Podrobno

Generiranje zgodbe v tekstovni igri z nevronskimi mrežami
ID HERCEG, ALEN (Avtor), ID Klemenc, Bojan (Mentor) Več o mentorju... Povezava se odpre v novem oknu

.pdfPDF - Predstavitvena datoteka, prenos (1007,35 KB)
MD5: 3190432DAA397C64540C8BC9B4E0DEFA

Izvleček
V diplomski nalogi smo se lotili uporabiti nevronske mreže za generiranje naključnih zgodb. Te zgodbe so se uporabile v tekstovni igri, kjer je lahko uporabnik igral skozi naključno zgodbo z neskončno načinov igranja. Model, ki je bil uporabljen za generiranje besedila, je GPT-2, sama tekstovna igrica pa je bila narejena v obliki spletne aplikacije, ki uporablja računalnik v oblaku za shranjevanje ter uporabljanje samega modela. Za učenje modela smo uporabili angleške knjige raznih zvrsti. Rezultati diplomske naloge je igra, kjer se za vsako igro generira nova zgodba ter uporabnik lahko igra od začetka do konca brez težav. S tem smo pokazali potencial nevronskih mrež v igricah ter možno uporabo.

Jezik:Slovenski jezik
Ključne besede:Generiranje zgodbe, tekstovna igrica, nevronske mreže
Vrsta gradiva:Diplomsko delo/naloga
Tipologija:2.11 - Diplomsko delo
Organizacija:FRI - Fakulteta za računalništvo in informatiko
Leto izida:2021
PID:20.500.12556/RUL-125364 Povezava se odpre v novem oknu
COBISS.SI-ID:54941443 Povezava se odpre v novem oknu
Datum objave v RUL:12.03.2021
Število ogledov:1625
Število prenosov:165
Metapodatki:XML DC-XML DC-RDF
:
HERCEG, ALEN, 2021, Generiranje zgodbe v tekstovni igri z nevronskimi mrežami [na spletu]. Diplomsko delo. [Dostopano 14 avgust 2025]. Pridobljeno s: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=slv&id=125364
Kopiraj citat
Objavi na:Bookmark and Share

Sekundarni jezik

Jezik:Angleški jezik
Naslov:Generating gameplay story with neural network in a text-based game
Izvleček:
In our thesis, we set out to use neural networks to generate random stories. These stories were used in a text game where the user could play through a random story with endless gameplay possibilities. The model used to generate the text is GPT-2, and the text adventure is made as a simple web application, which implements cloud services for computing and storage of the models. To train the model, we used English literature from various sources. The result of the thesis is a game where a new story is generated for each game and the user is able to play the game from start to finish without any problems. This use case shows the potential of neural networks in games.

Ključne besede:Generating gameplay story, text-based game, neural networks

Podobna dela

Podobna dela v RUL:
  1. Usefulness of immunohistochemically determined epidermal growth factor receptor mutations in lung cancer
  2. Review of the treatment of advanced non-small-cell lung cancer patients with mutation in the epidermal growth factor during the years of 2010 and 2014 at the University Clinic Golnik
  3. Design of novel constructs for expression of recombinant EGFR ectodomain in mammalian cell lines.
  4. Comparison of cartilage-associated genes expression in human primary chondrocytes cultured in different growth media
  5. Study of changes in expression of COL5A1, COL11A1, GREM1 and MAOA genes in human osteosarcoma cells exposed to bisphenols A, S and AF
Podobna dela v drugih slovenskih zbirkah:
  1. Multicenter evaluation of the fully automated PCR-based Idylla EGFR Mutation Assay on formalin-fixed, paraffin-embedded Q1 tissue of human lung cancer
  2. Adverse Effects of Treatment with EGFR Inhibitor – Cetuximab
  3. Epidermal growth factor receptor mutations and treatment of non-small-cell lung cancer
  4. Sequential treatment with afatinib and osimertinib in patients with EGFR mutation-positive non-small-cell lung cancer
  5. Expression of FGFR1-4 in malignant pleural mesothelioma tissue and corresponding cell lines and its relationship to patient survival and FGFR inhibitor sensitivity

Nazaj