Counteracting French Fake News on Climate Change Using Language Models

Meddeb, Paul; Ruşeţi, Ştefan; Dascălu, Mihai; Terian, Simina-Maria; Travadel, Sébastien

doi:10.3390/su141811724

Cited by 7 publications

(6 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Unlike most COVID-19 misinformation detection research utilising similar NLP models, that used either a four label annotation (Fake, Biased, Irrelevant, Other) [20], or a three label entailment task (Agree, Disagree, No stance) along with a bank of misconception claims [21], we intended to fully annotate each narrative that we deemed important to flag in a hierarchical fashion (Supernarrative, Narrative, Subnarrative) and build a model able to track the evolution of such narratives across time.…”

Section: Methodsmentioning

confidence: 99%

“…For (d), we developed an automatic narrative classifier, based on a multi-lingual pre-trained BERT model [27]. BERT-based models have been used before to monitor and detect misinformation both in the context of COVID-19 [20] and to conduct trend analysis of climate change denial claims over time [17]. Our classifier was built on our codebook of COVID-19 mis/disinformation narratives and the pre-trained model was fine-tuned with a subset of 30,000 annotated items.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Trend analysis of COVID-19 mis/disinformation narratives–A 3-year study

Kotseva,

Vianini,

Nikolaidis

et al. 2023

PLoS ONE

View full text Add to dashboard Cite

To tackle the COVID-19 infodemic, we analysed 58,625 articles from 460 unverified sources, that is, sources that were indicated by fact checkers and other mis/disinformation experts as frequently spreading mis/disinformation, covering the period from 1 January 2020 to 31 December 2022. Our aim was to identify the main narratives of COVID-19 mis/disinformation, develop a codebook, automate the process of narrative classification by training an automatic classifier, and analyse the spread of narratives over time and across countries. Articles were retrieved with a customised version of the Europe Media Monitor (EMM) processing chain providing a stream of text items. Machine translation was employed to automatically translate non-English text to English and clustering was carried out to group similar articles. A multi-level codebook of COVID-19 mis/disinformation narratives was developed following an inductive approach; a transformer-based model was developed to classify all text items according to the codebook. Using the transformer-based model, we identified 12 supernarratives that evolved over the three years studied. The analysis shows that there are often real events behind mis/disinformation trends, which unverified sources misrepresent or take out of context. We established a process that allows for near real-time monitoring of COVID-19 mis/disinformation. This experience will be useful to analyse mis/disinformation about other topics, such as climate change, migration, and geopolitical developments.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Trend analysis of COVID-19 mis/disinformation narratives–A 3-year study

Kotseva,

Vianini,

Nikolaidis

et al. 2023

PLoS ONE

View full text Add to dashboard Cite

show abstract

“…However, the field is experiencing an enormous revolution since the implementation of transformer models has been possible due to the rise of computing [25] -using a CamemBERT Transformer-based model, a transformer french language model, they classify french news on climate change, identifying those which are fake ones- [20,3,34,14,23]. Transformers are large language models that capture the dependencies between words, that are encoded in word embeddings whose space represents the meaning of the words.…”

Section: Introductionmentioning

confidence: 99%

Fine-tuning ClimateBert transformer with ClimaText for the disclosure analysis of climate-related financial risks

Garrido-Merchán¹,

González-Barthe²,

Vaca³

2023

Preprint

View full text Add to dashboard Cite

In recent years there has been a growing demand from financial agents, especially from particular and institutional investors, for companies to report on climate-related financial risks. A vast amount of information, in text format, can be expected to be disclosed in the short term by firms in order to identify these types of risks in their financial and non financial reports, particularly in response to the growing regulation that is being passed on the matter. To this end, this paper applies stateof-the-art NLP techniques to achieve the detection of climate change in text corpora. We use transfer learning to fine-tune two transformer models, BERT and ClimateBert -a recently published DistillRoBERTa-based model that has been specifically tailored for climate text classification-. These two algorithms are based on the transformer architecture which enables learning the contextual relationships between words in a text. We carry out the fine-tuning process of both models on the novel "Clima-Text" database, consisting of data collected from Wikipedia, 10K Files Reports and web-based claims. Our text classification model obtained from the ClimateBert fine-tuning process on ClimaText, outperforms the models created with BERT and the current state-of-the-art transformer in this particular problem. Our study is the first one to implement on the ClimaText database the recently published ClimateBert algorithm. Based on our results, it can be said that ClimateBert fine-tuned on Cli-maText is an outstanding tool within the NLP pre-trained transformer models that may and should be used by investors, institutional agents and companies themselves to monitor the disclosure of climate risk in financial reports, as well as in other textual sources, such as policies, new legislation or directives passed on this matter. In addition, our transfer learning methodology is cheap in computational terms, thus allowing any organization to perform it.

show abstract

“…The growing importance of these disclosures, with their intrinsic characteristic of heterogeneity and dispersed features, make the task of studying and analyzing these type of financial and non-financial reports worthy of automation. As a result, in recent years, a growing literature has emerged that relies on AI for the identification of climate-related information [16][17][18][19][20][21][22][23][24][25][26][27], among others.…”

Section: Introductionmentioning

confidence: 99%

“…However, the field is experiencing an enormous revolution since the implementation of transformer models has been possible due to the rise of computing Meddeb et al [17]-using a CamemBERT Transformer-based model, a transformer French language model, they classify French news on climate change, identifying those which are fake ones- [18,20,22,23,25]. Transformers are large language models that capture the dependencies between words, that are encoded in word embeddings whose space represents the meaning of the words.…”

Section: Introductionmentioning

confidence: 99%

Fine-tuning ClimateBert transformer with ClimaText for the disclosure analysis of climate-related financial risks

Garrido-Merchán,

González-Barthe,

Coronado-Vaca

2023

Preprint

View full text Add to dashboard Cite

In recent years there has been a growing demand from financial agents, especially from particular and institutional investors, for companies to report on climate-related financial risks. A vast amount of information, in text format, can be expected to be disclosed in the short term by firms in order to identify these types of risks in their financial and non-financial reports, particularly in response to the growing regulation that is being passed on the matter. To this end, this paper applies state-of-the-art NLP techniques to achieve the detection of climate change in text corpora. We use transfer learning to fine-tune two transformer models, BERT and ClimateBert -a recently published DistillRoBERTa-based model that has been specifically tailored for climate text classification-. These two algorithms are based on the transformer architecture which enables learning the contextual relationships between words in a text. We carry out the fine-tuning process of both models on the novel “ClimaText” database, consisting of data collected from Wikipedia, 10K Files Reports, and web-based claims. Our text classification model obtained from the ClimateBert fine-tuning process on ClimaText, outperforms the models created with BERT and the current state-of-the-art transformer in this particular problem. Our study is the first one to implement the recently published ClimateBert algorithm on the ClimaText database. Based on our results, it can be said that ClimateBert fine-tuned on ClimaText is an outstanding tool within the NLP pre-trained transformer models that may and should be used by investors, institutional agents, and companies themselves to monitor the disclosure of climate risk in financial reports, as well as in other textual sources, such as policies, new legislation or directives passed on this matter. In addition, our transfer learning methodology is cheap in computational terms, thus allowing any organization to perform it.

show abstract

Counteracting French Fake News on Climate Change Using Language Models

Cited by 7 publications

References 21 publications

Trend analysis of COVID-19 mis/disinformation narratives–A 3-year study

Trend analysis of COVID-19 mis/disinformation narratives–A 3-year study

Fine-tuning ClimateBert transformer with ClimaText for the disclosure analysis of climate-related financial risks

Fine-tuning ClimateBert transformer with ClimaText for the disclosure analysis of climate-related financial risks

Contact Info

Product

Resources

About