基于Apache Spark Streaming与Java的Twitter4J API：推文过滤效率对比

Great question—let’s break down each approach to figure out which gives you the most efficient mix of relevant results and minimal noise:

1. User Location Filtering #

• Let’s start with the least efficient option here. While targeting users in New York sounds logical, it’s full of holes. A huge portion of Twitter users don’t fill out their location profile at all, or use vague labels (like "USA" instead of "Manhattan"). Even if someone does have their location set to New York, their tweet might be about a Yankees game or a new coffee shop—not law and order services. You’ll waste time sifting through irrelevant posts and miss tons of relevant ones from folks who didn’t update their location settings.

2. Keyword Filtering (e.g., "New York" + "law and order") #

• This is a solid middle-ground option that balances coverage and relevance. By targeting tweets that mention both "New York" and terms tied to law and order services, you’ll capture posts from local businesses, service providers, and people seeking help—even if those users didn’t set their location. The catch? You’ll get a lot of noise. A huge chunk of tweets with "law and order" are about the TV show, not real-world services. To fix this, refine your keywords to include service-specific terms like "bail bonds", "legal defense", or "security company" alongside "New York". Even with tweaks, you’ll still have to sort through some non-target posts, but it’s way better than location filtering.

3. Hashtag Filtering (#NewYork, #Law, #Order) #

• Hashtags are great when users actually use them, but that’s the problem. Most small local law firms, bail bond services, or security companies don’t bother adding formal hashtags like #NewYork or #Law to their tweets. They might use niche tags (like #NYCLawyer) but those are hard to predict and cover a tiny fraction of relevant content. You’ll get super relevant tweets from the people who do use these hashtags, but you’ll miss the vast majority of posts that don’t include them. This makes it the least comprehensive option.

Final Takeaway #

If you had to pick just one of the three methods, keyword filtering is the most efficient—it gives you the widest coverage while still letting you target relevant content (especially when you add service-specific terms). But for the best results, combine it with a location filter: use keywords like "New York" AND ("bail bonds" OR "legal service" OR "security company") and limit results to users in New York. This cuts down on noise from non-local tweets and ensures you don’t miss posts from users who didn’t set their location.

内容的提问来源于stack exchange，提问作者Mingg Lex