ibatra / bert-keyword-extractor Goto Github PK

The convert Function that converts the phrases into B,H I tokens is not valid for repeated words in a keyphrase. As the function is made they will take the first occurance of the word and not the desired word in case of repetation.

Do you have a pre-trained model that we can use for a downstream system? It would be awesome if you can provide us with that!

Why using the flat_accuracy function, instand of the accuracy_score provied by seqeval package?

I am going to use BERT-based Keyword Extraction. But MAX_LENGTH config is set to 75, so is it okay to change this value? As far as I know, BERT has a MAX input length of 512, but I would like to ask you how to solve it if you want to put a much longer input.

argparse module not imported

i m really insipired by #6

but i m wondering even absent multiple keywords from paragraph?

Can the model be used to extract keyword from a paragraph, that consists of several sentences? Thank you.

RuntimeError: index out of range at /pytorch/aten/src/TH/generic/THTensorEvenMoreMath.cpp:191

Hello Sir,

can we use it for topic detection from a set of documents?

如题……

Can I use the fine-tuned model to extract keywords from any .txt file if I don't have the keys?

Running the repository without any modifications and getting the following error? Does anyone know how to fix this?

TypeError Traceback (most recent call last)
in
58 pred_tags = [tags_vals[p_i] for p in predictions for p_i in p]
59 valid_tags = [tags_vals[l_ii] for l in true_labels for l_i in l for l_ii in l_i]
---> 60 print("F1-Score: {}".format(f1_score(pred_tags, valid_tags)))

~/miniconda3/lib/python3.8/site-packages/seqeval/metrics/sequence_labeling.py in f1_score(y_true, y_pred, average, suffix, mode, sample_weight, zero_division, scheme)
357 suffix=suffix)
358 else:
--> 359 _, _, f, _ = precision_recall_fscore_support(y_true, y_pred,
360 average=average,
361 warn_for=('f-score',),

~/miniconda3/lib/python3.8/site-packages/seqeval/metrics/sequence_labeling.py in precision_recall_fscore_support(y_true, y_pred, average, warn_for, beta, sample_weight, zero_division, suffix)
128 return pred_sum, tp_sum, true_sum
129
--> 130 precision, recall, f_score, true_sum = _precision_recall_fscore_support(
131 y_true, y_pred,
132 average=average,

~/miniconda3/lib/python3.8/site-packages/seqeval/metrics/v1.py in _precision_recall_fscore_support(y_true, y_pred, average, warn_for, beta, sample_weight, zero_division, scheme, suffix, extract_tp_actual_correct)
120 raise ValueError('average has to be one of {}'.format(average_options))
121
--> 122 check_consistent_length(y_true, y_pred)
123
124 pred_sum, tp_sum, true_sum = extract_tp_actual_correct(y_true, y_pred, suffix, scheme)

~/miniconda3/lib/python3.8/site-packages/seqeval/metrics/v1.py in check_consistent_length(y_true, y_pred)
95 is_list = set(map(type, y_true)) | set(map(type, y_pred))
96 if not is_list == {list}:
---> 97 raise TypeError('Found input variables without list of list.')
98
99 if len(y_true) != len(y_pred) or len_true != len_pred:

TypeError: Found input variables without list of list.

I get some errors while running the code, some packages are not imported. There are some other error that may be fixed but I did not go through the details. Ishaan, would you update your code?

I have a requirement were I need to extract more than one keyword from the sentence. Is it possible to do with just passing the text of interest once?
Example:
Input text: "I don't like this coffee because it's too sweet after having it for few days I don't like it anymore"
Output keywords: "too sweet" and "don't like it".