J. Fernando Sánchez

J. Fernando Sánchez


Madrid, Madrid, ES
Spanish, English


Work Experience

Work Experience

  • Researcher - PhD StudentGSI UPM

    Jan, 2013 - Present

    The Ingelligent Systems Group is a research group at Universidad Politécnica de Madrid (UPM)

    • Semantic Technologies

    • Sentiment Analysis

    • Ontologies and vocabularies: Marl, Onyx

  • Undergraduate researcherGSI UPM

    Jun, 2008 - Dec, 20124 years 6 months

    • Web services: Node.js, Python, PHP, JSP/J2EE...

    • Semantic technologies: rdflib, easy-rdf, Protegé

    • Event middleware and messaging: XMPP, developed Maia



  • Data analysis, machine learning and NLP
  • Python
  • DevOps
  • Front-end
  • Go
  • Java
  • NodeJS


  • Telecommunications Engineering, Bachelor+Master, Tecnical University of Madrid (UPM)

    Oct, 2007 - Oct, 2012

    Computer NetworksSoftware EngineeringWeb Technologies
Volunteer Work

Volunteer Work

  • Vice-Chairman for External Affairs, EESTEC

    Apr, 2011 - May, 2012

    • Established connections with other Student Associations

    • Helped found new Observers (Local Groups)

    • Carried out International Board duties

  • IT Coordinator, EESTEC

    Apr, 2012 - May, 2013

    • Coordinated the work of a small international IT Team

    • In charge of the administration of the IT infrastructure of EESTEC: Plone portal, Mailman mailing lists, etc.

  • Oversight Committee, EESTEC

    May, 2012 - May, 2013

    • Supervised the work of the International Board



  • MixedEmotions: An Open-Source Toolbox for Multi-Modal Emotion Analysis,

    Published on: Oct 01, 2018

    Recently, there is an increasing tendency to embed the functionality of recognizing emotions from the user generated contents, to infer richer profile about the users or contents, that can be used for various automated systems such as call-center operations, recommendations, and assistive technologies. However, to date, adding this functionality was a tedious, costly, and time consuming effort, and one should look for different tools that suits one's needs, and should provide different interfaces to use those tools. The MixedEmotions toolbox leverages the need for such functionalities by providing tools for text, audio, video, and linked data processing within an easily integrable plug-and-play platform. These functionalities include: (i) for text processing: emotion and sentiment recognition, (ii) for audio processing: emotion, age, and gender recognition, (iii) for video processing: face detection and tracking, emotion recognition, facial landmark localization, head pose estimation, face alignment, and body pose estimation, and (iv) for linked data: knowledge graph. Moreover, the MixedEmotions Toolbox is open-source and free. In this article, we present this toolbox in the context of the existing landscape, and provide a range of detailed benchmarks on standardized test-beds showing its state-of-the-art performance. Furthermore, three real-world use-cases show its effectiveness, namely emotion-driven smart TV, call center monitoring, and brand reputation analysis.

  • Multimodal Multimodel Emotion Analysis as Linked Data, Proceedings of ACII 2017

    Published on: Oct 01, 2017

    The lack of a standard emotion representation model hinders emotion analysis due to the incompatibility of annota-tion formats and models from different sources, tools and an- notation services. This is also a limiting factor for multimodal analysis, since recognition services from different modalities (audio, video, text) tend to have different representation models (e. g., continuous vs. discrete emotions). This work presents a multi-disciplinary effort to alleviate this problem by formalizing conversion between emotion models. The specific contributions are: i) a semantic representation of emotion conversion; ii) an API proposal for services that perform automatic conversion; iii) a reference implementation of such a service; and iv) validation of the proposal through use cases that integrate different emotion models and service providers.

  • Enhancing Deep Learning Sentiment Analysis with Ensemble Techniques in Social Applications,

    Published on: Jun 01, 2017

    The appearance of new Deep Learning applications for Sentiment Analysis has motivated a lot of researchers, mainly because of their automatic feature extraction and representation capabilities, as well as their better performance compared to the previous feature based techniques. These traditional surface approaches are based on complex manually extracted features, and this extraction process is a fundamental question in feature driven methods. However, these long-established approaches can yield strong baselines on their own, and its predictive capabilities can be used in conjunction with the arising Deep Learning methods. In this paper we seek to improve the performance of these new Deep Learning techniques integrating them with more traditional surface approaches based on manually extracted features. The contributions of this paper are: first, we develop a Deep Learning based Sentiment classifier using the Word2Vec model and a linear machine learning algorithm. This classifier serves us as a baseline with which we can compare subsequent results. Second, we propose two ensemble techniques which aggregate our baseline classifier with other surface classifiers widely used in the field of Sentiment Analysis. Third, we also propose two models for combining deep features with both surface and deep features in order to merge the information from several sources. As fourth contribution, we introduce a taxonomy for classifying the different models we propose, as well as the ones found in the literature. Fifth, we conduct several reproducible experiments with the aim of comparing the performance of these models with the Deep Learning baseline. For this, we employ four public datasets that were extracted from the microblogging domain. Finally, as a result, the experiments confirm that the performance of these proposed models surpasses that of our original baseline using as metric the F1-Score, with improvements ranging from 0.21 to 3.62 %.

  • Modeling Social Influence in Social Networks with SOIL, a Python Agent-Based Social Simulator, Advances in Practical Applications of Cyber-Physical Multi-Agent Systems: The PAAMS Collection

    Published on: Jun 01, 2017

    The application of Agent-based Social Simulation (ABSS) for modeling social networks requires specific facilities for modeling, simulation and visualization of network structures. Moreover, ABSS can benefit from interactive shell facilities that can assist the model development process. We have addressed these problems through the development of a tool called SOIL, which provides a Python ABSS specifically designed for social networks. In this paper we present how this tool is applied to simulate viral marketing processes in a social network, and to evaluate the model with real data.

  • Soil: An Agent-Based Social Simulator in Python for Modelling and Simulation of Social Networks, Advances in Practical Applications of Cyber-Physical Multi-Agent Systems: The PAAMS Collection

    Published on: Jun 01, 2017

    Social networks have a great impact in our lives. While they started to improve and aid communication, nowadays they are used both in professional and personal spheres, and their popularity has made them attractive for developing a number of business models. Agent-based Social Simulation (ABSS) is one of the techniques that has been used for analysing and simulating social networks with the aim of understanding and even forecasting their dynamics. Nevertheless, most available ABSS platforms do not provide specific facilities for modelling, simulating and visualising social networks. This article aims at bridging this gap by introducing an ABSS platform specifically designed for modelling social networks. The main contributions of this paper are: (1) a review and characterisation of existing ABSS platforms; (2) the design of an ABSS platform for social network modelling and simulation; and (3) the development of a number of behaviour models for evaluating the platform for information, rumours and emotion propagation. Finally, the article is complemented by a free and open source simulator.

  • A modular architecture for intelligent agents in the evented web,

    Published on: Feb 01, 2017

    The growing popularity of public APIs and technologies such as web hooks is changing online services drastically. It is easier now than ever to interconnect services and access them as a third party. The next logical step is to use intelligent agents to provide a better user experience across services, connecting services with smart automatic behaviors or actions. In other words, it is time to start using agents in the so-called Evented Web. For this to happen, agent platforms need to seamlessly integrate external sources such as web services. As a solution, this paper introduces an event-based architecture for agent systems. This architecture has been designed in accordance with the new tendencies in web programming and with a Linked Data approach. The use of Linked Data and a specific vocabulary for events allows a smarter and more complex use of events. Two use cases have been implemented to illustrate the validity and usefulness of the architecture.

  • Applying Recurrent Neural Networks to Sentiment Analysis of Spanish Tweets,

    Published on: Jan 01, 2017

    This article presents the participation of the Intelligent Systems Group (GSI) at Universidad Polit ́ecnica de Madrid (UPM) in the Sentiment Analysis work- shop focused in Spanish tweets, TASS2017. We have worked on Task 1, aiming to classify sentiment polarity of Spanish tweets. For this task we propose a Recurrent Neural Network (RNN) architecture composed of Long Short-Term Memory (LSTM) cells followed by a feedforward network. The architecture makes use of two different types of features: word embeddings and sentiment lexicon values. The recurrent ar- chitecture allows us to process text sequences of different lengths, while the lexicon inserts directly into the system sentiment information. The results indicate that this feature combination leads to enhanced sentiment analysis performances.

  • Senpy: A Pragmatic Linked Sentiment Analysis Framework, Proceedings DSAA 2016 Special Track on Emotion and Sentiment in Intelligent Systems and Big Social Data Analysis (SentISData)

    Published on: Oct 01, 2016

    Sentiment and emotion analysis technologies have quickly gained momentum in industry and academia. This popularity has spawned a myriad of service and tools. Due to the lack of common interfaces and models, each of these services imposes specific interfaces and representation models. Heterogeneity makes it costly to integrate different services, evaluate them or switch between them. This work aims to remedy heterogeneity by providing an extensible framework and an API aligned with the NLP Interchange Format service specification. It also includes a reference implementation, a first step towards a successful and cost-effective adoption. The specific contributions in this paper are: (i) the Senpy framework; (ii) an architecture for the framework that follows a plug-in approach; (iii) a reference open source implementation of the architecture; (iv) the use and validation of the framework and architecture in a big data sentiment analysis European project. Our aim is to foster the development of a new generation of emotion aware services by isolating the development of new algorithms from the representation of results and the deployment of services.

  • Linked Data Models for Sentiment and Emotion Analysis in Social Networks, Sentiment Analysis in Social Networks

    Published on: Oct 01, 2016

    Language resource interoperability is still a major challenge in sentiment analysis. One of the current trends for solving this issue is the adoption of a linked data perspective for semantically modeling, interlinking, and publishing lexical and other linguistic resources. This chapter contributes to the development of the linguistic linked open data through a linked data model for sentiment and emotion analysis in social networks that is based on two vocabularies: Marl and Onyx for sentiment and emotion modeling respectively. These vocabularies are used for (1) affective corpus annotation, (2) affective lexicon annotation, and (3) sentiment and emotion services interoperability. Several aspects of the solution are discussed, such as the transformation of legacy resources, the generation of domain-specific sentiment lexicons, and the benefits of interlinking language resources for sentiment analysis with other resources such as WordNet or DBpedia.

  • Towards a Common Linked Data Model for Sentiment and Emotion Analysis, Proceedings of the LREC 2016 Workshop Emotion and Sentiment Analysis (ESA 2016)

    Published on: May 01, 2016

    The different formats to encode information currently in use in sentiment analysis and opinion mining are heterogeneous and often custom tailored to each application. Besides a number of existing standards, there are additionally still plenty of open challenges, such as representing sentiment and emotion in web services, integration of different models of emotions or linking to other data sources. In this paper, we motivate the switch to a linked data approach in sentiment and emotion analysis that would overcome these and other current limitations. This paper includes a review of the existing approaches and their limitations, an introduction of the elements that would make this change possible, and a discussion of the challenges behind that change.

  • Onyx: A Linked Data Approach to Emotion Representation,

    Published on: Jan 01, 2016

    Extracting opinions and emotions from text is becoming more and more important, especially since the advent of micro-blogging and social networking. Opinion mining has become particularly popular and now gathers many public services, datasets and lexical resources. Unfortunately, there are few available lexical and semantic resources for emotion recognition that could foster the development of new emotion aware services and applications. Some of the barriers for developing such resources are the diversity of emotion theories and the absence of a common vocabulary to express emotion. This article presents a semantic vocabulary, called Onyx, intended to provide support to represent emotions in lexical resources and emotion analysis services. It follows a linguistic Linked Data approach, it is aligned with the Provenance Ontology, and it has been integrated with lemon, an increasingly popular RDF model for representing lexical entries. This approach also means a new and interesting way to work with different theories of emotion. As part of our work, Onyx has been aligned with EmotionML and WordNet-Affect.

  • Aspect based Sentiment Analysis of Spanish Tweets, Proceedings of TASS 2015: Workshop on Sentiment Analysis at SEPLN co-located with 31st SEPLN Conference (SEPLN 2015)

    Published on: Sep 01, 2015

    This article presents the participation of the Intelligent Systems Group (GSI) at Universidad Polit´ecnica de Madrid (UPM) in the Sentiment Analysis workshop focused in Spanish tweets, TASS2015. This year two challenges have been proposed, which we have addressed with the design and development of a modular system that is adaptable to different contexts. This system employs Natural Language Processing (NLP) and machine-learning technologies, relying also in previously developed technologies in our research group. In particular, we have used a wide number of features and polarity lexicons for sentiment detection. With regards to aspect detection, we have relied on a graph-based algorithm. Once the challenge has come to an end, the experimental results are promising.

  • A Linked Data Model for Multimodal Sentiment and Emotion Analysis,

    Published on: Jul 01, 2015

    The number of tools and services for sentiment analysis is increasing rapidly. Unfortunately, the lack of standard formats hinders interoperability. To tackle this problem, previous works propose the use of the NLP Interchange Format (NIF) as both a common semantic format and an API for textual sentiment analysis. However, that approach creates a gap between textual and sentiment analysis that hampers multimodality. This paper presents a multimedia extension of NIF that can be leveraged for multimodal applications. The application of this extended model is illustrated with a service that annotates online videos with their sentiment and the use of SPARQL to retrieve results for different modes.

  • EUROSENTIMENT: Linked Data Sentiment Analysis, Proceedings of the ISWC 2014 Posters & Demonstrations Track a track within the 13th International Semantic Web Conference (ISWC 2014)

    Published on: Oct 01, 2014

    Sentiment and Emotion Analysis strongly depend on quality language resources, especially sentiment dictionaries. These resources are usually scattered, heterogeneous and limited to specific domains of application by simple algorithms. The EUROSENTIMENT project addresses these issues by 1) developing a common language resource representation model for sentiment analysis, and APIs for sentiment analysis services based on established Linked Data formats (lemon, Marl, NIF and ONYX) 2) by creating a Language Resource Pool (a.k.a. LRP) that makes available to the community existing scattered language resources and services for sentiment analysis in an interoperable way. In this paper we describe the available language resources and services in the LRP and some sample applications that can be developed on top of the EUROSENTIMENT LRP.

  • MAIA: An Event-based Modular Architecture for Intelligent Agents, Proceedings of 2014 IEEE/WIC/ACM International Conference on Intelligent Agent Technology

    Published on: Aug 01, 2014

    Online services are no longer isolated. The release of public APIs and technologies such as web hooks are allowing users and developers to access their information easily. Intelligent agents could use this information to provide a better user experience across services, connecting services with smart automatic behaviours or actions. However, agent platforms are not prepared to easily add external sources such as web services, which hinders the usage of agents in the so-called Evented or Live Web. As a solution, this paper introduces an event-based architecture for agent systems, in accordance with the new tendencies in web programming. In particular, it is focused on personal agents that interact with several web services. With this architecture, called MAIA, connecting to new web services does not involve any modification in the platform.

  • A Linked Data Approach to Sentiment and Emotion Analysis of Twitter in the Financial Domain, Second International Workshop on Finance and Economics on the Semantic Web (FEOSW 2014)

    Published on: May 01, 2014

    Sentiment analysis has recently gained popularity in the financial domain thanks to its capability to predict the stock market based on the wisdom of the crowds. Nevertheless, current sentiment indicators are still silos that cannot be combined to get better insight about the mood of different communities. In this article we propose a Linked data approach for modelling sentiment and emotions about financial entities. We aim at integrating sentiment information from different communities for providers, and complements existing initiatives such as FIBO. The approach has been validated in the semantic annotation of tweets of several stocks in the Spanish stock market, including its sentiment information.

  • Generating Linked-Data based Domain-Specific Sentiment Lexicons from Legacy Language and Semantic Resources, th International Workshop on Emotion, Social Signals, Sentiment & Linked Open Data, co-located with LREC 2014,

    Published on: May 01, 2014

    We present a methodology for legacy language resource adaptation that generates domain-specific sentiment lexicons organized around domain entities described with lexical information and sentiment words described in the context of these entities. We explain the steps of the methodology and we give a working example of our initial results. The resulting lexicons are modelled as Linked Data resources by use of established formats for Linguistic Linked Data (lemon, NIF) and for linked sentiment expressions (Marl), thereby contributing and linking to existing Language Resources in the Linguistic Linked Open Data cloud.

  • EuroLoveMap: Confronting feelings from News, Proceedings of Come Hack with OpeNER!” workshop at the 9th Language Resources and Evaluation Conference (LREC’14)

    Published on: Jan 01, 2014

  • Onyx: Describing Emotions on the Web of Data, Proceedings of the First International Workshop on Emotion and Sentiment in Social and Expressive Media: approaches and perspectives from AI (ESSEM 2013)

    Published on: Dec 01, 2013

    There are several different standardised and widespread formats to represent emotions. However, there is no standard semantic model yet. This paper presents a new ontology, called Onyx, that aims to become such a standard while adding concepts from the latest Semantic Web models. In particular, the ontology focuses on the representation of Emotion Analysis results. But the model is abstract and inherits from previous standards and formats. It can thus be used as a reference representation of emotions in any future application or ontology. To prove this, we have translated resources from EmotionML representation to Onyx. We also present several ways in which developers could benefit from using this ontology instead of an ad-hoc presentation. Our ultimate goal is to foster the use of semantic technologies for emotion Analysis while following the Linked Data ideals.

  • Linguistic Linked Data for Sentiment Analysis, 2nd Workshop on Linked Data in Linguistics (LDL-2013): Representing and linking lexicons, terminologies and other language data. Collocated with the Conference on Generative Approaches to the Lexicon

    Published on: Sep 01, 2013

    In this paper we describe the specification of a model for the semantically interoperable represen- tation of language resources for sentiment analysis. The model integrates ‘lemon’, an RDF-based model for the specification of ontology-lexica (Buitelaar et al. 2009), which is used increasingly for the representation of language resources as Linked Data, with 'Marl', an RDF-based model for the representation of sentiment annotations (Westerski et al., 2011; Sánchez-Rada et al., 2013).

  • Design and Implementation of an Agent Architecture Based on Web Hooks,

    Published on: Jan 01, 2012

    This project aims to introduce an event-based architecture for intelligent agents, inaccordance with the new tendencies in the Evented Web.The reason for this change is that agent communication is no longer suitable for theinmense amount of data generated nowadays and its nature. At least, not for their usein evolving scenarios where data sources interact without previous con guration. This isexactly what the precursors of the Live Web envision, and it is beginning to show in the newgeneration of evented applications, which enable customized interactions and a high level ofcommunication between di erent services.The proposed architecture shown in this document, called Maia, is based on a centralpiece or event router, which controls the ow of information/events to and from the connectedentities. These entities can be either event-aware agents or simply data sources andsubscribers. Thus giving a higher exibility than current technologies and easing the developmentof advanced systems by not requiring the complexity associated with agent systemsin all of the nodes.To demonstrate the feasibility and capabilities of the Maia architecture, a prototype hasbeen implemented which is also explained in detail in this document. It is based on the eventdrivenI/O server side JavaScript environment Node.js for the event routing components,and adapted Jason BDI agent platform as an example of a subscribed multi-agent system.Using this prototype, the bene ts of using Maia are illustrated by developing a personalagent capable of booking train tickets and that combines access to services, linked data andcommon sense reasoning.