Combining emojis with Arabic textual features for sentiment classification

.
Abstract:
With the significant growth of user-generated content on the Web, sentiment analysis has gained increasing importance to draw insights of online social data and turn it into a valuable asset for supporting decision making. Lots of efforts have adopted text mining techniques with linguistic features to detect and track people's opinions. Yet, the results are not satisfactory. In this paper, we aim at exploring the impact of combining emojis based features, which are pictographic symbols that are becoming more commonly used in social media, with various forms of textual features on the sentiment classification of dialectical Arabic tweets. We extract textual features using four different methods: Bag-of-Words (BoW), Latent Semantic Analysis, and two forms of Word Embedding. The effect of fusing emojis with textual features is analyzed using a support vector classifier with and without feature selection. It has been observed that simpler models can be constructed with much better results when emojis are merged with word embedding and the selection of the most relevant subset of features as input to the classifier.

.
https://ieeexplore.ieee.org/document/8355456

Emotion detection of tweets in Indonesian language using LDA and expression symbol conversion

.
Abstract:
Twitter is one of the social networks that attract many Indonesian people because it is considered as a medium to express opinions and feelings about certain topic. Twitter popularity can be used as an efficient source of sentiment data for marketing or social studies. Social studies that can be applied to the process of Twitter analysis is emotion detection. Emotion detection has a potency to be applied in a wide range of applications, ranging from health applications, counseling, business, to community population studies. This research utilizes one of the most popular and simplest topic modeling models, that is Latent Dirichlet Allocation (LDA), as well as conversion expression symbol (emoticon/ emoji), which shows the emotion or topic in a tweet to multiply the vocabulary that represents emotion. The advantage of the LDA method proposed is that it can detect some emotions on the tweet because the detection is not rigid and is able to show the proportion of emotion in the tweet. This research compares emotional detection using LDA and conversion expression symbol with emotional detection using LDA without conversion expression symbol. The result shows that emotional detection using LDA with conversion expression symbol is better with the reached average difference of accuracy 14.096%.
.
https://ieeexplore.ieee.org/document/8276371

Language-independent data set annotation for machine learning-based sentiment analysis

.
Abstract:
Social media platforms provide large amounts of user-generated text which can be utilized for text-based Sentiment Analysis in order to obtain insights about opinions on many aspects of life. Current approaches that are based on supervised learning require manually annotated data sets that are time-consuming to create and are specific to a single language. In this work, we present an approach to generate ground truth sentiment values for a data set from Twitter. We use a sentiment emoji lexicon and distribute known polarities of hashtags over neighbors in a graph we build on them. This approach is language-independent. Native speakers of five different languages evaluate the accuracy of the sentiment values assigned by our method on a corpus of Tweets. Our experiments show that the quality of our automatically assigned sentiment values is sufficiently high to be used for training of machine learning-based sentiment analysis.
.
https://ieeexplore.ieee.org/document/8122930

Automatic Construction of an Emoji Sentiment Lexicon

.
ABSTRACT
Emojis have been frequently used to express users' sentiments, emotions, and feelings in text-based communication. To facilitate sentiment analysis of users' posts, an emoji sentiment lexicon with positive, neutral, and negative scores has been recently constructed using manually labeled tweets. However, the number of emojis listed in the lexicon is smaller than that of currently existing emojis, and expanding the lexicon manually requires time and effort to reconstruct the labeled dataset. This paper presents a simple and efficient method for automatically constructing an emoji sentiment lexicon with arbitrary sentiment categories. The proposed method extracts sentiment words from WordNet-Affect and calculates the cooccurrence frequency between the sentiment words and each emoji. Based on the ratio of the number of occurrences of each emoji among the sentiment categories, each emoji is assigned a multidimensional vector whose elements indicate the strength of the corresponding sentiment. In experiments conducted on a collection of tweets, we show a high correlation between the conventional lexicon and our lexicon for three sentiment categories. We also show the results for a new lexicon constructed with additional sentiment categories.
.
https://dl.acm.org/doi/10.1145/3110025.3110139

The Effects of Emoji in Sentiment Analysis

.
Abstract: This study investigates the usage of Emoji characters on social networks and the effects of Emoji in text mining and sentiment analysis. As it provides live access to text based public opinions, we chose Twitter as our information source in our analysis. We collected text data for some global positive and negative events to analyze the impact of Emoji characters in sentiment analysis. In our analysis, we noticed that the utilization of Emoji characters in sentiment analysis results in higher sentiment scores.
Furthermore, we observed that the usage of Emoji characters in sentiment analysis appeared to have higherimpact on overall sentiments of the positive opinions in comparison to the negative opinions.

Key words: Emoji, opinion mining, sentiment analysis, twitter.
.

Multi-sentiment Modeling with Scalable Systematic Labeled Data Generation via Word2Vec Clustering

.
Abstract:
Social networks are now a primary source for news and opinions on topics ranging from sports to politics. Analyzing opinions with an associated sentiment is crucial to the success of any campaign (product, marketing, or political). However, there are two significant challenges that need to be overcome. First, social networks produce large volumes of data at high velocities. Using traditional (semi-) manual methods to gather training data is, therefore, impractical and expensive. Second, humans express more than two emotions, therefore, the typical binary good/bad or positive/negative classifiers are no longer sufficient to address the complex needs of the social marketing domain. This paper introduces a hugely scalable approach to gathering training data by using emojis as proxy for user sentiments. This paper also introduces a systematic Word2Vec based clustering method to generate emoji clusters that arguably represent different human emotions (multi-sentiment). Finally, this paper also introduces a threshold-based formulation to predicting one or two class labels (multi-label) for a given document. Our scalable multi-sentiment multi-label model produces a cross-validation accuracy of 71.55% (± 0.22%). To compare against other models in the literature, we also trained a binary (positive vs. negative) classifier. It produces a cross-validation accuracy of 84.95% (± 0.17%), which is arguably better than several results reported in literature thus far.
.
https://ieeexplore.ieee.org/document/7836770

Emotion Analysis Of Twitter Data That Use Emoticons And Emoji Ideograms

.

0. Abstract

Twitter is an online social networking service on which users worldwide publish their opinions on a variety of topics, discuss current issues, complain, and express many kinds of emotions. Therefore, Twitter is a rich source of data for opinion mining, sentiment and emotion analysis. This paper focuses on this issue by analysing symbols called emotion tokens, including emotion symbols (e.g. emoticons and emoji ideograms). According to observations, emotion tokens are commonly used in many tweets. They directly express one’s emotions regardless of his/her language, hence they have become a useful signal for sentiment analysis in multilingual tweets. The paper describes the approach to extending existing binary sentiment classification approaches using a multi-way emotions classification.

Keywords: Twitter, data mining, sentiment analysis, emotion analysis.1.

1. Introduction

Microblogging websites such as Twitter (www.twitter.com) have evolved to become a great source of various kinds of information. This is due to the nature of microblogs on which people post real-time messages regarding their opinions on a variety of topics, discuss current issues,complain, and express many kinds of emotions. As the audience of microblogging platforms and social networks grows every day, data from these sources can be used in opinion mining,sentiment and emotion analysis tasks. Opinions and related concepts such as sentiments and emotions are the subjects of study of sentiment analysis and opinion mining. The inception and rapid growth of the field coincide with those of the social media on the Web, e.g., reviews,forum discussions, blogs, microblogs, Twitter, and social networks. Most Natural LanguageProcessing (NLP) methods perform without particular success in the social media. Almost all forms of social media are very noisy and full of all kinds of spelling, grammatical, and punctuation errors.

Current sentiment analysis methods typically focus on the polarity of like/dislike emotions.Sometimes neutral emotions can be detected in between. Human emotions are far beyond these simple metrics and are much more diverse. This implies that such polarity analysis gives only limited information on the actual intent of the author of the message. Defining either positive or negative emotions only is relatively simple, yet defining a complete and clear set of emotions is much more difficult. Researchers have thus created a wide range of research tools on the identification of basic emotions.

2. Related Research

Sentiment analysis is a growing area of the Natural Language Processing task at many levels of granularity. Starting from being a document level classification task ([33], [24]), it has been handled at the sentence level ([16], [18]) and more recently at the phrase level ([34], [6]), or even at the polarity of words and phrases ([15], [12])
However, the informal and specialised language that is used in tweets as well as the nature of the microblogging domain make sentiment analysis in Twitter a very different task. With the growing number of blogs and social networks, opinion mining and sentiment analysis have become fields of interest to many researches. A very broad overview of the existing work was presented in [25]. J. Read, in [27], used emoticons such as “:-)” and “:- (” to form a training set for sentiment classification. For this purpose, the authors collected texts containing emoticons from Usenet newsgroups. The dataset was divided into “positive” (texts with happy emoticons) and “negative” (texts with sad or angry emoticons) samples.
Researchers have also begun to investigate various ways of automatically collecting training data. Several researchers have relied on emoticons to define training data ([23], [9]). Barbosa and Feng in [7] exploited existing Twitter sentiment sites to collect training data. Davidov, Tsur and Rappoport [11] also used hashtags to create training data but they limited their experiments to sentiment/non-sentiment classification rather than the multi-way emotion classification that is presented in this article.
Extending research to many different kinds of emotions is a very new concept and has not been extensively studied yet. There are currently few examples in which researchers have gone beyond the polarity of sentiment analysis. Socher et al. [30] predicted five different dimensions of sentiments. Tromp and Pechenizkiy [32] used Pluchnik’s wheel of emotions model and a rule-based approach for emotion detection in the text. A slightly different but also interesting approach was presented by Mohammad [20], who detected emotion on twitter posts by using emotion-word hashtags.

3. From Sentiment to Emotion Analysis

Sentiment analysis, which is also known as opinion mining, focuses on discovering patterns in the text that can be analysed to classify sentiment in that text. The term sentiment analysis probably first appeared in [21], and the term opinion mining first appeared in [10]. However, research on sentiments and opinions appeared earlier.

According to Liu “sentiment analysis is the field of study that analyses people’s opinions,sentiments, evaluations, appraisals, attitudes, and emotions towards entities such as products,services, organizations, and their attributes. It represents a large problem space. There are also many names and slightly different tasks, e.g., sentiment analysis, opinion mining, opinion extraction, sentiment mining, subjectivity analysis, affect analysis, emotion analysis, review mining, etc.” [19]. Sentiment analysis has grown to be one of the most active research fields in natural language processing. It is also widely studied in data mining, Web mining and text mining. In fact, it has spread from computer science to management sciences and social sciences due to its importance to business and society.

Sentiment analysis is predominantly implemented in software which can autonomously ex-tract emotions and opinions from a text. It has many real-world applications, e.g. it allows companies to analyse how their products or brand is being perceived by their consumers, and politicians may be interested in knowing how people plan to vote in elections, etc. It is difficult to classify sentiment analysis as one specific field of study as it incorporates many different areas, such as linguistics, Natural Language Processing, and Machine Learning or Artificial In-telligence. Since the majority of sentiment that is uploaded to the Internet is of an unstructured nature, it is a difficult task for computers to process it and to extract meaningful information from it. Some of the most effective machine learning algorithms, e.g. support vector machines,naïve Bayes and conditional random fields, often produce no human understandable results.

Emotions are closely related to sentiments. Emotions can be defined as subjective feel-ings and thoughts. People’s emotions have been categorised into distinct categories. Emotionan alysis can thus be done as an additional layer on top of the (relatively) simpler sentiment clas-sification. However, there is still no set of agreed basic emotions among researchers. Based on 26], people have six primary emotions, i.e., love, joy, surprise, anger, sadness, and fear, which can be sub-divided into many secondary and tertiary emotions. Each emotion can also have different intensities.

Emotions in virtual communication differ in a variety of ways from those in face-to-face interactions due to the characteristics of computer-mediated communication, which may lack many of the auditory and visual cues that are normally associated with the emotional aspects of interactions. While text-based communication eliminates audio and visual cues, there are other methods for adding emotion. Emoticons, or emotional icons, can be used to display various types of emotions. Ortony and Turner [22] collated a wide range of research on the identification of basic emotions (Table 1)
 

There are many common categories in the different research studies as presented in Table 1,but there are still too many very different emotions for effective analysis. Some concepts aimto minimise the number of basic emotions. Jack, Garrod and Schyns [17] analysed the 42 facialmuscles that shape emotions on the face and they came up with only four basic emotions, yetthere is still no consensus on the basic set of emotions that would be generally accepted andcould be objectively verified. For the purposes of this work, emotions can be classified intoemoticon types similar to those in Wikipedia [5] (Table 2):
 

4. Twitter Data Description

Twitter has its own conventions that renders it distinct from other textual data; Twitter messages are called tweets. Twitter also has its own conventions that renders it distinct from other textual data. There are some particular features that can be used to compose a tweet 1.The first pieces of information, UE Katowice and @UE_Katowice are the twitter name for the University of Economics in Katowice; #UEKatowice, #international, and #Erasmus are tags provided by the user for this message, the so-called hashtags. Users of Twitter use the “@” symbol to refer to other users. Referring to other users in this manner automatically alerts them.Users usually use hashtags to mark topics. This is primarily done to increase the visibility of their tweets. These symbols provide an easy way of identifying Twitter user names and topics and thus allows to search for and filter information on any subject. In the tweet the following emoticons, – :), and emoji characters — were used.

Twitter messages have many unique attributes, which differentiates twitter analysis from other fields of research. The first attribute is length. The maximum length of a Twitter message is 140 characters. The average length of a tweet is 14 words [14]. This is very different from the domains of other research studies, which were mostly focused on reviews which consisted of multiple sentences. The second attribute is the availability of data. With the Twitter API or other tools it is much easier to collect millions of tweets for training.

4.1. Emoticons

There are two fundamental data mining tasks that can be considered in conjunction with Twitter data, i.e. text analysis and symbol analysis. Due to the nature of this microblogging service(quick and short messages), people use acronyms, make spelling mistakes, and use emoticons and other characters that express special meanings. Emoticons constitute a meta communicative pictorial representation of a facial expression that is pictorially represented by using punctuation and letters or pictures; they express the user’s mood.The use of emoticons can be traced back to the 19th century. The first documented person to have used the emoticons :-) and :-( on the Internet was Scott Fahlman from Carnegie Mellon University in a message dated 19 September 1982 [1].Some emoticons as characters are included in the Unicode standard, three in the Miscellaneous Symbols block, and over sixty in the Emoticons block [4]. More symbols and meanings which can be used to determine emotional state can be found on the Wikipedia website [5]. Thetop 20 emoticons collected from 96 269 892 tweets is presented in [8].

4.2. Emoji ideograms

Emoji were originally used in Japanese electronic messages and spread outside of Japan. The characters are used much like emoticons, although a wider range is provided. The rise in the popularity of emoji is due to its being incorporated into sets of characters available in mo-bile phones. Apple in IOS, Android and other mobile operating systems included some emoji character sets. Emoji characters are also included in the Unicode standard [4]. Emoji can be categorised into similar categories as emoticons. Emoji can even be translated into English by using http://emojitranslate.com website.

5. Architecture of Emotion Classification System

The main problem is how to extract the rich information that is available on Twitter and how to use it to draw meaningful insight. To achieve this, at first stage of works an accurate analyser for tweets was build. Twitter allows developers to collect data via Twitter REST API [2] and The Streaming API [3]. Twitter has numerous regulations and rate limits imposed on its API, and for this reason it requires that all users must register an account and provide authentication details when they query the API. One of the best ways of connecting to Twitter Streaming API and downloading the data is by using Python and a library called Tweepy [28].

Twitter API exports data only in JSON format, which should be translated into readable for databases or an analytical software format. A combination of Twitter API, scripts for converting JSON to CSV [29], SAS Macro [13] or Excel Macro [31] was used to extract information from Twitter and to create an input dataset for the analysis. The entire process of data acquisition can be fully automated by scheduling the run of VBA or SAS macros.

Data can also be stored directly in JSON format in an NoSQL Database, which provides a mechanism for the storage and retrieval of data which is modelled in means other than the tabular relations used in relational databases. MongoDB (https://www.mongodb.org/)allows to store data in JSON-like documents with dynamic schemas (MongoDB calls the format BSON), thus making the integration of data in certain types of applications easier and faster.

Since opinions have targets, further preprocessing and filtering of collected data was done using @twitter_names and #hashtags as targets in the way described in [2]. This method is more precise and provides better results than other text mining approaches. Software used for data analysis can be SAS Text Miner, SAS Visual Analytics or other tools. SAS Visual Analytics allows for direct import of Twitter data, but in order to use SAS Text Miner and other tools, the data have to be downloaded and converted.

Instead of using commercial software, all sentiment and emotion analysis tasks was solved using python programming language. Python language has many machine learning and data mining extensions that are suitable for this kind of work. The challenge remains to fetch customised Tweets and to clean data before any text or symbol mining takes place.

A classification of tweets was made in unsupervised learning method by using the lexicon-based approach. The sentiment lexicon contains a list of emoticons and emoji ideograms based on Table 2. Data was gathered by searching Twitter posts using Twitter API. An assumption must be made in order to use this method, this assumption is that the emoticon in the tweet represents the overall emotion contained in that tweet. This assumption is quite reasonable as the maximum length of a tweet is 140 characters, so in the majority of cases the emoticon will correctly represent the overall sentiment of that tweet. This kind of evaluation is commonly known as the document-level sentiment classification because it considers the whole document as a basic information unit. The model can be developed on a sample of data; then can be used to classify the emotions of the tweet.

First verification of the method used here indicates that the most recognisable are only the basic emotions. The obtained results are similar to work [8], in which Nick Berry reveals that the top 20 emoticons accounted for 90% of all 96,269,892 Tweets. Therefore, for effective emotion analysis we could use only a limited subset of emoticons and recognised only basic, common emotions, such as:

This set of emoticons and the corresponding emotions covers nearly 90% of all occurrences.

However objective verification of emotions is very difficult. Further work should go in the direction of carrying out supervised check what part of emoticons does not reflect correctly emotions of tweet.

6. Conclusions

Microblogging such as on Twitter has today become one of the major types of communication.The large amount of information contained in these websites makes them an attractive source of data for opinion mining and sentiment analysis. Most text-based methods of analysis maynot always be useful for sentiment analysis in these domains. We as researchers still need novel ideas to make significant progress in this area. Using symbol analysis that makes use of emoticons and emoji characters can significantly increase precision in recognising many kinds of emotions. Also, applying Twitter names and hashtags to filter collected training data can provide better results. The most successful algorithms will probably be integration of natural language processing methods and symbol analysis
.

References 

1. :-) turns 25. https://web.archive.org/web/20071012051803/http://www.cnn.com/2007/TECH/09/18/emoticon.anniversary.ap/index.html, 09 2007.
2. Twitter rest api. the search api. https://dev.twitter.com/rest/public/search, 05 2015.
3. Twitter the streaming apis. https://dev.twitter.com/streaming/overview, 05 2015.
4. Unicode miscellaneous symbols. http://www.unicode.org/charts/PDF/U2600.pdf, 05 2015.
5. List of emoticons. https://en.wikipedia.org/wiki/List_of_emoticons, 04 2016.
6. Apoorv Agarwal, Fadi Biadsy, and Kathleen R. Mckeown. Contextual phrase-levelpolarity analysis using lexical affect scoring and syntactic n-grams. In Proceedingsof the 12th Conference of the European Chapter of the Association for ComputationalLinguistics, EACL ’09, pages 24–32, Stroudsburg, PA, USA, 2009. Association forComputational Linguistics.
7. Luciano Barbosa and Junlan Feng. Robust sentiment detection on twitter from biasedand noisy data. In Proceedings of the 23rd International Conference on ComputationalLinguistics: Posters, COLING ’10, pages 36–44, Stroudsburg, PA, USA, 2010. Associ-ation for Computational Linguistics.
8. N. Berry. Datagenetics. http://www.datagenetics.com/blog/october52012/index.html, 05 2015.
9. Albert Bifet and Eibe Frank. Sentiment knowledge discovery in twitter streaming data.In Proceedings of the 13th International Conference on Discovery Science, DS’10,pages 1–15, Berlin, Heidelberg, 2010. Springer-Verlag.
10. Kushal Dave, Steve Lawrence, and David M. Pennock. Mining the peanut gallery:Opinion extraction and semantic classification of product reviews. In Proceedings ofthe 12th International Conference on World Wide Web, WWW ’03, pages 519–528,New York, NY, USA, 2003. ACM.
11. Dmitry Davidov, Oren Tsur, and Ari Rappoport. Enhanced sentiment learning usingtwitter hashtags and smileys. In Proceedings of the 23rd International Conference onComputational Linguistics: Posters, COLING ’10, pages 241–249, Stroudsburg, PA,USA, 2010. Association for Computational Linguistics.
12. Andrea Esuli and Fabrizio Sebastiani. Sentiwordnet: A publicly available lexical re-source for opinion mining. In In Proceedings of the 5th Conference on Language Re-sources and Evaluation (LREC’06, pages 417–422, 2006.
13. S. Garla and G. Chakraborty. tweets. Paper 324-2011. Oklahoma State University,Stillwater, OK, 2001.
14. Alec Go, Richa Bhayani, and Lei Huang. Twitter sentiment classification using distant supervision. Processing, pages 1–6, 2009.
15. Vasileios Hatzivassiloglou and Kathleen R. McKeown. Predicting the semantic orien-tation of adjectives. In Proceedings of the 35th Annual Meeting of the Association forComputational Linguistics and Eighth Conference of the European Chapter of the As-sociation for Computational Linguistics, ACL ’98, pages 174–181, Stroudsburg, PA,USA, 1997. Association for Computational Linguistics.
16. Minqing Hu and Bing Liu. Mining and summarizing customer reviews. In Proceedingsof the Tenth ACM SIGKDD International Conference on Knowledge Discovery andData Mining, KDD ’04, pages 168–177, New York, NY, USA, 2004. ACM.
17. Rachael E. Jack, Oliver G.B. Garrod, and Philippe G. Schyns. Dynamic facial expres-sions of emotion transmit an evolving hierarchy of signals over time. Current Biology,24(2):187–192, 2014.
18. Soo-Min Kim and Eduard Hovy. Determining the sentiment of opinions. In Proceed-ings of the 20th International Conference on Computational Linguistics, COLING ’04,Stroudsburg, PA, USA, 2004. Association for Computational Linguistics.
19. Bing Liu. Sentiment analysis and opinion mining. Synthesis Lectures on Human Lan-guage Technologies, 5(1):1–167, 2012.
20. Saif Mohammad. #emotional tweets. In *SEM 2012: The First Joint Conference onLexical and Computational Semantics – Volume 1: Proceedings of the main conferenceand the shared task, and Volume 2: Proceedings of the Sixth International Workshopon Semantic Evaluation (SemEval 2012), pages 246–255, Montréal, Canada, 7-8 June2012. Association for Computational Linguistics.
21. Tetsuya Nasukawa and Jeonghee Yi. Sentiment analysis: Capturing favorability usingnatural language processing. In Proceedings of the 2Nd International Conference onKnowledge Capture, K-CAP ’03, pages 70–77, New York, NY, USA, 2003. ACM.
22. Andrew Ortony and Terence J Turner. What’s basic about basic emotions? Psycholog-ical review, 97(3):315, 1990.
23. Alexander Pak and Patrick Paroubek. Twitter as a corpus for sentiment analysis andopinion mining. In Nicoletta Calzolari (Conference Chair), Khalid Choukri, Bente Mae-gaard, Joseph Mariani, Jan Odijk, Stelios Piperidis, Mike Rosner, and Daniel Tapias,editors, Proceedings of the Seventh International Conference on Language Resourcesand Evaluation (LREC’10), Valletta, Malta, may 2010. European Language ResourcesAssociation (ELRA).
24. Bo Pang and Lillian Lee. A sentimental education: Sentiment analysis using subjectiv-ity summarization based on minimum cuts. In Proceedings of the 42Nd Annual Meetingon Association for Computational Linguistics, ACL ’04, Stroudsburg, PA, USA, 2004.Association for Computational Linguistics.
25. Bo Pang and Lillian Lee. Opinion mining and sentiment analysis. Found. Trends Inf.Retr., 2(1-2):1–135, January 2008.
26. W Parrott. Emotions in social psychology: Essential readings. Psychology Press, 2001.
27. Jonathon Read. Using emoticons to reduce dependency in machine learning techniquesfor sentiment classification. In Proceedings of the ACL Student Research Workshop,ACLstudent ’05, pages 43–48, Stroudsburg, PA, USA, 2005. Association for Computa-tional Linguistics.
28. Matthew A. Russell. Mining the Social Web, 2nd Edition Data Mining Facebook, Twit-ter, LinkedIn, Google+, GitHub, and More. O’Reilly Media, 2013.
29. Falko Shultz. How to import twitter tweets in sas data step using oauth 2 authenti-cation style. http://blogs.sas.com/content/sascom/2013/12/12/how-to-import-twitter- tweets-in-sas-data-step- using-oauth/-2-authentication-style/, 12 2013.
30. Richard Socher, Jeffrey Pennington, Eric H. Huang, Andrew Y. Ng, and Christopher D.Manning. Semi-supervised recursive autoencoders for predicting sentiment distribu-tions. In Proceedings of the Conference on Empirical Methods in Natural LanguageProcessing, EMNLP ’11, pages 151–161, Stroudsburg, PA, USA, 2011. Associationfor Computational Linguistics.
31. Analytics Tools. Can sas be used to gather sentiments fromtwitter? http://www.analytics-tools.com/2012/06/social-media-analytics-twitter- text.html, 07 2012.
32. Erik Tromp and Mykola Pechenizkiy. Pattern-based emotion classification on socialmedia. In Advances in Social Media Analysis, volume 602, pages 1–20. Springer, 2015.
33. Peter D. Turney. Thumbs up or thumbs down?: Semantic orientation applied to unsupervised classification of reviews. In Proceedings of the 40th Annual Meeting onAssociation for Computational Linguistics, ACL ’02, pages 417–424, Stroudsburg, PA,USA, 2002. Association for Computational Linguistics.
34. Theresa Wilson, Janyce Wiebe, and Paul Hoffmann. Recognizing contextual polarity inphrase-level sentiment analysis. In Proceedings of the Conference on Human LanguageTechnology and Empirical Methods in Natural Language Processing, HLT ’05, pages347–354, Stroudsburg, PA, USA, 2005. Association for Computational Linguistics.

.
https://www.researchgate.net/publication/308413240_TWITTER_SENTIMENT_ANALYSIS_USING_EMOTICONS_AND_EMOJI_IDEOGRAMS

.

Sentiment of Emojis

.
Abstract
There is a new generation of emoticons, called emojis, that is increasingly being used in mobile communications and social media. In the past two years, over ten billion emojis were used on Twitter. Emojis are Unicode graphic symbols, used as a shorthand to express concepts and ideas. In contrast to the small number of well-known emoticons that carry clear emotional contents, there are hundreds of emojis. But what are their emotional contents? We provide the first emoji sentiment lexicon, called the Emoji Sentiment Ranking, and draw a sentiment map of the 751 most frequently used emojis. The sentiment of the emojis is computed from the sentiment of the tweets in which they occur. We engaged 83 human annotators to label over 1.6 million tweets in 13 European languages by the sentiment polarity (negative, neutral, or positive). About 4% of the annotated tweets contain emojis. The sentiment analysis of the emojis allows us to draw several interesting conclusions. It turns out that most of the emojis are positive, especially the most popular ones. The sentiment distribution of the tweets with and without emojis is significantly different. The inter-annotator agreement on the tweets with emojis is higher. Emojis tend to occur at the end of the tweets, and their sentiment polarity increases with the distance. We observe no significant differences in the emoji rankings between the 13 languages and the Emoji Sentiment Ranking. Consequently, we propose our Emoji Sentiment Ranking as a European language-independent resource for automated sentiment analysis. Finally, the paper provides a formalization of sentiment and a novel visualization in the form of a sentiment bar.
.

Query-by-Emoji Video Search

.
ABSTRACT
This technical demo presents Emoji2Video, a query-by-emoji interface for exploring video collections. Ideogram-based video search and representation presents an opportunity for an intuitive, visual interface and concise non-textual summary of video contents, in a form factor that is ideal for small screens. The demo allows users to build search strings comprised of ideograms which are used to query a large dataset of YouTube videos. The system returns a list of the top-ranking videos for the user query along with an emoji summary of the video contents so that users may make an informed decision whether to view a video or refine their search terms. The ranking of the videos is done in a zero-shot, multi-modal manner that employs an embedding space to exploit semantic relationships between user-selected ideograms and the video's visual and textual content.
.
https://dl.acm.org/doi/10.1145/2733373.2807961

Microblog Sentiment Analysis with Emoticon Space Model

.
Abstract
Emoticons have been widely employed to express different types of moods, emotions, and feelings in microblog environments. They are therefore regarded as one of the most important signals for microblog sentiment analysis. Most existing studies use several emoticons that convey clear emotional meanings as noisy sentiment labels or similar sentiment indicators. However, in practical microblog environments, tens or even hundreds of emoticons are frequently adopted and all emoticons have their own unique emotional meanings. Besides, a considerable number of emoticons do not have clear emotional meanings. An improved sentiment analysis model should not overlook these phenomena. Instead of manually assigning sentiment labels to several emoticons that convey relatively clear meanings, we propose the emoticon space model (ESM) that leverages more emoticons to construct word representations from a massive amount of unlabeled data. By projecting words and microblog posts into an emoticon space, the proposed model helps identify subjectivity, polarity, and emotion in microblog environments. The experimental results for a public microblog benchmark corpus (NLP&CC 2013) indicate that ESM effectively leverages emoticon signals and outperforms previous state-of-the-art strategies and benchmark best runs.
.
https://link.springer.com/article/10.1007%2Fs11390-015-1587-1

Exploiting emoticons in sentiment analysis

.
ABSTRACT
As people increasingly use emoticons in text in order to express, stress, or disambiguate their sentiment, it is crucial for automated sentiment analysis tools to correctly account for such graphical cues for sentiment. We analyze how emoticons typically convey sentiment and demonstrate how we can exploit this by using a novel, manually created emoticon sentiment lexicon in order to improve a state-of-the-art lexicon-based sentiment classification method. We evaluate our approach on 2,080 Dutch tweets and forum messages, which all contain emoticons and have been manually annotated for sentiment. On this corpus, paragraph-level accounting for sentiment implied by emoticons significantly improves sentiment classification accuracy. This indicates that whenever emoticons are used, their associated sentiment dominates the sentiment conveyed by textual cues and forms a good proxy for intended sentiment.
.
https://dl.acm.org/doi/10.1145/2480362.2480498