Abstract
In this paper we study the combined use of four different NLP toolkits-Stanford CoreNLP, GATE, OpenNLP and Twitter NLP tools-in the context of social media posts. Previous studies have shown performance comparisons between these tools, both on news and social media corporas. In this paper, we go further by trying to understand how differently these toolkits predict Named Entities, in terms of their precision and recall for three different entity types, and how they can complement each other in this task in order to achieve a combined performance superior to each individual one. Experiments on two publicly available datasets from the workshops WNUT-2015 and #MSM2013 show that using an ensemble of toolkits can improve the recognition of specific entity types - up to 10.62% for the entity type Person, 1.97% for the type Location and 1.31% for the type Organization, depending on the dataset and the criteria used for the voting. Our results also showed improvements of 3.76% and 1.69%, in each dataset respectively, on the average performance of the three entity types.

Abstract
Current Learning Management Systems (LMS) generically provide virtual places to conduct interactions between students and educators. Chats, forums and other communication mechanisms usually are present in any LMS. In this paper, we propose a tool that can be embedded in any LMS that features some sort of hierarchical communication mechanisms. The proposed system is capable of depicting and analyzing online interactions in an easy to understand social graph. The vertex positioning algorithm is based on social network analysis statistics, taken from the collected interactions, and is able to smoothly present the temporal evolution in order to find communicational patterns and report them to the educator and the students.

Abstract
We present a system for a dynamic graphical representation of the interactions captured in educational online environments. The system goes beyond interaction between students and teachers, also addressing resource usage or any other entity for which it is possible to create a relation which binds two entities. By defining these relationships between pairs of entities in an online learning environment (Moodle, in our case) our tool creates a graph, where it is possible to apply techniques of social network analysis. This system brings up new possibilities for e-learning as a tool capable of helping the teacher assorting and illustrating the degree of participation and to find the implicit relations between participants, or participants and resources or events. © 2017 Association for Computing Machinery.

Abstract
Research concerning mining data from learning management systems have been consistently been appearing in the literature. However, in many situations there is not a clear path on the data mining procedures that lead to solid conclusions. Therefore, many studies result in ad-hoc conclusions with insufficient generalization capabilities. In this article, we describe a methodology and report our findings in an experiment which one online course which involved more than 150 students. We used the Moodle LMS during the period of one academic semester, collecting all the interactions between the students and the system. These data scales up to more than 33K records of interactions where we applied data mining tools following the procedure for data extraction, cleaning, feature identification and preparation. We then proceeded to the creation of automatic learning models based on decision trees, we assessed the models and validate the results by assessing the accuracy of the predictions using traditional metrics and draw our conclusions on the validity of the process and possible alternatives1. © 2017 Association for Computing Machinery.

Abstract
Predicting whether a student will pass or fail is one of the most important actions to take while giving lectures. Usually, the experienced teacher is able to detect problematic situations at early stages. However, this is only true for classes up to a hundred students. For bigger ones, automatic methods are needed. In this paper, we present a predictive system based on three criteria retrieved and computed from the logs of the learning management system. We built fast frugal decision trees to help predict and prevent student failures, using data retrieved from their resource usage patterns. Evaluation of the decision system shows that the system's accuracy is very high both in train and test phases, surpassing logistic regression and CART. © 2017 IEEE.

Abstract
Annotated data is one of the most important components for supervised learning tasks. To ensure the reliability of the models, this data is usually labeled by several human annotators through volunteering or using Crowdsourcing platforms. However, such approaches are unfeasible (regarding time and cost) in datasets with an enormous number of entries, which in the specific case of journalistic relevance detection in social media posts, is necessary due to the wide scope of topics that can be considered relevant. Therefore, with the goal of building a relevance detection model, we propose an architecture to build a large scale annotated dataset regarding the journalistic relevance of Twitter posts (i.e. tweets). This methodology is based on the predictability of the content in Twitter accounts. Next, we used the retrieved dataset and build relevance detection models, combining text, entities, and sentiment features. Finally, we validated the best model through a smaller manually annotated dataset with posts from Facebook and Twitter. The F1-measure achieved in the validation dataset was 63% which is still far from excellent. However, given the characteristics of the validation data, these results are encouraging since 1) our model is not affected by content from other social networks and 2) our validation dataset was restrained to a specific time interval and specific keywords (which can affect the performance of the model). © 2017 IEEE.