Tempest Speech Portal drastically reduces the complexities related to the advanced use of speech technologies (speech recognition, text-to-speech, barge-in, etc.), while taking full advantage of its potential.

It does not require the knowledge of programming languages.

The graphical interface allows a simple and fast design of telephone call flows.

Tools for dynamic data access (XML, Odbc, Web Services, Middleware) offer different approaches and levels of integration.

 

Design of speech enabled flows, using the graphic environment

A speech portal service enables an automatic interaction system for mobile and fixed-line users, which reacts and responds to the requests of the latter, according to a predefined logical flow.

To realize this interaction, Tempest Speech Portal has several functionalities (e.g. database queries, access to external applications with HTTP + XML protocol, execution of external processes, execution of VB script within the platform, Web Services) which are made available through simple procedures.

Building a service involves the definition of an operational flow, made with a sequence of appropriately configured objects.

The configurator allows to graphically build this flow: each functionality is represented by an icon and the flow is represented by the interconnection of such icons.

Tempest Speech Portal: Key features

There are a lot of features that cover different needs and with different levels of complexity, depending on the type of service that must be achieved:

• Multi Service – Multi Tenant: Simultaneously executes a plurality of services on several voice channels; the call is routed according to routing rules definitions, based on:
  • Called number (DID);
  • Caller number (CID);
  • Physical called channel;
  • Dates, time and other parameters.
• Multi Language and Multi Vendor Speech Recognition: The speech recognition module of the Tempest platform can use several speech recognition engines and supports dozens of languages.
• Barge In: The interaction mode based on speech activated selections requires the support of the “barge-in” capability, in order to stop the system messages as soon as the user says something, a name or any voice command (with no wait for a “beep” at the end of the message).
• Multi Language and Multi Vendor Text To Speech: Voice messages are dynamic contents that are generated in real time by the speech synthesis software (TTS), in dozens of languages and with different voices. The speech synthesis module of the Tempest platform can use several TTS engines, from different producers, in paid versions as well as in free versions.
• Speech transcription: Thanks to the transcription service of Tempest platform, it is possible to transcribe the content of the audio messages which are recorded during the calls (e.g. VoiceMail) with a very high level of recognition accuracy.
• Data Access: Integration with the Back End can be done through standard interfaces such as ODBC, SQL, XML, Web Services. Simple Wizards guide the user to the integration with the main Database Engine (Oracle, Sql Server, MySql, Db2 etc.).
• PBX: The integration with the telephone PBX is possible in Voip, through the Trunk, as Sip User or even via traditional lines (Analog, BRI, PRI). It supports the Asterisk based Pbx and almost all the existing Pbx on the market.
• Accurate statistics: For each telephone service it is possible to automatically generate dedicated analytics on the specific service, which can be viewed through the browser.

Voice activated selection and Speech Recognition

The navigation with the voice, based on speech recognition of Tempest Speech Portal, is performed through the definition of dedicated resources called “Vocal Selections”.

The vocal selections created through the Tempest graphical interface include

• “Simple” grammars such as lists of commands and words;
• “Complex” grammars which contain regular expressions;
• Specialized grammars “automatic attendant” like, automatically generated by the tables: companies, offices, users, etc.

Within the vocal selections it is possible to define “rejection” grammars, which allow to manage possible words not covered by the defined grammars, background noises, breaths, hesitations and line disturbances.

A voice service based on simple interactions can be designed using only the graphical interface to define the grammars.

Instead, a voice portal based on complex speech interactions (e.g., composed sentences from which the concepts must be extracted) is realized by fully exploiting the recognition engines potential; for this it may be necessary to use standard GRXML grammars on files or dynamics through web server.

The (“parsing”) interpretation of the recognition engine output can take place within “VbScript” like resources. From it, it is possible to extract “concepts” and data to be treated within the application flow.