search

Babelscape / rebel

REBEL is a seq2seq model that simplifies Relation Extraction (EMNLP 2021).
502 stars 73 forks source link

Financial relations #73

Open vmehera123 opened 10 months ago

vmehera123 commented 10 months ago

Hi there! I tried to extract financial information from a bunch of financial articles summaries, but I didn't get expected results (such as funding info, investment etc). Does your model able to extract that kind of relations, or I need to fine-tune it to do so?

Text example:

Stocks rose on Monday amid a slew of mergers and acquisitions news and ahead of a flood of earnings reports this week. Macy's shares jumped more than 20% after the department store chain reportedly received a $5.8 billion buyout offer. Shares of other retailers also got a boost, with Nordstrom and Kohl's rising more than 5%. Cigna's stock rose after a report said the insurer dropped its attempt to acquire Humana. Shares of both companies rose on the news, with Cigna up more than 15% and Humana down around 1.5%. Shake Shack rose 7.4% after the burger chain announced a timeline for its CEO transition.

Exctracted relations:

{(8, 3): {'relation': 'point in time',
  'head_span': mergers and acquisitions news,
  'tail_span': Monday},
 (23, 33): {'relation': 'instance of',
  'head_span': Macy's,
  'tail_span': department store},
 (57, 55): {'relation': 'parent organization',
  'head_span': Kohl's,
  'tail_span': Nordstrom},
 (80, 65): {'relation': 'parent organization',
  'head_span': Humana,
  'tail_span': Cigna},
 (105, 112): {'relation': 'product or material produced',
  'head_span': Shake Shack,
  'tail_span': burger}}

I expected to receive relation regarding received money ($5.8b)