Hugging Face's logo

Datasets:
WINGNUS
/
ACL-OCL

Tasks:
Token Classification
Languages:
English
Size:
10K<n<100K
Tags:
research papers
acl
License:
Dataset card Files Community
5
ACL-OCL / Base_JSON /prefixR /json /R15 /R15-1036.json
Benjamin Aw
Add updated pkl file v3
6fa4bc9
raw
history blame contribute delete
129 kB
 {
"paper_id": "R15-1036",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T14:57:44.176349Z"
},
"title": "Fine-Grained Sentiment Analysis for Movie Reviews in Bulgarian",
"authors": [
{
"first": "Borislav",
"middle": [],
"last": "Kapukaranov",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Sofia University",
"location": {
"country": "Bulgaria"
}
},
"email": "[email protected]"
},
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": "",
"affiliation": {
"laboratory": "",
"institution": "Qatar Computing Research Institute HBKU",
"location": {
"country": "Qatar"
}
},
"email": "[email protected]"
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "We present a system for fine-grained sentiment analysis in Bulgarian movie reviews. As this is pioneering work for this combination of language and sentiment granularity, we create suitable, freely available resources: a dataset of movie reviews with fine-grained scores, and a sentiment polarity lexicon. We further compare experimentally the performance of classification, regression and ordinal regression in a 3-way, 5-way and 11-way classification setups, using as features not only the text from the reviews, but also contextual information in the form of metadata, e.g., movie length, director, actors, genre, country, and various scores: IMDB, Cinexio, and user-average. The results show that adding contextual information yields strong performance gains.",
"pdf_parse": {
"paper_id": "R15-1036",
"_pdf_hash": "",
"abstract": [
{
"text": "We present a system for fine-grained sentiment analysis in Bulgarian movie reviews. As this is pioneering work for this combination of language and sentiment granularity, we create suitable, freely available resources: a dataset of movie reviews with fine-grained scores, and a sentiment polarity lexicon. We further compare experimentally the performance of classification, regression and ordinal regression in a 3-way, 5-way and 11-way classification setups, using as features not only the text from the reviews, but also contextual information in the form of metadata, e.g., movie length, director, actors, genre, country, and various scores: IMDB, Cinexio, and user-average. The results show that adding contextual information yields strong performance gains.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "With the recent explosion in the popularity of Web forums and social media, sentiment analysis has emerged as a hot research topic. As sentiment-annotated data became readily available, researchers tapped into it and started developing various models for sentiment polarity prediction. Nowadays, there are many applications for sentiment analysis, e.g., businesses getting automatically classified feedback from customers, automated review scoring in retail Web sites, exploration of positive and negative trends, etc.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Movie reviews are a popular and widely available source of sentiment-annotated data. Unlike reviews produced by critics, those contributed by users are typically short and serve primarily to provide brief justification of a user's rating. An important characteristic of movie reviews compared to other sentiment sources is that they are commonly scored on a 5-star scale. This is very different from sentiment analysis on Twitter, where three-way sentiment classification schemes (positive, negative, neutral) have been preferred, e.g., at SemEval 2013-2015 (Nakov et al., 2013; Rosenthal et al., 2014; . In contrast, the star system makes the task more fine-grained, thus allowing to capture user opinion better.",
"cite_spans": [
{
"start": 558,
"end": 578,
"text": "(Nakov et al., 2013;",
"ref_id": "BIBREF20"
},
{
"start": 579,
"end": 602,
"text": "Rosenthal et al., 2014;",
"ref_id": "BIBREF32"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Here, we present experiments in predicting fine-grained stars, including halves, for Bulgarian movie reviews. This is a challenging task, that can be seen as (a) multi-way classification, i.e., choosing one out of eleven classes, (b) regression, i.e., predicting a real number, or (c) something in between, namely ordinal regression, i.e., predicting eleven values, but taking ordering into account, e.g., predicting 4 when the actual value is 3.5 would be better than predicting 1.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "While sentiment classification in movie reviews has been extensively studied for English (movie reviews datasets were among the earliest to use for this task), it has not been tried for Bulgarian so far. Moreover, most research has focused on positive/negative/neutral classification and finergrained schemes have been less popular (as they are harder). Even when used, the focus has typically been on having just five categories, not allowing halves. Thus, our contributions in this paper can be summarised as follows:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "\u2022 We create a new dataset for movies in Bulgarian, 1 where each review is associated with an 11-scale star rating: 0, 0.5, 1, ..., 4.5, 5.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "\u2022 We prepare a new sentiment lexicon for Bulgarian, which is also freely available.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "\u2022 Most importantly, we present the first work for Bulgarian on predicting fine-grained sentiment.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "The remainder of this paper is organized as follows: Section 2 introduces related work, Section 3 describes the dataset and teh lexicon we prepared, Section 4 presents the features we experiment with, Section 5 describes our experiments, and Section 6 discusses the results. Finally, Section 7 concludes and points to some possible directions for future work. Pang et al. (2002) were the first to look into text classification not in terms of topics, but focusing on how sentiment polarity is distributed in a document. They tried several machine learning algorithms on an English movie reviews dataset, and evaluated the performance of basic features such as n-grams and part of speech (POS) tags.",
"cite_spans": [
{
"start": 360,
"end": 378,
"text": "Pang et al. (2002)",
"ref_id": "BIBREF27"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Introduction",
"sec_num": "1"
},
{
"text": "Movie reviews were one of the first research domains for sentiment analysis as they (i) have the properties of a short message, and (ii) are already manually annotated by the author, as the score generally reflects sentiment polarity. Popular features for score/sentiment prediction include POS tags, word n-grams, word lemmata, and various context features based on the distance from a topic word. The challenge with movie reviews is that only some of the words are relevant for sentiment analysis. In fact, often the review is just a short narrative of the movie plot. One way to approach the problem is to use a subjectivity classifier (Pang and Lee, 2004) , which can be used to filter out objective sentences from the reviews, thus allowing the classifier then to focus on the subjective sentences only.",
"cite_spans": [
{
"start": 639,
"end": 659,
"text": "(Pang and Lee, 2004)",
"ref_id": "BIBREF24"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Early researchers realized the importance of external sentiment lexicons, e.g., Turney (2002) proposed an unsupervised approach to learn the sentiment orientation of words/phrases: positive vs. negative. Later work looked into the linguistic aspects of how opinions, evaluations, and speculations are expressed in text (Wiebe et al., 2004) , into the role of context for determining the sentiment orientation (Wilson et al., 2005) , of deeper linguistic processing such as negation handling (Pang and Lee, 2008) , of finer-grained sentiment distinctions (Pang and Lee, 2005) , of positional information (Raychev and Nakov, 2009) , etc. Moreover, it was recognized that in many cases, it was crucial to know not just the sentiment, but also the topic towards which this sentiment was expressed (Stoyanov and Cardie, 2008) .",
"cite_spans": [
{
"start": 80,
"end": 93,
"text": "Turney (2002)",
"ref_id": "BIBREF38"
},
{
"start": 319,
"end": 339,
"text": "(Wiebe et al., 2004)",
"ref_id": "BIBREF41"
},
{
"start": 409,
"end": 430,
"text": "(Wilson et al., 2005)",
"ref_id": "BIBREF42"
},
{
"start": 491,
"end": 511,
"text": "(Pang and Lee, 2008)",
"ref_id": "BIBREF26"
},
{
"start": 554,
"end": 574,
"text": "(Pang and Lee, 2005)",
"ref_id": "BIBREF25"
},
{
"start": 603,
"end": 628,
"text": "(Raychev and Nakov, 2009)",
"ref_id": "BIBREF30"
},
{
"start": 793,
"end": 820,
"text": "(Stoyanov and Cardie, 2008)",
"ref_id": "BIBREF37"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Fine-grained sentiment analysis tries to predict sentiment in a text using a finer scale, e.g., 5-stars; Pang and Lee (2005) pioneered this sub-field. In their work, they looked at the problem from two perspectives: as one vs. all classification, and as a regression by putting the 5-star ratings on a metric scale. An interesting observation in their research is that humans are not very good at doing such kinds of highly granular judgments and are often off the target mark by a full star.",
"cite_spans": [
{
"start": 105,
"end": 124,
"text": "Pang and Lee (2005)",
"ref_id": "BIBREF25"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Naturally, most research in sentiment analysis was done for English, and very little efforts were devoted to other languages. We are not aware of other work on fine-grained sentiment analysis for Bulgarian. There is work on sentiment analysis by Bulgarian scolars (Raychev, 2009; Raychev and Nakov, 2009; Kraychev and Koychev, 2012; Kraychev, 2014) .",
"cite_spans": [
{
"start": 264,
"end": 279,
"text": "(Raychev, 2009;",
"ref_id": "BIBREF31"
},
{
"start": 280,
"end": 304,
"text": "Raychev and Nakov, 2009;",
"ref_id": "BIBREF30"
},
{
"start": 305,
"end": 332,
"text": "Kraychev and Koychev, 2012;",
"ref_id": "BIBREF13"
},
{
"start": 333,
"end": 348,
"text": "Kraychev, 2014)",
"ref_id": "BIBREF14"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "We are aware of three publications for the closely-related Macedonian language, 2 which is mutually intelligible with Bulgarian. Gajduk and Kocarev (2014) experimented with 800 posts from the Kajgana forum (260 positive, 260 negative, and 280 objective), using Support Vector Machines (SVM) and Na\u00efve Bayes classifiers, and features such as bag of words, rules for negation, and stemming.",
"cite_spans": [
{
"start": 129,
"end": 154,
"text": "Gajduk and Kocarev (2014)",
"ref_id": "BIBREF6"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "More closely related to our work, Uzunova and Kulakov (2015) experimented with 400 movie reviews (200 positive + 200 negative), and a Na\u00efve Bayes classifier, using a small manually annotated sentiment lexicon of unknown size, and various preprocessing techniques such as negation handling and spelling/character translation.",
"cite_spans": [
{
"start": 34,
"end": 60,
"text": "Uzunova and Kulakov (2015)",
"ref_id": "BIBREF39"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Finally, Jovanoski et al. (2015) presented work on sentiment analysis of Macedonian tweets (8,583 for training + 1,139 for testing) using a 3-way tweet-level sentiment polarity classification scheme: positive, negative, and neutral/objective. They used standard features but variety of preprocessing steps, including morphological processing and POS tagging for Macedonian, negation handling, text standardization, tweet-specific processing, etc. More imporantly, they made use of several lexicons, some translated from other languages, 3 which they augmented with bootstrapping, ultimately achieving results that are on par with the state of the art for English.",
"cite_spans": [
{
"start": 9,
"end": 32,
"text": "Jovanoski et al. (2015)",
"ref_id": "BIBREF12"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Given the lack of previously developed datasets or sentiment polarity lexicons for Bulgarian, we had to create them ourselves. In addition to preparing a dataset of annotated movies, we further focused on building a sentiment polarity lexicon for Bulgarian. This is because lexicons are crucial for sentiment analysis. Since the very beginning, researchers have realized that sentiment analysis was quite different from standard document classification (Sebastiani, 2002) , e.g., into categories such as business, sport, and politics, and that sentiment analysis crucially needed external knowledge in the form of suitable sentiment polarity lexicons. For further detail, see the surveys by Pang and Lee (2008) and Liu and Zhang (2012) .",
"cite_spans": [
{
"start": 453,
"end": 471,
"text": "(Sebastiani, 2002)",
"ref_id": "BIBREF34"
},
{
"start": 691,
"end": 710,
"text": "Pang and Lee (2008)",
"ref_id": "BIBREF26"
},
{
"start": 715,
"end": 735,
"text": "Liu and Zhang (2012)",
"ref_id": "BIBREF15"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Until recently, such sentiment polarity lexicons were manually crafted, and were thus of small to moderate size, e.g., LIWC (Pennebaker et al., 2001) , General Inquirer (Stone et al., 1966) , Bing Liu's lexicon (Hu and Liu, 2004) , and MPQA (Wilson et al., 2005) , all have 2000-8000 words. Early efforts in building them automatically also yielded lexicons of moderate sizes (Esuli and Sebastiani, 2006; Baccianella et al., 2010) .",
"cite_spans": [
{
"start": 124,
"end": 149,
"text": "(Pennebaker et al., 2001)",
"ref_id": "BIBREF28"
},
{
"start": 169,
"end": 189,
"text": "(Stone et al., 1966)",
"ref_id": "BIBREF36"
},
{
"start": 211,
"end": 229,
"text": "(Hu and Liu, 2004)",
"ref_id": "BIBREF10"
},
{
"start": 241,
"end": 262,
"text": "(Wilson et al., 2005)",
"ref_id": "BIBREF42"
},
{
"start": 376,
"end": 404,
"text": "(Esuli and Sebastiani, 2006;",
"ref_id": "BIBREF5"
},
{
"start": 405,
"end": 430,
"text": "Baccianella et al., 2010)",
"ref_id": "BIBREF1"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "However, recent results have shown that automatically extracted large-scale lexicons (e.g., up to a million words and phrases) offer important performance advantages, as confirmed at shared tasks on Sentiment Analysis on Twitter at Se-mEval 2013-2015 (Nakov et al., 2013; Rosenthal et al., 2014; . These lexicons were crucial for the top-performing teams in the competition in all three years.",
"cite_spans": [
{
"start": 251,
"end": 271,
"text": "(Nakov et al., 2013;",
"ref_id": "BIBREF20"
},
{
"start": 272,
"end": 295,
"text": "Rosenthal et al., 2014;",
"ref_id": "BIBREF32"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Similar observations were made in the Aspect-Based Sentiment Analysis task at SemEval 2014-2015 (Pontiki et al., 2014) . In both tasks, the winning systems benefited from building and using massive sentiment polarity lexicons (Mohammad et al., 2013; Zhu et al., 2014) .",
"cite_spans": [
{
"start": 96,
"end": 118,
"text": "(Pontiki et al., 2014)",
"ref_id": "BIBREF29"
},
{
"start": 226,
"end": 249,
"text": "(Mohammad et al., 2013;",
"ref_id": "BIBREF19"
},
{
"start": 250,
"end": 267,
"text": "Zhu et al., 2014)",
"ref_id": "BIBREF43"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Related Work",
"sec_num": "2"
},
{
"text": "Our dataset consist of 347 movies with a total of 10,198 Bulgarian reviews, which we crawled from the ticket-booking website Cinexio. 4 We chose only movies for which scored reviews in Bulgarian were present on the website. For each movie, we include a set of user reviews, each annotated with a score on an 11-point scale: 0, 0.5, 1, ..., 4.5, 5 stars. More detailed statistics about our movie reviews dataset can be found in Figure 1 shows a distribution of the user ratings in our movie reviews dataset. We can see that the distribution is generally skewed towards full scores, while scores with halves are much less frequent: people seem to prefer a 5-point scale, and would not take full advantage of an 11-point one. Moreover, the distribution is also skewed towards high scores, and quite heavily towards a 5-star rating in particular.",
"cite_spans": [
{
"start": 134,
"end": 135,
"text": "4",
"ref_id": null
}
],
"ref_spans": [
{
"start": 427,
"end": 435,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "In addition to the movie reviews dataset, we further automatically generated a sentiment polarity lexicon for Bulgarian, using that dataset and pointwise mutual information (PMI) with respect to the positive and to the negative class, following the idea presented in (Turney, 2002) :",
"cite_spans": [
{
"start": 267,
"end": 281,
"text": "(Turney, 2002)",
"ref_id": "BIBREF38"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "pmi (w , class) = log p(w & class) p(w) p(class)",
"eq_num": "(1)"
}
],
"section": "Data",
"sec_num": "3"
},
{
"text": "Then, we calculated a sentiment polarity score:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "EQUATION",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [
{
"start": 0,
"end": 8,
"text": "EQUATION",
"ref_id": "EQREF",
"raw_str": "polarity = pmi(w, pos) \u2212 pmi(w, neg)",
"eq_num": "(2)"
}
],
"section": "Data",
"sec_num": "3"
},
{
"text": "Words with high positive/negative polarity were included in our sentiment polarity lexicon; this included 5,016 positive and 2,415 negative words.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "Here, we list some examples of movie reviews (with English translations in footnotes):",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "\u2022 \"\u0434\u043e\u0431\u0440\u0435, \u0447\u0435 \u0433\u043e \u0431\u0430\u0441\u0442\u0438\u0441\u0430\u0445\u0430 \u043d\u0430\u043a\u0440\u0430\u044f, \u0447\u0435 \u0434\u0430 \u0441\u0435 \u0441\u043f\u0440\u0435 \u0441 \u0442\u0430\u044f \u043f\u0440\u043e\u043f\u0430\u0433\u0430\u043d\u0434\u0430.. ;)\" 5 with score 3.5",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "This example is very interesting as it expresses cheerful and light mood, as indicated by the winking emoticon. Yet, it also mentions propaganda, which hints slight irritation about the way the movie plot developed.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "\u2022 \"\u041c\u043d\u043e\u0433\u043e \u0434\u043e\u0431\u044a\u0440 \u0444\u0438\u043b\u043c\" 6 with score 5.0",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "Nothing really surprising here: typical positive comment, without going into specifics.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "\u2022 \"\u041c\u043d\u043e\u0433\u043e \u0434\u043e\u0431\u044a\u0440 \u0444\u0438\u043b\u043c. \u041f\u0440\u043e\u0441\u0442\u043e \u043d\u0438 \u0435 \u0434\u0430\u043b\u0435\u0447\u0435\u043d \u0410\u043c\u0435\u0440\u0438\u043a\u0430\u043d\u0441\u043a\u0438\u044f \u043f\u0430\u0442\u0440\u0438\u043e\u0442\u0438\u0437\u044a\u043c.\" 7 with score 4.0",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "Here, despite having the same text as the max-scored previous example, the review author has given a slightly lower score. From the text alone, we can conclude that the author likes the movie very much, but still makes a general remark on how the movie will be accepted culturally by Bulgarian viewers.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "\u2022 \"\u041d\u0435 \u043c\u0435 \u0432\u043f\u0435\u0447\u0430\u0442\u043b\u0438.\" 8 with score 3.0",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "A typical mid-score comment: direct, clear.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "\u2022 \"\u0414\u043e\u0441\u0442\u0430 \u0442\u044a\u0436\u043d\u043e :(\" 9 with score 5.0",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "This is a perfect example of why this problem is hard. The text alone shows clearly negative emotions both indicated by text and by the crying emoticon, but it seems that the author actually liked the movie very much and gave it a maximum score.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "Negative reviews from other movies:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "\u2022 \"\u0434\u043e\u0441\u0442\u0430 \u0434\u044a\u043b\u044a\u0433 \u0438 \u043f\u043e\u043d\u0435 \u0434\u0430 \u0441\u0435 \u0441\u043b\u0443\u0447\u0432\u0430\u0448\u0435 \u043d\u0435\u0449\u043e...\" 10 with 1.5 score",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "On the low end of the scale, the scores become highly subjective, and often the same wording can be annotated with a full star difference in the score.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "\u2022 \"\u0424\u0438\u043b\u043c\u0430 \u0435 \u0441\u0430\u043c\u043e \u0447\u0430\u0441\u0442 \u043e\u0442 \u0442\u0440\u0438\u043b\u043e\u0433\u0438\u044f, \u043f\u043e\u043c\u043d\u0435\u0442\u0435 \u0447\u0435 \u0438\u0441\u0442\u043e\u0440\u0438\u044f\u0442\u0430 \u0441\u0432\u044a\u0440\u0448\u0432\u0430 \u043d\u0430\u0439\u0438\u043d\u0442\u0435\u0440\u0435\u0441\u043d\u043e\u0442\u043e\" 11 with score 2.0 This is another confusing example. Did the author actually like the movie? Or was s/he affected by somebody else's opinion?",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "\u2022 \"\u0414\u043e\u0441\u0442\u0430 \u043f\u043e\u0432\u0435\u0447\u0435 \u0435\u043a\u0448\u044a\u043d \u043e\u0442 \u043f\u044a\u0440\u0432\u0430\u0442\u0430 \u0447\u0430\u0441\u0442\" 12 with score 3.0",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "This is a great example showing that the perception of movies in a multipart series is influenced by earlier parts. It is not clear what people are scoring: the entire series or just the current (latest) part of the movie? People naturally try to compare with earlier series, which influences their scores.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "In general, scores could be heavily biased, and also relative: if one has recently watched a bad movie, the following movie, even if just slightly better, could get an inflated score.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Data",
"sec_num": "3"
},
{
"text": "In this section, we describe the features we experimented with: textual and contextual.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Features",
"sec_num": "4"
},
{
"text": "We used the following textual features:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Textual Features",
"sec_num": "4.1"
},
{
"text": "\u2022 words: binary feature for each word;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Textual Features",
"sec_num": "4.1"
},
{
"text": "\u2022 emoticons: binary feature for each positive/negative emoticon;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Textual Features",
"sec_num": "4.1"
},
{
"text": "\u2022 n-grams: binary feature for each n-gram (we only used bigrams).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Textual Features",
"sec_num": "4.1"
},
{
"text": "\u2022 lexicon: We further included two features based on our automatically generated movie reviews lexicon. They represent the positive and the negative overall score of the movie review, obtained by aggregating the lexicon scores of each word in the review text.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Textual Features",
"sec_num": "4.1"
},
{
"text": "Note that our dataset lacks enough relevant instances to use features such as all-caps and punctuation, and thus we did not use them here.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Textual Features",
"sec_num": "4.1"
},
{
"text": "Moreover, we found that using bigram features did not make much difference for this particular dataset, therefore the final feature set for the baseline system only used bag of words, emoticons, and the lexicon features.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Textual Features",
"sec_num": "4.1"
},
{
"text": "In addition to the above textual features, we further added some contextual (metadata) features:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Contextual Features",
"sec_num": "4.2"
},
{
"text": "\u2022 movie length: numeric feature indicating the run-length of the movie;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Contextual Features",
"sec_num": "4.2"
},
{
"text": "\u2022 country: binary feature indicating the country the movie comes from;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Contextual Features",
"sec_num": "4.2"
},
{
"text": "\u2022 genres: indicator feature for each genre;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Contextual Features",
"sec_num": "4.2"
},
{
"text": "\u2022 actors: indicator feature for each actor;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Contextual Features",
"sec_num": "4.2"
},
{
"text": "\u2022 director: indicator feature for each director;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Contextual Features",
"sec_num": "4.2"
},
{
"text": "\u2022 average user rating: numeric feature with the user's average movie review score;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Contextual Features",
"sec_num": "4.2"
},
{
"text": "\u2022 IMDB score: numeric feature, current average score for this movie in IMDB;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Contextual Features",
"sec_num": "4.2"
},
{
"text": "\u2022 Cinexio score: numeric feature, current average score for this movie in Cinexio.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Contextual Features",
"sec_num": "4.2"
},
{
"text": "Below we describe the class granularities we experimented with, the learning algorithms we used, and the evaluation results.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Experiments and Evaluation",
"sec_num": "5"
},
{
"text": "In the original formulation, we have eleven classes: 0, 0.5, 1, ..., 4.5, 5. For model comparison purposes, we further experimented with aggregated classes. Thus, we ended up with three class inventories of various sizes:",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Class Granularity",
"sec_num": "5.1"
},
{
"text": "\u2022 11-way: includes all labels, both integer and half-star;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Class Granularity",
"sec_num": "5.1"
},
{
"text": "\u2022 5-way: includes only the full stars;",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Class Granularity",
"sec_num": "5.1"
},
{
"text": "\u2022 3-way: divides the scores into three classes, positive \u2265 3.5 > neutral \u2265 2 > negative.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Class Granularity",
"sec_num": "5.1"
},
{
"text": "We performed experiments with three machine learning approaches: (i) classification, (ii) regression, and (iii) ordinal regression. We evaluated using a 5-fold cross validation. For scoring, we used the same metric for all class inventories and for all learning approaches, namely Mean Squared Error (MSE), which is standard for a task asking to predict ordinal values as in our case.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Learning Algorthms",
"sec_num": "5.2"
},
{
"text": "Classification. For classification, we used SVM with a linear kernel and L2-regularized L2loss, as implemented in LibSVM (Chang and Lin, 2011) . We used a one vs. all model, which we applied for each class inventory size: 3, 5, 11.",
"cite_spans": [
{
"start": 121,
"end": 142,
"text": "(Chang and Lin, 2011)",
"ref_id": "BIBREF3"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Learning Algorthms",
"sec_num": "5.2"
},
{
"text": "Regression. For regression, we used the same SVM tool and the same features and parameters as for classification, but we predicted a numerical value; this is known as support vector regression (Smola and Sch\u00f6lkopf, 2004) Ordinal Regression. For this scenario, we used ordinal logistic regression. This model is also known as proportional odds and was introduced by McCullagh (1980). 13 The use of ordinal regression for sentiment analysis, is not very common, mostly because the ordinal formulation of the task is not very common, even though it was used by some researchers (Pang and Lee, 2005; Goldberg and Zhu, 2006; Baccianella et al., 2009 ). Yet, it makes a lot of sense to use it as it tries to fit the data into thresholded regions as a classification task would do, and at the same time tries to predict values with an established order and position in the label space. This makes it interesting especially in the 5-class setup, where we have a small number of labels and there is ordering between them.",
"cite_spans": [
{
"start": 193,
"end": 220,
"text": "(Smola and Sch\u00f6lkopf, 2004)",
"ref_id": "BIBREF35"
},
{
"start": 383,
"end": 385,
"text": "13",
"ref_id": null
},
{
"start": 575,
"end": 595,
"text": "(Pang and Lee, 2005;",
"ref_id": "BIBREF25"
},
{
"start": 596,
"end": 619,
"text": "Goldberg and Zhu, 2006;",
"ref_id": "BIBREF9"
},
{
"start": 620,
"end": 644,
"text": "Baccianella et al., 2009",
"ref_id": "BIBREF0"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Learning Algorthms",
"sec_num": "5.2"
},
{
"text": "Our preliminary cross-validation experiments have shown that not all features that we have introduced above were really relevant; thus, we created a selected set of highly-relevant features: words, emoticons, lexicons, Cinexio score, and average user rating. We used this feature set when comparing the three machine learning algorithms (classification, regression, and ordinal regression), for the three class sizes (3, 5, and 11). The results are shown in Table 2 . Table 3 : Impact of individual contextual features when added to the baseline. Shown is MSE for the regression model with 11 classes.",
"cite_spans": [],
"ref_spans": [
{
"start": 458,
"end": 465,
"text": "Table 2",
"ref_id": "TABREF2"
},
{
"start": 468,
"end": 475,
"text": "Table 3",
"ref_id": null
}
],
"eq_spans": [],
"section": "Results",
"sec_num": "5.3"
},
{
"text": "We can see in Table 2 that the best results are achieved for regression, where the mean squared error is within half a point away for the 11-way and the 5-way class inventories, and it is about four times lower for the 3-way one. The secondbest performing machine learning approach is classification; its performance is very close to that of regression on the 3-way class inventory, but the gap widens with 5 classes (about 50% difference), and becomes huge with 11 classes (100% difference), where the predictions are on average a full point off from the target. Finally comes the worstperforming approach, ordinal regression, which consistently performs about four times worse than the standard regression.",
"cite_spans": [],
"ref_spans": [
{
"start": 14,
"end": 21,
"text": "Table 2",
"ref_id": "TABREF2"
}
],
"eq_spans": [],
"section": "Results",
"sec_num": "5.3"
},
{
"text": "Interestingly, while classification performs badly compared to regression on the 11-way class inventory, it quickly catches up for smaller numbers of classes, and the two learning approaches get quite close on the 3-way class inventory. This is expected, as classification usually struggles with too many class labels, especially in the case of uneven class distribution, and this is indeed our case, as we have seen in Figure 1 .",
"cite_spans": [],
"ref_spans": [
{
"start": 420,
"end": 428,
"text": "Figure 1",
"ref_id": "FIGREF0"
}
],
"eq_spans": [],
"section": "Results",
"sec_num": "5.3"
},
{
"text": "However, the low performance of ordinal regression is quite surprising; the expectation was that it would perform the best. In future work, we plan to have a closer look at the reasons for these results. At this point, we can only note that we used SVM as the basic underlying classifier in our classification and regression experiments, but we used logistic regression as the basis for our ordinal regression. It is unclear whether this alone could explain the difference in performance, though. Table 3 , shows the impact of the individual context features (and some feature combinations) when added to the baseline textual features. We report results for 11-way classification with the regression model; and the last column shows the difference in MSE compared to the baseline. We can see that each of the features yields improvements, which means that they all are indeed relevant. The most important features turn out to be movie length, actors, and user average score.",
"cite_spans": [],
"ref_spans": [
{
"start": 497,
"end": 504,
"text": "Table 3",
"ref_id": null
}
],
"eq_spans": [],
"section": "Results",
"sec_num": "5.3"
},
{
"text": "Yet, some features might be redundant, i.e., having one feature might mean that we do not need to have some other ones. In order to study this, we performed experiments excluding features one at a time from the full set of features, both textual and contextual. The results are shown in Table 4 . As before, we study 11-way classification with the regression model. The relative change in MSE compared to the full model is shown in the last column of the table. We can see that lexicons have the biggest impact, which is to be expected, as we know from previous work that they are among the most important resources for sentiment analysis. Another strong feature turns out to be the user average score, which also makes sense: a user who has been giving high scores in the past is likely to give high scores in the future. We can further see that many contextual features, e.g., movie length, actors, director, genres and country, made almost no difference. This is surprising as the first two yielded the largest improvements over the baseline features in Table 3 ; we believe this reflects feature interaction, but we plan closer investigation.",
"cite_spans": [],
"ref_spans": [
{
"start": 287,
"end": 294,
"text": "Table 4",
"ref_id": "TABREF3"
},
{
"start": 1057,
"end": 1064,
"text": "Table 3",
"ref_id": null
}
],
"eq_spans": [],
"section": "Results",
"sec_num": "5.3"
},
{
"text": "We have seen in our experiments above that the best-performing model used regression and contextual features, in addition to textual ones. We believe that the kind of context we model, primarily metadata, is indeed important as, while it is not present in the text of the review, it has been taken into account when the author rated the movie.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "Interestingly, we have found that factual information was not very useful. This is a good sign as it suggests that Cinexio users seem not to have prejudice about the expected quality of a movie based on its country of origin, director(s), or genre; however, actors playing do have impact.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "One of the most useful contextual features was the user average score. Some users tend to give consistently high/low scores regardless of the movie, and thus knowing their average scores allows us to take this into account.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "A related useful feature was the Cinexio score of the target movie. The idea is that if a movie has a high/low overall score, we should expect a new user also to give it a high/low score. While IMDB scores are quite similar, we had mixed results for them: they were quite helpful compared to the baseline, but were harmful with respect to the full set of features.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "Given the difference between Cinexio and IMDB scores, we decided to have a closer look at how they relate to each other. This is shown in Figure 2. The blue line connects the corresponding Cinexio-IMDB scores, while the red line shows how perfect correlation would look like. Note that IMDB scores are in the 0-10 range. This is an interesting plot as it reflects how viewers (a) in Bulgaria and (b) worldwide feel about the same movie. We can see that the general correlation is there, especially for the midhigh scores. However, there is a lot of discrepancy with the extreme scores, i.e., what Bulgarian viewers see as extremely good is regarded as average at IMDB, and what they consider extremely bad, actually has an above-average score at IMDB. This discrepancy in IMDB vs. Cinexio scores explains the mixed results we got when using the IMDB score as a feature. One way to fix this could be to split the IMDB feature into several features, each responsible for just a sub-interval of the possible values of the original feature. This might be useful for some other features with numerical values, which could show non-linearity, e.g., Cinexio score, average user score, or movie length.",
"cite_spans": [],
"ref_spans": [
{
"start": 138,
"end": 144,
"text": "Figure",
"ref_id": null
}
],
"eq_spans": [],
"section": "Discussion",
"sec_num": "6"
},
{
"text": "We presented the first research on fine-grained sentiment analysis for Bulgarian. As this is pioneering work for this language, we created a suitable dataset and a sentiment polarity lexicon, which we made freely available for research purposes; this should enable further research.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Work",
"sec_num": "7"
},
{
"text": "We further compared experimentally the performance of classification, regression and ordinal regression in a 3-way, 5-way and 11-way classification setups, using as features not only the text from the reviews, but also contextual information in the form of metadata, e.g., movie length, director, actors, genre, country, and various scores: IMDB, Cinexio, and user-average. The experimental results have shown that adding contextual information yields strong performance gains.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Work",
"sec_num": "7"
},
{
"text": "In future work, we plan to investigate the low performance of ordinal regression. We further want to experiment with more features, e.g., summary of the plot, subtitles, information from other websites such as IMDB, as well as with more linguistic processing of the text, e.g., stemming (Nakov, 2003b; Nakov, 2003a) , POS tagging (Georgiev et al., 2012) , and named entity recognition (Georgiev et al., 2009) . We also want to see the impact of earlier comments on the sentiment of newer comments (Vanzo et al., 2014; . Finally, we would like to apply our system to help other tasks, e.g., finding trolls in Web forums (Mihaylov et al., 2015a; Mihaylov et al., 2015b) .",
"cite_spans": [
{
"start": 287,
"end": 301,
"text": "(Nakov, 2003b;",
"ref_id": "BIBREF23"
},
{
"start": 302,
"end": 315,
"text": "Nakov, 2003a)",
"ref_id": "BIBREF22"
},
{
"start": 330,
"end": 353,
"text": "(Georgiev et al., 2012)",
"ref_id": "BIBREF8"
},
{
"start": 385,
"end": 408,
"text": "(Georgiev et al., 2009)",
"ref_id": "BIBREF7"
},
{
"start": 497,
"end": 517,
"text": "(Vanzo et al., 2014;",
"ref_id": "BIBREF40"
},
{
"start": 619,
"end": 643,
"text": "(Mihaylov et al., 2015a;",
"ref_id": "BIBREF17"
},
{
"start": 644,
"end": 667,
"text": "Mihaylov et al., 2015b)",
"ref_id": "BIBREF18"
}
],
"ref_spans": [],
"eq_spans": [],
"section": "Conclusion and Future Work",
"sec_num": "7"
},
{
"text": "The dataset is freely available for research purposes at http://bkapukaranov.github.io/",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "Some linguists consider Macedonian a dialect of Bulgarian; this is also the position of the Bulgarian government.3 In fact, they used, without translation, the Bulgarian lexicon that we present in this work.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "\"it is good that they got him in the end, so the propaganda could finally be over.. ;)\"6 \"Very good movie\" 7 \"Very good movie, we are just a little bit off on the American partriotic message\"8 \"Not impressed\" 9 \"Quite sad :(\" 10 \"quite long, on top of that nothing actually happens...\"",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "\"The movie is just the first part of a series, keep in mind the story ends in the most interesting part\"12 \"Definitely more action compared to the first part\"",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
},
{
"text": "There are several alternative machine learning approaches to ordinal regression, e.g., support vector ordinal regression(Chu and Keerthi, 2007).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "Multi-facet rating of product reviews",
"authors": [
{
"first": "Stefano",
"middle": [],
"last": "Baccianella",
"suffix": ""
},
{
"first": "Andrea",
"middle": [],
"last": "Esuli",
"suffix": ""
},
{
"first": "Fabrizio",
"middle": [],
"last": "Sebastiani",
"suffix": ""
}
],
"year": 2009,
"venue": "Proceedings of the 31th European Conference on IR Research on Advances in Information Retrieval, ECIR '09",
"volume": "",
"issue": "",
"pages": "461--472",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Stefano Baccianella, Andrea Esuli, and Fabrizio Sebas- tiani. 2009. Multi-facet rating of product reviews. In Proceedings of the 31th European Conference on IR Research on Advances in Information Retrieval, ECIR '09, pages 461-472, Toulouse, France.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "SentiWordNet 3.0: An enhanced lexical resource for sentiment analysis and opinion mining",
"authors": [
{
"first": "Stefano",
"middle": [],
"last": "Baccianella",
"suffix": ""
},
{
"first": "Andrea",
"middle": [],
"last": "Esuli",
"suffix": ""
},
{
"first": "Fabrizio",
"middle": [],
"last": "Sebastiani",
"suffix": ""
}
],
"year": 2010,
"venue": "Proceedings of the International Conference on Language Resources and Evaluation, LREC '10",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Stefano Baccianella, Andrea Esuli, and Fabrizio Sebas- tiani. 2010. SentiWordNet 3.0: An enhanced lexi- cal resource for sentiment analysis and opinion min- ing. In Proceedings of the International Conference on Language Resources and Evaluation, LREC '10, Valletta, Malta.",
"links": null
},
"BIBREF2": {
"ref_id": "b2",
"title": "Threadlevel information for comment classification in community question answering",
"authors": [
{
"first": "Alberto",
"middle": [],
"last": "Barr\u00f3n-Cede\u00f1o",
"suffix": ""
},
{
"first": "Simone",
"middle": [],
"last": "Filice",
"suffix": ""
},
{
"first": "Giovanni",
"middle": [],
"last": "Da San",
"suffix": ""
},
{
"first": "Shafiq",
"middle": [],
"last": "Martino",
"suffix": ""
},
{
"first": "Llu\u00eds",
"middle": [],
"last": "Joty",
"suffix": ""
},
{
"first": "Preslav",
"middle": [],
"last": "M\u00e0rquez",
"suffix": ""
},
{
"first": "Alessandro",
"middle": [],
"last": "Nakov",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Moschitti",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, ACL-IJCNLP '15",
"volume": "",
"issue": "",
"pages": "687--693",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Alberto Barr\u00f3n-Cede\u00f1o, Simone Filice, Giovanni Da San Martino, Shafiq Joty, Llu\u00eds M\u00e0rquez, Preslav Nakov, and Alessandro Moschitti. 2015. Thread- level information for comment classification in com- munity question answering. In Proceedings of the 53rd Annual Meeting of the Association for Compu- tational Linguistics and the 7th International Joint Conference on Natural Language Processing, ACL- IJCNLP '15, pages 687-693, Beijing, China.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "LIB-SVM: A library for support vector machines",
"authors": [
{
"first": "Chih-Chung",
"middle": [],
"last": "Chang",
"suffix": ""
},
{
"first": "Chih-Jen",
"middle": [],
"last": "Lin",
"suffix": ""
}
],
"year": 2011,
"venue": "ACM Transactions on Intelligent Systems and Technology",
"volume": "2",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Chih-Chung Chang and Chih-Jen Lin. 2011. LIB- SVM: A library for support vector machines. ACM Transactions on Intelligent Systems and Technology, 2:27:1-27:27.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Support vector ordinal regression",
"authors": [
{
"first": "Wei",
"middle": [],
"last": "Chu",
"suffix": ""
},
{
"first": "S",
"middle": [],
"last": "Sathiya Keerthi",
"suffix": ""
}
],
"year": 2007,
"venue": "Neural Comput",
"volume": "19",
"issue": "3",
"pages": "792--815",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Wei Chu and S. Sathiya Keerthi. 2007. Support vector ordinal regression. Neural Comput., 19(3):792-815, March.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "SENTI-WORDNET: A publicly available lexical resource for opinion mining",
"authors": [
{
"first": "Andrea",
"middle": [],
"last": "Esuli",
"suffix": ""
},
{
"first": "Fabrizio",
"middle": [],
"last": "Sebastiani",
"suffix": ""
}
],
"year": 2006,
"venue": "Proceedings of the International Conference on Language Resources and Evaluation, LREC '06",
"volume": "",
"issue": "",
"pages": "417--422",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Andrea Esuli and Fabrizio Sebastiani. 2006. SENTI- WORDNET: A publicly available lexical resource for opinion mining. In Proceedings of the Interna- tional Conference on Language Resources and Eval- uation, LREC '06, pages 417-422, Genoa, Italy.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "Opinion mining of text documents written in Macedonian language",
"authors": [
{
"first": "Andrej",
"middle": [],
"last": "Gajduk",
"suffix": ""
},
{
"first": "Ljupco",
"middle": [],
"last": "Kocarev",
"suffix": ""
}
],
"year": 2014,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {
"arXiv": [
"arXiv:1411.4472"
]
},
"num": null,
"urls": [],
"raw_text": "Andrej Gajduk and Ljupco Kocarev. 2014. Opin- ion mining of text documents written in Macedonian language. arXiv preprint arXiv:1411.4472.",
"links": null
},
"BIBREF7": {
"ref_id": "b7",
"title": "Featurerich named entity recognition for Bulgarian using conditional random fields",
"authors": [
{
"first": "Georgi",
"middle": [],
"last": "Georgiev",
"suffix": ""
},
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
},
{
"first": "Kuzman",
"middle": [],
"last": "Ganchev",
"suffix": ""
},
{
"first": "Petya",
"middle": [],
"last": "Osenova",
"suffix": ""
},
{
"first": "Kiril",
"middle": [],
"last": "Simov",
"suffix": ""
}
],
"year": 2009,
"venue": "Proceedings of the International Conference Recent Advances in Natural Language Processing, RANLP '09",
"volume": "",
"issue": "",
"pages": "113--117",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Georgi Georgiev, Preslav Nakov, Kuzman Ganchev, Petya Osenova, and Kiril Simov. 2009. Feature- rich named entity recognition for Bulgarian using conditional random fields. In Proceedings of the In- ternational Conference Recent Advances in Natural Language Processing, RANLP '09, pages 113-117, Borovets, Bulgaria.",
"links": null
},
"BIBREF8": {
"ref_id": "b8",
"title": "Feature-rich part-of-speech tagging for morphologically complex languages: Application to Bulgarian",
"authors": [
{
"first": "Georgi",
"middle": [],
"last": "Georgiev",
"suffix": ""
},
{
"first": "Valentin",
"middle": [],
"last": "Zhikov",
"suffix": ""
},
{
"first": "Petya",
"middle": [],
"last": "Osenova",
"suffix": ""
},
{
"first": "Kiril",
"middle": [],
"last": "Simov",
"suffix": ""
},
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the 13th Conference of the European Chapter of the Association for Computational Linguistics, EACL '12",
"volume": "",
"issue": "",
"pages": "492--502",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Georgi Georgiev, Valentin Zhikov, Petya Osenova, Kiril Simov, and Preslav Nakov. 2012. Feature-rich part-of-speech tagging for morphologically complex languages: Application to Bulgarian. In Proceed- ings of the 13th Conference of the European Chap- ter of the Association for Computational Linguistics, EACL '12, pages 492-502, Avignon, France.",
"links": null
},
"BIBREF9": {
"ref_id": "b9",
"title": "Seeing stars when there aren't many stars: Graph-based semi-supervised learning for sentiment categorization",
"authors": [
{
"first": "B",
"middle": [],
"last": "Andrew",
"suffix": ""
},
{
"first": "Xiaojin",
"middle": [],
"last": "Goldberg",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Zhu",
"suffix": ""
}
],
"year": 2006,
"venue": "Proceedings of the First Workshop on Graph Based Methods for Natural Language Processing, TextGraphs '06",
"volume": "",
"issue": "",
"pages": "45--52",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Andrew B. Goldberg and Xiaojin Zhu. 2006. See- ing stars when there aren't many stars: Graph-based semi-supervised learning for sentiment categoriza- tion. In Proceedings of the First Workshop on Graph Based Methods for Natural Language Processing, TextGraphs '06, pages 45-52, Sydney, Australia.",
"links": null
},
"BIBREF10": {
"ref_id": "b10",
"title": "Mining and summarizing customer reviews",
"authors": [
{
"first": "Minqing",
"middle": [],
"last": "Hu",
"suffix": ""
},
{
"first": "Bing",
"middle": [],
"last": "Liu",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the 10th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD '04",
"volume": "",
"issue": "",
"pages": "168--177",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Minqing Hu and Bing Liu. 2004. Mining and summa- rizing customer reviews. In Proceedings of the 10th ACM SIGKDD International Conference on Knowl- edge Discovery and Data Mining, KDD '04, pages 168-177, Seattle, WA, USA.",
"links": null
},
"BIBREF11": {
"ref_id": "b11",
"title": "Global thread-level inference for comment classification in community question answering",
"authors": [
{
"first": "Shafiq",
"middle": [],
"last": "Joty",
"suffix": ""
},
{
"first": "Alberto",
"middle": [],
"last": "Barr\u00f3n-Cede\u00f1o",
"suffix": ""
},
{
"first": "Giovanni",
"middle": [],
"last": "Da San",
"suffix": ""
},
{
"first": "Simone",
"middle": [],
"last": "Martino",
"suffix": ""
},
{
"first": "Llu\u00eds",
"middle": [],
"last": "Filice",
"suffix": ""
},
{
"first": "Alessandro",
"middle": [],
"last": "M\u00e0rquez",
"suffix": ""
},
{
"first": "Preslav",
"middle": [],
"last": "Moschitti",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Nakov",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the Conference on Empirical Methods in Natural Language Processing, EMNLP '15",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Shafiq Joty, Alberto Barr\u00f3n-Cede\u00f1o, Giovanni Da San Martino, Simone Filice, Llu\u00eds M\u00e0rquez, Alessandro Moschitti, and Preslav Nakov. 2015. Global thread-level inference for comment clas- sification in community question answering. In Proceedings of the Conference on Empirical Meth- ods in Natural Language Processing, EMNLP '15, Lisbon, Portugal.",
"links": null
},
"BIBREF12": {
"ref_id": "b12",
"title": "Sentiment analysis in Twitter for Macedonian",
"authors": [
{
"first": "Dame",
"middle": [],
"last": "Jovanoski",
"suffix": ""
},
{
"first": "Veno",
"middle": [],
"last": "Pachovski",
"suffix": ""
},
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the International Conference on Recent Advances in Natural Language Processing, RANLP '15",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Dame Jovanoski, Veno Pachovski, and Preslav Nakov. 2015. Sentiment analysis in Twitter for Macedo- nian. In Proceedings of the International Confer- ence on Recent Advances in Natural Language Pro- cessing, RANLP '15, Hissar, Bulgaria.",
"links": null
},
"BIBREF13": {
"ref_id": "b13",
"title": "Computationally effective algorithm for information extraction and online review mining",
"authors": [
{
"first": "Boris",
"middle": [],
"last": "Kraychev",
"suffix": ""
},
{
"first": "Ivan",
"middle": [],
"last": "Koychev",
"suffix": ""
}
],
"year": 2012,
"venue": "Proceedings of the 2nd International Conference on Web Intelligence, Mining and Semantics, WIMS '12",
"volume": "64",
"issue": "",
"pages": "1--64",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Boris Kraychev and Ivan Koychev. 2012. Computa- tionally effective algorithm for information extrac- tion and online review mining. In Proceedings of the 2nd International Conference on Web Intelligence, Mining and Semantics, WIMS '12, pages 64:1-64:4, Craiova, Romania.",
"links": null
},
"BIBREF14": {
"ref_id": "b14",
"title": "Extraction and Analysis of Opinions and Sentiments from Online Text",
"authors": [
{
"first": "Boris",
"middle": [],
"last": "Kraychev",
"suffix": ""
}
],
"year": 2014,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Boris Kraychev. 2014. Extraction and Analysis of Opinions and Sentiments from Online Text. Ph.D. thesis, Sofia University, January. (in Bulgarian).",
"links": null
},
"BIBREF15": {
"ref_id": "b15",
"title": "A survey of opinion mining and sentiment analysis",
"authors": [
{
"first": "Bing",
"middle": [],
"last": "Liu",
"suffix": ""
},
{
"first": "Lei",
"middle": [],
"last": "Zhang",
"suffix": ""
}
],
"year": 2012,
"venue": "Mining Text Data",
"volume": "",
"issue": "",
"pages": "415--463",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bing Liu and Lei Zhang. 2012. A survey of opin- ion mining and sentiment analysis. In Charu C. Ag- garwal and ChengXiang Zhai, editors, Mining Text Data, pages 415-463. Springer.",
"links": null
},
"BIBREF16": {
"ref_id": "b16",
"title": "Regression models for ordinal data",
"authors": [
{
"first": "Peter",
"middle": [],
"last": "Mccullagh",
"suffix": ""
}
],
"year": 1980,
"venue": "J. Roy. Statist. Soc. B",
"volume": "42",
"issue": "",
"pages": "109--142",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Peter McCullagh. 1980. Regression models for ordinal data. J. Roy. Statist. Soc. B, 42:109-142.",
"links": null
},
"BIBREF17": {
"ref_id": "b17",
"title": "Finding opinion manipulation trolls in news community forums",
"authors": [
{
"first": "Todor",
"middle": [],
"last": "Mihaylov",
"suffix": ""
},
{
"first": "Georgi",
"middle": [],
"last": "Georgiev",
"suffix": ""
},
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the Conference on Computational Natural Language Learning",
"volume": "",
"issue": "",
"pages": "310--314",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Todor Mihaylov, Georgi Georgiev, and Preslav Nakov. 2015a. Finding opinion manipulation trolls in news community forums. In Proceedings of the Confer- ence on Computational Natural Language Learning, pages 310-314, Beijing, China.",
"links": null
},
"BIBREF18": {
"ref_id": "b18",
"title": "Exposing paid opinion manipulation trolls",
"authors": [
{
"first": "Todor",
"middle": [],
"last": "Mihaylov",
"suffix": ""
},
{
"first": "Ivan",
"middle": [],
"last": "Koychev",
"suffix": ""
},
{
"first": "Georgi",
"middle": [],
"last": "Georgiev",
"suffix": ""
},
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the Conference on Computational Natural Language Learning",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Todor Mihaylov, Ivan Koychev, Georgi Georgiev, and Preslav Nakov. 2015b. Exposing paid opinion ma- nipulation trolls. In Proceedings of the Confer- ence on Computational Natural Language Learning, Hissar, Bulgaria.",
"links": null
},
"BIBREF19": {
"ref_id": "b19",
"title": "NRC-Canada: Building the state-of-theart in sentiment analysis of tweets",
"authors": [
{
"first": "Saif",
"middle": [],
"last": "Mohammad",
"suffix": ""
},
{
"first": "Svetlana",
"middle": [],
"last": "Kiritchenko",
"suffix": ""
},
{
"first": "Xiaodan",
"middle": [],
"last": "Zhu",
"suffix": ""
}
],
"year": 2013,
"venue": "Proceedings of the Seventh international workshop on Semantic Evaluation Exercises, SemEval '13",
"volume": "",
"issue": "",
"pages": "321--327",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Saif Mohammad, Svetlana Kiritchenko, and Xiaodan Zhu. 2013. NRC-Canada: Building the state-of-the- art in sentiment analysis of tweets. In Proceedings of the Seventh international workshop on Semantic Evaluation Exercises, SemEval '13, pages 321-327, Atlanta, GA, USA.",
"links": null
},
"BIBREF20": {
"ref_id": "b20",
"title": "SemEval-2013 task 2: Sentiment analysis in Twitter",
"authors": [
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
},
{
"first": "Sara",
"middle": [],
"last": "Rosenthal",
"suffix": ""
},
{
"first": "Zornitsa",
"middle": [],
"last": "Kozareva",
"suffix": ""
},
{
"first": "Veselin",
"middle": [],
"last": "Stoyanov",
"suffix": ""
},
{
"first": "Alan",
"middle": [],
"last": "Ritter",
"suffix": ""
},
{
"first": "Theresa",
"middle": [],
"last": "Wilson",
"suffix": ""
}
],
"year": 2013,
"venue": "Proceedings of the Seventh International Workshop on Semantic Evaluation, SemEval '13",
"volume": "",
"issue": "",
"pages": "312--320",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Preslav Nakov, Sara Rosenthal, Zornitsa Kozareva, Veselin Stoyanov, Alan Ritter, and Theresa Wilson. 2013. SemEval-2013 task 2: Sentiment analysis in Twitter. In Proceedings of the Seventh Interna- tional Workshop on Semantic Evaluation, SemEval '13, pages 312-320, Atlanta, GA, USA.",
"links": null
},
"BIBREF21": {
"ref_id": "b21",
"title": "Developing a successful SemEval task in sentiment analysis of Twitter and other social media texts. Language Resources and Evaluation",
"authors": [
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
},
{
"first": "Sara",
"middle": [],
"last": "Rosenthal",
"suffix": ""
},
{
"first": "Svetlana",
"middle": [],
"last": "Kiritchenko",
"suffix": ""
},
{
"first": "Saif",
"middle": [],
"last": "Mohammad",
"suffix": ""
},
{
"first": "Zornitsa",
"middle": [],
"last": "Kozareva",
"suffix": ""
},
{
"first": "Alan",
"middle": [],
"last": "Ritter",
"suffix": ""
},
{
"first": "Veselin",
"middle": [],
"last": "Stoyanov",
"suffix": ""
},
{
"first": "Xiaodan",
"middle": [],
"last": "Zhu",
"suffix": ""
}
],
"year": 2015,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Preslav Nakov, Sara Rosenthal, Svetlana Kiritchenko, Saif Mohammad, Zornitsa Kozareva, Alan Ritter, Veselin Stoyanov, and Xiaodan Zhu. 2015. Devel- oping a successful SemEval task in sentiment anal- ysis of Twitter and other social media texts. Lan- guage Resources and Evaluation.",
"links": null
},
"BIBREF22": {
"ref_id": "b22",
"title": "Building an inflectional stemmer for Bulgarian",
"authors": [
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
}
],
"year": 2003,
"venue": "Proceedings of the 4th International Conference on Computer Systems and Technologies, CompSysTech '03",
"volume": "",
"issue": "",
"pages": "419--424",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Preslav Nakov. 2003a. Building an inflectional stem- mer for Bulgarian. In Proceedings of the 4th In- ternational Conference on Computer Systems and Technologies, CompSysTech '03, pages 419-424, Sofia, Bulgaria.",
"links": null
},
"BIBREF23": {
"ref_id": "b23",
"title": "BulStem: Design and evaluation of an inflectional stemmer for Bulgarian",
"authors": [
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
}
],
"year": 2003,
"venue": "Proceedings of the Workshop on Balkan Language Resources and Tools",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Preslav Nakov. 2003b. BulStem: Design and evalu- ation of an inflectional stemmer for Bulgarian. In Proceedings of the Workshop on Balkan Language Resources and Tools, Thessaloniki, Greece.",
"links": null
},
"BIBREF24": {
"ref_id": "b24",
"title": "A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts",
"authors": [
{
"first": "Bo",
"middle": [],
"last": "Pang",
"suffix": ""
},
{
"first": "Lillian",
"middle": [],
"last": "Lee",
"suffix": ""
}
],
"year": 2004,
"venue": "Proceedings of the 42nd Annual Meeting on Association for Computational Linguistics, ACL '04",
"volume": "",
"issue": "",
"pages": "271--278",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bo Pang and Lillian Lee. 2004. A sentimental educa- tion: Sentiment analysis using subjectivity summa- rization based on minimum cuts. In Proceedings of the 42nd Annual Meeting on Association for Com- putational Linguistics, ACL '04, pages 271-278, Barcelona, Spain.",
"links": null
},
"BIBREF25": {
"ref_id": "b25",
"title": "Seeing stars: Exploiting class relationships for sentiment categorization with respect to rating scales",
"authors": [
{
"first": "Bo",
"middle": [],
"last": "Pang",
"suffix": ""
},
{
"first": "Lillian",
"middle": [],
"last": "Lee",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the Annual Meeting of the Association for Computational Linguistics, ACL '05",
"volume": "",
"issue": "",
"pages": "115--124",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bo Pang and Lillian Lee. 2005. Seeing stars: Ex- ploiting class relationships for sentiment categoriza- tion with respect to rating scales. In Proceedings of the Annual Meeting of the Association for Compu- tational Linguistics, ACL '05, pages 115-124, Ann Arbor, MI, USA.",
"links": null
},
"BIBREF26": {
"ref_id": "b26",
"title": "Opinion mining and sentiment analysis. Foundations and trends in information retrieval",
"authors": [
{
"first": "Bo",
"middle": [],
"last": "Pang",
"suffix": ""
},
{
"first": "Lillian",
"middle": [],
"last": "Lee",
"suffix": ""
}
],
"year": 2008,
"venue": "",
"volume": "2",
"issue": "",
"pages": "1--135",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bo Pang and Lillian Lee. 2008. Opinion mining and sentiment analysis. Foundations and trends in infor- mation retrieval, 2(1-2):1-135.",
"links": null
},
"BIBREF27": {
"ref_id": "b27",
"title": "Thumbs up?: Sentiment classification using machine learning techniques",
"authors": [
{
"first": "Bo",
"middle": [],
"last": "Pang",
"suffix": ""
},
{
"first": "Lillian",
"middle": [],
"last": "Lee",
"suffix": ""
},
{
"first": "Shivakumar",
"middle": [],
"last": "Vaithyanathan",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the Conference on Empirical Methods in Natural Language Processing, EMNLP '02",
"volume": "",
"issue": "",
"pages": "79--86",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Bo Pang, Lillian Lee, and Shivakumar Vaithyanathan. 2002. Thumbs up?: Sentiment classification us- ing machine learning techniques. In Proceedings of the Conference on Empirical Methods in Natu- ral Language Processing, EMNLP '02, pages 79- 86, Philadelphia, PA, USA.",
"links": null
},
"BIBREF28": {
"ref_id": "b28",
"title": "Linguistic Inquiry and Word Count",
"authors": [
{
"first": "James",
"middle": [
"W"
],
"last": "Pennebaker",
"suffix": ""
},
{
"first": "Martha",
"middle": [
"E"
],
"last": "Francis",
"suffix": ""
},
{
"first": "Roger",
"middle": [
"J"
],
"last": "Booth",
"suffix": ""
}
],
"year": 2001,
"venue": "Lawerence Erlbaum Associates",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "James W. Pennebaker, Martha E. Francis, and Roger J. Booth. 2001. Linguistic Inquiry and Word Count. Lawerence Erlbaum Associates, Mahwah, NJ.",
"links": null
},
"BIBREF29": {
"ref_id": "b29",
"title": "SemEval-2014 task 4: Aspect based sentiment analysis",
"authors": [
{
"first": "Maria",
"middle": [],
"last": "Pontiki",
"suffix": ""
},
{
"first": "Harris",
"middle": [],
"last": "Papageorgiou",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the 8th International Workshop on Semantic Evaluation, SemEval '14",
"volume": "",
"issue": "",
"pages": "27--35",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Maria Pontiki, Harris Papageorgiou, Dimitrios Gala- nis, Ion Androutsopoulos, John Pavlopoulos, and Suresh Manandhar. 2014. SemEval-2014 task 4: Aspect based sentiment analysis. In Proceedings of the 8th International Workshop on Semantic Evalu- ation, SemEval '14, pages 27-35, Dublin, Ireland.",
"links": null
},
"BIBREF30": {
"ref_id": "b30",
"title": "Languageindependent sentiment analysis using subjectivity and positional information",
"authors": [
{
"first": "Veselin",
"middle": [],
"last": "Raychev",
"suffix": ""
},
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
}
],
"year": 2009,
"venue": "Proceedings of the International Conference on Recent Advances in Natural Language Processing, RANLP '09",
"volume": "",
"issue": "",
"pages": "360--364",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Veselin Raychev and Preslav Nakov. 2009. Language- independent sentiment analysis using subjectivity and positional information. In Proceedings of the International Conference on Recent Advances in Natural Language Processing, RANLP '09, pages 360-364, Borovets, Bulgaria.",
"links": null
},
"BIBREF31": {
"ref_id": "b31",
"title": "Automatic Recognition of Positive/Negative Subjectivity in Text",
"authors": [
{
"first": "",
"middle": [],
"last": "Veselin Raychev",
"suffix": ""
}
],
"year": 2009,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Veselin Raychev. 2009. Automatic Recognition of Positive/Negative Subjectivity in Text. Ph.D. thesis, Sofia University, March. (in Bulgarian).",
"links": null
},
"BIBREF32": {
"ref_id": "b32",
"title": "SemEval-2014 Task 9: Sentiment analysis in Twitter",
"authors": [
{
"first": "Sara",
"middle": [],
"last": "Rosenthal",
"suffix": ""
},
{
"first": "Alan",
"middle": [],
"last": "Ritter",
"suffix": ""
},
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
},
{
"first": "Veselin",
"middle": [],
"last": "Stoyanov",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the 8th International Workshop on Semantic Evaluation, SemEval '14",
"volume": "",
"issue": "",
"pages": "73--80",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sara Rosenthal, Alan Ritter, Preslav Nakov, and Veselin Stoyanov. 2014. SemEval-2014 Task 9: Sentiment analysis in Twitter. In Proceedings of the 8th International Workshop on Semantic Evaluation, SemEval '14, pages 73-80, Dublin, Ireland.",
"links": null
},
"BIBREF33": {
"ref_id": "b33",
"title": "SemEval-2015 task 10: Sentiment analysis in Twitter",
"authors": [
{
"first": "Sara",
"middle": [],
"last": "Rosenthal",
"suffix": ""
},
{
"first": "Preslav",
"middle": [],
"last": "Nakov",
"suffix": ""
},
{
"first": "Svetlana",
"middle": [],
"last": "Kiritchenko",
"suffix": ""
},
{
"first": "Saif",
"middle": [],
"last": "Mohammad",
"suffix": ""
},
{
"first": "Alan",
"middle": [],
"last": "Ritter",
"suffix": ""
},
{
"first": "Veselin",
"middle": [],
"last": "Stoyanov",
"suffix": ""
}
],
"year": 2015,
"venue": "Proceedings of the 9th International Workshop on Semantic Evaluation, SemEval '15",
"volume": "",
"issue": "",
"pages": "450--462",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sara Rosenthal, Preslav Nakov, Svetlana Kiritchenko, Saif Mohammad, Alan Ritter, and Veselin Stoyanov. 2015. SemEval-2015 task 10: Sentiment analysis in Twitter. In Proceedings of the 9th International Workshop on Semantic Evaluation, SemEval '15, pages 450-462, Denver, CO, USA.",
"links": null
},
"BIBREF34": {
"ref_id": "b34",
"title": "Machine learning in automated text categorization",
"authors": [
{
"first": "Fabrizio",
"middle": [],
"last": "Sebastiani",
"suffix": ""
}
],
"year": 2002,
"venue": "ACM Comput. Surv",
"volume": "34",
"issue": "1",
"pages": "1--47",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Fabrizio Sebastiani. 2002. Machine learning in au- tomated text categorization. ACM Comput. Surv., 34(1):1-47, March.",
"links": null
},
"BIBREF35": {
"ref_id": "b35",
"title": "A tutorial on support vector regression",
"authors": [
{
"first": "Alex",
"middle": [
"J"
],
"last": "Smola",
"suffix": ""
},
{
"first": "Bernhard",
"middle": [],
"last": "Sch\u00f6lkopf",
"suffix": ""
}
],
"year": 2004,
"venue": "Statistics and Computing",
"volume": "14",
"issue": "3",
"pages": "199--222",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Alex J. Smola and Bernhard Sch\u00f6lkopf. 2004. A tu- torial on support vector regression. Statistics and Computing, 14(3):199-222, August.",
"links": null
},
"BIBREF36": {
"ref_id": "b36",
"title": "The General Inquirer: A Computer Approach to Content Analysis",
"authors": [
{
"first": "Philip",
"middle": [
"J"
],
"last": "Stone",
"suffix": ""
},
{
"first": "Dexter",
"middle": [
"C"
],
"last": "Dunphy",
"suffix": ""
},
{
"first": "Marshall",
"middle": [
"S"
],
"last": "Smith",
"suffix": ""
},
{
"first": "Daniel",
"middle": [
"M"
],
"last": "Ogilvie",
"suffix": ""
}
],
"year": 1966,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Philip J. Stone, Dexter C. Dunphy, Marshall S. Smith, and Daniel M. Ogilvie. 1966. The General In- quirer: A Computer Approach to Content Analysis. MIT Press.",
"links": null
},
"BIBREF37": {
"ref_id": "b37",
"title": "Topic identification for fine-grained opinion analysis",
"authors": [
{
"first": "Veselin",
"middle": [],
"last": "Stoyanov",
"suffix": ""
},
{
"first": "Claire",
"middle": [],
"last": "Cardie",
"suffix": ""
}
],
"year": 2008,
"venue": "Proceedings of the 22nd International Conference on Computational Linguistics, COLING '08",
"volume": "",
"issue": "",
"pages": "817--824",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Veselin Stoyanov and Claire Cardie. 2008. Topic identification for fine-grained opinion analysis. In Proceedings of the 22nd International Conference on Computational Linguistics, COLING '08, pages 817-824, Manchester, United Kingdom.",
"links": null
},
"BIBREF38": {
"ref_id": "b38",
"title": "Thumbs up or thumbs down?: Semantic orientation applied to unsupervised classification of reviews",
"authors": [
{
"first": "D",
"middle": [],
"last": "Peter",
"suffix": ""
},
{
"first": "",
"middle": [],
"last": "Turney",
"suffix": ""
}
],
"year": 2002,
"venue": "Proceedings of the Annual Meeting of the Association for Computational Linguistics, ACL '02",
"volume": "",
"issue": "",
"pages": "417--424",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Peter D. Turney. 2002. Thumbs up or thumbs down?: Semantic orientation applied to unsupervised classi- fication of reviews. In Proceedings of the Annual Meeting of the Association for Computational Lin- guistics, ACL '02, pages 417-424, Philadelphia, PA.",
"links": null
},
"BIBREF39": {
"ref_id": "b39",
"title": "Sentiment analysis of movie reviews written in Macedonian language",
"authors": [
{
"first": "Vasilija",
"middle": [],
"last": "Uzunova",
"suffix": ""
},
{
"first": "Andrea",
"middle": [],
"last": "Kulakov",
"suffix": ""
}
],
"year": 2014,
"venue": "ICT Innovations",
"volume": "",
"issue": "",
"pages": "279--288",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Vasilija Uzunova and Andrea Kulakov. 2015. Sen- timent analysis of movie reviews written in Mace- donian language. In ICT Innovations 2014, pages 279-288. Springer.",
"links": null
},
"BIBREF40": {
"ref_id": "b40",
"title": "A context-based model for sentiment analysis in Twitter",
"authors": [
{
"first": "Andrea",
"middle": [],
"last": "Vanzo",
"suffix": ""
},
{
"first": "Danilo",
"middle": [],
"last": "Croce",
"suffix": ""
},
{
"first": "Roberto",
"middle": [],
"last": "Basili",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the 25th International Conference on Computational Linguistics, COLING '14",
"volume": "",
"issue": "",
"pages": "2345--2354",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Andrea Vanzo, Danilo Croce, and Roberto Basili. 2014. A context-based model for sentiment anal- ysis in Twitter. In Proceedings of the 25th Inter- national Conference on Computational Linguistics, COLING '14, pages 2345-2354, Dublin, Ireland.",
"links": null
},
"BIBREF41": {
"ref_id": "b41",
"title": "Learning subjective language",
"authors": [
{
"first": "Janyce",
"middle": [],
"last": "Wiebe",
"suffix": ""
},
{
"first": "Theresa",
"middle": [],
"last": "Wilson",
"suffix": ""
},
{
"first": "Rebecca",
"middle": [],
"last": "Bruce",
"suffix": ""
},
{
"first": "Matthew",
"middle": [],
"last": "Bell",
"suffix": ""
},
{
"first": "Melanie",
"middle": [],
"last": "Martin",
"suffix": ""
}
],
"year": 2004,
"venue": "Comput. Linguist",
"volume": "30",
"issue": "3",
"pages": "277--308",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Janyce Wiebe, Theresa Wilson, Rebecca Bruce, Matthew Bell, and Melanie Martin. 2004. Learning subjective language. Comput. Linguist., 30(3):277- 308, September.",
"links": null
},
"BIBREF42": {
"ref_id": "b42",
"title": "Recognizing contextual polarity in phraselevel sentiment analysis",
"authors": [
{
"first": "Theresa",
"middle": [],
"last": "Wilson",
"suffix": ""
},
{
"first": "Janyce",
"middle": [],
"last": "Wiebe",
"suffix": ""
},
{
"first": "Paul",
"middle": [],
"last": "Hoffmann",
"suffix": ""
}
],
"year": 2005,
"venue": "Proceedings of the Conference on Human Language Technology and Empirical Methods in Natural Language Processing, HLT-EMNLP '05",
"volume": "",
"issue": "",
"pages": "347--354",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Theresa Wilson, Janyce Wiebe, and Paul Hoffmann. 2005. Recognizing contextual polarity in phrase- level sentiment analysis. In Proceedings of the Con- ference on Human Language Technology and Em- pirical Methods in Natural Language Processing, HLT-EMNLP '05, pages 347-354, Vancouver, BC, Canada.",
"links": null
},
"BIBREF43": {
"ref_id": "b43",
"title": "NRC-Canada-2014: Detecting aspects and sentiment in customer reviews",
"authors": [
{
"first": "Xiaodan",
"middle": [],
"last": "Zhu",
"suffix": ""
},
{
"first": "Svetlana",
"middle": [],
"last": "Kiritchenko",
"suffix": ""
},
{
"first": "Saif",
"middle": [
"M"
],
"last": "Mohammad",
"suffix": ""
}
],
"year": 2014,
"venue": "Proceedings of the Workshop on Semantic Evaluation, SemEval '14",
"volume": "",
"issue": "",
"pages": "437--442",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Xiaodan Zhu, Svetlana Kiritchenko, and Saif M. Mo- hammad. 2014. NRC-Canada-2014: Detecting as- pects and sentiment in customer reviews. In Pro- ceedings of the Workshop on Semantic Evaluation, SemEval '14, pages 437-442, Dublin, Ireland.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"type_str": "figure",
"text": "User rating distribution in our dataset.",
"num": null,
"uris": null
},
"FIGREF1": {
"type_str": "figure",
"text": "Cinexio vs. IMDB scores.",
"num": null,
"uris": null
},
"TABREF0": {
"text": "http://www.cinexio.com",
"html": null,
"content": "<table><tr><td>Characteristic</td><td>Count</td></tr><tr><td>unique words</td><td>8,406</td></tr><tr><td>unique users</td><td>3,395</td></tr><tr><td>unique movie genres</td><td>23</td></tr><tr><td>unique movie countries</td><td>49</td></tr><tr><td>unique movie actors</td><td>1,668</td></tr><tr><td>unique movie directors</td><td>317</td></tr><tr><td colspan=\"2\">Table 1: Statistics about our dataset.</td></tr></table>",
"type_str": "table",
"num": null
},
"TABREF2": {
"text": "Evaluation using the selected features.",
"html": null,
"content": "<table><tr><td>Shown is MSE for the three machine learning al-</td></tr><tr><td>gorithms and for the three class sizes. (Lower</td></tr><tr><td>scores are better.)</td></tr></table>",
"type_str": "table",
"num": null
},
"TABREF3": {
"text": "Impact of individual features when excluded from the full feature set. Shown is MSE for the regression model with 11 classes.",
"html": null,
"content": "<table/>",
"type_str": "table",
"num": null
}
}
}
}