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.

The unified sentiment lexicon using GPUs

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. A sentiment lexicon is a critical and essential resource for tagging subjective corpora on the web or elsewhere. In many situations, the multilingual property of the sentiment lexicon is important because the writer is using two languages alternately in the same text, message or post. 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 the UnifiedMetrics procedure for CPU and GPU, respectively.

Hacia un Lexicón Unificado de Sentimientos basado en Unidades de Procesamiento Gráfico

Esta tesis presenta un modelo, una metodología, una arquitectura, varios algoritmos y programas para crear un lexicón de sentimientos unificado (LSU) que cubre cuatro lenguas: inglés, español, portugués y chino. El objetivo principal es alinear, unificar, y expandir el conjunto de lexicones de sentimientos disponibles en Internet y los desarrollados a lo largo de esta investigación. Así, el principal problema a resolver es la tarea de unificar de forma automatizada los diferentes lexicones de sentimientos obtenidos por el crawler CSR, porque la unidad de medida para asignar la intensidad de los valores de la polaridad (de forma manual, semiautomática y automática) varía de acuerdo con las diferentes metodologías utilizadas para la construcción de cada lexicón. La representación codificada de la estructura de datos de los términos presenta también una variación en la estructura de lexicón a lexicón. Por lo que al unificar en un lexicón de sentimientos se hace posible la reutilización del conocimiento recopilado por los diferentes grupos de investigación y se incrementa, a la vez, el alcance, la calidad y la robustez de los lexicones. Nuestra metodología LSU calcula un valor unificado de la intensidad de la polaridad para cada entrada léxica que está presente en al menos dos de los lexicones de sentimientos que forman parte de este estudio. En contraste, las entradas léxicas que no son comunes en al menos dos de los lexicones conservan su valor original. El coeficiente de Pearson resultante permite medir la correlación existente entre las entradas léxicas asignándoles un rango de valores de uno a menos uno, donde uno indica que los valores de los términos están perfectamente correlacionados, cero indica que no existe correlación y menos uno significa que están inversamente correlacionados. Este procedimiento se lleva acabo con la función de MetricasUnificadas tanto en la CPU como en la GPU. Otro problema a resolver es el tiempo de procesamiento que se requiere para realizar la tarea de unificación de la intensidad de la polaridad y con ello alcanzar una cobertura mayor de lemas en los lexicones de sentimientos existentes. Asimismo, la metodología LSU utiliza el procesamiento paralelo para unificar los 155 802 términos. El algoritmo LSU procesa mediante cargas iguales el subconjunto de entradas léxicas en cada uno de los 1344 núcleos en la GPU. Los resultados de nuestro análisis arrojaron un total de 95 430 entradas léxicas donde 35 201 obtuvieron valores positivos, 22 029 negativos y 38 200 neutrales. Finalmente, el tiempo de ejecución fue de 2,506 segundos para el total de las entradas léxicas, lo que permitió reducir el procesamiento de cómputo hasta en una tercera parte con respecto al algoritmo secuencial. De estos resultados se concluye que al lograr un lexicón de sentimientos unificado que permite homogeneizar la intensidad de la polaridad de las unidades léxicas (con valores positivos, negativos y neutrales) deriva no sólo en el análisis semántico del corpus basado en los términos con una mayor carga de polaridad, o del resumen de las valoraciones o las tendencias de neuromarketing, sino también en aplicaciones como el etiquetado subjetivo de sitios web o de portales sintácticos y semánticos, por mencionar algunas. ABSTRACT This thesis 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, N, Z 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 155,802 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 2.505 seconds for 95,430 lexical entries; this allows a reduction of the time computing for the UnifiedMetrics by 3 times with respect to the sequential implementation. A key contribution of this work is that we preserve the use of a unified sentiment lexicon for all tasks. Such lexicon is used to define resources and resource-related properties that can be verified based on the results of the analysis and is powerful, general and extensible enough to express a large class of interesting properties. Some applications of this work include merging, aligning, pruning and extending the current sentiment lexicons.

DAEDALUS at SemEval-2014 Task 9: Comparing approaches for sentiment analysis in twitter

This paper describes our participation at SemEval- 2014 sentiment analysis task, in both contextual and message polarity classification. Our idea was to com- pare two different techniques for sentiment analysis. First, a machine learning classifier specifically built for the task using the provided training corpus. On the other hand, a lexicon-based approach using natural language processing techniques, developed for a ge- neric sentiment analysis task with no adaptation to the provided training corpus. Results, though far from the best runs, prove that the generic model is more robust as it achieves a more balanced evaluation for message polarity along the different test sets.

The Spanish Travel Subjective Lexicon (STSL).

This paper presents a proposal for a recognition model for the appraisal value of sentences. It is based on splitting the text into independent sentences (full stops) and then analysing the appraisal elements contained in each sentence according to the previous value in the appraisal lexicon. In this lexicon, positive words are assigned a positive coefficient (+1) and negative words a negative coefficient (-1). We take into account word such as ?too?, ?little? (when it is not ?a bit?), ?less?, and ?nothing? than can modify the polarity degree of lexical unit when appear in the nearby environment. If any of these elements are present, then the previous coefficient will be multiplied by (-1), that is, they will change their sign. Our results show a nearly theoretical effectiveness of 90%, despite not achieving the recognition (or misrecognition) of implicit elements. These elements represent approximately 4% of the total of sentences analysed for appraisal and include the errors in the recognition of coordinated sentences. On the one hand, we found that 3.6 % of the sentences could not be recognized because they use different connectors than those included in the model; on the other hand, we found that in 8.6% of the sentences despite using some of the described connectors could not be applied the rules we have developed. The percentage relative to the whole group of appraisal sentences in the corpus was approximately of 5%.

Linguistic linked data for sentiment analysis

In this paper we describe the specification of amodel for the semantically interoperable representation of language resources for sentiment analysis. The model integrates "lemon", an RDF-based model for the specification of ontology-lexica (Buitelaar et al. 2009), which is used increasinglyfor the representation of language resources asLinked Data, with Marl, an RDF-based model for the representation of sentiment annotations (West-erski et al., 2011; Sánchez-Rada et al., 2013)

A linked data approach to sentiment and emotion analysis of twitter in the financial domain

Sentiment analysis has recently gained popularity in the financial domain thanks to its capability to predict the stock market based on the wisdom of the crowds. Nevertheless, current sentiment indicators are still silos that cannot be combined to get better insight about the mood of different communities. In this article we propose a Linked Data approach for modelling sentiment and emotions about financial entities. We aim at integrating sentiment information from different communities or providers, and complements existing initiatives such as FIBO. The ap- proach has been validated in the semantic annotation of tweets of several stocks in the Spanish stock market, including its sentiment information.

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.

EUROSENTIMENT: Linked Data Sentiment Analysis

Sentiment and Emotion Analysis strongly depend on quality language resources, especially sentiment dictionaries. These resources are usually scattered, heterogeneous and limited to specific domains of appli- cation by simple algorithms. The EUROSENTIMENT project addresses these issues by 1) developing a common language resource representation model for sentiment analysis, and APIs for sentiment analysis services based on established Linked Data formats (lemon, Marl, NIF and ONYX) 2) by creating a Language Resource Pool (a.k.a. LRP) that makes avail- able to the community existing scattered language resources and services for sentiment analysis in an interoperable way. In this paper we describe the available language resources and services in the LRP and some sam- ple applications that can be developed on top of the EUROSENTIMENT LRP.

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.

Expansión supervisada de léxicos polarizados adaptable al contexto

El análisis de opiniones es un área en la cual múltiples disciplinas han otorgado diferentes enfoques para elaborar modelos que sean capaces de extraer la polaridad de los textos analizados. En función del dominio o categoría del texto analizado, donde ejemplos de categorías son Deportes o Banca, estos modelos deben ser modificados para obtener un análisis de opinión de calidad. En esta tesis se presenta un modelo que pretende elaborar un análisis de opiniones independiente de la categoría a analizar y un extenso estado del arte sobre análisis de opiniones. Se propone un enfoque cuantitativo que haría uso de un léxico polarizado semilla como único recurso cualitativo del modelo. El enfoque propuesto hace uso de un corpus anotado de textos por polaridad y categoría y el léxico polarizado semilla para producir un modelo capaz de elaborar un análisis de opinión de calidad en las distintas categorías analizadas y expandir el léxico polarizado semilla con términos que se adecúan a las categorías procesadas.---ABSTRACT---Sentiment analysis is an area in which multiple disciplines have given diferent approaches to make models that are able to extract the polarity of the analyzed texts. Depending on the domain or category of the analyzed text, where examples of categories are Sports or Banking, these models should be modified to obtain a good opinion analysis. This thesis presents a model that aims to develop a category independent opinion analysis model and a extensive sentiment analysis state of the art. A quantitative approach is proposed that will use a polarized lexicon as the only qualitative resource. The proposed approach uses an annotated corpus by polarity and category and a polarized lexicon seed to produce a model able to develop a good opinion analysis in the various categories analyzed and to expand the polarized lexicon seed with terms that fit the processed categories.

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.