matteo casu exploring the betrothed lovers hamburg
TRANSCRIPT
Exploring The Betrothed Lovers
ANDREA BOLIOLI – CROSS LIBRARY SRL MATTEO CASU – CELI SRL
MAURIZIO LANA -‐ UNIVERSITÀ DEGLI STUDI DEL PIEMONTE ORIENTALE
RENATO RODA – UNIVERSITÀ DI TORINO
Sèduco (Sharing EDUcaDonal COntent) aims to study, develop and test digital tools for the creaDon, management and sharing of educaDonal contents in the italian upper secondary school (age 14 to 18).
The Promessi Sposi App in the context of Sèduco
CELI srl – Turin, Italy
Cross Library srl – Trento, Italy spinoff Celi/FBK (Bruno Kessler FoundaDon)
• Educa,onal objec,ve: classics are perceived as boring! – and o[en they are not …
• Business objec,ve: going beyond the concept of e-‐book
Provide an applicaDon for: • The explora,on of rich contents • The discovery of hidden informaDon
Apply semanDcs and SNA to a classic Why?
How?
Divise an architecture to make it repeatable with low costs
With a constraint
• “I Promessi Sposi” (1827) is a classic of italian literature by Alessandro Manzoni (1785 – 1873) – historical novel set in 1628, during the Spanish dominaDon (parallel to Austrian dominaDon during Manzoni’s Dme)
• Historical novel represenDng the cultural and linguisDc unity of Italy
• It is studied (mandatory) in the second year of italian high schools
What The Betrothed is
Curiosi,es • The translaDon of the Dtle in “The Betrothed Lovers” comes from the first review by E.A.Poe (1835)
• I Promessi Sposi was inspired by Scok’s Ivanhoe (1819)
The Betrothed as a semanDc mashup
• segmentaDon at the level of paragraphs (indexed in the Celi semanDc search engine)
• annotaDon of the novel for enDty menDons (persons and locaDons, automated task) and narraDve sequences (unit of Dme/place/acDon, manual task)
• two NER tools -‐-‐ TextPro by FBK and Sophia Seman,c Engine by CELI
• the automaDc annotaDon was manually corrected
CreaDon of the applicaDon
• During the creaDon of I Promessi Sposi we learned some lessons…
Structural annota,ons: XML-‐based -‐-‐ there are limits (e.g. non-‐overlapping annotaDons) Named-‐en,,es annota,ons: poor -‐-‐ only recently TEI introduced the use of keys Indexing: choose a level of granularity (paragraphs)
Architecture
We want the architecture to be reusable for any novel: • arbitrary annotaDon/segmentaDon of porDons of text
(chapters, sequences, dialogue lines, “like” … ) • Link text annotaDons with externally-‐defined enDDes • The index does not mirror a parDcular annotaDon
Generaliza,on
Tradi,onal methodology
Architecture
Ontological Model
We reused Linked Open Vocabularies (foaf, geoNames ontology, SPAR ontologies).
• FRBR’s disDncDon between a Work, its Expressions (different revisions) and its ManifestaDons (ediDons) to enable cross-‐ediDon analyDcs
• Discourse elements (chapters, paragraphs) are fixed (by the author) • Sequences: narraDve (incl. dialogues and monologues), descripDve or
reflexive -‐ in many-‐to-‐many relaDon with paragraphs. NarraDve sequences are linked to characters which act in them, and to speakers
• Characters are persons or groups , with links between groups and their members – are actors or speakers in sequences. The noDon of character is not a class, but a role played by an Agent in a Work.
• Loca,ons: can be municipaliDes (such as Milan, Lecco), relaDve locaDons (e.g. Lucia’s house), indefinite locaDons ("near Lecco”), journeys -‐-‐ change status over ,me (e.g. the home of Lucia was in Acquate, formerly a municipality, now a neighboroud in Lecco)
• Contribu,ons (images, web documents) can be added by students – they can be linked to any enDty in the model.
Learning and Social Aspects
• Students and teachers contributed in the creaDon
process: • Searching for images • SegmentaDon in narraDve sequences • Research on different maps for the visualizaDon of
the locaDons in different Dmes • And in its maintaining: • contribute with images, works about characters,
etc.
Social Network AnalyDcs
1. ConversaDonal network • Nodes and edges are weighted: weight = number
of dialogues in which the characters are involved 2. NarraDve network • Edges idenDfy co-‐occurrence of characters in the
same narra,ve sequence
Extrac,on of two interac,onal networks
Detail: NarraDve Sequences
Detail: Characters
Detail: LocaDons
Detail: CollocaDons
Detail: ConversaDonal Network
• Renzo is central w.r.t. dialogue interacDons.
• Clusters emerge: heroes / villains /supporDng characters
Detail: NarraDve Network
• Lucia is a more acDon-‐oriented figure
• Again, clusters emerge
Next Steps
• We are enhancing the search experience of I Promessi Sposi with structure expected by our model (e.g. locaDons hierarchy)
• We are applying the process from scratch to new novels
• Next? Ø Unleash sem web techniques: use a triple store
with inference enabled, link enDDes to the LOD cloud …
I Promessi Sposi was born together with the design and refinement of the architecture
Thank you