{"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/interpretation-of-semantic-tweet","title":"Interpretation of Semantic Tweet Representations","arxiv_id":"1704.00898","date":"2017-04-04","proceeding":null,"authors":["J Ganesh","Manish Gupta","Vasudeva Varma"],"abstract":"Research in analysis of microblogging platforms is experiencing a renewed\nsurge with a large number of works applying representation learning models for\napplications like sentiment analysis, semantic textual similarity computation,\nhashtag prediction, etc. Although the performance of the representation\nlearning models has been better than the traditional baselines for such tasks,\nlittle is known about the elementary properties of a tweet encoded within these\nrepresentations, or why particular representations work better for certain\ntasks. Our work presented here constitutes the first step in opening the\nblack-box of vector embeddings for tweets. Traditional feature engineering\nmethods for high-level applications have exploited various elementary\nproperties of tweets. We believe that a tweet representation is effective for\nan application because it meticulously encodes the application-specific\nelementary properties of tweets. To understand the elementary properties\nencoded in a tweet representation, we evaluate the representations on the\naccuracy to which they can model each of those properties such as tweet length,\npresence of particular words, hashtags, mentions, capitalization, etc. Our\nsystematic extensive study of nine supervised and four unsupervised tweet\nrepresentations against most popular eight textual and five social elementary\nproperties reveal that Bi-directional LSTMs (BLSTMs) and Skip-Thought Vectors\n(STV) best encode the textual and social properties of tweets respectively.\nFastText is the best model for low resource settings, providing very little\ndegradation with reduction in embedding size. Finally, we draw interesting\ninsights by correlating the model performance obtained for elementary property\nprediction tasks with the highlevel downstream applications.","url_abs":"http://arxiv.org/abs/1704.00898v2","url_pdf":"http://arxiv.org/pdf/1704.00898v2.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":"interpretation-of-semantic-tweet","repo_url":"https://github.com/ganeshjawahar/fine-tweet","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"feature-engineering","task_name":"Feature Engineering"},{"task_slug":"property-prediction","task_name":"Property Prediction"},{"task_slug":"representation-learning","task_name":"Representation Learning"},{"task_slug":"semantic-textual-similarity","task_name":"Semantic Textual Similarity"},{"task_slug":"sentiment-analysis","task_name":"Sentiment Analysis"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}