Using Large Language Models

This assignment constitutes 30% of your final grade. Please work on the task individually. Unless specifically requested otherwise, your report may later be posted on-line with a CC-BY license.

Deadline: see the main page.

Create linguistic resources using one or more large language models

Create the following 3 resources (which would be useful for creating a sentiment analysis system like Vader)
1. A list of evaluative words in Czech, with English Translations, with a valence between +1 and -1, formatted as TSV (tab separated values).
```
	
skvělý	truly exceptional, outstanding	+1
dobrý	good, satisfactory, of a good standard	+0.7
fajn	okay, good enough, not bad	+0
špatný	bad, poor, of poor quality	-0.6
zlý	bad, evil, wicked	-0.7
otřesný	terrible, shocking, causing great distress	-0.9
```
2. A list of boosters in Czech, with English translations, formatted as TSV
```
opravdu   really
velmi	  very
hodně     a lot
	
```
3. A list of diminishers in Czech, with English translations, formatted as TSV
```
docela     quite
trochu     a bit
poněkud    somewhat
celkem     fairly
    
```

You should use at least two different LLMs

You should try multiple different queries

You should discuss which queries worked better with which systems and why (you need to date the query/response, as they change often)

You should try to get as many useful words (i.e. those with a valence not equal to zero) as you can

You can combine the results of different queries by hand, or try to use an LLM to do this

You can use the LLM to write part of the report, if you do, you should make clear which part and what the prompt was

You will be evaluated both on the quality of the resource you are able to build and the discussion of how you used the LLMs to build it.

Deliverable

The deliverable is be a paper of no more than twelve pages including diagrams, prompts and references
- Along with three files valence.tsv, booster.tsv, diminisher.tsv.
- zip them up in a file surname.zip
Make sure you include page numbers
Use single spacing.
Submit only softcopy

References

Ethan Mollick (2023-12) An Opinionated Guide to Which AI to Use: ChatGPT Anniversary Edition, One Useful Thing
Hutto, C.J. & Gilbert, E.E. (2014). VADER: A Parsimonious Rule-based Model for Sentiment Analysis of Social Media Text. Eighth International Conference on Weblogs and Social Media (ICWSM-14). Ann Arbor, MI, June 2014. (Software)

Assignment Three for LTI