How do you hide low-quality tweets from Twitter searches?

I created my current Twitter account in 2009. Back then, the service was still relatively new and no one really knew what to use it for. Consequently nearly all of the “content” produced by individuals was crap: most people probably used it to share dumb status updates about literally everything, as if they were trying to implement some sort of digital real-life version of event sourcing.

Things have improved a lot since then. There’s a lot more interesting content on Twitter, especially for developers… provided that you can find it.

You can follow individual accounts or lists that have been created by other users of course. Or you could subscribe to certain topics of interest.

Tweets about specific things made by casual users (whose tweets have very few likes and retweets) can be found using Twitter’s search functionality. It works for the most part. But it sorely lacks an easy way to filter out (what I think are) uninteresting tweets.

Because everyone probably has their own preferences and definitions of “good” and “bad” tweets, here’s what I mean by uninteresting tweets.

Tweets that only show up in search results because they contain an obscene number of hashtags that aren’t relevant at all:

As far as I know, there’s no easy way to filter out tweets that contain more hashtags than actual content.

Keywords don’t have to be hashtags however. Most people who regularly search for tweets about a popular programming language will likely have seen something like this:

It gets worse when a programming language uses a fairly generic name that has other more common meanings, like PHP.

I wish there was an easy way to filter out accounts from the Philippines (especially those with Korean avatars 😅), as its currency is also abbreviated using PHP:

While you can filter by tweet location, it doesn’t seem possible to exclude locations. Also, very few tweets actually have location data.

Then there are also very personal tweets about partial hospitalisation programmes (also PHP) for patients with mental illnesses. I’m not providing an example of such a tweet here for reasons that should be fairly obvious…

PHP is far from the only language or concept that has this problem. java, go, nlp, and (to a lesser extent) python also have other, more common uses.

How I (try to) filter out bad tweets

I mostly use Twitter’s search for personal (rather than commercial or professional) reasons. I therefore simply “subscribe” to search queries using Twitter’s own freely available TweetDeck web app.

My queries usually look like this:

• I start with the keywords that I’m interested in.
• I append a lang:en so that most of the search results will be in English
• Virtually all advertisements will include links, which I sometimes (but not always) remove from my results using -filter:links.
• Because there are no easy ways to filter out spam, I then add a list of words that I want to exclude:

As you can see the list is pretty long. This means that there are inevitably going to be interesting tweets that I never see because they are excluded by my queries.

What about you?

This approach mostly works for me, but I have the feeling there are much better ways to do this.

How do you separate the (t)wheat from the chaff? Do you use extensions? Alternative Twitter clients? Share your strategies in the comments!