Search results for: BERT fine-tuning

Authors: Qitao Xie, Qingquan Zhang, Xiaofei Zhang, Di Tian, Ruixuan Wen, Ting Zhu, Ping Yi, Xin Li

Abstract:

Inspired by the recent movement of digital currency, we are building a question answering system concerning the subject of cryptocurrency using Bidirectional Encoder Representations from Transformers (BERT). The motivation behind this work is to properly assist digital currency investors by directing them to the corresponding knowledge bases that can offer them help and increase the querying speed. BERT, one of newest language models in natural language processing, was investigated to improve the quality of generated responses. We studied different combinations of hyperparameters of the BERT model to obtain the best fit responses. Further, we created an intelligent chatbot for cryptocurrency using BERT. A chatbot using BERT shows great potential for the further advancement of a cryptocurrency market tool. We show that the BERT neural networks generalize well to other tasks by applying it successfully to cryptocurrency.

Keywords: BERT, chatbot, cryptocurrency, deep learning.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 981

Authors: Josiel Delgadillo, Johnson Kinyua, Charles Mutigwe

Abstract:

The purpose of sentiment analysis is to determine the sentiment strength (e.g., positive, negative, neutral) from a textual source for good decision-making. Natural Language Processing (NLP) in domains such as financial markets requires knowledge of domain ontology, and pre-trained language models, such as BERT, have made significant breakthroughs in various NLP tasks by training on large-scale un-labeled generic corpora such as Wikipedia. However, sentiment analysis is a strong domain-dependent task. The rapid growth of social media has given users a platform to share their experiences and views about products, services, and processes, including financial markets. StockTwits and Twitter are social networks that allow the public to express their sentiments in real time. Hence, leveraging the success of unsupervised pre-training and a large amount of financial text available on social media platforms could potentially benefit a wide range of financial applications. This work is focused on sentiment analysis using social media text on platforms such as StockTwits and Twitter. To meet this need, SkyBERT, a domain-specific language model pre-trained and fine-tuned on financial corpora, has been developed. The results show that SkyBERT outperforms current state-of-the-art models in financial sentiment analysis. Extensive experimental results demonstrate the effectiveness and robustness of SkyBERT.

Keywords: BERT, financial markets, Twitter, sentiment analysis.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 716

Authors: Ebipatei Victoria Tunyan, T. A. Cao, Cheol Young Ock

Abstract:

Detecting subjectively biased statements is a vital task. This is because this kind of bias, when present in the text or other forms of information dissemination media such as news, social media, scientific texts, and encyclopedias, can weaken trust in the information and stir conflicts amongst consumers. Subjective bias detection is also critical for many Natural Language Processing (NLP) tasks like sentiment analysis, opinion identification, and bias neutralization. Having a system that can adequately detect subjectivity in text will boost research in the above-mentioned areas significantly. It can also come in handy for platforms like Wikipedia, where the use of neutral language is of importance. The goal of this work is to identify the subjectively biased language in text on a sentence level. With machine learning, we can solve complex AI problems, making it a good fit for the problem of subjective bias detection. A key step in this approach is to train a classifier based on BERT (Bidirectional Encoder Representations from Transformers) as upstream model. BERT by itself can be used as a classifier; however, in this study, we use BERT as data preprocessor as well as an embedding generator for a Bi-LSTM (Bidirectional Long Short-Term Memory) network incorporated with attention mechanism. This approach produces a deeper and better classifier. We evaluate the effectiveness of our model using the Wiki Neutrality Corpus (WNC), which was compiled from Wikipedia edits that removed various biased instances from sentences as a benchmark dataset, with which we also compare our model to existing approaches. Experimental analysis indicates an improved performance, as our model achieved state-of-the-art accuracy in detecting subjective bias. This study focuses on the English language, but the model can be fine-tuned to accommodate other languages.

Keywords: Subjective bias detection, machine learning, BERT–BiLSTM–Attention, text classification, natural language processing.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 830

Authors: A. M. John-Otumu, M. M. Rahman, O. C. Nwokonkwo, M. C. Onuoha

Abstract:

Online social media networks have long served as a primary arena for group conversations, gossip, text-based information sharing and distribution. The use of natural language processing techniques for text classification and unbiased decision making has not been far-fetched. Proper classification of these textual information in a given context has also been very difficult. As a result, a systematic review was conducted from previous literature on sentiment classification and AI-based techniques. The study was done in order to gain a better understanding of the process of designing and developing a robust and more accurate sentiment classifier that could correctly classify social media textual information of a given context between hate speech and inverted compliments with a high level of accuracy using the knowledge gain from the evaluation of different artificial intelligence techniques reviewed. The study evaluated over 250 articles from digital sources like ACM digital library, Google Scholar, and IEEE Xplore; and whittled down the number of research to 52 articles. Findings revealed that deep learning approaches such as Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Bidirectional Encoder Representations from Transformer (BERT), and Long Short-Term Memory (LSTM) outperformed various machine learning techniques in terms of performance accuracy. A large dataset is also required to develop a robust sentiment classifier. Results also revealed that data can be obtained from places like Twitter, movie reviews, Kaggle, Stanford Sentiment Treebank (SST), and SemEval Task4 based on the required domain. The hybrid deep learning techniques like CNN+LSTM, CNN+ Gated Recurrent Unit (GRU), CNN+BERT outperformed single deep learning techniques and machine learning techniques. Python programming language outperformed Java programming language in terms of development simplicity and AI-based library functionalities. Finally, the study recommended the findings obtained for building robust sentiment classifier in the future.

Keywords: Artificial Intelligence, Natural Language Processing, Sentiment Analysis, Social Network, Text.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 594

Authors: Tomás Speziale

Abstract:

This paper aims to analyse how Ian Hacking states the theoretical basis of his research on the classification of people. Although all his early philosophical education had been based in Foucault, it is also true that Erving Goffman’s perspective provided him with epistemological and methodological tools for understanding face-to-face relationships. Hence, all his works must be thought of as social science texts that combine the research on how the individuals are constituted ‘top-down’ (as in Foucault), with the inquiry into how people renegotiate ‘bottom-up’ the classifications about them. Thus, Hacking´s proposal constitutes a middle ground between the French Philosopher and the American Sociologist. Placing himself between both authors allows Hacking to build a frame that is expected to adjust to Social Sciences’ main particularity: the fact that they study interactive kinds. These are kinds of people, which imply that those who are classified can change in certain ways that prompt the need for changing previous classifications themselves. It is all about the interaction between the labelling of people and the people who are classified. Consequently, understanding the way in which Hacking uses Foucault’s and Goffman’s theories is essential to fully comprehend the social dynamic between individuals and concepts, what Bert Hansen had called dialectical realism. His theoretical proposal, therefore, is not only valuable because it combines diverse perspectives, but also because it constitutes an utterly original and relevant framework for Sociological theory and particularly for Criminology.

Keywords: Classification of people, Foucault`s archaeology, Goffman`s interpersonal sociology, interactive kinds.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2050

Authors: Shahad Nagoor, Lucy Hederman, Kevin Koidl, Annalina Caputo

Abstract:

Doctors’ notes reflect their impressions, attitudes, clinical sense, and opinions about patients’ conditions and progress, and other information that is essential for doctors’ daily clinical decisions. Despite their value, clinical notes are insufficiently researched within the language processing community. Automatically extracting information from unstructured text data is known to be a difficult task as opposed to dealing with structured information such as physiological vital signs, images and laboratory results. The aim of this research is to investigate how Natural Language Processing (NLP) techniques and machine learning techniques applied to clinician notes can assist in doctors’ decision making in Intensive Care Unit (ICU) for coronavirus disease 2019 (COVID-19) patients. The hypothesis is that clinical outcomes like survival or mortality can be useful to influence the judgement of clinical sentiment in ICU clinical notes. This paper presents two contributions: first, we introduce COVID_ICU_BERT, a fine-tuned version of a clinical transformer model that can reliably predict clinical sentiment for notes of COVID patients in ICU. We train the model on clinical notes for COVID-19 patients, ones not previously seen by Bio_ClinicalBERT or Bio_Discharge_Summary_BERT. The model which was based on Bio_ClinicalBERT achieves higher predictive accuracy than the one based on Bio_Discharge_Summary_BERT (Acc 93.33%, AUC 0.98, and Precision 0.96). Second, we perform data augmentation using clinical contextual word embedding that is based on a pre-trained clinical model to balance the samples in each class in the data (survived vs. deceased patients). Data augmentation improves the accuracy of prediction slightly (Acc 96.67%, AUC 0.98, and Precision 0.92).

Keywords: BERT fine-tuning, clinical sentiment, COVID-19, data augmentation.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 276

Authors: Vishal Raj, Noorhan Abbas

Abstract:

Ambiguity in NLP (Natural Language Processing) refers to the ability of a word, phrase, sentence, or text to have multiple meanings. This results in various kinds of ambiguities such as lexical, syntactic, semantic, anaphoric and referential. This study is focused mainly on solving the issue of Lexical ambiguity. Word Sense Disambiguation (WSD) is an NLP technique that aims to resolve lexical ambiguity by determining the correct meaning of a word within a given context. Most WSD solutions rely on words for training and testing, but we have used lemma and Part of Speech (POS) tokens of words for training and testing. Lemma adds generality and POS adds properties of word into token. We have designed a method to create an affinity matrix to calculate the affinity between any pair of lemma_POS (a token where lemma and POS of word are joined by underscore) of given training set. Additionally, we have devised an algorithm to create the sense clusters of tokens using affinity matrix under hierarchy of POS of lemma. Furthermore, three different mechanisms to predict the sense of target word using the affinity/similarity value are devised. Each contextual token contributes to the sense of target word with some value and whichever sense gets higher value becomes the sense of target word. So, contextual tokens play a key role in creating sense clusters and predicting the sense of target word, hence, the model is named Contextual SenSe Model (CSM). CSM exhibits a noteworthy simplicity and explication lucidity in contrast to contemporary deep learning models characterized by intricacy, time-intensive processes, and challenging explication. CSM is trained on SemCor training data and evaluated on SemEval test dataset. The results indicate that despite the naivety of the method, it achieves promising results when compared to the Most Frequent Sense (MFS) model.

Keywords: Word Sense Disambiguation, WSD, Contextual SenSe Model, Most Frequent Sense, part of speech, POS, Natural Language Processing, NLP, OOV, out of vocabulary, ELMo, Embeddings from Language Model, BERT, Bidirectional Encoder Representations from Transformers, Word2Vec, lemma_POS, Algorithm.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 384