Below you'll find frequently asked questions about the Source Matters platform regarding source identification.
How Source Matters identifies sources
The Source Matters platform uses advanced natural language processing (NLP) algorithms to “read” each story and identify which people are quoted or attributed.
Using this process, we can extract thousands of sources from thousands of stories — in just minutes. This saves the time journalists would have spent remembering and manually recording their sources for every story.
This makes your source tracking more efficient and thorough, but like all things, it’s not always perfect. There may be times when you notice that some of your sources have not been identified on the Source Matters dashboard.
Why a source might not be identified in Source Matters
Source Matters identifies entities (people, places, things) in the story headline and body text. These are retrieved based on the text given in the RSS feed during setup or page scraping settings in your account. If the combined text is less than 20 words then we will not attempt to identify sources.
Source Matters searches for sources by searching for keywords. For English, keywords include the following:
'said', 'says', 'claims', 'claimed', 'tells', 'told', 'according to', 'reports', 'reported', 'responded', 'responds', 'reiterated', 'reiterates', 'sang,' ‘explains’ and ‘explained.’
See the example below:
It’s important to remember that Source Matters only captures sources for a story when it is initially published and the system is unaware of new updates and sources added later on. If sources are added to a story after it is published, you can manually add the source into Source Matters.
How to manually add sources in Source Matters
You can manually add a source into Source Matters using the "Add a missing source" link under a story on the “Categorize Sources” dashboard.
See the example below:
It’s also possible the sources were pulled in but could be “hidden.” If a story shows it’s “hiding” any sources on a story in the tag sources dashboard then just click “Show all sources” to see all sources pulled in. From there, you can tag/add source categories for a source.
Identifying Spanish-language sources
Source Matters does consider the Spanish language. It is still required that a potential source is a high-salient entity and then it is paired with one of the following phrases: dijo afirmó, respondió, reportada, reportado, informó, cantó, canto, canta. If a source is not identified, you can manually add one using the "Add a missing source" link under a story on the “Categorize Sources” dashboard.
Need help? Send support tickets here.