The aim of the project

The aim of this project is to collect knowledge of a poem (we choose as an example “I Felt a Funeral, in my Brain” by Emily Dickinson) and convert it in a linked open data environment, as the RDF graph. This process has been based on the development of an on-purpose Ontology (Krepo). Then, from the RDF graph has been extracted most of the knowledge, that has been finally presented in a usable interface. The final product is a graphic interface of the poem related to its additional information and interpretation, that a casual user can easily exploit to read, analyze and understand the poem’s general information, its literal and symbolic meaning. The whole process is totally expressed by the project’s name: KREPoLiSMA.

Choice of the poem

Our main idea was to have a poem with a manageable number of lines, enough to have different kind of rhetorical devices but not too long-winded because we thought it was better to get more information on a smaller scale than to get sporadic information on a huge content. Moreover, we tried to avoid both widespread poems with overloads of information and numerous project already done over them and also niche works that didn’t have any critical analysis or general information or references that we could exploit in our project. Considering all the criteria previously cited we ended up choosing “I felt a Funeral, in my Brain” by Emily Dickinson.

Gathering of information (Manual)

Research on the critics who analysed the poem and on its rhetorical devices
As a starting point for our project’s development (RDF Graph) we looked for information on the poem’s background. In particular we focused on some basic information on the poem and the author. Then, we searched for Critical Analysis on “I Felt a Funeral in my Brain” from several authors. In this way we could summarily and clearly present poem’s interpretation through reliable sources. So, from these critics’ analysis we extracted just brief “snippets”, which could explain a particular aspect or of the whole poem, or a particular stanza/line/rhetorical figure/line.

Gathering of information (Automatic)

Use of Aylien Library to automatically get sentiment analysis
Extracting knowledge over the poems and its relation automatically was one of the key features that we wanted to implement in our project. In particular, we focused on sentiment analysis, automatic poetic analysis and links to external resources. We choose to use the Aylien (link) library, that can be called and installed with multiple programming languages (we chose the python one) to automatically extract information about a text. Over the various kind of information that we could get from our poem using Aylien we chose to display the sentiment analysis and we decided to try to get it line by line to create a graph that could show how the polarity of the poem evolved through its intercourse. We used P.A.N software to automatically analyse our poem and its meaning. In particular we used the function of general meaning and themes, the word stress and syllabic emphasis and emotional fragments analysis.

Development of the Ontology

Drafting of our Ontology
Following the aim of our project we wanted our ontology to focus mostly on the meaning of the poem itself, in particular on the difference between the literal and symbolic meaning of its stanzas, lines and words. In addition to that we wanted to make sure the basic information about the poem could be inserted and that every critic that we choose to include was correctly linked to its analysis. We’ve also taken into consideration the addition of external sources such as editions, manuscripts and other related sources.

The analysis on the state of art ontologies
Once we drafted an ontology prototype we tried to analyse if everything we needed had already been created and developed in existing ontologies. We found some of the aspects noted in our draft in the schema ontology (Dates), cidoc-crm (Author), Remetca (general classes / properties for poems) and frbrer (properties related to the place and date of publication).

Filling the gaps / simplifying the structure of the existing ontologies
Although some existing ontologies seemed to work perfectly with our initial idea, because our project was on a smaller scale we decided to simplify the structure of these existing ontologies. In particular we decided to not use some of the classes of the Remetca ontologies because, in our mind, some subproperties/classes that referred to the poem were not reflected in the ontology structure. We decided to create on our own all the classes, predicates and data properties related to the meaning, both literal and symbolic, because are the most important aspect of our project and we thought it was important that we created these parts on our own. Development of new classes/predicates/data properties using Protégé We used Protégé to import existing ontologies and to develop our own. Once we finished creating it we published the ontology on our webspace (krepolisma.altervista.org).

Development of the RDF and the Rendered Graph

Creation of the individuals based on our previous research on the Poem
We translated the previous knowledge acquired by our research on the poem into our ontology by creating the individuals and the RDF file to reach our goal to get the information on a LOD environment. Exporting our rdf file and the creation of a knowledge graph using rhizomik.net Once we had the RDF ready we proceeded to convert into a graph in SVG format (to make sure it could be scalable and zoomable given the huge dimensions) with the help rhzomik.net website.

We also exploited the possibility given by Protégé’s Plug-in, OntoGraf, to export a graph of all the classes and individuals to show a different approach/view to our ontology and its instances.

Creation of a secondary graph dedicated to the author
To extract knowledge about the poet, we have used the DbPedia and Wikidata databases, which store their data in the labeled graph data format : RDF. Hence the proper query method was the SPARQL - Query Language for RDF (https://www.w3.org/TR/rdf-sparql-query/). The information gathered by the SPARQL Queries are reflected on the mind map for the poet, manually.

Development of the RDF and Graph

General structure of the website
We gathered our information a website divided into some sections, including these ones. The other sections are about our ontology, graphs, the interface to explore the poem in multiple ways and a page dedicated to Emily Dickinson. The general Aesthetic is based on the Avana template (Available for free on the Web).

User Experience Focus
Most of our attention has been given on the User experience and usability of the website, especially the “Interactive Poem” Section. The section is a really simple interface that summarize what is in the RDF graph and it also includes information about the automatic analysis.
The semantic web can be quite complicated for casual users, and the development of a user-friendly interface was necessary in order to simplify it.