Browse Prior Art Database

Automatic voice service personalisation

IP.com Disclosure Number: IPCOM000099001D
Original Publication Date: 2005-Mar-08
Included in the Prior Art Database: 2005-Mar-08
Document File: 4 page(s) / 63K

Publishing Venue

IBM

Abstract

Currently, a speech-enabled web service can only be personalised in terms of content and presentation to a speaker user on the basis of a profile which they initiate and which requires manual maintenance. Personalisation of such services is known within the area of value-added services to be a significant commercial benefit in promoting 'stickiness' - ie., preventing the subscriber from moving to another service provider and increasing access to the specific service. The proposal seeks to build up information dynamically about the user based on three key characteristics of how they interact with the service by voice: (1) prosody (the pitch and duration characteristics of their input); (2) the language they use (in terms of lexical content and syntactic structure); and (3) accent (deviation from the standard target accent of the speech recogniser). The information will be used to: 1. to filter and order the presentation of search results or other content 2. to reword the output so that the language type matches the user’s 3. to establish a persona for the outgoing information/content transmission These are precisely the main features currently set and managed manually for personalisation of current web-based services, as stated above. With this proposal, these can be set automatically and dynamically per user access via the steps as described here.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 27% of the total text.

Page 1 of 4

Automatic voice service personalisation

Currently, a speech-enabled web service can only be personalised in terms of content and presentation to a speaker user on the basis of a profile which they initiate and which requires manual maintenance. Personalisation of such services is known within the area of value-added services to be a significant commercial benefit in promoting 'stickiness' - ie., preventing the subscriber from moving to another service provider and increasing access to the specific service. However, without a user-specific profile to define the characteristics of presentation and content that the user would like, there is currently no automated method to support personalisation.

    The proposal described in full below seeks to build up information dynamically about the user based on three key characteristics of how they interact with the service by voice: (1) prosody (the pitch and duration characteristics of their input); (2) the language they use (in terms of lexical content and syntactic structure); and (3) accent (deviation from the standard target accent of the speech recogniser). All three parameters are important and need to be cross-correlated with eachother to establish the user's personalisation profile automatically.

    This proposal relates to the automatic extraction of specific speech and language features during voice access to a service such that the logical flow of the service as well as the associated data content can be personalised the better to match the user's requirements, without manual intervention.

The objectives of current personalisation, from a user perspective, include:

· Filtering and ordering content/information appropriate for the specific user; this is currently handled via the set up of a personal portal;
· Gearing presentation language and style to the user's preferences;
· Setting a persona , which means the personality used for audio prompting and output, for instance Linda Croft versus Brian Willis and so forth.

    This is assumed to be commercially attractive, in much the same way as users can personalise their home page and so forth on the internet, in providing increased user satisfaction and prompting "stickiness" (ensuring the user does not move to another service provider).

Prior Art

    Personalisation of web-based services, for instance, is usually handled by explicit intervention either by the user themselves to provide the look and feel that they want, or through operator/service provider configuration. This process remains manual and to some degree intrusive, especially for those services which are infrequently accessed by the user.

Once this process has been followed, however, the service(s) can become more attractive to users because they are both more comfortable with the user interface and are also able to block or remove information and content which they never use.

    In addition, for many web-services, a "history" is retained at the service platform of how the user proceeds...