Data processing: speech signal processing – linguistics – language – Speech signal processing – Application
Reexamination Certificate
1997-08-14
2003-09-30
S{haeck over (m)}its, T{overscore (a)}livaldis Ivars (Department: 2654)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Application
Reexamination Certificate
active
06629074
ABSTRACT:
TECHNICAL FIELD
The present invention relates in general to a data processing system, and more particularly, to a graphical user interface for indicating utilization of resources within the data processing system, implementing a speech dictation application.
BACKGROUND INFORMATION
With new and exciting advances in technology, speech command and control systems are being deployed in telephone and desktop data processing systems. As such speech command and control systems are developed, the introduction of a speech-to-text, or dictation, system has become feasible. For example, refer to the IBM Voice Type Dictation products and the IBM Simply Speaking products. This dictation system, available from IBM (International Business Machines, Inc.) is primarily used for applications which require short text entry, such as a memo of a few pages, a note for e-mail, or a medical report. As dictation systems become more robust and as dictation capability is added to full-fledged word processing systems, an amount of text entered will grow. Therefore, associated with such increased dictation capability, is the increased utilization of memory required to implement the dictation system within the data processing system.
An important aspect of such dictation systems is a correction mechanism. Speech recognition inherently involves a relatively high error rate, attributable to both the technology and the vocabulary. As text correction is a time-intensive process which requires oversight by an external human user, it is helpful to implement a correction system using a data processing system which allows the user or an associate to be involved during the correction process. To implement such a system, a speech-aware text processor must maintain persistent results information values which are obtained at the time each word is recognized. These results information values include an audio recording of the user as they are speaking that word, as well as alternative words which may be associated with a “best guess” word recognized by a voice dictation system.
Thus, when a user proofreads the dictated text, the user may utilize the results information to efficiently correct the misrecognized words. For example, when the system is notified by the external user that an error occurred during the dictation process, the system may use the results information to play back the user's voice for that word. Additionally, when the system is notified that an error occurred during the dictation process, the results information may be used to present a list of alternatives for the erroneous word, where the list of alternatives was determined by recognition algorithms at the time of recognition. Such recognition algorithms are well-known in the data processing art and, therefore, will not be described in greater detail. It should be noted that when a word is corrected via a selection of a word from the list of alternatives, a crucial learning process of the speech recognition system is implemented. Through the use of this correction process, the system's understanding of the user's speech and usage patterns may be updated. It should be noted that in this context, “usage” refers to the sequences of words that the user employs. A statistical model of n-word sequences (n-grams) is kept to predict spoken words.
The results information required to provide this learning process and an easily usable system can be very costly in terms of system resources. For example, the results information comprise a tag or a handle into a recognition results data base that is maintained by a speech recognition system and is returned to a text system along with the best alternative. This tag is typically thirty-two bits long. Furthermore, the recognition results database that is accessed by the tag includes an audio recording of the user's voice for each word, as well as the alternative guesses for each word. The alternative guesses represent at least a multiple of five or six times the storage space needed for the “best guess” word. Additionally, the audio recording is typically captured at 11 Kbytes per second. The recording capture rate may also be as much as twice that rate, depending on the needed fidelity. Thus, it may be seen that a substantial amount of memory is required for the audio recording, as well as storage space for each of the identified alternative words. One embodiment of the system which stores the audio recording of a voice which is being provided to a speech-to-text, or dictation, system as disclosed in U.S. Pat. No. 5,031,113 by Höllerbauer. Additional information is disclosed in “Multimedia Presentation Techniques and Technologies,” International Technical Support Center, IBM, Boca Raton, Fla., 1993, IBM Order No. GG243975, which is hereby incorporated by reference herein.
As the results information consumes such a substantial amount of memory within a data processing system, an amount of results information which may be stored is limited. Typically, where small amounts of text are dictated, the use of such substantial memory is not an issue. However, as voice dictation systems are advancing technologically, a need has arisen for voice dictation systems which will support more substantial documents, especially where said documents are composed in multiple sessions and the results information persists across sessions. In such situations, the capacity of resources within a data processing system may be exceeded if an amount of data stored within the results information value is not minimized.
Therefore, as dictation is used for longer documents, a mechanism is needed to prune the resources consumed by the dictated document and its persistent recognition data, also referred to as persistent results information. Furthermore, a mechanism is needed to involve an end user in the pruning process, since only the user knows when they have completed some text entry, and no longer require information for subsequent correction of the text. Automatic mechanisms, such as least recently used algorithms, are not appropriate for this task.
SUMMARY OF THE INVENTION
The previously mentioned needs are fulfilled with the present invention. Accordingly, there is provided, in a first form, a data processing system. The data processing system includes a central processing unit for detecting an availability of a resource during execution of the data processing application. The availability of the resource is selectively modified in response to execution of the data processing application. The data processing system also includes a display device for providing an indication of the availability of the resource. The indication of the availability is selectively modified in response to execution of the data processing application.
Additionally, there is provided, in a second form, a program storage device readable by a machine, tangibly embodying a program of instructions executable by the machine to perform method steps for executing a data processing application. The method steps include receiving a first plurality of user inputs at an input circuit and translating the first plurality of user inputs to a first plurality of translated data values. The method steps also include storing the first plurality of translated data values in a first portion of a memory having a predetermined storage space and modifying a presentation of an indicator in response to a remaining portion of the predetermined storage space of the memory.
Furthermore, there is provided, in a third form, a data processing system. The data processing system includes an input circuit for receiving a first plurality of data values and commit input control signal. A memory stores the first plurality of data values in a first portion of the memory. A central processing unit selectively enables the memory to make the first portion of the memory available for storing a second plurality of translated data values in response to a first logic state of the commit input control signal.
The foregoing has outlined rather broadly the features and technical advantages of
LaBaw Jeffrey S.
Newberger Barry S.
S{haeck over (m)}its T{overscore (a)}livaldis Ivars
Winstead Sechrest & Minick P.C.
LandOfFree
Resource utilization indication and commit mechanism in a... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Resource utilization indication and commit mechanism in a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Resource utilization indication and commit mechanism in a... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3031063