1. This demo is based on a publicly available language model. For production use, a model customized and trained for your use case must be used. FAQ query matching can be applied to a larger passage, this is a less extensive demo. If you want to learn more about it, don't hesitate to contact us.

2. The AI in the demo quantifies the similarity between the FAQ questions and the user query, the FAQ answer is not included in the calculation. The FAQ pair with the highest question-query-similarity is returned as the result.

3. For use in productive systems, the FAQ answer would also be included in the result. For this purpose, an additional neural language model would be trained, which calculates how relevant a FAQ answer is for the user query (see this paper). The combination of question-query-similarity and answer-query-relevance would then provide more accurate results.

4. The FAQ examples were copied from the corresponding websites of the companies (easyJet, Zalando, Tier Mobility).