Adding WordNet as synonym generator

[HonzaCuhel]

This PR includes:

• creating an abstract class for synonym generation
• adding WordNet synonym generator

Comparison of Mistral synonym generator vs WordNet:

Object	Mistral Generated Synonyms	WordNet Generated Synonyms
astronaut	['astronaut', 'cosmonaut', 'spaceman']	['spaceman', 'astronaut', 'cosmonaut']
cat	['feline', 'kitty', 'puss']	['throw up', 'bozo', 'kat']
dog	['cat', 'pup', 'hound']	['chase', 'weenie', 'wienerwurst']
person	['individual', 'human', 'being']	['mortal', 'individual', 'soul']
horse	['horse', 'equine', 'nag']	['sawhorse', 'knight', 'gymnastic horse']
car	['automobile', 'vehicle', 'transportation']	['cable car', 'automobile', 'elevator car']
alien	['Extraterrestrial', 'extraneous', 'foreign']	['disaffect', 'extraterrestrial', 'foreigner']
city	['town', 'urban area', 'metropolis']	['metropolis', 'urban center', 'city']
plane	['level', 'flat', 'even']	['aeroplane', 'planer', 'shave']
airport	['airport', 'terminal', 'aviation']	['aerodrome', 'airport', 'drome']
robot	['machine', 'automaton', 'mechanism']	['golem', 'robot', 'automaton']
tractor	['Tractor', 'Truck', 'Vehicle']	['tractor']
bus	['1 vehicle', '2 coach', '3 publictransport']	['motorbus', 'passenger vehicle', 'autobus']
bicycle	['bicycle', 'bike', 'bicycle']	['pedal', 'cycle', 'wheel']

Latency (measured on device with 30GB RAM, L4 24GB gpu) of generation 3 synonyms for 14 objects:

• Mistral: 8 seconds
• WordNet: 1 second

[github-actions]

☂️ Python Coverage

current status: ✅

Overall Coverage

Lines	Covered	Coverage	Threshold	Status
919	444	48%	0%	🟢

New Files

File	Coverage	Status
datadreamer/prompt_generation/lm_synonym_generator.py	32%	🟢
datadreamer/prompt_generation/wordnet_synonym_generator.py	81%	🟢
TOTAL	56%	🟢

Modified Files

File	Coverage	Status
datadreamer/pipelines/generate_dataset_from_scratch.py	44%	🟢
datadreamer/prompt_generation/init.py	100%	🟢
datadreamer/prompt_generation/synonym_generator.py	84%	🟢
TOTAL	76%	🟢

updated for commit: 8b05c99 by action🐍

[sokovninn]

@HonzaCuhel is it possible to configure wordnet to always return nouns, because for the word "bear" it returns verbs.

[HonzaCuhel]

Yes, it's possible

[github-actions]

Test Results

  6 files  ± 0    6 suites  ±0   45m 51s ⏱️ - 2m 44s
 81 tests + 5   33 ✅ + 3   48 💤 + 2  0 ❌ ±0 
486 runs  +30  198 ✅ +18  288 💤 +12  0 ❌ ±0 

Results for commit 8b05c99. ± Comparison against base commit bd40543.

[HonzaCuhel]

@sokovninn changed the code, so the generator only returns noun synonyms