Yeah, probably meant probability. This is different from embeddings thought. Here, you can just think of it as a multi-class problem where each word represents a class. Like in a multi-class setting, you get a score or probability for each class, and you usually pick the highest one as the predicted one. An embedding is a different concept, although you could consider the values before going to the softmax function as an embedding so it is somewhat related.
Hi Sebastian , I have one quick question . for multi class text classification for short text (max 3-4 words) . What's the best possible approach . Am sure like embeddings is not meant for that . NER not entirely sure . Waiting for your valuable feedback
You could try a bag-of-words and n-gram-based approach: github.com/rasbt/machine-learning-book/blob/main/ch08/ch08.ipynb Or, as a baseline, a rule-based / dictionary-based approach is maybe also a good idea: github.com/rasbt/stat453-deep-learning-ss21/blob/main/L15/0_rule-based-baseline.ipynb
Excellent tutorial!
Possible word percentage is probability? Is it same like embedings?
Yeah, probably meant probability. This is different from embeddings thought. Here, you can just think of it as a multi-class problem where each word represents a class. Like in a multi-class setting, you get a score or probability for each class, and you usually pick the highest one as the predicted one. An embedding is a different concept, although you could consider the values before going to the softmax function as an embedding so it is somewhat related.
Hi Sebastian , I have one quick question .
for multi class text classification for short text (max 3-4 words) . What's the best possible approach .
Am sure like embeddings is not meant for that . NER not entirely sure .
Waiting for your valuable feedback
You could try a bag-of-words and n-gram-based approach: github.com/rasbt/machine-learning-book/blob/main/ch08/ch08.ipynb
Or, as a baseline, a rule-based / dictionary-based approach is maybe also a good idea: github.com/rasbt/stat453-deep-learning-ss21/blob/main/L15/0_rule-based-baseline.ipynb
nice.