madhurjindal

/

autonlp-Gibberish-Detector-492513457

like 52

Hi madhurjindal,
I tried to use your code to detect Gibberish, but I the output I received seemed strange and I didn't understand it. What I did was the following:

• installed transformers, etc.
• from transformers import AutoModelForSequenceClassification, AutoTokenizer
  model = AutoModelForSequenceClassification.from_pretrained("madhurjindal/autonlp-Gibberish-Detector-492513457")
  tokenizer = AutoTokenizer.from_pretrained("madhurjindal/autonlp-Gibberish-Detector-492513457")
• inputs = tokenizer("I like apples.", return_tensors="pt")
  outputs = model(**inputs)
  outputs

The output is: SequenceClassifierOutput(loss=None, logits=tensor([[ 2.0615, 0.1996, -2.1773, -0.5643]], grad_fn=), hidden_states=None, attentions=None)

Now, if I understood correctly, tensor gives you the weight/number representing [clean, mild gibberish, word salad, noise] ? What I don't understand is the meaning of the positive and negative number plus its size. What is more, I wanted to ask if with this code one can identify a Gibberish word in a sentence, document, instead of getting some general numbers related to the full sentence?

Thank you in advance for you reply.
Kind regards,
M