3 ms·

Show HN: We Turned the HN FAQs into an FAQ-Bot

9 points by yunusabd 7y ago

yunusabd 7y ago
We're building an automatic FAQ-bot and we wanted to try it out with the hacker news FAQs. The bot uses a model trained on FAQ pages and can answer basic user questions without any setup or additional training. To analzye the questions that have been asked and get a clustering of the most common questions, you can log in with a demo account at https://dialogbar.com/dashboard/analytics/clustering https://dialogbar.com/dashboard/analytics/clustering email: info@dialogbar.com password: hackernews42
ftreml 7y ago
wonder if this works in real life without additional training phrases for an FAQ ?
- yunusabd 7y ago
  That's exactly what we're trying to do, to make it as general as possible, so that it can be used on any FAQ without additional training. The model used in the example was trained on a custom dataset that we made (that didn't include the HN FAQs). The only thing that we added by hand are the suggestions in the widget, because we saw people asking a lot of unrelated questions.
gitgud 7y ago
The animations are pretty cool, but I was a bit confused and thought I could ask it any question... Maybe a little more information on the page would help, like a graphic showing an FAQ page turning into a bot or something...
- yunusabd 7y ago
  Thanks for the feedback, really appreciated! Just to clarify, when saying 'any question', do you mean any question in a global context (e.g. "how's the weather", like Siri or Google Assistant) or any question within the scope of your website ("How do I do X on this website")?
  - gitgud 7y ago
    It just wasn't immediately clear that the product turns FAQ's into a bot. I was on mobile, and I don't think you can see the FAQ image in the background like you can on desktop, so it was just a bit confusing.
    - yunusabd 7y ago
      I see, that's kind of what we gathered from looking at the queries too. People are asking all kind of general questions, so we either have to make it more clear that this only relates to the FAQs, or actually add functionality to answer more general questions, e.g. by using more data from the website. We're looking into the second option right now. Thanks again for the follow-up!

9 points by yunusabd 7y ago

We're building an automatic FAQ-bot and we wanted to try it out with the hacker news FAQs. The bot uses a model trained on FAQ pages and can answer basic user questions without any setup or additional training. To analzye the questions that have been asked and get a clustering of the most common questions, you can log in with a demo account at https://dialogbar.com/dashboard/analytics/clustering https://dialogbar.com/dashboard/analytics/clustering email: info@dialogbar.com password: hackernews42

ftreml 7y ago

wonder if this works in real life without additional training phrases for an FAQ ?

yunusabd 7y ago
That's exactly what we're trying to do, to make it as general as possible, so that it can be used on any FAQ without additional training. The model used in the example was trained on a custom dataset that we made (that didn't include the HN FAQs). The only thing that we added by hand are the suggestions in the widget, because we saw people asking a lot of unrelated questions.

gitgud 7y ago

The animations are pretty cool, but I was a bit confused and thought I could ask it any question... Maybe a little more information on the page would help, like a graphic showing an FAQ page turning into a bot or something...

yunusabd 7y ago
Thanks for the feedback, really appreciated! Just to clarify, when saying 'any question', do you mean any question in a global context (e.g. "how's the weather", like Siri or Google Assistant) or any question within the scope of your website ("How do I do X on this website")?
- gitgud 7y ago
  It just wasn't immediately clear that the product turns FAQ's into a bot. I was on mobile, and I don't think you can see the FAQ image in the background like you can on desktop, so it was just a bit confusing.
  - yunusabd 7y ago
    I see, that's kind of what we gathered from looking at the queries too. People are asking all kind of general questions, so we either have to make it more clear that this only relates to the FAQs, or actually add functionality to answer more general questions, e.g. by using more data from the website. We're looking into the second option right now. Thanks again for the follow-up!