RESEARCHER


I’m PhD candidate since 2013 in the Music Technology Group of the Universitat Pompeu Fabra in Barcelona, Spain. The central idea of my PhD is to semantically exploit unstructured text sources from the music domain. The Web is full of editorial data and user generated content talking about music, and the vast majority of this knowledge is not being exploited. Therefore, the first objective is to extract structured knowledge from unstructured text sources in order to create knowledge representations. Then, exploit this new structured knowledge in typical problems of the Music Information Retrieval field (e.g. music recommendation, artist similarity, retrieval in music collections). Initial ideas about this were presented in the doctoral symposium of the ESWC conference in 2014 [6].

Following this research idea, I have developed an Information Extraction system specific for the music domain [2]. Using this system a music knowledge graph is built and it is used to feed a recommendation system [1]. The main contribution of the approach is that recommendations can be explained to users by means of natural language text. The same extraction system has been applied to a music digital library, and a navigation system over a knowledge layer has been developed [4]. In addition, I have explored other ways of exploiting implicit semantics in texts using entity linking and ontologies to feed a hybrid recommender system that outperforms pure collaborative approaches [3]. Moreover, I have been working on the problem of annotation of music and audio collections. To this end, a methodology for extending tagging ontologies with domain specific knowledge has been developed [5].

Currently I’m working in a semantic approach for artist similarity. Results show that our approach outperforms other pure text-based approaches. In addition, I’m working in the creation of a domain specific knowledge base about flamenco music, by combining a process of content curation and entity resolution from different data sources (knowledge bases, databases, ontologies, specialized websites), with a process of knowledge extraction from unstructured text sources.

[1] Sordo M., Oramas S., Espinosa-Anke L. (2015). Extracting Relations from Unstructured Text Sources for Music Recommendation. 20th International Conference on Applications of Natural Language to Information Systems.

[2] Oramas S., Sordo M., Espinosa-Anke L. (2015). A Rule-based Approach to Extracting Relations from Music Tidbits. 2nd Workshop on Knowledge Extraction from Text (KET 2015), at WWW 2015

[3] Ostuni V. C., Oramas S., Di Noia T., Serra, X., & Di Sciascio E. (2015). A Semantic Hybrid Approach for Sound Recommendation. 24th International World Wide Web Conference (WWW 2015)

[4] Oramas S., Sordo M., & Serra X. (2014). Automatic Creation of Knowledge Graphs from Digital Musical Document Libraries. Conference in Interdisciplinary Musicology (CIM 2014)

[5] Font F., Oramas S., Fazekas G., & Serra X. (2014). Extending Tagging Ontologies with Domain Specific Knowledge. International Semantic Web Conference (ISWC 2014)

[6] Oramas S. (2014). Harvesting and Structuring Social Data in Music Information Retrieval. PhD Symposium of the Extended Semantic Web Conference (ESWC 2014)

SOFTWARE ENGINEER


I’m also a software developer, with more than 10 years of experience with multiple languages and platforms, more specifically in web development. Currently I’m involved in the development team of freesound.org, an online audio sharing repository with more than 4 million registered users. I’m also an entrepreneur, and I developed and manage my own start up compartemesa.com.