Method for preparing information for a speech dialogue system
View Patent ↗In many application environments, it is desirable to provide voice access to tables on Internet pages, where the user asks a subject-related question in a natural language and receives an adequate answer from the table read out to him in a natural language. A method is disclosed for preparing information presented in a tabular form for a speech dialogue system so that the information of the table can be consulted in a user dialogue in a targeted manner.
1. A device for editing information for a speech-dialog system, comprising:
means for providing information presented in tabular form,
means for standardizing the information presented in tabular form and/or its representation in accordance with predefined criteria and a means for accessibly storing it/them,
means for assigning a first grammar to horizontal and/or vertical rows and a second grammar to the table elements from the respective row, the grammars describing structural and conceptual rules for spoken inputs by means of which the assigned row and the assigned table elements of the respective row can be recognized so that the information presented in tabular form has been edited for a speech-dialog system on the basis of the assigned first and second grammars.
2. A computer-implemented method for providing speech access to a table of data, the table comprising a plurality of elements organized as a plurality of horizontal rows of the elements and a plurality of vertical columns of the elements, each element storing respective element contents, the method comprising:
reading the contents of the plurality of elements;
automatically defining a first speech-recognition grammar corresponding to the plurality of elements of the table, wherein the first speech-recognition grammar is defined based at least in part on the contents of the respective elements of the table; and
automatically defining a second speech-recognition grammar corresponding to the plurality of rows of the table, wherein the second speech-recognition grammar is defined based at least in part on the contents of the elements of the respective rows of the table.
3. A method according to claim 2 , wherein defining the second speech-recognition grammar comprises, for each row, defining a trigger grammar and a filter grammar.
4. A method according to claim 2 , further comprising automatically:
ascertaining orientation of the table, wherein the orientation comprises one of horizontal and vertical; and
if the table is determined to be vertically oriented, transforming the table into a horizontally oriented table.
5. A method according to claim 4 , where ascertaining the orientation of the table comprises ascertaining the orientation of the table based at least in part on character formatting of the contents of at least two of the elements of the table.
6. A method according to claim 2 , further comprising automatically:
for each element, assigning at least one class to the element, wherein each of the at least one class is assigned to the element based at least in part based on the contents of the element; and
for each row, assigning a class to the row, wherein the class is assigned to the row based at least in part on the classes assigned to the elements of the row.
7. A method according to claim 6 , wherein defining the second speech-recognition grammar comprises, for each row, defining a trigger grammar and a filter grammar based at least in part on the class of the row.
8. A method according to claim 6 , wherein defining the second speech-recognition grammar comprises, for each row, defining a trigger grammar and a filter grammar based at least in part on the class of the row and headings of the row.
9. A method according to claim 6 , wherein defining the second speech-recognition grammar comprises, for each row, defining a trigger grammar and a filter grammar based at least in part on the class of the row, headings of the row and minimum and maximum values of the contents of elements of the row.
10. A method according to claim 6 , wherein assigning the at least one class to the element comprises:
comparing the contents of the element to contents of a plurality of predefined lists of words, each of the predefined list of words corresponding to a respective predefined class; and
if the contents of the element correspond to a word in one of the predefined lists of words, assigning the class corresponding to the list to the element.
11. A method according to claim 10 , wherein assigning the at least one class to the element further comprises:
if the contents of the element do not correspond to a word in any of the predefined lists of words, assigning a default class to the element.
12. A method according to claim 10 , wherein assigning the class to the row comprises:
assigning to the row a class selected from a plurality of possible classes, the plurality of possible classes comprising the classes assigned to the elements of the row, wherein the assigned class simultaneously optimizes at least two criteria, including:
(a) the class is assigned to as large a number of the elements in the row as possible; and
(b) the predefined list of words corresponding to the class has as small a number of words as possible.
13. A method according to claim 2 , further comprising automatically normalizing the contents of at least a portion of the elements of the table.
14. A method according to claim 2 , wherein reading the contents of the plurality of elements comprises reading contents of at least a portion of a web page.
15. A method according to claim 2 , further comprising:
generating a structure from the first speech-recognition grammar and the second speech-recognition grammar; and
providing the generated structure to a speech-dialog application.
16. A system for providing speech access to a table of data, the table comprising a plurality of elements organized as a plurality of horizontal rows of the elements and a plurality of vertical columns of the elements, each element storing respective element contents, the system comprising:
a table transformer configured to read the contents of the plurality of elements; and
a grammar guesser coupled to the table transformer and configured to:
automatically define a first speech-recognition grammar corresponding to the plurality of elements of the table, wherein the first speech-recognition grammar is defined based at least in part on the contents of the respective elements of the table; and
automatically define a second speech-recognition grammar corresponding to the plurality of rows of the table, wherein the second speech-recognition grammar is defined based at least in part on the contents of the elements of the respective rows of the table.
17. A system according to claim 16 , wherein the grammar guesser is configured to automatically define, for each row, a trigger grammar and a filter grammar.
18. A system according to claim 16 , wherein the table transformer is configured to automatically:
ascertain orientation of the table, wherein the orientation comprises one of horizontal and vertical; and
if the table is determined to be vertically oriented, transform the table into a horizontally oriented table.
19. A system according to claim 18 , where the table transformer is configured to ascertain the orientation of the table based at least in part on character formatting of the contents of at least two of the elements of the table.
20. A system according to claim 16 , wherein the grammar guesser is configured to automatically:
for each element, assign at least one class to the element, wherein each of the at least one class is assigned to the element based at least in part based on the contents of the element; and
for each row, assign a class to the row, wherein the class is assigned to the row based at least in part on the classes assigned to the elements of the row.
21. A system according to claim 20 , wherein the grammar guesser is configured to define, for each row, a trigger grammar and a filter grammar based at least in part on the class of the TOW.
22. A system according to claim 20 , wherein the grammar guesser is configured to define, for each row, a trigger grammar and a filter grammar based at least in part on the class of the row and headings of the row.
23. A system according to claim 20 , wherein the grammar guesser is configured to define, for each row, a trigger grammar and a filter grammar based at least in part on the class of the row, headings of the row and minimum and maximum values of the contents of elements of the row.
24. A system according to claim 20 , wherein the grammar guesser is configured to automatically:
compare the contents of the element to contents of a plurality of predefined lists of words, each of the predefined list of words corresponding to a respective predefined class; and
if the contents of the element correspond to a word in one of the predefined lists of words, assign the class corresponding to the list to the element.
25. A system according to claim 24 , wherein the grammar guesser is configured to:
if the contents of the element do not correspond to a word in any of the predefined lists of words, assign a default class to the element.
26. A system according to claim 24 , wherein the grammar guesser is configured to:
assign to the row a class selected from a plurality of possible classes, the plurality of possible classes comprising the classes assigned to the elements of the row, wherein the assigned class simultaneously optimizes at least two criteria, including:
(a) the class is assigned to as large a number of the elements in the row as possible; and
(b) the predefined list of words corresponding to the class has as small a number of words as possible.
27. A system according to claim 16 , wherein the table transformer is configured to automatically normalize the contents of at least a portion of the elements of the table.
28. A system according to claim 16 , wherein the table transformer is configured to read the table from contents of at least a portion of a web page.
29. A system according to claim 16 , further comprising an application generator coupled to the grammar guesser and configured to:
generate a structure from the first speech-recognition grammar and the second speech-recognition grammar; and
provide the generated structure to a speech-dialog application.