Example is not question classification

[zzj0402]

In the second example of Universal Sentence Encoder, the link directs to a sentiment classification instead of a "question" classification.

https://tfhub.dev/google/universal-sentence-encoder/2 https://colab.research.google.com/github/tensorflow/hub/blob/master/docs/tutorials/text_classification_with_tf_hub.ipynb#scrollTo=N6ZDpd9XzFeN

[arnoegw]

Indeed, the section titled "Classification" on page https://tfhub.dev/google/universal-sentence-encoder/2 shows a picture about binary classificaton of sentences as question or not, but the linked colab notebook https://colab.research.google.com/github/tensorflow/hub/blob/master/docs/tutorials/text_classification_with_tf_hub.ipynb is about classifying positive vs negative sentiment.

To my understanding, that's because the two pictures in sections "Semantic Similarity" and "Classification" show the same inputs but a different use of their embeddings. The actual colabs cannot run on just those three examples pictured, and instead use larger, well-known datasets for the respective kind of problem. The reader is being asked to abstract from the concrete problem to its underlying structure.

To me, this looks "working as intended".

[zzj0402]

To me that is an illustration error.