Learning domain-specific sentiment lexicons with applications to recommender systems

Search is now going beyond looking for factual information, and people wish to search for the opinions of others to help them in their own decision-making. Sentiment expressions or opinion expressions are used by users to express their opinion and embody important pieces of information, particularly in online commerce. The main problem that the present dissertation addresses is how to model text to find meaningful words that express a sentiment. In this context, I investigate the viability of automatically generating a sentiment lexicon for opinion retrieval and sentiment classification applications. For this research objective we propose to capture sentiment words that are derived from online users’ reviews. In this approach, we tackle a major challenge in sentiment analysis which is the detection of words that express subjective preference and domain-specific sentiment words such as jargon. To this aim we present a fully generative method that automatically learns a domain-specific lexicon and is fully independent of external sources. Sentiment lexicons can be applied in a broad set of applications, however popular recommendation algorithms have somehow been disconnected from sentiment analysis. Therefore, we present a study that explores the viability of applying sentiment analysis techniques to infer ratings in a recommendation algorithm. Furthermore, entities’ reputation is intrinsically associated with sentiment words that have a positive or negative relation with those entities. Hence, is provided a study that observes the viability of using a domain-specific lexicon to compute entities reputation. Finally, a recommendation system algorithm is improved with the use of sentiment-based ratings and entities reputation.

A sentiment-aware approach to community formation in social media

Participating in a community exemplifies the aspect of sharing, networking and interacting in a social media system. There has been extensive work on characterising on-line communities by their contents and tags using topic modelling tools. However, the role of sentiment and mood has not been studied. Arguably, mood is an integral feature of a text, and becomes more significant in the context of social media: two communities might discuss precisely the same topics, yet within an entirely different atmosphere. Such sentiment-related distinctions are important for many kinds of analysis and applications, such as community recommendation. We present a novel approach to identification of latent hyper-groups in social communities based on users’ sentiment. The results show that a sentiment-based approach can yield useful insights into community formation and metacommunities, having potential applications in, for example, mental health—by targeting support or surveillance to communities with negative mood—or in marketing—by targeting customer communities having the same sentiment on similar topics.

Hyper-community detection in the blogosphere

Most existing work on learning community structure in social network is graph-based whose links among the members are often represented as an adjacency matrix, encoding direct pairwise associations between members. In this paper, we propose a method to group online communities in blogosphere based on the topics learnt from the content blogged. We then consider a different type of online community formulation - the sentiment-based grouping of online communities. The problem of sentiment-based clustering for community structure discovery is rich with many interesting open aspects to be explored. We propose a novel approach for addressing hyper-community detection based on users' sentiment. We employ a nonparametric clustering to automatically discover hidden hyper-communities and present the results obtained from a large dataset.

Emotional reactions to real-world events in social networks

A convergence of emotions among people in social networks is potentially resulted by the occurrence of an unprecedented event in real world. E.g., a majority of bloggers would react angrily at the September 11 terrorist attacks. Based on this observation, we introduce a sentiment index, computed from the current mood tags in a collection of blog posts utilizing an affective lexicon, potentially revealing subtle events discussed in the blogosphere. We then develop a method for extracting events based on this index and its distribution. Our second contribution is establishment of a new bursty structure in text streams termed a sentiment burst. We employ a stochastic model to detect bursty periods of moods and the events associated. Our results on a dataset of more than 12 million mood-tagged blog posts over a 4-year period have shown that our sentiment-based bursty events are indeed meaningful, in several ways.

Mood sensing from social media texts and its applications

We present a large-scale mood analysis in social media texts. We organise the paper in three parts: (1) addressing the problem of feature selection and classification of mood in blogosphere, (2) we extract global mood patterns at different level of aggregation from a large-scale data set of approximately 18 millions documents (3) and finally, we extract mood trajectory for an egocentric user and study how it can be used to detect subtle emotion signals in a user-centric manner, supporting discovery of hyper-groups of communities based on sentiment information. For mood classification, two feature sets proposed in psychology are used, showing that these features are efficient, do not require a training phase and yield classification results comparable to state of the art, supervised feature selection schemes, on mood patterns, empirical results for mood organisation in the blogosphere are provided, analogous to the structure of human emotion proposed independently in the psychology literature, and on community structure discovery, sentiment-based approach can yield useful insights into community formation.

COMPENDIUM: a text summarisation tool for generating summaries of multiple purposes, domains, and genres

In this paper, we present a Text Summarisation tool, compendium, capable of generating the most common types of summaries. Regarding the input, single- and multi-document summaries can be produced; as the output, the summaries can be extractive or abstractive-oriented; and finally, concerning their purpose, the summaries can be generic, query-focused, or sentiment-based. The proposed architecture for compendium is divided in various stages, making a distinction between core and additional stages. The former constitute the backbone of the tool and are common for the generation of any type of summary, whereas the latter are used for enhancing the capabilities of the tool. The main contributions of compendium with respect to the state-of-the-art summarisation systems are that (i) it specifically deals with the problem of redundancy, by means of textual entailment; (ii) it combines statistical and cognitive-based techniques for determining relevant content; and (iii) it proposes an abstractive-oriented approach for facing the challenge of abstractive summarisation. The evaluation performed in different domains and textual genres, comprising traditional texts, as well as texts extracted from the Web 2.0, shows that compendium is very competitive and appropriate to be used as a tool for generating summaries.

Improved lexicon-based sentiment analysis for social media analytics

Social media channels, such as Facebook or Twitter, allow for people to express their views and opinions about any public topics. Public sentiment related to future events, such as demonstrations or parades, indicate public attitude and therefore may be applied while trying to estimate the level of disruption and disorder during such events. Consequently, sentiment analysis of social media content may be of interest for different organisations, especially in security and law enforcement sectors. This paper presents a new lexicon-based sentiment analysis algorithm that has been designed with the main focus on real time Twitter content analysis. The algorithm consists of two key components, namely sentiment normalisation and evidence-based combination function, which have been used in order to estimate the intensity of the sentiment rather than positive/negative label and to support the mixed sentiment classification process. Finally, we illustrate a case study examining the relation between negative sentiment of twitter posts related to English Defence League and the level of disorder during the organisation’s related events.

New markov chain based methods for single and cross-domain sentiment classification

Nowadays communication is switching from a centralized scenario, where communication media like newspapers, radio, TV programs produce information and people are just consumers, to a completely different decentralized scenario, where everyone is potentially an information producer through the use of social networks, blogs, forums that allow a real-time worldwide information exchange. These new instruments, as a result of their widespread diffusion, have started playing an important socio-economic role. They are the most used communication media and, as a consequence, they constitute the main source of information enterprises, political parties and other organizations can rely on. Analyzing data stored in servers all over the world is feasible by means of Text Mining techniques like Sentiment Analysis, which aims to extract opinions from huge amount of unstructured texts. This could lead to determine, for instance, the user satisfaction degree about products, services, politicians and so on. In this context, this dissertation presents new Document Sentiment Classification methods based on the mathematical theory of Markov Chains. All these approaches bank on a Markov Chain based model, which is language independent and whose killing features are simplicity and generality, which make it interesting with respect to previous sophisticated techniques. Every discussed technique has been tested in both Single-Domain and Cross-Domain Sentiment Classification areas, comparing performance with those of other two previous works. The performed analysis shows that some of the examined algorithms produce results comparable with the best methods in literature, with reference to both single-domain and cross-domain tasks, in $2$-classes (i.e. positive and negative) Document Sentiment Classification. However, there is still room for improvement, because this work also shows the way to walk in order to enhance performance, that is, a good novel feature selection process would be enough to outperform the state of the art. Furthermore, since some of the proposed approaches show promising results in $2$-classes Single-Domain Sentiment Classification, another future work will regard validating these results also in tasks with more than $2$ classes.

Towards a unified sentiment lexicon based on graphics processing units

This paper presents an approach to create what we have called a Unified Sentiment Lexicon (USL). This approach aims at aligning, unifying, and expanding the set of sentiment lexicons which are available on the web in order to increase their robustness of coverage. One problem related to the task of the automatic unification of different scores of sentiment lexicons is that there are multiple lexical entries for which the classification of positive, negative, or neutral {P, Z, N} depends on the unit of measurement used in the annotation methodology of the source sentiment lexicon. Our USL approach computes the unified strength of polarity of each lexical entry based on the Pearson correlation coefficient which measures how correlated lexical entries are with a value between 1 and -1, where 1 indicates that the lexical entries are perfectly correlated, 0 indicates no correlation, and -1 means they are perfectly inversely correlated and so is the UnifiedMetrics procedure for CPU and GPU, respectively. Another problem is the high processing time required for computing all the lexical entries in the unification task. Thus, the USL approach computes a subset of lexical entries in each of the 1344 GPU cores and uses parallel processing in order to unify 155802 lexical entries. The results of the analysis conducted using the USL approach show that the USL has 95.430 lexical entries, out of which there are 35.201 considered to be positive, 22.029 negative, and 38.200 neutral. Finally, the runtime was 10 minutes for 95.430 lexical entries; this allows a reduction of the time computing for the UnifiedMetrics by 3 times.

Generating Linked-Data based Domain-Specific Sentiment Lexicons from Legacy Language and Semantic Resources

We present a methodology for legacy language resource adaptation that generates domain-specific sentiment lexicons organized around domain entities described with lexical information and sentiment words described in the context of these entities. We explain the steps of the methodology and we give a working example of our initial results. The resulting lexicons are modelled as Linked Data resources by use of established formats for Linguistic Linked Data (lemon, NIF) and for linked sentiment expressions (Marl), thereby contributing and linking to existing Language Resources in the Linguistic Linked Open Data cloud.

Linked-data based domain-specific sentiment lexicons

In this paper we present a dataset componsed of domain-specific sentiment lexicons in six languages for two domains. We used existing collections of reviews from Trip Advisor, Amazon, the Stanford Network Analysis Project and the OpinRank Review Dataset. We use an RDF model based on the lemon and Marl formats to represent the lexicons. We describe the methodology that we applied to generate the domain-specific lexicons and we provide access information to our datasets.

Development of a Dashboard for Sentiment Analysis of Football in Twitter based on Web Components and D3.js

This thesis is the result of a project whose objective has been to develop and deploy a dashboard for sentiment analysis of football in Twitter based on web components and D3.js. To do so, a visualisation server has been developed in order to present the data obtained from Twitter and analysed with Senpy. This visualisation server has been developed with Polymer web components and D3.js. Data mining has been done with a pipeline between Twitter, Senpy and ElasticSearch. Luigi have been used in this process because helps building complex pipelines of batch jobs, so it has analysed all tweets and stored them in ElasticSearch. To continue, D3.js has been used to create interactive widgets that make data easily accessible, this widgets will allow the user to interact with them and �filter the most interesting data for him. Polymer web components have been used to make this dashboard according to Google's material design and be able to show dynamic data in widgets. As a result, this project will allow an extensive analysis of the social network, pointing out the influence of players and teams and the emotions and sentiments that emerge in a lapse of time.

Learning domain-specific sentiment lexicons for predicting product sales

Generic sentiment lexicons have been widely used for sentiment analysis these days. However, manually constructing sentiment lexicons is very time-consuming and it may not be feasible for certain application domains where annotation expertise is not available. One contribution of this paper is the development of a statistical learning based computational method for the automatic construction of domain-specific sentiment lexicons to enhance cross-domain sentiment analysis. Our initial experiments show that the proposed methodology can automatically generate domain-specific sentiment lexicons which contribute to improve the effectiveness of opinion retrieval at the document level. Another contribution of our work is that we show the feasibility of applying the sentiment metric derived based on the automatically constructed sentiment lexicons to predict product sales of certain product categories. Our research contributes to the development of more effective sentiment analysis system to extract business intelligence from numerous opinionated expressions posted to the Web

Product reputation model : an opinion mining based approach

Product rating systems are very popular on the web, and users are increasingly depending on the overall product ratings provided by websites to make purchase decisions or to compare various products. Currently most of these systems directly depend on users’ ratings and aggregate the ratings using simple aggregating methods such as mean or median [1]. In fact, many websites also allow users to express their opinions in the form of textual product reviews. In this paper, we propose a new product reputation model that uses opinion mining techniques in order to extract sentiments about product’s features, and then provide a method to generate a more realistic reputation value for every feature of the product and the product itself. We considered the strength of the opinion rather than its orientation only. We do not treat all product features equally when we calculate the overall product reputation, as some features are more important to customers than others, and consequently have more impact on customers buying decisions. Our method provides helpful details about the product features for customers rather than only representing reputation as a number only.

Pattern-based topic models for information filtering

Topic modelling, such as Latent Dirichlet Allocation (LDA), was proposed to generate statistical models to represent multiple topics in a collection of documents, which has been widely utilized in the fields of machine learning and information retrieval, etc. But its effectiveness in information filtering is rarely known. Patterns are always thought to be more representative than single terms for representing documents. In this paper, a novel information filtering model, Pattern-based Topic Model(PBTM) , is proposed to represent the text documents not only using the topic distributions at general level but also using semantic pattern representations at detailed specific level, both of which contribute to the accurate document representation and document relevance ranking. Extensive experiments are conducted to evaluate the effectiveness of PBTM by using the TREC data collection Reuters Corpus Volume 1. The results show that the proposed model achieves outstanding performance.