Re-annotate of UD_Turkish_Atis with slots and intents. This treebank is a translation of the original ATIS dataset. Hence Atis Turkish is an intent and entity detection dataset for Turkish.
Intents and slots are extracted by transfer learning. For the explanation of the work I'll compile a paper soon. Here are the all intent names and slots, same with the original ATIS.
All intent names
abbreviation
aircraft
airfare
airline
airport
capacity
city
distance
flight
flight_no
flight_time
ground_fare
ground_service
meal
quantity
restriction
All slot names
aircraft_code
airline_code
airline_name
airport_code
airport_name
arrive_date.date_relative
arrive_date.day_name
arrive_date.day_number
arrive_date.month_name
arrive_date.today_relative
arrive_end_time.time
arrive_start_time.time
arrive_time.day_number
arrive_time.period_of_day
arrive_time.time
arrive_time.time_relative
city_name
class_type
connect
cost_relative
date_relative
day_name
day_number
depart_date.date_relative
depart_date.day_name
depart_date.day_number
depart_date.month_name
depart_date.today_relative
depart_date.year
depart_day.day_number
depart_end_time.time
depart_start_time.time
depart_time.period_of_day
depart_time.time
depart_time.time_relative
fare_amount
fare_basis_code
flight_mod
flight_number
flight_stop
flight_time
fromloc.airport_code
fromloc.airport_name
fromloc.city_name
fromloc.state_code
fromloc.state_name
meal
meal_code
meal_description
mod
month_name
or
part_date.day_number
period_of_day
restriction_code
return_date.date_relative
return_date.day_name
return_date.day_number
return_date.month_name
return_date.today_relative
return_end_time.time
return_time.period_of_day
return_time.time
round_trip
state_code
state_name
stoploc.airport_name
stoploc.city_name
stoploc.ity_name
stoploc.state_code
today_relative
toloc.airport_code
toloc.airport_name
toloc.city_name
toloc.country_name
toloc.state_code
toloc.state_name
transport_type
Dataset is presented both in conll and json formats. Here is an example conll annotation for you:
# sent_id = 0001.dev
# text = Pittsburgh'tan Atlanta'ya 25 Nisan'da kalkan ve 6 Mayıs'ta dönen en ucuz uçuşu istiyorum
# intent: flight
1 Pittsburgh'tan pittsburgh PROPN B-fromloc.city_name
2 Atlanta'ya atlanta PROPN B-toloc.city_name
3 25 25 NUM B-depart_date.day_number
4 Nisan'da nisan PROPN B-depart_date.month_name
5 kalkan kalk ADJ O
6 ve ve CCONJ O
7 6 6 NUM B-return_end_time.time
8 Mayıs'ta mayıs PROPN B-return_date.month_name
9 dönen dön ADJ O
10 en en ADV B-cost_relative
11 ucuz ucuz ADJ I-cost_relative
12 uçuşu uç NOUN O
13 istiyorum iste VERB O
and a json annotation:
{
"text": "Cleveland'dan Miami'ye uçuşlar",
"intent": "flight",
"entities": [
{
"start": 0,
"end": 13,
"value": "Cleveland'dan",
"entity": "fromloc.city_name"
},
{
"start": 14,
"end": 22,
"value": "Miami'ye",
"entity": "toloc.city_name"
}
]
},
If you want to find out more about how I compiled this dataset you can visit my Youtube channel and watch the playlist Veriseti nasıl derlenir?. For finest Turkish NLP, you can our website for Turkish NLP or visit my personal website.
This work is supported by Google Developer Experts Program. Part of Duygu 2022 Fall-Winter collection, "Turkish NLP with Duygu"/ "Duygu'yla Türkçe NLP". All rights reserved.
Shield:
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
