Intricate Twitternlp Modelling

Abstract

In this research, we introduce TweetNLP, a platform for social media Natural Language Processing (NLP). An extensive range of NLP tasks are supported by TweetNLP, including standard focus areas like sentiment analysis and named entity recognition as well as social media-specific tasks like emoticon prediction and offensive language detection. Task-specific systems run on moderately small Transformer-based language models that are focused on social media text, particularly Twitter, and don\'t require specialized hardware or cloud services to operate. TweetNLP\'s major contributions are: (1) an integrated Python library for a contemporary toolkit supporting social media analysis using various task-specific models tailored to the social domain; (2) an interactive online demo for codeless experimentation using our models; and (3) a tutorial covering a wide range of typical social media applications.

Introduction

I. INTRODUCTION

Social media is characterised by its connectivity, accessibility, and content creation. It is a useful tool for sharing, creating, and disseminating information as well as for communicating with people locally and globally. Social media usage has evolved into a regular activity in today's world. Twitter, Instagram, Youtube and other social media platforms have all emerged as major informational resources.

It has been discovered that, by extracting and analysing data from social networking sites, an understanding of contemporary society can be developed. Online users communicate with each other by sending text-only messages or enhancing them with multimedia content like images, audio, or video. This has led to the usage of these platforms to comprehend user, group, and organisational behaviour. Particularly, twitter, the primary medium examined in this work, has long been a valuable tool for comprehending society as a whole. Twitter is a crucial research and practical resource for natural language processing (nlp) because of its relevance and accessibility.

Twitter is intriguing for nlp because it embodies many characteristics that come naturally in fast-paced, impromptu conversation. The improvement of results on benchmark datasets with roughly independent and identically distributed (iid) training, validation, and testing sections, drawn from data that was gathered or validated by open sourcing, has been the focus of a significant and influential thread of research on natural language understanding (nlu).

Additionally there are significant flaws in allegedly high-performing systems, and they nonetheless lack human-level task competence. In fact, it has been demonstrated that even conventional nlp systems perform poorly when applied to social media, particularly when performing tasks like normalisation, part-of-speech tagging, sentiment analysis, or named entity recognition because of problems like noise, length restrictions for messages related to platforms, jargon, emoticon, colloquial language and multilinguality.

Tweetnlp (tweetnlp.org) provides a library tailored to twitter. Transformer-based language models that have been trained on twitter make up the core of tweetnlp (barbieri et al., 2020, 2022; loureiro et al., 2022). These specialised language models have then undergone additional fine-tuning for particular nlp tasks on twitter data.

All of these resources are consolidated into one platform by tweetnlp. Tweetnlp provides a simple python api that makes it simple to use social media models.

Despite the tendency toward progressively larger language models (shoeybi et al., 2019; brown et al., 2020), tweetnlp is more concerned with the general user and applicability and hence include base models that are simple to operate on standard computers or on free cloud services. The ability to test models and conduct real-time analysis on twitter is provided by an interactive online demo that provides access to all models.

II. SUPPORTING TASKS AND EMBEDDINGS

Discussing the tasks supported by TweetNLP.For classification tasks, we simply fine-tune the models which are described in the TweetEval library, and for refining named entity recognition, we depend on the T-NER library, which is also integrated into TweetNLP.

1. Sentiment Analysis- Sentiment analysis is the process of identifying and categorizing the emotions represented in a text source. When analyzed, tweets may produce a significant quantity of sentiment data. These statistics help us understand how individuals feel about a range of issues.  Aims to forecast the feeling of a tweet that has one of the three classifications: positive, negative, or neutral. 
2. Emotion Recognition- Emotions are considered of utmost importance as they have a key responsibility in human interaction. The goal of this activity is to match the most relevant emotion to a tweet, boiled down to surprise, love, hate, boredom, anger, happiness, sadness, and empty.
3. Emoji Prediction- The goal of emoji prediction is to predict the final emoji of a tweet, including 20 emoji labels.It is important to take into account various selection heuristics, such as choosing the first or most prevalent emoji inside a tweet, while attempting to identify the genuine label from a variety of emojis in a tweet.
4. Hate Speech Detection- The hate speech dataset consists of identifying tweets that are hostile toward immigrants or women. The categories of the dataset in which this model was fine-tuned are hate speech, offensive but not hate speech, or neither offensive nor hate speech.
5. Offensive Language Identification- It is crucial to research the detection of abusive language on social media to prevent conflicts brought on by the usage of such language and to make social media platforms safer for kids and teenagers.The assignment is to find any derogatory words in a tweet.

1. Embeddings

1. Word Embeddings- Social media has the actual ability to injure individuals and may serve as a platform for the spread of racism, misogyny, and other hateful ideologies. Although having the right to free expression is crucial, there are instances when it's necessary to spot such stuff and stop it from being published. In the NLP approach known as word embedding, each word is represented as an n-dimensional vector that depicts the word's projection in vector space. A particular word's position in the model is determined during training by the words that surround it. Word2Vec and GloVe are the two most popular techniques used to build this kind of model.
2. Tweet Embeddings- The syntactic and semantic components of a word are both captured by its embedding. Tweets vary from other forms of text in that they are brief, loud, and have particular lexical and semantic characteristics. Word embeddings must be specially learnt from tweets as a result. We selected one response at random for tweets that received several replies. Each mini-batch in training consists of a list of tweet-reply pairings. The enumeration of all other conceivable combinations of tweet-reply, tweet-tweet, and tweet-reply pairings is considered a negative sample; the tweet-reply pairs are considered positive samples. In this paper, we also present experiments demonstrating how to use the data sets in some NLP tasks, such as tweet sentiment analysis and hate recognition, emotion recognition, emoji prediction, and offensive language detection.

A. METHODOLOGY

A. Sentiment Analysis

The data input is a test and train dataset containing various tweets and comments and the tweets are of mixed sentiments, such as positive, negative, and neutral. The distribution of Training and Testing data is depicted through a histogram using visualization tools such as Seaborn and Matplotlib.

The training data is used to train the model in order for it to understand the different words and related contextual sentiment for further analysis with the test data. According to the analysis based on the training data, the most repetitive words within the dataset are the following. By the use of a few different packages within the python environment, we can find out the vocabulary and different sentiment-related words.

C. Emotion Recognition

The ability to recognize emotions has several uses, including the ability to identify psychological problems like anxiety or sadness in people or gauge how a community feels about a certain issue. In human-computer interaction systems and their applications, emotion recognition is essential. These days, the majority of individuals use social media sites like Twitter, Facebook, Instagram, and others extensively to express their feelings or opinions about a certain subject. As a result, these sites serve as enormous data warehouses for emotional information.

The following is the graphical representation of emotion that we classified from the dataset we used

E. Offensive Language Identification

Social networks have been increasingly popular in recent years. The idea behind social media was to allow us to express our opinions online, stay in touch with loved ones, and share happy moments. However, as reality is not so ideal, there are others who share hate speech-related messages, use it to abuse particular people, for example, or even build robots whose sole purpose is to attack particular circumstances or individuals. It is difficult to determine who created such content, but there are numerous approaches that might be used, such as natural language processing or machine learning algorithms that can look into the text and make predictions using the related meta-data.

Conclusion

In this paper, we have introduced TwitterNLP, an NLP platform with a focus on social media. The software uses very simple language models that were trained on Twitter and adjusted for many prominent NLP tasks on social media, including sentiment analysis, identifying objectionable language, emotion recognition, emoji prediction, detecting hate speech, and named entity recognition. Additionally, TwitterNLP makes it simple for non-programmers to analyze the models, which can assist in discovering negative biases or flaws and ultimately lead to future model improvement. While this first released version of TwitterNLP is autarchic and complete, we want to continuously add more models and tasks to it. We intend to create additional datasets and models for social media tasks because TwitterNLP\'s foundation is social media data. In particular, future expansion can be beyond the tweet categorization task, which is currently sufficed by TwitterNLP in depth. Part-Of-Speech tagging, stopword removal, syntactic parsing has always proven challenging in a gigantic setting like social media. In addition, we want to support languages other than English in a larger range of activities and expand TwitterNLP to include other social networking sites like Reddit, LinkedIn, and Instagram. The results of this experiment indicate that TwitterNLP has multifold advantages and can be used explicitly.

References

[1] S. M. Metev and V. P. Veiko, Laser Assisted Microtechnology, 2nd ed., R. M. Osgood, Jr., Ed. Berlin, Germany: Springer-Verlag, 1998. [2] [TweetEval: Unified Benchmark and Comparative Evaluation for Tweet Classification](https://aclanthology.org/2020.findings-emnlp.148) (Barbieri et al., Findings 2020) [3] https://www.analyticsvidhya.com/blog/2021/06/twitter-sentiment-analysis-a-nlp-use-case-for-beginners/ [4] Pak, Alexander & Paroubek, Patrick. (2010). Twitter as a Corpus for Sentiment Analysis and Opinion Mining. Proceedings of LREC. 10. [5] A. Yousaf et al., \"Emotion Recognition by Textual Tweets Classification Using Voting Classifier (LR-SGD),\" in IEEE Access, vol. 9, pp. 6286-6295, 2021, doi: 10.1109/ACCESS.2020.3047831. [6] V. N. Durga Pavithra Kollipara, V. N. Hemanth Kollipara and M. D. Prakash, \"Emoji Prediction from Twitter Data using Deep Learning Approach,\" 2021 Asian Conference on Innovation in Technology (ASIANCON), 2021, pp. 1-6, doi: 10.1109/ASIANCON51346.2021.9544680. [7] G. A. De Souza and M. Da Costa-Abreu, \"Automatic offensive language detection from Twitter data using machine learning and feature selection of metadata,\" 2020 International Joint Conference on Neural Networks (IJCNN), 2020, pp. 1-6, doi: 10.1109/IJCNN48605.2020.9207652. [8] Camacho-Collados, José & Rezaee, Kiamehr & Riahi, Talayeh & Ushio, Asahi & Loureiro, Daniel & Antypas, Dimosthenis & Boisson, Joanne & Espinosa-Anke, Luis & Liu, Fangyu & Martínez-Cámara, Eugenio & Medina, Gonzalo & Buhrmann, Thomas & Neves, Leonardo & Barbieri, Francesco. (2022). TweetNLP: Cutting-Edge Natural Language Processing for Social Media. 10.48550/arXiv.2206.14774. [9] [Feature-Rich Twitter Named Entity Recognition and Classification](https://aclanthology.org/W16-3922) (Sikdar & Gambäck, 2016). [10] https://paperswithcode.com/dataset/hate-speech-and-offensive-language [11] M. Krommyda, A. Rigos, K. Bouklas and A. Amditis, \"Emotion detection in Twitter posts: a rule-based algorithm for annotated data acquisition,\" 2020 International Conference on Computational Science and Computational Intelligence (CSCI), 2020, pp. 257-262, doi: 10.1109/CSCI51800.2020.00050. [12] https://www.ijstr.org/final-print/mar2020/Emotion-Recognition-Of-Twitter-Posts-In-Real-time-A-Survey.pdf [13] https://www.researchgate.net/publication/339980709_Sentiment_Analysis_with_NLP_on_Twitter_Data [14] Anupama B S , Rakshith D B , Rahul Kumar M , Navaneeth M, 2020, Real Time Twitter Sentiment Analysis using Natural Language Processing, INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY (IJERT) Volume 09, Issue 07 (July 2020) [15] Manju Venugopalan and Deepa Gupta, Exploring Sentiment Analysis on Twitter Data, IEEE 2015 [16] Kishori K. Pawar, Pukhraj P Shrishrimal, R. R. Deshmukh, Twitter Sentiment Analysis: A Review International Journal of Scientific & Engineering Research, Volume 6, Issue 4, April-2015 [17] B. Pariyani, K. Shah, M. Shah, T. Vyas and S. Degadwala, \"Hate Speech Detection in Twitter using Natural Language Processing,\" 2021 Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV), 2021, pp. 1146-1152, doi: 10.1109/ICICV50876.2021.9388496. [18] Pitsilis, Georgios & Ramampiaro, Heri & Langseth, Helge. (2018). Effective hate-speech detection in Twitter data using recurrent neural networks. Applied Intelligence. 48. in press.. 10.1007/s10489-018-1242-y. [19] Park, Ji & Fung, Pascale. (2017). One-step and Two-step Classification for Abusive Language Detection on Twitter. [20] V. N. Durga Pavithra Kollipara, V. N. Hemanth Kollipara and M. D. Prakash, \"Emoji Prediction from Twitter Data using Deep Learning Approach,\" 2021 Asian Conference on Innovation in Technology (ASIANCON), 2021, pp. 1-6, doi: 10.1109/ASIANCON51346.2021.9544680. [21] Wolny, Wieslaw. (2016). TWITTER SENTIMENT ANALYSIS USING EMOTICONS AND EMOJI IDEOGRAMS. [22] Pitsilis, Georgios & Ramampiaro, Heri & Langseth, Helge. (2018). Detecting Offensive Language in Tweets Using Deep Learning. [23] M. Kanakaraj and R. M. R. Guddeti, \"NLP based sentiment analysis on Twitter data using ensemble classifiers,\" 2015 3rd International Conference on Signal Processing, Communication and Networking (ICSCN), 2015, pp. 1-5, doi: 10.1109/ICSCN.2015.7219856. [24] Garg, Y., Chatterjee, N. (2014). Sentiment Analysis of Twitter Feeds. In: Srinivasa, S., Mehta, S. (eds) Big Data Analytics. BDA 2014. Lecture Notes in Computer Science, vol 8883. Springer, Cham. [25] TWEETEVAL: Unified Benchmark and Comparative Evaluation for Tweet Classification Francesco Barbieri, Jose Camacho-Collados Leonardo Neves, Luis Espinosa-Anke Snap Inc., Santa Monica, CA 90405, USA School of Computer Science and Informatics, Cardiff University, United Kingdom. [26] [TimeLMs: Diachronic Language Models from Twitter](https://aclanthology.org/2022.acl-demo.25) (Loureiro et al., ACL 2022) [27] [T-NER: An All-Round Python Library for Transformer-based Named Entity Recognition](https://aclanthology.org/2021.eacl-demos.7) (Ushio & Camacho-Collados, EACL 2021) [28] https://paperswithcode.com/paper/xlm-t-a-multilingual-language-model-toolkit [29] Kalyan KS, Rajasekharan A, Sangeetha S. AMMU: A survey of transformer-based biomedical pretrained language models. J Biomed Inform. 2022 Feb;126:103982. doi: 10.1016/j.jbi.2021.103982. Epub 2021 Dec 31. PMID: 34974190. [30] Yang F, Wang X, Ma H, Li J. Transformers-sklearn: a toolkit for medical language understanding with transformer-based models. BMC Med Inform Decis Mak. 2021 Jul 30;21(Suppl 2):90. doi: 10.1186/s12911-021-01459-0. PMID: 34330244; PMCID: PMC8323195. [31] https://www.kaggle.com/code/mangipudiprashanth/twitter-sentiment-analysis-using-ml-nlp

Copyright

Copyright © 2022 Aditi Ashish Gawande, S Karthikeyan, Sriram Balasubramanian, S Ajay. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.