{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/combining-similarity-features-and-deep","title":"Combining Similarity Features and Deep Representation Learning for Stance Detection in the Context of Checking Fake News","arxiv_id":"1811.00706","date":"2018-11-02","proceeding":null,"authors":["Luís Borges","Bruno Martins","Pável Calado"],"abstract":"Fake news are nowadays an issue of pressing concern, given their recent rise\nas a potential threat to high-quality journalism and well-informed public\ndiscourse. The Fake News Challenge (FNC-1) was organized in 2017 to encourage\nthe development of machine learning-based classification systems for stance\ndetection (i.e., for identifying whether a particular news article agrees,\ndisagrees, discusses, or is unrelated to a particular news headline), thus\nhelping in the detection and analysis of possible instances of fake news. This\narticle presents a new approach to tackle this stance detection problem, based\non the combination of string similarity features with a deep neural\narchitecture that leverages ideas previously advanced in the context of\nlearning efficient text representations, document classification, and natural\nlanguage inference. Specifically, we use bi-directional Recurrent Neural\nNetworks, together with max-pooling over the temporal/sequential dimension and\nneural attention, for representing (i) the headline, (ii) the first two\nsentences of the news article, and (iii) the entire news article. These\nrepresentations are then combined/compared, complemented with similarity\nfeatures inspired on other FNC-1 approaches, and passed to a final layer that\npredicts the stance of the article towards the headline. We also explore the\nuse of external sources of information, specifically large datasets of sentence\npairs originally proposed for training and evaluating natural language\ninference methods, in order to pre-train specific components of the neural\nnetwork architecture (e.g., the RNNs used for encoding sentences). The obtained\nresults attest to the effectiveness of the proposed ideas and show that our\nmodel, particularly when considering pre-training and the combination of neural\nrepresentations together with similarity features, slightly outperforms the\nprevious state-of-the-art.","url_abs":"http://arxiv.org/abs/1811.00706v1","url_pdf":"http://arxiv.org/pdf/1811.00706v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"combining-similarity-features-and-deep","repo_url":"https://github.com/LuisPB7/fnc-msc","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null},{"paper_slug":"combining-similarity-features-and-deep","repo_url":"https://github.com/yinghao1019/imdb_prac","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"combining-similarity-features-and-deep","repo_url":"https://github.com/imran3180/pytorch-nli","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"document-classification","task_name":"Document Classification"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"natural-language-inference","task_name":"Natural Language Inference"},{"task_slug":"representation-learning","task_name":"Representation Learning"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"stance-detection","task_name":"Stance Detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/fake-news-detection-on-fnc-1","task":"Fake News Detection","dataset":"FNC-1","model":"Bi-LSTM (max-pooling, attention)","rank_in_archive_order":4,"of":10,"metrics":{"Per-class Accuracy (Agree)":"51.34","Per-class Accuracy (Disagree)":"10.33","Per-class Accuracy (Discuss)":"81.52","Per-class Accuracy (Unrelated)":"96.74","Weighted Accuracy":"82.23"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-multinli","task":"Natural Language Inference","dataset":"MultiNLI","model":"Stacked Bi-LSTMs (shortcut connections, max-pooling)","rank_in_archive_order":50,"of":67,"metrics":{"Matched":"71.4","Mismatched":"72.2"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-multinli","task":"Natural Language Inference","dataset":"MultiNLI","model":"Bi-LSTM sentence encoder (max-pooling)","rank_in_archive_order":52,"of":67,"metrics":{"Matched":"70.7","Mismatched":"71.1"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-multinli","task":"Natural Language Inference","dataset":"MultiNLI","model":"Stacked Bi-LSTMs (shortcut connections, max-pooling, attention)","rank_in_archive_order":53,"of":67,"metrics":{"Matched":"70.7","Mismatched":"70.5"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-snli","task":"Natural Language Inference","dataset":"SNLI","model":"Stacked Bi-LSTMs (shortcut connections, max-pooling)","rank_in_archive_order":75,"of":98,"metrics":{"% Test Accuracy":"84.8"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-snli","task":"Natural Language Inference","dataset":"SNLI","model":"Bi-LSTM sentence encoder (max-pooling)","rank_in_archive_order":79,"of":98,"metrics":{"% Test Accuracy":"84.5"},"uses_additional_data":false},{"leaderboard":"/sota/natural-language-inference-on-snli","task":"Natural Language Inference","dataset":"SNLI","model":"Stacked Bi-LSTMs (shortcut connections, max-pooling, attention)","rank_in_archive_order":80,"of":98,"metrics":{"% Test Accuracy":"84.4"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1811.00706","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}