Abstract

Abstract

Abstract
Sentiment lexicon plays a vital role in lexicon-based sentiment analysis. The lexicon-based method is often preferred because it leads to more explainable answers in comparison with many machine learning-based methods. But, semantic orientation of a word depends on its domain. Hence, a general-purpose sentiment lexicon may gives sub-optimal performance compare with a domain-specific lexicon. However, it is challenging to manually generate a domain-specific sentiment lexicon for each domain. Still, it is impractical to generate complete sentiment lexicon for a domain from a single corpus. To this end, we propose an approach to automatically generate a domain-specific sentiment lexicon using a vector model enriched by weights. Importantly, we propose an incremental approach for updating an existing lexicon to either the same domain or different domain (domain-adaptation). Finally, we discuss how to incorporate sentiment lexicons information in neural models (word embedding) for better performance.

Abstract

Abstract
Progress in the field of Natural Language Processing (NLP) has been closely followed by applications in the medical domain. Recent advancements in Neural Language Models (NLMs) have transformed the field and are currently motivating numerous works exploring their application in different domains. In this paper, we explore how NLMs can be used for Medical Entity Linking with the recently introduced MedMentions dataset, which presents two major challenges: (1) a large target ontology of over 2M concepts, and (2) low overlap between concepts in train, validation and test sets. We introduce a solution, MedLinker, that addresses these issues by leveraging specialized NLMs with Approximate Dictionary Matching, and show that it performs competitively on semantic type linking, while improving the state-of-the-art on the more fine-grained task of concept linking (+4 F1 on MedMentions main task).

Abstract
The task of Sentence Compression aims at producing a shorter version of a given sentence. This task may assist many other applications, as Automatic Summarization and Text Simplification. In this paper, we investigate methods for Sentence Compression for Portuguese. We focus on machine learning-based algorithms and propose new strategies. We also create reference corpora/datasets for the area, allowing to train and to test the methods of interest. Our results show that some of our methods outperform previous initiatives for Portuguese and produce competitive results with a state of the art method in the area.