Automatic handling of Frequently Asked Questions using Latent Semantic Analysis

We present results from using Latent Semantic Analysis (LSA) for automatic handling of FAQs (Frequently Asked Questions). FAQs have a high language variability and include a mixture of technical and non-technical terms. LSA has a potential to be useful for automatic handling of FAQ as it reduces the linguistic variability and capture semantically related concept. It is also easy to adapt for FAQ. LSA does not require any sophisticated linguistic analyses and merely involves various vector operations. We evaluate LSA for FAQ on a corpus comprising 4905 FAQ items from a collection of 65000 mail conversations. Our results show that Latent Semantic Analysis, without linguistic analyses, gives results that are on par other methods for automatic FAQ.