In recent years, there has been a growing research interest focused on identifying traces of mental disorders through social media analysis. These disorders significantly impair millions of individuals' cognitive and behavioral functions worldwide. Our study aims to advance the understanding of four prevalent mental disorders: Anorexia, Depression, Gambling, and Self-harm. We present a comprehensive framework designed for the domain adaptation of models to analyze and identify signs of these conditions on social media posts. The language models' adapting strategy consisted of three key stages. First, we gathered and enriched substantial data on the four psychological disorders. Second, we adapted the different models to the language used to discuss mental health concerns on social media. Finally, we employed an adapter to fine-tune the models for multiple classification tasks (specific to each mental health condition). The intuitive idea is to adapt a language model smoothly to each domain. Our work includes a comparative study of different language models under in- and cross-domain conditions. This allows us to, for example, assess the ability of a depression-based language model to detect signs of disorders such as anorexia or self-harm. We show that the resulting mental health models perform well in early risk detection tasks. Additionally, we thoroughly analyze the linguistic qualities of these models by testing their predictive abilities using conventional clinical tools, such as specialized questionnaires. We rigorously examine the models across multiple predictive tasks to provide evidence of the adaptation approach's robustness and effectiveness. Our evaluation results are promising. They demonstrate that our framework enhances classification performance and competes favorably with state-of-the-art models.

The use of attention mechanisms in deep learning solutions has become popular within natural language processing tasks. The use of these mechanisms allows managing the relevance of the elements of a sequence in accordance with their context, however, this relevance has been observed independently between the pairs of elements of a sequence (self-attention) or between the application domain of a sequence (contextual attention), leading to the loss of relevant information and limiting the representation of the sequences. To tackle these particular issues, we propose a dual attention mechanism, which trades off the previous limitations, by considering the internal and contextual relationships between the elements of the sequence. Additionally, we propose the extension of the dual attention mechanism into a multi-layer perspective, through the weighted fusion of the different encoding layers of deep architectures. As the interpretation of abusive language is highly context-dependent, its identification is an ideal task to evaluate the proposed attention mechanism. Accordingly, we considered six standard collections for the abusive language identification task. The obtained results are encouraging; the proposed hierarchical attention mechanism outperformed the current self-attention and contextual attention mechanisms coupled with recurrent neural networks and Transformers, as well as, state-of-the-art approaches in detecting abusive language.

Mario Aragon, Adrian Pastor Lopez Monroy, Luis Gonzalez, David E. Losada, Manuel Montes. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Millions of people around the world are affected by one or more mental disorders that interfere in their thinking and behavior. A timely detection of these issues is challenging but crucial, since it could open the possibility to offer help to people before the illness gets worse. One alternative to accomplish this is to monitor how people express themselves, that is for example what and how they write, or even a step further, what emotions they express in their social media communications. In this article, we analyze two computational representations that aim to model the presence and changes of the emotions expressed by social media users. In our evaluation we use two recent public data sets for two important mental disorders: Depression and Anorexia. The obtained results suggest that the presence and variability of emotions, captured by the proposed representations, allow to highlight important information about social media users suffering from depression or anorexia. Furthermore, the fusion of both representations can boost the performance, equalling the best reported approach for depression and barely behind the top performer for anorexia by only 1 percent. Moreover, these representations open the possibility to add some interpretability to the results.

Passage retrieval is the task of identifying text snippets that are valid answers for a natural language posed question. One way to address this problem is to look at it as a metric learning problem, where we want to induce a metric between questions and passages that assign smaller distances to more relevant passages. In this work, we present a novel method for passage retrieval that learns a metric for questions and passages based on their internal semantic interactions. The method uses a similar approach to that of triplet networks, where the training samples are composed of one anchor (the question) and two positive and negative samples (passages). However,and in contrast with triplet networks, the proposed method uses a novel deep architecture that better exploits the particularities of text and takes into consideration complementary relatedness measures. Besides, the paper presents a sampling strategy that selects both easy and hard negative samples which improves the accuracy of the trained model. The method is particularly well suited for domain-specific passage retrieval where it is very important to take into account different sources of information. The proposed approach was evaluated in a biomedical passage retrieval task, the BioASQ challenge, outperforming standard triplet loss substantially by 10%,and state-of-the-art performance by 26%.

Likability prediction of books has many uses. Readers, writers, as well as the publishing industry, can all benefit from automatic book likability prediction systems. In order to make reliable decisions, these systems need to assimilate information from different aspects of a book in a sensible way. We propose a novel multimodal neural architecture that incorporates genre supervision to assign weights to individual feature types. Our proposed method is capable of dynamically tailoring weights given to feature types based on the characteristics of each book. Our architecture achieves competitive results and even outperforms state-of-the-art for this task.

Adrian Pastor López-Monroy, Fabio A. González, Manuel Montes, Hugo Jair Escalante, Thamar Solorio. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers). 2018.

Suraj Maharjan, John Arevalo, Manuel Montes, Fabio A. González, Thamar Solorio. Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, Long Papers. 2017.

Upendra Sapkota, Steven Bethard, Manuel Montes, Thamar Solorio. Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2015.

Abstract This paper introduces a novel ranking refinement approach based on relevance feedback for the task of document retrieval. We focus on the problem of ranking refinement since recent evaluation results from Information Retrieval (IR) systems indicate that current methods are effective retrieving most of the relevant documents for different sets of queries, but they have severe difficulties to generate a pertinent ranking of them. Motivated by these results, we propose a novel method to re-rank the list of documents returned by an IR system. The proposed method is based on a Markov Random Field (MRF) model that classifies the retrieved documents as relevant or irrelevant. The proposed MRF combines: (i) information provided by the base IR system, (ii) similarities among documents in the retrieved list, and (iii) relevance feedback information. Thus, the problem of ranking refinement is reduced to that of minimising an energy function that represents a trade-off between document relevance and inter-document similarity. Experiments were conducted using resources from four different tasks of the Cross Language Evaluation Forum (CLEF) forum as well as from one task of the Text Retrieval Conference (TREC) forum. The obtained results show the feasibility of the method for re-ranking documents in IR and also depict an improvement in mean average precision compared to a state of the art retrieval machine.

This paper elaborates on the benefits of using particle swarm model selection (PSMS) for building effective ensemble classification models. PSMS searches in a toolbox for the best combination of methods for preprocessing, feature selection and classification for generic binary classification tasks. Throughout the search process PSMS evaluates a wide variety of models, from which a single solution (i.e. the best classification model) is selected. Satisfactory results have been reported with the latter formulation in several domains. However, many models that are potentially useful for classification are disregarded for the final model. In this paper we propose to re-use such candidate models for building effective ensemble classifiers. We explore three simple formulations for building ensembles from intermediate PSMS solutions that do not require of further computation than that of the traditional PSMS implementation. We report experimental results on benchmark data as well as on a data set from object recognition. Our results show that better models can be obtained with the ensemble version of PSMS, motivating further research on the combination of candidate PSMS models. Additionally, we analyze the diversity of the classification models, which is known to be an important factor for the construction of ensembles.

In this paper a novel approach for improving automatic image annotation methods is proposed. The approach is based on the fact that accuracy of current image annotation methods is low if we look at the most confident label only. Instead, accuracy is improved if we look for the correct label within the set of the top−k candidate labels. We take advantage of this fact and propose a Markov random field (MRF) based on word co-occurrence information for the improvement of annotation systems. Through the MRF structure we take into account spatial dependencies between connected regions. As a result, we are considering semantic relationships between labels. We performed experiments with iterated conditional modes and simulated annealing as optimization strategies in a subset of the Corel benchmark collection. Experimental results of the proposed method together with a k−nearest neighbors classifier as our annotation method show important error reductions. 1

Artificial neural networks have been proven to be effective learning algorithms since their introduction. These methods have been widely used in many domains, including scientific, medical, and commercial applications with great success. However, selecting the optimal combination of preprocessing methods and hyperparameters for a given data set is still a challenge. Recently a method for supervised learning model selection has been proposed: Particle Swarm Model Selection (PSMS). PSMS is a reliable method for the selection of optimal learning algorithms together with preprocessing methods, as well as for hyperparameter optimization. In this paper we applied PSMS for the selection of the (pseudo) optimal combination of preprocessing methods and hyperparameters for a fixed neural network on benchmark data sets from a challenging competition: the (IJCNN 2007) agnostic vs prior knowledge challenge. A forum for the evaluation of methods for model selection and data representation discovery. In this paper we further show that the use of PSMS is useful for model selection when we have no knowledge about the domain we are dealing with. With PSMS we obtained competitive models that are ranked high in the official results of the challenge.

Spoken language identification consists in recognizing a language based on a sample of speech from an unknown speaker. The traditional approach for this task mainly considers the phonothactic information of languages. However, for marginalized languages –languages with few speakers or oral languages without a fixed writing standard–, this information is practically not at hand and consequently the usual approach is not applicable. In this paper, we present a method that only considers the acoustic features of the speech signal and does not use any kind of linguistic information. This method applies a wavelet transform to extract the acoustic features of the speech signal. The experimental results on a pairwise discrimination task among nine languages demonstrated that this approach considerably outperforms other previous methods based on the sole use of acoustic features. Index Terms: spoken language identification, acoustic features, wavelet transform.

The Mexico Research Centre of the Web Intelligence Consortium was established in 2003 motivated by Mexico being selected as the host of the 2nd Atlantic Web Intelligence Conference. It currently has 18 members including faculty and doctoral students from 7 different institutions. The WIC-Mexico includes groups working in the areas of Intelligent Web Information Retrieval, Web Mining and Farming, Knowledge Management, and Agents in Ubiquitous Computing.

A method for conceptual clustering of a collection of texts represented with conceptual graphs is presented. It uses an incremental strategy to construct the cluster hierarchy and incorporates some characteristics attractive for text mining purposes. For instance, it considers the structural information of the graphs, uses domain knowledge to detect the clusters with generalized descriptions, and uses a user-defined similarity measure between the graphs.