European Dimension and Cross-Cultural Issues

FAQs produced in original language.
FAQs translated to English.
FAQs translated to other languages.
Partner in each language region responsible for texts and content in their language.

Medical Quality, Risks

Intelligent Natural-Language Question-Answering	Number of answers	Precision
Web4Health Natural-Language Question-Answering system	12 plus more via link to link page	100 %
Alkaline (conventional free text search engine, uses OR between search words)	939 answers	about 10 %
Alkaline query "+causes +eating +disorders"	82 answers	about 30 %
Google with "site:web4health.info"	90 answers	about 20 %
Google with no site restriction	437 000 answers	not measured

No answer exists in our data base	The question is within the intended scope of our data base	The question is added to a list of FAQ topics which should be written by our medical experts
No	No	Usually not answered. Sometimes we provide answers to common such questions, or references to where an answer can be found.
No	Yes	The question is put on a list of new topics, which our medical experts should write FAQs about.
Yes	Yes	The matching templates are modified, so that future similar questions will get the existing answer(s).

Forums: One on health issues, one for testing auto-translated forums.

Chats: Only used internally between project partners.

Ask-the-expert: About 2500 questions in the first year, answered about half of them.

Work-flow-tools implement to support answering steps for English questions:

Search Engine Optimization (SEO).
MEDNET 2003 (World conference on Internet and Medicine HON Geneva).
12th AEP (Association of European Psychiatrists) Congress 2004.
Content Village.
High-Level European eHealth Conference and Exhibition (Cork, Ireland, Finalist).
AACE E-learning conference.

User observation and questionnaires.
Questionnaire on the web site.
System logs, including logs of questions asked by users and the response they get.
Analysis of questions and answers for the natural-language question-answering system (random sample of actual questions, February 2004):

No.	%	No. in scope	% of in scope	Quality of answer
8	10%	8	13%	Several direct answers to the question.
8	10%	8	13%	One direct answer to the question.
30	37%	30	48%	Information probably of interest, but not direct answer to the question.
19	23%		0%	No answer because the question was out of scope for the web site (like questions about somatic problems).
9	11%	9	15%	No answer to the question, even though an answer existed in the FAQ data base, but it was not found by the question-answering engine.
6	7%	6	10%	No answer to the question, because there did not exist any answer to this question in the FAQ data base.
1	1%	1	2%	We do not want to answer.
81	100%	62	100%	Total

top Questionnaire Answers Was the content of the site useful to you? Please rate on a scale from 1 to 5:
1	6 %	not at all useful
2	5 %
3	31 %
4	21 %
5	37 %	very useful