could you provide the topic entity in each sentence?

[ToneLi]

No description provided.

[Wuyxin]

Hi, are you referring to the paper topic (field) of STaRK-MAG?

[ToneLi]

the topic entity in the input sentence (query) of stark-mag, STaRK-prime, and stark-amazon

[ToneLi]

thanks for that

[ToneLi]

The entity's surface name that appears in the sentence, such as 'Yuriy Norshteyn,' is shown below.

[Wuyxin]

Got it! Thanks for the explanations.

For now we see the topic entity ids as part of the meta data of the natual language query. We prefer not to release them for now since it may be eaiser to conduct retrieval while they may not always be available in the real-world cases.

Another note if you want to try to use NER model to recognize the entities: Some entities (like products) may have the same name. We also found that a proportion of authors in STaRK-MAG are "Unknown".

While the tasks could be much more challenging due to the unavailable topic entities and ambiguity as mentioned, we hope to keep the tasks close to real-world situations. Let me know if these make sense to you!