When was the last time you had to search for some new hashtags? For example, because you wanted to reach a wider audience, or just to show you’re always in the know about the new trends in your field?
If you’re at all like me, that was not long ago. And again, if you’re at all like me, you feel that having to sift through the millions of hashtags people use these days is a pretty daunting task.
If this is the case, you’ll be glad to meet Hashtagify’s latest hire, the Research Assistant. I introduced him a week ago; you might remember that he boasts of knowing all the 38 million hashtags we classified at hashtagify.me, to be able to find many new targeted hashtags just by taking a look at your Twitter account, and that he’ll start working next week.
I thought you might want to know a bit more about him. He’s a reclusive guy, so, to give a better introduction, I went to his laboratory and interviewed him for you. This is the transcript of our talk.
Dan: Hi Assistant. So, the rumors say that you know everything about more than 38 million hashtags. Is that true?
Assistant: Hi Dan. Well, to be honest with you, I don’t actually know everything about those 38 million hashtags. But I know a lot about them. And I’m always on the lookout for new ones to study. As a matter of fact, just during the last 10 minutes, I found 38 more. Right now, the exact number of hashtags I know about is 38,476,189.
OK, the quantity is impressive. But what about the quality? Are all those hashtags really good?
This is an interesting point. Actually, it turns out that lots of hashtags are too generic for any targeted message; others are too loved by spammers; others still were great a few months ago, but now they’re dead in the water. That’s just why I’m always learning new ways to classify and filter them, using information about the hashtag users, associations, languages, timing, and many other interesting signals. It’s no easy task, but they tell me I’m getting better and better at this.
Let’s talk about personalization. After you filter out the hashtags that may look popular, but aren’t really useful, how do you decide which ones are the most targeted for a specific Twitter user?
First of all, I study up to 3,200 of the last tweets that the user sent, and analyze the hashtags they use, the language, and some other technical stuff. I then try to understand if the user has different areas of interest to talk about, using a technique called clustering. For each cluster of hashtags, I then find the most interesting hashtags that the user has never used.
When you say “interesting hashtags”, what do you actually mean? And usually, how many interesting hashtags do you find for a user?
A hashtag can be interesting for many different reasons. For example, it could be a newly trending hashtag in a related field. It may have a popularity that isn’t very high, but is very targeted. I try to create a good mix and show the user around 200 options to choose from; the user can sort them by popularity, trend, and correlation, and also look at lots of details to better understand the hashtag if it looks interesting. Users can also discard hashtags that they know aren’t good for them; this really helps me to learn and be even more targeted in the future.
Great, I’m sure all our advanced users will like the possibility to start from an already sorted and filtered pool of candidates, and still be able to choose by themselves. But what about beginners? This still looks a bit difficult for them, couldn’t you make things even easier?
Come on Dan! You already know that I’m much better at helping users who already know their way around hashtags. Isn’t that the reason why you also hired Hashtagify Tutor? I just do the analysis; then, for those who don’t want to check lots of hashtags and meddle with all the finer details, he can take over and guide them step by step…
Dan: Don’t worry, I remember perfectly well our deal. I just thought that this could be the perfect question to finish this interview, and link to our next one with Hashtagify Tutor. So, Assistant, thank you for your time, and I’ll leave you to your analyses now. Maybe, before the great launch next week, you’ll even learn some new tricks!
Assistant: Thank you Dan. You know that I’m a lab rat and I’m not much into interviews, but I really hope I could answer some of our readers’ curiosity about my work. And, while thanking them for staying with us until the end, I’d like to remind them that sharing is caring. And remember… I’m going to read all your tweets!