philipericsson.github.io/lyrics.html at main · philipericsson/philipericsson.github.io · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <meta name="description" content="Philip Ericsson - Lyrics Generation AI Project">
    <link rel="icon" type="image/x-icon" href="/summit.ico">
    <link rel="stylesheet" href="css/main.css">
    <title>Philip Ericsson</title>
    <link rel="stylesheet" href="css/project-pages.css">
</head>
<body class="project-page">
  <header class="intro-logo">
    <p id="intro_paragraph">
      <span class="squiggle-word"><a href="index.html">Philip Ericsson</a></span>
    </p>
  </header>

  <main>
    <section class="select-projects">
      <h2 id="select_projects"><a">CORNELL UNIVERSITY</a></h2>
    </section>

    <section class="footer-vertical text-heavy">
      <p id="second_paragraph">
        "Spittin' Bars: Fine-tuning Large Language Models for Lyric Generation" demonstrates notable differences in text generation between pre and post fine-tuned large language models. The fine-tuned model consistently produced text aligning with the stylistic and thematic patterns found in song lyrics. Utilizing a proprietary dataset created by scraping lyrics from thirty songs spanning diverse genres—including classic rock, reggae, drill, psychedelic, indie, country, emo rap, soul, pop, electronic, bluegrass, R&B, new wave, and hip hop—the model's fine-tuning process demonstrates significant impact on its ability to generate text resembling song lyrics.
      </p>
      <img src="images/landscape.jpg" alt="Landscape" loading="lazy">
    </section>
  </main>

</body>
</html>