Vidyarth E.S

Project: Blogwall

Part: Poetry Analysis

The younger generation today is fast embracing popular culture and this has inadvertently endangered the preservation of intangible cultural sources. Creating awareness in intangible cultural resources, such as traditional poetry, by making it an integral part of the consuming popular culture, we can introduce the topic of preservation to the younger generation. The BlogWall provides a means of expression that is inherently familiar to young people today. However, it is difficult for most of us to actually create a poem. The BlogWall is an attempt to bridge this gap. By blending media art and poetry, we have developed a poetry mixer called to extend text messages to a new level of self expression and public communication. From a single text message, the system is capable of creating a new poem by drawing from an existing body of poetry. The technique integrates a number of ideas from different disciplines such as artificial intelligence, supervised learning, information retrieval and natural language understanding including word sense disambiguation, topic summarizing, super sense matching, genetic algorithms etc and augments the system with basic emotional intelligence to create a model for poetry generation and remixing. Thus, it allows for new forms of cultural computing.

However, it is difficult for computers to actually create a poem. What it has is just a database of poems that it can work on. This research study aims to come up with intelligent algorithms to come up with proper user understandable poems based on user inputs. Simple word matching & term frequencies might not suffice for such a system.  This research specifically studies a few novel Natural Language Processing approaches to understand the user inputs and the poems. Specifically it studies the various industry standard word sense disambiguation methods and come up with the best method (experimentally) to understand the poems and based on that come up with a basic topic summarizing system (for nouns, verbs, adjectives and adverbs) which in turn helps in matching user input and the poem lines and then use genetic algorithms to come up with remixed poems that are coherent.