Analysis of words: algorithms for biological sequences, music and texts

单词分析：生物序列、音乐和文本的算法

• 批准号: RGPIN-2016-03661
• 负责人: Brown, Daniel
• 金额: $ 1.6万
• 依托单位: University of Waterloo
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2016
• 资助国家: 加拿大
• 起止时间: 2016-01-01 至 2017-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=614536
• 关键词: Analysis; words; algorithms; biological; sequences

We will develop new methods in three interrelated areas of computer science: bioinformatics, music information retrieval and computational creativity. All three projects will develop the skills of young Canadian researchers to work in the new economy. For bioinformatics, our focus is on building better methods for reconstructing evolutionary trees. We have previously designed extremely fast algorithms for this problem that depend on sequences being pre-aligned, and that depend on the high quality of those alignments. In this project, we will focus on the faster, but traditionally less accurate, methods that don't start with alignments, and build a hybrid system that merges some alignment approaches with existing alignment-free methods, to join high speed and high accuracy in this important evolutionary biology task. In music information retrieval, our focus is on using bioinformatics algorithms to enhance similarity search in symbolically-represented music. We will develop methods that look for similar motifs and common patterns, as a way of identifying musical influence and historical changes in style. Our approach, derived from DNA sequence alignment, allows us to speed up what has traditionally been a slow area of music information retrieval; we may also be able to develop algorithms that work on recorded music. Another focus is on music lyrics, where we will continue our work on analysis of patterns: rhyme, meter, wordplay, authorial voice, and other features. Previously, we have mostly focused on rap music, but we will also see to what extent we can characterize what makes a successful lyric in other genres, such as pop music. Preliminary work has suggested that highly complex rhymes predict song success and popularity on music sites like Shazam, and we will continue to explore the power and quality of lyrics. Finally, in computational creativity, we will build a cyborg poet that combines current event news tweets with structure and style rules of successful poetry to build a "deadline poet" that responds to the news. We will investigate which features of tweets make them more and less successful in such a poet, and also analyze which other features are necessary to separate successful poems from unsuccessful ones. Along the way, we will also critique what it means to create artistic, creative objects from human-generated ones, and what this dichotomy has to say about whether computers can be creative agents.

我们将在计算机科学的三个相互关联的领域开发新方法：生物信息学，音乐信息检索和计算创造力。这三个项目都将培养加拿大年轻研究人员在新经济中工作的技能。 对于生物信息学，我们的重点是建立更好的方法来重建进化树。我们以前已经设计了非常快速的算法，这个问题取决于序列被预先对齐，并依赖于这些对齐的高质量。在这个项目中，我们将专注于更快，但传统上不太准确的方法，这些方法不从比对开始，并建立一个混合系统，将一些比对方法与现有的无比对方法合并，以在这个重要的进化生物学任务中加入高速度和高准确性。 在音乐信息检索，我们的重点是使用生物信息学算法，以提高相似性搜索的象征性代表的音乐。我们将开发寻找相似主题和共同模式的方法，作为识别音乐影响和风格历史变化的一种方式。我们的方法来自DNA序列比对，使我们能够加快传统上缓慢的音乐信息检索领域;我们也可以开发出适用于录制音乐的算法。 另一个重点是音乐歌词，我们将继续分析模式：押韵，节拍，文字游戏，作者声音和其他特征。以前，我们主要关注说唱音乐，但我们也将看到我们在多大程度上可以描述什么使其他类型的歌词成功，如流行音乐。初步研究表明，高度复杂的押韵可以预测歌曲在Shazam等音乐网站上的成功和受欢迎程度，我们将继续探索歌词的力量和质量。 最后，在计算创造力中，我们将构建一个电子诗人，将时事新闻推文与成功诗歌的结构和风格规则相结合，构建一个响应新闻的“最后期限诗人”。我们将调查哪些功能的推文，使他们更成功和更不成功的这样一个诗人，也分析了其他功能是必要的，以区分成功的诗歌从不成功的。沿着，我们还将批判从人类生成的对象中创造艺术的、创造性的对象意味着什么，以及这种二分法对计算机是否可以成为创造性代理人的说法。