search

huggingface / neuralcoref

✨Fast Coreference Resolution in spaCy with Neural Networks
https://huggingface.co/coref/
MIT License
2.83k stars 474 forks source link

Results completely differ from web-demo #347

Open chris-aeviator opened 2 years ago

chris-aeviator commented 2 years ago

When using neuralcoref master with Space==2.1.0 I can use neuralcoref just fine. However the results drastically differ from the version deployed to huggingface.co/neuralcoref

"She is close to the habour" yields: grafik

Whereas the same text executed via examples/server.py yields an empty reply

❯ curl --data-urlencode "text=she is close to the habour" -G localhost:8000
{}%

I can confirm that my curl call succeeds with other prompts.

▽ {mentions: […], clusters: […], resolved: "she is close to the habour. where might she be heading?"}
  ▽ mentions: [{…}, {…}]
    ▽ [0]: {start: 0, end: 3, text: "she", resolved: "she"}
        start: 0
        end: 3
        text: "she"
        resolved: "she"
    ▽ [1]: {start: 40, end: 43, text: "she", resolved: "she"}
        start: 40
        end: 43
        text: "she"
        resolved: "she"
  ▽ clusters: [["she", "she"]]
    ▽ [0]: ["she", "she"]
        [0]: "she"
        [1]: "she"
    resolved: "she is close to the habour. where might she be heading?"

It seems like NOMINAL is missing somehow.