If you are a researcher and use Twitter's search API, it is important to know that the API will also match search terms against screen names. This may make timeline aggregations misleading depending on what you are attempting to show.
Conversation
If you are only interested in matching against the tweet text, you will need to do post-ingest processing to filter out unwanted matches.
3
1
8
Replying to
Yup. Generally all keywords one wants to search (or stream) will match some user handles. So it's important to have a post-collection clean up for any data collection task.
1
1
Replying to
I have seen matches against tweets where the search term didn't match the tweet text or the screen name. Do you know of any other external fields it might match against (like URL text)?
I have to do some additional testing. Wondering if you saw that as well.
Replying to
I had assumed that it was because people change their usernames. A good % of users do it very frequently. But that was an assumption, not something I have researched. Typically, for our prescription drug research, there are profiles that use the drug names & later change. So,
1
1
when we later use the numeric IDs to collect past data, the earlier user handle disappears in our db. This did have us confused for a bit.
1