The application of generative AI in education has shown significant potential to enhance learning outcomes by providing personalized, adaptive feedback to students. In this work, we present a novel Retrieval-Augmented Generation (RAG) system designed to improve the explanations and feedback provided to students during self-assessment activities. The system we developed is grounded in the course's reference material, ensuring that the feedback remains accurate, consistent, and contextually relevant to the student's curriculum. The system retrieves information directly from the textbook, reducing ambiguity and interpretation errors, and generates responses tailored to the specific needs of each student. The feedback is not only designed to correct misconceptions but also to reinforce key concepts, making the system a valuable tool for self-guided learning. In this study, we also explore the importance of prompt engineering in creating effective AI-generated feedback. We detail the iterative process used to optimize the prompts and the strategies employed to ensure high-quality, interpretable responses. The findings from this work suggest that generative AI, when integrated with subject-specific textbooks and careful prompt engineering, can significantly enhance the educational experience by providing dynamic, and contextually accurate feedback. This approach opens new possibilities for AI-driven education tools, contributing to more personalized and effective learning experiences.

Assigning a numerical value to a temporal expression (TE), known as temporal expression normalization, is a crucial process for tasks like timeline creation and temporal reasoning. Rule-based and classical deep-learning normalization systems lack versatility because they are limited to specific domains and languages, while current Large Language Models (LLMs) solutions are relatively unexplored. To overcome the current limitations in adaptability, we suggest utilizing five of the latest generative Large Language Models (LLMs) - Mistral 7B, Gemma 7B, Gemma 2B, Phi-2, and Llama-3 8B. We have explored various performance enhancement strategies, including using different prompts, contexts, and training techniques like Neftune. Our proposed models demonstrate the ability to adapt to diverse domains (news and biomedical) and multiple languages (Spanish, English, Italian, French, Portuguese, Catalan, and Basque) simultaneously. These models can handle expressions in various domains and languages, making them more versatile and useful for a wide range of applications. As a result, our approach offers significant performance improvements when compared to existing LLM-based and rule-based solutions for TE normalization and a promising solution for the challenges of temporal normalization.

Personalization of self-assessment educational tools plays a key role in enhancing the learning process. By allowing students to tailor the content and pace of their learning according to their individual needs, these tools encourage a more learner-centered approach. Our proposal consists in defining different personalization mechanisms for a self-assessment tool. This tool allows students to navigate the hierarchy of concepts of the subject and suggests them exercises covering the different concepts. After the student selects one of the possible answers, the tool informs if the student is right or wrong, The study has been carried out in a Computer Science course devoted to the teaching of algorithms and advanced data structures. However, the proposed tool and mechanisms can be easily exported to other subjects and topics. The tool has been implemented as a Telegram bot, which facilitates access to the tool from different platforms. The personalization mechanisms introduced in the tool on the one hand, allow to present the questions on each topic in increasing order of difficulty. This is achieved by analyzing the results of the exams in which the questions were originally proposed. In addition, the tool records the user's history, avoiding repeating questions already answered correctly and providing data on the user's performance. According to a questionnaire answered by the students, the tool, and in particular the personalization mechanisms, has been very useful for them in the preparation of the subject.

Algorithms and data structures are one of the most difficult parts of Computer Science for students to learn. It is therefore essential to develop approaches and tools that facilitate the acquisition of this crucial part of a computer scientist’s training. Based on the hypothesis that making the access to information and self-assessment easier improves the learning of the concepts, in this work we propose a graphical tool that allows the student to navigate through the topics of the subject, and to reach collections of self-assessment exercises on each of the topic. For this purpose, the information of the subject has been organized in a navigable hierarchy of topics that allows students to access exercises in a simple and visual way, on the one hand, but also to easily learn the relationships between the different concepts of the subject. These exercises allow students to self-evaluate and solve their doubts, as well as to find analogies between different topics in the hierarchy. This process is complemented by a web page recommender for the topics of the hierarchy. Results according to a questionnaire about the tool answered by the students indicate that it is considered of high utility and interest.

Keyphrases represent the main topics a text is about. In this article, we introduce SemGraph, an unsupervised algorithm for extracting keyphrases from a collection of texts based on a semantic relationship graph. The main novelty of this algorithm is its ability to identify semantic relationships between words whose presence is statistically significant. Our method constructs a co‐occurrence graph in which words appearing in the same document are linked, provided their presence in the collection is statistically significant with respect to a null model. Furthermore, the graph obtained is enriched with information from WordNet. We have used the most recent and standardized benchmark to evaluate the system ability to detect the keyphrases that are part of the text. The result is a method that achieves an improvement of 5.3% and 7.28% in F measure over the two labeled sets of keyphrases used in the evaluation of SemEval‐2010.

Abstract We have developed a heuristic method for unsupervised parsing of unrestricted text. Our method relies on detecting certain patterns of part-of-speech tag sequences of words in sentences. This detection is based on statistical data obtained from the corpus and allows us to classify part-of-speech tags into classes that play specific roles in the parse trees. These classes are then used to construct the parse tree of new sentences via a set of deterministic rules. Aiming to asses the viability of the method on different languages, we have tested it on English, Spanish, Italian, Hebrew, German, and Chinese. We have obtained a significant improvement over other unsupervised approaches for some languages, including English, and provided, as far as we know, the first results of this kind for others.

Abstract In this paper, we present a new method based on co-occurrence graphs for performing Cross-Lingual Word Sense Disambiguation (CLWSD). The proposed approach comprises the automatic generation of bilingual dictionaries, and a new technique for the construction of a co-occurrence graph used to select the most suitable translations from the dictionary. Different algorithms that combine both the dictionary and the co-occurrence graph are then used for performing this selection of the final translations: techniques based on sub-graphs (communities) containing clusters of words with related meanings, based on distances between nodes representing words, and based on the relative importance of each node in the whole graph. The initial output of the system is enhanced with translation probabilities, provided by a statistical bilingual dictionary. The system is evaluated using datasets from two competitions: task 3 of SemEval 2010, and task 10 of SemEval 2013. Results obtained by the different disambiguation techniques are analysed and compared to those obtained by the systems participating in the competitions. Our system offers the best results in comparison with other unsupervised systems in most of the experiments, and even overcomes supervised systems in some cases.

The natural mate-selection behavior of preferring individuals which are somewhat (but not too much) different has been proved to increase the resistance to infection of the resulting offspring, and thus fitness. Inspired by these results we have investigated the improvement obtained from diversity induced by differences between individuals sent and received and the resident population in an island model, by comparing different migration policies, including our proposed multikulti methods, which choose the individuals that are going to be sent to other nodes based on the principle of multiculturality; the individual sent should be different enough to the target population, which will be represented through a proxy string (computed in several possible ways) in the emitting population. We have checked a set of policies following these principles on two discrete optimization problems of diverse difficulty for different sizes and number of nodes, and found that, in average or in median, multikulti policies outperform the usual policy of sending the best or a random individual; however, the size of this advantage changes with the number of nodes involved and the difficulty of the problem, tending to be greater as the number of nodes increases. The success of this kind of policies will be explained via the measurement of entropy as a representation of population diversity for the policies tested.

Retrieving precise information from large collections of documents or from the web is an important task in our world. The specification of the information needed is done in form of a sequence of terms or query, which is frequently too short or unspecific to allow selecting a set of relevant documents small enough to be inspected by the user. This problem can be alleviated by expanding the query with other terms that make it more specific. The selection of these possible expansion terms is the problem addressed in this work. We have developed a classifier which has been trained for distinguishing good expansion terms. The identification of good terms to train the classifier has been achieved with a genetic algorithm whose fitness function is based on users' relevance judgements on a set of documents. Results show that the training performed by the genetic algorithm is able to improve the quality of the query expansion results.

Unsupervised grammar induction is one of the most difficult works of language processing. Its goal is to extract a grammar representing the language structure using texts without annotations of this structure. We have devised an evolutionary algorithm which for each sentence evolves a population of trees that represent different parse trees of that sentence. Each of these trees represent a part of a grammar. The evaluation function takes into account the contexts in which each sequence of Part-Of-Speech tags (POSseq) appears in the training corpus, as well as the frequencies of those POSseqs and contexts. The grammar for the whole training corpus is constructed in an incremental manner. The algorithm has been evaluated using a well known Annotated English corpus, though the annotation have only been used for evaluation purposes. Results indicate that the proposed algorithm is able to improve the results of a classical optimization algorithm, such as EM (Expectation Maximization), for short grammar constituents (right side of the grammar rules), and its precision is better in general.

Web spam is a serious problem for search engines because the quality of their results can be severely degraded by the presence of this kind of page. In this paper, we present an efficient spam detection system based on a classifier that combines new link-based features with language-model (LM)-based ones. These features are not only related to quantitative data extracted from the Web pages, but also to qualitative properties, mainly of the page links. We consider, for instance, the ability of a search engine to find, using information provided by the page for a given link, the page that the link actually points at. This can be regarded as indicative of the link reliability. We also check the coherence between a page and another one pointed at by any of its links. Two pages linked by a hyperlink should be semantically related, by at least a weak contextual relation. Thus, we apply an LM approach to different sources of information from a Web page that belongs to the context of a link, in order to provide high-quality indicators of Web spam. We have specifically applied the Kullback-Leibler divergence on different combinations of these sources of information in order to characterize the relationship between two linked pages. The result is a system that significantly improves the detection of Web spam using fewer features, on two large and public datasets SUchasWEBSPAM-UK2006 and WEBSPAM-UK2007.

Migration policies in distributed evolutionary algorithms are bound to have, as much as any other evolutionary operator, an impact on the overall performance. However, they have not been an active area of research until recently, and this research has concentrated on the migration rate. In this paper we compare different migration policies, including our proposed multikulti methods, which choose the individuals that are going to be sent to other nodes based on the principle of multiculturalism: the individual sent should be as different as possible to the receiving population (represented in several possible ways). We have checked this policy on two discrete optimization problems for different number of nodes, and found that, in average or in median, multikulti policies outperform others like sending the best or a random individual; however, their advantage changes with the number of nodes involved and the difficulty of the problem. The success of these kind of policies is explained via the measurement of entropies, which are known to have an impact in the performance of the evolutionary algorithm.

In this paper we compare different policies to select individuals to migrate in an island model. Our thesis is that choosing individuals in a way that exploits differences between populations can enhance diversity, and improve the system performance. This has lead us to propose a family of policies that we call multikulti, in which nodes exchange individuals different "enough" among them. In this paper we present a policy according to which the receiver node chooses the most different individual among the sample received from the sending node. This sample is randomly built but only using individuals with a fitness above a threshold. This threshold is previously established by the receiving node. We have tested our system in two problems previously used in the evaluation of parallel systems, presenting different degree of difficulty. The multikulti policy presented herein has been proved to be more robust than other usual migration policies, such as sending the best or a random individual.

This paper presents how it is possible to increase the Genetic Programming (GP) Computing Power (CP) for free, via Volunteer Computing (VC), using the well known framework BOINC plus a new ``virtualization'' layer which adds all the benefits from the virtualization paradigm. Two different experiments, employing a standard GP tool and a complex GP system, are performed --with distributed PCs over several cities-- to show the free achieved CP by means of VC, without the necessity of modifying or adapting the original GP source code. The methodology can be easily extended to Evolutionary Algorithms (EAs).

This paper presents an evolutionary algorithm for modeling the arrival dates of document streams, which is any time-stamped collection of documents, such as newscasts, e-mails, scientific journals archives and weblog postings. The goal is to find a frequency curve that fits the data circumventing the unavoidable noise. Classical dynamic programming algorithms are limited by memory and efficiency requirements, which can be a problem when dealing with long streams. This suggests to explore alternative search methods which although do not guarantee optimality, are far more efficient. Experiments have shown that the designed evolutionary algorithm is able to reach high quality solutions in a short time. We have also explored different approaches to infer whether new arrivals increase or decrease interest in the topic the document stream is about. In particular, we present a variant of the evolutionary algorithm, which is able to very quickly fit a stream extended with new data, by taking advantage of the fit obtained for the original substream. These mechanisms can be used for real time detection of changes in the trend of interest in a topic, an important application of this kind of models.

Noun phrases of a document usually are the main information bearers. Thus, the detection of these units is crucial in many applications related to information retrieval, such as collecting relevant documents by search engines according to a user query, text summarizing, etc. We present an evolutionary algorithm for obtaining a probabilistic finite-state automaton, able to recognize valid noun phrases defined as a sequence of lexical categories. This approach is highly flexible in the sense that the automaton is able to recognize noun phrases similar enough to the ones given by the inferred noun phrase grammar. This flexibility can be allowed thanks to the very accurate set of probabilities provided by the evolutionary algorithm. It works with both, positive and negative examples of the language, thus improving the system coverage, while maintaining its precision. Experimental results show a clear improvement of the performance with respect to others systems

Presents some applications of evolutionary programming to different tasks of natural language processing (NLP). First of all, the work defines a general scheme of application of evolutionary techniques to NLP, which gives the mainstream for the design of the elements of the algorithm. This scheme largely relies on the success of probabilistic approaches to NLP. Secondly, the scheme has been illustrated with two fundamental applications in NLP: tagging, i.e., the assignment of lexical categories to words and parsing, i.e., the determination of the syntactic structure of sentences. In both cases, the elements of the evolutionary algorithm are described in detail, as well as the results of different experiments carried out to show the viability of this evolutionary approach to deal with tasks as complex as those of NLP.

A novel approach for the integration of evolution programs and constraint-solving techniques over finite domains is presented. This integration provides a problem-independent optimization strategy for large-scale constrained optimization problems over finite domains. In this approach, genetic operators are based on an arc-consistency algorithm, and chromosomes are arc-consistent portions of the search space of the problem. The paper describes the main issues arising in this integration: chromosome representation and evaluation, selection and replacement strategies, and the design of genetic operators. We also present a parallel execution model for a distributed memory architecture of the previous integration. We have adopted a global parallelization approach that preserves the properties, behavior, and fundamentals of the sequential algorithm. Linear speedup is achieved since genetic operators are coarse grained as they perform a search in a discrete space carrying out arc consistency. The implementation has been tested on a GRAY T3E multiprocessor using a complex constrained optimization problem.