Data processing: speech signal processing – linguistics – language – Speech signal processing – Application
Reexamination Certificate
1999-01-28
2002-06-04
Banks-Harold, Marsha D. (Department: 2654)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Application
C704S231000
Reexamination Certificate
active
06401067
ABSTRACT:
BACKGROUND
1. Technical Field
The present application relates generally to machine recognition and, more particularly, to system and methods for providing user-directed recognition constraints, wherein a user interface allows a writer to select one or more recognition constraints which temporarily modify default recognition parameters so as to decode uncharacteristic and/or special data with increased accuracy.
2. Description of the Related Art
In general, conventional machine recognition systems recognize input data by decoding the data using a plurality of trained models. For instance, speech recognition systems decode input speech utterances using trained speech models to recognize spoken words. Likewise, handwriting recognition systems recognize input handwriting data using trained character models to convert the handwriting data to machine printed text.
One example of a handwriting recognition system is the CROSSPAD™ personal digital notepad (PDN) device (marketed by the A.T. Cross Company) which is sold with International Business Machines Corporation's INK MANAGER™ handwriting text recognition software package. The CROSSPAD device allows a user to produce handwritten records on a standard paper notepad (which is placed over a digitizing tablet) using an electronic inking stylus (i.e., pen). A data stream representing the handwriting strokes is generated by the digitizing tablet (simultaneously with the inking) by detecting RF emissions from the stylus, thereby capturing and storing an electronic carbon copy of the handwritten record. The handwriting data that is electronically recorded by the CROSSPAD device can be subsequently uploaded to a computer and processed by the INK MANAGER system which is stored in the computer.
Typically, machine recognition of input data (e.g., speech or handwriting) is performed using predetermined default recognition parameters which are automatically applied by the recognition engine for decoding the input data. These default recognition parameters may be those default parameters which are originally programmed in the system (i.e., “out-of-the-box” default recognition parameters). In addition, the default recognition parameters may also comprise user-selected default recognition parameters (i.e., user preferences) which modify and/or replace the original out-of-the-box default parameters. For instance, handwriting recognition systems as the INK MANAGER system allow the user to initially indicate, for example, the user's normal style of writing (e.g., a user preference) prior to using the system. In particular, before using the INK MANAGER system, the user will be prompted to select (if the user desires) one of the following writing style preferences: pure discrete, pure upper case, and cursive (which includes all other handwriting such as printing and script mixed). During recognition, the INK MANAGER system will automatically apply any default parameters associated with a previously selected user preference. The INK MANAGER system includes other default settings to determine which, if any, word lists will be used for decoding handwriting.
In most situations, utilizing the default recognition parameters will result in optimal data recognition. In certain circumstances, however, utilizing the default recognition parameters may result in sub-optimal recognition accuracy. For example, with regard to handwriting recognition, such special circumstances occur when a user writes in an uncharacteristic manner (e.g., when a user writes an address using all capital letters rather than the user's typical style of writing addresses with printed mixed case). Decoding such uncharacteristic handwriting data using default recognition parameters can result in poor recognition accuracy.
There is a need, therefore, for a system and methods for use with machine recognition that allow a user to temporarily modify default recognition parameters by selecting one or more recognition constraints to decode uncharacteristic and/or special data so as to obtain increased recognition accuracy.
SUMMARY
The present application is directed to system and methods which can be implemented in a machine recognition device to provide a user (via a user interface) the option of selecting one of a plurality of recognition modes including a “default recognition” mode and a “constrained recognition” mode. In the default recognition mode, the recognition engine utilizes predetermined default recognition parameters (e.g., out-of-the-box and/or user-selected default parameters) to decode selected data. In the constrained recognition mode, the user can select one or more of a plurality of recognition constraints which are temporarily applied by a recognition engine for decoding uncharacteristic and/or special data. When the user selects a particular recognition constraint, default recognition parameters are temporarily modified to reflect the selected constraint. The recognition parameters associated with the selected constraint enable the recognition engine to temporarily utilize specific information for decoding the data, thereby providing increased recognition accuracy.
These and other features and advantages of the present system and method will become apparent from the following detailed description of preferred embodiments, which is to be read in connection with the accompanying drawings.
REFERENCES:
patent: 5086385 (1992-02-01), Launey et al.
patent: 5386494 (1995-01-01), White
patent: 5632002 (1997-05-01), Hashimoto et al.
patent: 5819225 (1998-10-01), Eastwood et al.
patent: 5943395 (1999-08-01), Hansen
patent: 6167376 (2000-12-01), Ditzik
patent: 6177931 (2001-01-01), Alexander et al.
Lewis James R.
Perrone Michael P.
Pitrelli John F.
Ratzlaff Eugene H.
Subrahmonia Jayashree
Banks-Harold Marsha D.
F. Chau & Associates LLP
International Business Machines - Corporation
McFadden Susan
LandOfFree
System and method for providing user-directed constraints... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for providing user-directed constraints..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for providing user-directed constraints... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2935492