Cross-Sum-News-Aligned

Continually updated repository housing cross lingual and mono lingual summarization data for different language pairs.

Updates

• Added English-English v0.1.

Language Pairs Available:

• English - English

Data description

• document: the document text.
• document_lang: predominant langcode of the document text.
• summary: the summary text.
• summary_lang: predominant langcode of the summary text.
• cosine_similarity: cosine distance between the distUSE embeddings for both the document and summary.
• unigram_overlap: lexical overlap (0-1) or translated lexical overlap (for cross lingual pairs) quotient. Higher means more unigram overlap between the summary and the document.

References

• TBD

P.S.: More language pairs will be added soon.