Web site or directory search using speech recognition of letters

• US 8,185,539 B1
• Filed: 11/06/2008
• Issued: 05/22/2012
• Est. Priority Date: 08/12/2008
• Status: Expired due to Fees

• Current Assignee: Foneweb Incorporated
• Original Assignee: Foneweb Incorporated
• Inventors: Bhardwaj, Vinod K.
• Primary Examiner(s): Alam, Hosain
• Assistant Examiner(s): Johnson, Johnese
• Application Number:
  US12/266,191
• Time in Patent Office: 1,293 Days
• Field of Search: 707/756, 707/758
• US Class Current: 707/756
• CPC Class Codes: G06F 16/90344 by using string matching techniques
• CPC Class Codes: H04M 2201/40 using speech recognition speech recognition per se G10L15/00
• CPC Class Codes: H04M 2203/252 where a voice mode is enhanced with visual information
• CPC Class Codes: H04M 2203/357 Autocues for dialog assistance
• CPC Class Codes: H04M 3/4936 Speech interaction details speech recognition per se G10L15/00

Abstract

A mobile telephone is used to search the web site (or computer database) of an entity to find information such as in an FAQ. The user speaks the first few letters of a word (or the first few letters of each of a string of words) into the telephone. The spoken letters are converted into text, search words are formed and the search is sent to the web site. A match results in the answer to the question being returned to the user. Once the first letters are spelled out, the latter letters are essentially redundant, and software can figure out what the intended word is. Error correction is used if the user or speech engine makes a mistake. Confidence values are output by the speech engine when recognizing a spoken letter. The location of each letter and the number of letters spoken is taken into account when attempting a match. The search technique not only matches recognized letters, but also matches their determined location within a search word. The user is provided with a choice if there is more than one match. The user provides their preferred delivery means (voice, SMS, e-mail, facsimile, etc.) on input. Also, a Web server implements a telephone directory that may be citywide, statewide, countrywide or even global. A user speaks a few letters in order find a telephone number of a person or business.

Claims

1. A method of searching a website using error correction, said method comprising:

accepting a call over an initial communications network from a user using an access device;

receiving a plurality of individually spoken characters over said initial communications network, said spoken characters having originated with said user at said access device, and wherein each of said spoken characters holds a position within said spoken characters;

converting said spoken characters into individual text characters using a speech-to-text engine, said text characters including a least one incorrect character;

converting, for one of said positions, a single spoken character into two text characters each having a confidence value output from said speech-to-text engine;

searching a website of a Web server using said individual text characters;

matching said individual text characters with a corresponding word on said website, said text characters not providing a perfect match with a portion of said word due to said incorrect character, wherein said text characters are matched with said word based in part upon said confidence values; and

returning information corresponding to said word to said user over a return communications network.

2. A method as recited in claim 1 wherein said incorrect character occurs due to an error by said user or by an error by said speech-to-text engine.

3. A method as recited in claim 1 wherein said text characters provide a total length that is the same length as said word.

4. A method as recited in claim 1 wherein said matching occurs when at least one of said text characters does not match with the corresponding character of said word.

5. A method as recited in claim 1 wherein each text character has a confidence value output from said speech-to-text engine.

6. A method of searching a website using error correction, said method comprising:

accepting a call over an initial communications network from a user using an access device;

receiving a plurality of individually spoken characters over said initial communications network, said spoken characters having originated with said user at said access device, and wherein each of said spoken characters holds a position within said spoken characters;

converting said spoken characters into individual text characters using a speech-to-text engine;

converting, for one of said positions, a single spoken character into two text characters;

assigning a confidence value to each of said text characters by said speech-to-text engine;

searching a website of a Web server using said individual text characters;

matching said individual text characters with a corresponding word on said website by taking into account said confidence values of said text characters;

returning information corresponding to said word to said user over a return communications network.

7. A method as recited in claim 6 wherein said word is matched using the higher of said confidence values of said two text characters.

8. A method as recited in claim 6 wherein said spoken characters include at least one incorrect character due to a mistake of the user or due to a mistake of the speech-to-text engine.

9. A method as recited in claim 6 wherein said matched word includes a first word character having a position in said word corresponding to a first one of said text characters, said first word character being different from said first text character.