Configuration 1: xlnet on fasttext with cosine distance
Configuration 2: bert on fasttext with cosine distance
Testing on: combined queries for the online_delivary schema
Entries below have greater num_guesses with a threshold of 3
--------------------------------------------------
--------------------------------------------------

Query (#1): predict the total order where the cuisine is chinese for the next week
Ground Truth (filter): PERFERENCE(P1)

Annotation 1: [{'text': 'cuisine', 'confidence': 0.8491058945655823}, {'text': 'chinese', 'confidence': 0.9999929964542389}]
Annotation 2: [{'text': 'chinese', 'confidence': 0.6809283196926117}]

Configuration 1 took 15 attempt(s) to get the correct answer
Configuration 2 took 15 attempt(s) to get the correct answer

--------------------------------------------------

Query (#2): predict the total order where the preferred medium of order is online within tomorrow
Ground Truth (filter): MEDIUM (P1)

Annotation 1: [{'text': 'preferred medium of order is online within', 'confidence': 0.8779428601264954}]
Annotation 2: [{'text': 'preferred medium of order', 'confidence': 0.876037523150444}, {'text': 'online', 'confidence': 0.7616453766822815}]

Configuration 1 took 11 attempt(s) to get the correct answer
Configuration 2 took 11 attempt(s) to get the correct answer

--------------------------------------------------

