I have a query relating to the sharing of newspaper content as research data. The researcher has used newspaper articles from 5 different news sites on which to carry out their analysis. They would now like to share their data to support reproducability and follow-on research. However, as they don't own the copyright of the original articles we're wondering how/if they can share the data. 

The data include the full text of the articles, the articles with words stemmed (e.g. charg = charger, charging etc.) together with some metadata, including URL, derived keywords and phrases, headline, source, date etc. Is it possible to share a portion of the data, for example the metadata and derived keywords, can the headlines be shared, is there a certain percentage of the text that can be shared without breaching copyright?

Some of the solutions we discussed include only releasing the URLs for the articles, but this causes issues as the URLs may not be stable or the articles may have been updated in the meantime. If a database such as Proquest were used then only researchers with institutional access to these databases could access the data. The researcher could of course contact the news organisations to request permission to share but if even one says no then the resulting dataset is not a complete dataset. And this is also not necessarily scalable if the research were broadened to 100 news organisations.