptable word combinations for speech recognition, wherein certain constraints are switchable constraints, and switchable constraints are enabled during the speech recognition process and disabled during the teaching process
• a receiving module configured to receive exemplary clauses containing permissible word combinations for speech recognition;
• a parser configured to parse the exemplary clauses using applicable lexical property tests;
• and
• an updating module configured to update the lexicon database with additional lexical properties for selected words in the lexicon database corresponding to words in the received exemplary clauses.

A speech recognition method using a lexicon database containing words and associated lexical properties, including word types and features, and a grammar database with lexical property tests defining rules and related constraints for combining words in recognizable word combinations, and wherein switchable constraints are disabled during a teaching process, the method comprising the steps, performed by a processor, of:

• receiving input speech patterns containing a word sequence;
• and
• parsing the word sequence, wherein the parsing step comprises the substeps of
• identifying word types associated with the words of the word sequence,
• selecting lexical property tests applicable to the word sequence based on the identified word types, and
• applying the selected lexical property tests with any associated switchable constraints in the enabled state to limit the recognizable word combinations based on features associated with the words in the word sequence.

A speech recognition apparatus having a lexicon database containing words and associated lexical properties, including word types and features, and a grammar database with lexical property tests defining rules and related constraints for combining words in recognizable word combinations, and wherein switchable constraints are disabled during a teaching process, the apparatus comprising:

• a receiver configured to receive input speech patterns containing a word sequence;
• and
• a parser configured to parse the word sequence, wherein the parser comprises
• a component configured to identify word types associated with the words of the word sequence,
• a component configured to select lexical property tests applicable to the word sequence based on the identified word types, and
• a component configured to apply the selected lexical property tests with any associated switchable constraints in the enabled state to limit the recognizable word combinations based on features associated with the words in the word sequence.

A computer program product comprising:

• a computer usable medium having computer readable code embodied therein, comprising:
• a lexicon database containing words and associated lexical properties, including word types and features;
• a grammar database with lexical property tests defining rules and related constraints for combining words in recognizable word combinations, wherein switchable constraints are disabled during a teaching process;
• a receiving module configured to receive input speech patterns containing a word sequence;
• and
• a parsing module configured to parse the word sequence, wherein the parsing module comprises
• a module configured to identify word types associated with the words of the word sequence,
• a module configured to select lexical property tests applicable to the word sequence based on the identified word types, and
• a module configured to apply the selected lexical property tests with any associated switchable constraints in the enabled state to limit the recognizable word combinations based on features associated with the words in the word sequence.

A computer readable medium containing instructions for causing a computer system to perform a teaching process associated with (modify) a lexicon database containing words and associated lexical properties, by:

• providing a grammar database comprised of lexical property tests having rules and related constraints for combining words in acceptable word combination$ for speech recognition;
• identifying certain constraints as switchable constraints, wherein switchable constraints are enabled during speech recognition and disabled during the teaching process;
• receiving exemplary clauses containing permissible word combinations for speech recognition;
• parsing the exemplary clauses using applicable lexical property tests;
• updating the lexicon database with additional lexical properties for selected words in the lexicon database corresponding to words in the received exemplary clauses.

A computer readable medium containing instructions for causing a computer system having a grammar database with lexical property tests defining rules and related constraints for combining words in recognizable word combinations, wherein switchable constraints are disabled during a teaching process, to recognize input speech patterns, by:

• receiving input speech patterns containing a word sequence;
• and
• parsing the word sequence by
• identifying word types associated with the words of the word sequence,
• selecting lexical property tests applicable to the word sequence based on the identified word types, and
• applying the selected lexical property tests with any associated switchable constraints in the enabled state to limit the recognizable word combinations based on features associated with the words in the word sequence.

A method for developing a grammar database having lexical property tests for use in a speech recognition process using a lexicon database containing words and associated lexical properties, the method comprising the steps, performed by a processor, of:

• receiving exemplary clauses containing types of word sequences to be understood during a speech recognition process;
• identifying additional lexical properties for selected words in the lexicon database corresponding to words in the received exemplary clauses using the lexical property tests of the grammar database by
• parsing the exemplary clauses with switchable constraints of the lexical property tests disabled to produce an index of the lexical properties for words in the exemplary clauses;
• updating the lexicon database with the additional lexical properties by
• assigning the additional lexical properties to the corresponding words of the lexicon database;
• and
• compiling the lexicon database and the grammar database with the switchable constraints of the lexical property tests enabled to produce a new grammar database that embodies constraints of the lexical property tests and the lexical properties of the lexicon database including the additional lexical properties.