The bots are pro-active. After 20 seconds, they will open the web widget and greet the user with the “Hi! We'll be happy to answer your questions” message. You may edit this message via the Edit tab -- see the "Greeting" special class. You may also change the timeout or disable the active pop-up altogether. ...

You can combine your main FAQ corpus with pre-defined common corpora. The latter include "universal" question – answer pairs, the ones that are not specific to your FAQ file. There are two common corpora available: chitchat and garb_info. You can add them via the Settings tab of FAQ Editor. More here

Chitchat common corpus contains dialogs that would enable the bot to answer small talk questions like "how are you?", "are you a bot or a human?", "thank you, good bye", etc. User queries related to these topics are independent of your knowledge base. This engine comes with a collection of such common queries. The answers, however, are under your control -- you can edit the default values as you think fit, via the Edit tab of FAQ Editor. You can add the Chitchat corpus via the Settings tab of FAQ Editor. ...

"grab_info" corpus covers the "contact me" user intents. Specifically, if the user message contains the user's email address the bot will reply with the “Thank you. We will get back to you shortly.” message. (You may edit it via the Edit tab of FAQ Editor.) ...

You may export your data via the Settings tab of FAQ Editor. More here

Unlike most Deep Learning systems ours doesn't require millions of examples. A dozen paraphrase questions (queries) per class (answer/article) shall be enough. The system is pre-trained on generic English. You only have to add your jargon. Still, the bigger the training corpus the better. ...

AI is dangerous. The danger is that companies will waste billions on useless crap.

Deep Learning is a subfield of Machine Learning concerned with algorithms inspired by the structure and function of the brain called artificial neural networks. We are extensively using it in all our products.

word2vec is a word embedding algorithm. It maps words to vectors in a ~100-dimensional space, so as semantically similar words cluster together. It is powered by an artificial neural network, and was introduced by Mikolov et.al. It has a sister algorithm, GloVe , and a few similar ones. We built our phrase2vec on top of them.

One of our products, Colibri , is a speech recognition app. It records online meetings, transcribes them in real time and generates concise searchable meeting notes that can be easily shared with teammates. Our other product, Alterra Answers and NLP Engine do not include voice recognition. They work with written text.