search

AIPHES / emnlp19-moverscore

MoverScore: Text Generation Evaluating with Contextualized Embeddings and Earth Mover Distance
MIT License
192 stars 31 forks source link

why the score you compute using emd function is always between 0 and 1? Does it contain some normalization process? #8

Closed eyuansu62 closed 4 years ago

andyweizhao commented 4 years ago

Indeed, one can see negative values returning from that python package, meaning that the final score is not normalized.

andyweizhao commented 4 years ago
  1. Check out more information in MT evaluation metrics: http://www.statmt.org/wmt17/metrics-task.html
  2. See l.17 in mt_utils.py where I have packed all data used in MT evaluation
  3. The test set in WMT17 metrics shared task is about 560 for each language pair.
andyweizhao commented 4 years ago

You can reach out to me by email for additional questions (andyweizhao1@gmail.com)

eyuansu62 commented 4 years ago

really really thanks😁😁

---Original--- From: "Wei Zhao"<notifications@github.com> Date: Thu, Dec 12, 2019 21:46 PM To: "AIPHES/emnlp19-moverscore"<emnlp19-moverscore@noreply.github.com>; Cc: "Author"<author@noreply.github.com>;"eyuansu62"<772468951@qq.com>; Subject: Re: [AIPHES/emnlp19-moverscore] why the score you compute using emd function is always between 0 and 1? Does it contain some normalization process? (#8)

You can reach out to me by email for additional questions (andyweizhao1@gmail.com)

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.