City University of Hong Kong
The Dialogue Systems Group

Department of Chinese, Translation and Linguistics

RAMCORP Biomedical Expert Dialogue System
Thursday, 10 June 2010 08:32

This dialogue system was developed as part of the RAMCORP (A Remote-Access Multilingual Corpus-Based System for Linguistic Applications) project. This project is dedicated to developing and maintaining a service-oriented knowledge base that facilitates information retrieval through integration of various language resources. At present, the demo system can provide voice services in the domain of biomedicine, as follows:

  • Term definition
  • Preferred term use
  • Synonym term


Future developments of the system will provide additional services, such as:

  • Term translation
  • Related terms
  • Article retrieval
  • Abstract retrieval


Access to the system requires a third-party SIP softphone and a stable internet connection.

Installing and configuring an SIP softphone

We recommend X-Lite for use in communicating with the dialogue system. Install the appropriate version for your operating system, available at

Create a new SIP Account:

  • The Display Name, User name and Password fields can be any value
  • The Domain should be set to
  • Uncheck the "Register with domain and receive incoming calls" option

Create a new contact:

  • The Name and Display as fields can be any value
  • Softphone should be set to " This e-mail address is being protected from spambots. You need JavaScript enabled to view it "


Test your configuration by calling the newly-created contact. The system will answer the call with "Welcome to Interactive Biomedical Information System".

You can view a dynamic transcript of your ongoing dialogue at This web page will also display term lists and other extended information.

How to communicate with the dialogue system

This section describes possible system prompts and your options for responses during the human-machine dialogue, so that you can successfully communicate with the system.

Number indication

At the beginning, a caller is required to indicate “How many terms would you like to know?”. This is used to set the size of recognition space. A caller is free to say a number belonging to the following range:

  • [1-99]
  • (100, 200, 300, 400, 500, 600, 700, 800, 900)

For experiencing the demo system, "100" is recommended here.

Query expressions

In order to make their speech input by recognised smoothly, a caller is required to follow some pre-defined expression patterns when requesting information from the dialogue system. For example, a caller is interested in the term body regions and wants to know its definition:

  • Indicating one term/service per expression
    • Detailed expressions
      • I'm interested in body regions; or
      • Tell me about body regions; or
      • I want that definition service
    • Concise expressions
      • body regions
      • definition
  • Indicate both the term and service in one expression
    • Detailed expressions
      • Give me the definition of body regions; or
      • I want to know about the definition of body regions
    • Concise expression
      • the definition of body regions


  • [I need to | I wanna | I'd like to | I would like to ] = "I want to"
  • [check | retrieve | find] = "know"
  • [tell] = "give"

System prompts

In order to fulfil the caller's informational requirements the system will prompt for further information when appropriate. In this section, we list possible system prompts and detail the responses expected from the user.


System Prompts Expected Response
An open question
What's your query? Any of the patterns below
One slot pattern:
What term do you require information about? / What term are you interested in? A term. Please read an entry from the term list we provide
What service to you require? A service, e.g. "definition", "synonym"
What do you want to know about [term]? A service, e.g. "definition", "synonym"
Two slot pattern:
What's your query? e.g. "the definition of elbow."
What's your query? e.g. "synonym of elbow"

When the caller's speech input is recognised by the ASR engine with low confidence one of the following system prompts will be generated to confirm the caller's utterance:

System Prompts Expected Response
Am I right? Yes/No
Is that correct? Yes/No
Right Yes/No
true or false? Yes/No, true/false
Right or wrong? Yes/No, right/wrong

Note: In any case, you can use the softphone dialing pad to enter 1 for yes or 0 for no.

Joomla 1.5 Templates by Joomlashack