Different output than the one shown on the README #2
Comments
|
Hello Daniel,
Indeed, the code does not provide the intended result. I update the git
repository and added the postprocessing function.
You need to update your repository in order to have the latest code.
I did not update the transformer model. Should be done in a few months.
Thank you for your feedback!
Have a nice day
Mohamed Hatmi
…On Sun, Sep 18, 2022 at 5:48 PM Daniel Izham ***@***.***> wrote:
Hello there,
I have found that the model is not giving the results I expect based on my
reading of the documentation on GitHub and Hugging Face.
When I tried the sentence 'رغم الهدنة .. معارك قره باغ متواصلة وأذربيجان
تعلن سيطرتها على مزيد من القرى', I got:
[[], [{'entity': 'B-LOCATION', 'score': 0.99871314, 'index': 2, 'word': 'قر', 'start': 7, 'end': 9}, {'entity': 'B-LOCATION', 'score': 0.998519, 'index': 3, 'word': '##ه', 'start': 9, 'end': 10}, {'entity': 'I-LOCATION', 'score': 0.9986701, 'index': 4, 'word': 'باغ', 'start': 11, 'end': 14}]]
when what is expected is the following:
{"text": "رغم الهدنة . . معارك قره باغ متواصلة و أذربيجان تعلن سيطرتها على مزيد من القرى", "entities": [{"type": "LOCATION", "entity": "قره باغ", "start_offset": 21, "end_offset": 28}, {"type": "ORGANIZATION", "entity": "أذربيجان", "start_offset": 39, "end_offset": 47}]}
In terms of the format, I can modify my code or the source code so that I
can process the data as I want to. But the output of the entities are not
as good and some common entities have been missed (i.e. أذربيجان). I have
ran this with multiple examples and the results are the same. Has the model
changed?
—
Reply to this email directly, view it on GitHub
<#2>, or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AFLPK3FTQYU2TO7T5BEFUXTV642T7ANCNFSM6AAAAAAQPP3YY4>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
|
Thank you for your response. The output looks nicer and as expected now. However, I noticed that the words that are not named entities are being detected as PERSON and the named entities are labelled wrongly: Is this due to the outdated model as you mentioned? |
|
Hello Daniel,
Indeed, something went wrong. I think it is related to the mapping between
the output of the model and the labels. I will check it and get back to you.
Thank you for your feedback
Have a nice day
…On Tue, Sep 20, 2022 at 8:24 AM Daniel Izham ***@***.***> wrote:
Thank you for your response. The output looks nicer and as expected now.
However, I noticed that the words that are not named entities are being
detected as PERSON and the named entities are labelled wrongly:
{"text": "رغم الهدنة . . معارك قره باغ متواصلة و أذربيجان تعلن سيطرتها على مزيد من القرى", "entities": [{"type": "PERSON", "entity": "رغم", "start_offset": 1, "end_offset": 3}, {"type": "PERSON", "entity": "الهدنة", "start_offset": 4, "end_offset": 10}, {"type": "PERSON", "entity": ".", "start_offset": 11, "end_offset": 12}, {"type": "PERSON", "entity": ".", "start_offset": 13, "end_offset": 14}, {"type": "PERSON", "entity": "معارك", "start_offset": 15, "end_offset": 20}, {"type": "LOCATION", "entity": "قره", "start_offset": 21, "end_offset": 24}, {"type": "DATE", "entity": "باغ", "start_offset": 25, "end_offset": 28}, {"type": "PERSON", "entity": "متواصلة", "start_offset": 29, "end_offset": 36}, {"type": "PERSON", "entity": "و", "start_offset": 37, "end_offset": 38}, {"type": "PERSON", "entity": "أذربيجان", "start_offset": 39, "end_offset": 47}, {"type": "PERSON", "entity": "تعلن", "start_offset": 48, "end_offset": 52}, {"type": "PERSON", "entity": "سيطرتها", "start_offset": 53, "end_offset": 60}, {"type": "PERSON", "entity": "على", "start_offset": 61, "end_offset": 64}, {"type": "PERSON", "entity": "مزيد", "start_offset": 65, "end_offset": 69}, {"type": "PERSON", "entity": "من", "start_offset": 70, "end_offset": 72}, {"type": "PERSON", "entity": "القرى", "start_offset": 73, "end_offset": 78}]}
Is this due to the outdated model as you mentioned?
—
Reply to this email directly, view it on GitHub
<#2 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AFLPK3B26WWWUN557QULPNTV7FKABANCNFSM6AAAAAAQPP3YY4>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
|
Hello Daniel,
I updated the Arabic NER model. It should be working correctly now.
Have a nice day
On Thu, Sep 22, 2022 at 12:20 PM Mohamed Hatmi ***@***.***>
wrote:
… Hello Daniel,
Indeed, something went wrong. I think it is related to the mapping between
the output of the model and the labels. I will check it and get back to you.
Thank you for your feedback
Have a nice day
On Tue, Sep 20, 2022 at 8:24 AM Daniel Izham ***@***.***>
wrote:
> Thank you for your response. The output looks nicer and as expected now.
> However, I noticed that the words that are not named entities are being
> detected as PERSON and the named entities are labelled wrongly:
>
> {"text": "رغم الهدنة . . معارك قره باغ متواصلة و أذربيجان تعلن سيطرتها على مزيد من القرى", "entities": [{"type": "PERSON", "entity": "رغم", "start_offset": 1, "end_offset": 3}, {"type": "PERSON", "entity": "الهدنة", "start_offset": 4, "end_offset": 10}, {"type": "PERSON", "entity": ".", "start_offset": 11, "end_offset": 12}, {"type": "PERSON", "entity": ".", "start_offset": 13, "end_offset": 14}, {"type": "PERSON", "entity": "معارك", "start_offset": 15, "end_offset": 20}, {"type": "LOCATION", "entity": "قره", "start_offset": 21, "end_offset": 24}, {"type": "DATE", "entity": "باغ", "start_offset": 25, "end_offset": 28}, {"type": "PERSON", "entity": "متواصلة", "start_offset": 29, "end_offset": 36}, {"type": "PERSON", "entity": "و", "start_offset": 37, "end_offset": 38}, {"type": "PERSON", "entity": "أذربيجان", "start_offset": 39, "end_offset": 47}, {"type": "PERSON", "entity": "تعلن", "start_offset": 48, "end_offset": 52}, {"type": "PERSON", "entity": "سيطرتها", "start_offset": 53, "end_offset": 60}, {"type": "PERSON", "entity": "على", "start_offset": 61, "end_offset": 64}, {"type": "PERSON", "entity": "مزيد", "start_offset": 65, "end_offset": 69}, {"type": "PERSON", "entity": "من", "start_offset": 70, "end_offset": 72}, {"type": "PERSON", "entity": "القرى", "start_offset": 73, "end_offset": 78}]}
>
>
> Is this due to the outdated model as you mentioned?
>
> —
> Reply to this email directly, view it on GitHub
> <#2 (comment)>,
> or unsubscribe
> <https://github.com/notifications/unsubscribe-auth/AFLPK3B26WWWUN557QULPNTV7FKABANCNFSM6AAAAAAQPP3YY4>
> .
> You are receiving this because you commented.Message ID:
> ***@***.***>
>
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hello there,
I have found that the model is not giving the results I expect based on my reading of the documentation on GitHub and Hugging Face.
When I tried the sentence 'رغم الهدنة .. معارك قره باغ متواصلة وأذربيجان تعلن سيطرتها على مزيد من القرى', I got:
when what is expected is the following:
In terms of the format, I can modify my code or the source code so that I can process the data as I want to. But the output of the entities are not as good and some common entities have been missed (i.e. أذربيجان). I have ran this with multiple examples and the results are the same. Has the model changed?
The text was updated successfully, but these errors were encountered: