Search

Correctly identifying multiword expressions (MWEs) is an important task for most natural language processing systems since their misidentification can result in ambiguity and misunderstanding of the underlying text. In this work, we evaluate the performance of the mBERT model for MWE identification in a multilingual context by training it on all 14 languages available in version 1.2 of the PARSEME corpus. We also incorporate lateral inhibition and language adversarial training into our methodology to create language-independent embeddings and improve its capabilities in identifying multiword expressions. The evaluation of our models shows that the approach employed in this work achieves better results compared to the best system of the PARSEME 1.2 competition, MTLB-STRUCT, on 11 out of 14 languages for global MWE identification and on 12 out of 14 languages for unseen MWE identification. Additionally, averaged across all languages, our best approach outperforms the MTLB-STRUCT system by 1.23% on global MWE identification and by 4.73% on unseen global MWE identification., Comment: Accepted at Mathematics 2023

Reading comprehension requires readers to connect ideas within and across texts to produce a coherent mental representation. One important factor in that complex process regards the cohesion of the document(s). Here, we tackle the challenge of providing researchers and practitioners with a tool to visualize text cohesion both within (intra) and between (inter) texts. This tool, Multi-document Cohesion Network Analysis (MD-CNA), expands the structure of a CNA graph with lexical overlap links of multiple types, together with coreference links to highlight dependencies between text fragments of different granularities. We introduce two visualizations of the CNA graph that support the visual exploration of intratextual and intertextual links. First, a "hierarchical view" displays a tree-structure of discourse as a visual illustration of CNA links within a document. Second, a "grid view" available at paragraph or sentence levels displays links both within and between documents, thus ensuring ease of visualization for links spanning across multiple documents. Two use cases are provided to evaluate key functionalities and insights for each type of visualization. [This is a chapter in: Bittencourt I., Cukurova M., Muldner K., Luckin R., Millán E. (Eds) "Artificial Intelligence in Education. AIED 2020. Lecture Notes in Computer Science," v12164 (p80-85). Cham, Switzerland: Springer Nature Switzerland.]

Users from the online environment can create different ways of expressing their thoughts, opinions, or conception of amusement. Internet memes were created specifically for these situations. Their main purpose is to transmit ideas by using combinations of images and texts such that they will create a certain state for the receptor, depending on the message the meme has to send. These posts can be related to various situations or events, thus adding a funny side to any circumstance our world is situated in. In this paper, we describe the system developed by our team for SemEval-2020 Task 8: Memotion Analysis. More specifically, we introduce a novel system to analyze these posts, a multimodal multi-task learning architecture that combines ALBERT for text encoding with VGG-16 for image representation. In this manner, we show that the information behind them can be properly revealed. Our approach achieves good performance on each of the three subtasks of the current competition, ranking 11th for Subtask A (0.3453 macro F1-score), 1st for Subtask B (0.5183 macro F1-score), and 3rd for Subtask C (0.3171 macro F1-score) while exceeding the official baseline results by high margins., Comment: Accepted at SemEval-2020, 7 pages, 1 figure, 2 tables

Intelligence Tutoring Systems (ITSs) focus on promoting knowledge acquisition, while providing relevant feedback during students' practice. Self-explanation practice is an effective method used to help students understand complex texts by leveraging comprehension. Our aim is to introduce a deep learning neural model for automatically scoring student self-explanations that are targeted at specific sentences. The first stage of the processing pipeline performs an initial text cleaning and applies a set of predefined rules established by human experts in order to identify specific cases (e.g., students who do not understand the text, or students who simply copy and paste their self-explanations from the given input text). The second step uses a Recurrent Neural Network with pre-trained Glove word embeddings to predict self-explanation scores on a scale of 1 to 3. In contrast to previous SVM models trained on the same dataset of 4109 self-explanations, we obtain a significant increase of accuracy from 59% to 73%. Moreover, the new pipeline can be integrated in learning scenarios requiring near real-time responses from the ITS, thus addressing a major limitation in terms of processing speed exhibited by the previous approach. [This paper was published in: M. Scheffel et al. (Eds.), "EC-TEL 2019" (pp. 659-663). Switzerland: Springer.]