• Korean Journal of Psychosomatic Medicine
• /
• v.11 no.1
• /
• pp.69-76
• /
• 2003

Objectives: Many researches have been done to compare psychopathology of functional gastrointestinal disorder and inflammatory bowel disease which involves structural change. Recently, many studies focused on the topic of alexithymia. The results from these studies were questionable for lack of valid measures of alexithymia and valid diagnostic criteria of functional gastrointestinal disorders. Therefore, we tried to overcome these two problems and to assess alexithymia, personality characteristics, and other psychopathology. Methods: The subjects consisted of ulcerative colitis group(N=28) who were diagnosed by colonoscopy and biopsy, irritable bowel syndrome group(N=27) who were diagnosed by Rome II criteria and normal control group(N=22). All patients were diagnosed at outpatient department of Kyungpook National University Hospital. All these groups completed three psychological tests, including MMPI, Rorschach test, and well validated TAS-20K(The Korean Version of the 20-Item Toronto Alexithymia Scale). Results: Twenty-five percent of the ulcerative colitis group and 22% of the irritable bowel syndrome group scored in the alexithymia range, compared with 0% of the normal group. In Rorschach test, irritable bowel syndrome group showed high levels of weighted Sum C and EA. Most of clinical scales of MMPI were higher in two gastrointestinal groups than the normal control group. And two gastrointestinal groups showed low ego strength level, but there was no statistical significant difference between them. Conclusion: Two gastrointestinal groups showed high rate of alexithymia, other psychopathological profiles, and low ego strength but there was no significant difference between two groups.

• Investigative Magnetic Resonance Imaging
• /
• v.18 no.4
• /
• pp.341-351
• /
• 2014

Purpose : In this study, the medication effects of Milnacipran and Pregabalin, as well known as fibromyalgia treatment medicine, in fibromyalgia syndrome patients were compared through the change of BOLD signal in pain related functional MRI. Materials and Methods: Twenty fibromyalgia syndrome patients were enrolled in this study and they were separated into two groups according to the treatment medicine: 10 Milnacipran (MLN) treatment group and 7 Pregabalin (PGB) treatment group. For accurate diagnosis, all patients underwent several clinical tests. Pre-treated and post-treated fMRI image with block-designed pressure-pain stimulation for each group were obtained to conduct the statistical analysis of paired t-test and two sample t-test. All statistical significant level was less than 0.05. Results: In clinical tests, the clinical scores of the two groups were not significantly different at pre-treatment stage. But, PGB treatment group had lower Widespread Pain Index (WPI) and Brief Fatigue Inventory (BFI) score than those of MLN treatment group at post-treatment stage. In functional image analysis, BOLD signal of PGB treatment group was higher BOLD signal at several regions including anterior cingulate and insula than MLN treatment group at post-treatment stage. Also, paired t-test values of the BOLD signal in MLN group decreased in several regions including insula and thalamus as known as 'pain network'. In contrast, size and number of regions in which the BOLD signal decreased in PGB treatment group were smaller than those of MLN treatment group. Conclusion: This study showed that MLN group and PGB group have different medication effects. It is not surprising that MLN and PGB have not the same therapeutic effects since these two drugs have different medicinal mechanisms such as antidepressants and anti-seizure medication, respectively, and different detailed target of fibromyalgia syndrome treatment. Therefore, it is difficult to say which medicine will work better in this study.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.8 no.3
• /
• pp.52-63
• /
• 2009

Recently in Korea, in the case of metropolis, the urban signalized intersections are controlled by traffic information center or ITS center. Cheon-an City also established traffic information center through the 1st.-$\sim$3rd. ITS public construction and has managed this center that includes bus information service, traffic information collection and providing service, parking information service, and traffic responsive control system. In the Cheon-an metropolitan traffic signal operation, traffic signal controllers were grouped by the each main traffic flow axes and performed with coordinated signal timing for the signalized arterials, and also cycle and split changed by realtime traffic demands. Cheon-an urban traffic responsive control system was evaluated by intersection delay and speed, then it was verified that the delay decreased and vehicle speed improved. However, the rural signal control system to connect adjacency town was evaluated to have lower status than urban area due to the unimproved TOD (Time of day) plan. Therefore actuated signal control was examined for substitutive control system in isolated signal intersection. The aim of this article is to compare actuated signal control with TOD mode in the rural intersection of Cheon-an and to fine superiority of these two control mode, with evaluation of vehicle delay by using HCM(2000) method and by micro-simulation CORSlM. The result of field test show that actuated signal control gave better performance in delay comparison than the existing TOD signal control. And simulation outcome verified that non-optimized TOD has higher delay than optimized TOD mode, non-optimal actuated mode, and optimal actuated signal control mode. Particularly, these three modes delays had not different values according to the paired sample t-test. This is because small traffic demands were loaded in each links. This suggested actuated signal control is expected to be more effective than TOD mode in some rural isolated intersections which frequently need to survey for traffic volume.

• Geophysics and Geophysical Exploration
• /
• v.10 no.4
• /
• pp.383-392
• /
• 2007

Karstic features and mining-related cavities not only lead to severe restrictions in land utilizations, but also constitute serious concern about geohazard and groundwater contamination. A microgravity survey was applied for detecting, mapping and monitoring karstic cavities in the test site at Muan prepared by KIGAM. The gravity data were collected using an AutoGrav CG-3 gravimeter at about 800 stations by 5 m interval along paddy paths. The density distribution beneath the profiles was drawn by two dimensional inversion based on the minimum support stabilizing functional, which generated better focused images of density discontinuities. We also imaged three dimensional density distribution by growing body inversion with solution from Euler deconvolution as a priori information. The density image showed that the cavities were dissolved, enlarged and connected into a cavity network system, which was supported by drill hole logs. A time-lapse microgravity was executed on the road in the test site for monitoring the change of the subsurface density distribution before and after grouting. The data were adjusted for reducing the effects due to the different condition of each survey, and inverted to density distributions. They show the change of density structure during the lapsed time, which implies the effects of grouting. This case history at the Muan test site showed that the microgravity with accuracy and precision of ${\mu}Gal$ is an effective and practical tool for detecting, mapping and monitoring the subsurface cavities.

• Journal of Korean Society of Environmental Engineers
• /
• v.32 no.10
• /
• pp.916-927
• /
• 2010

In general water treatment process, the disinfection process by chlorine is used to prevent water borne disease and microbial regrowth in water distribution system. Because chlorines were reacted with organic matter, carcinogens such as disinfection by-products (DBPs) were produced in drinking water. Therefore, a suitable injection of chlorine is need to decrease DBPs. Rechlorination in water pipelines or reservoirs are recently increased to secure the residual chlorine in the end of water pipelines. EPANET 2.0 developed by the U.S. Environmental Protection Agency (EPA) is used to compute the optimal chlorine injection in water treatment plant and to predict the dosage of rechlorination into water distribution system. The bulk decay constant ($k_{bulk}$) was drawn by bottle test and the wall decay constant ($k_{wall}$) was derived from using systermatic analysis method for water quality modeling in target region. In order to predict water quality based on hydraulic analysis model, residual chlorine concentration was forecasted in water distribution system. The formation of DBPs such as trihalomethanes (THMs) was verified with chlorine dosage in lab-scale test. The bulk decay constant ($k_{bulk}$) was rapidly decreased with increasing temperature in the early time. In the case of 25 degrees celsius, the bulk decay constant ($k_{bulk}$) decreased over half after 25 hours later. In this study, there were able to calculate about optimal rechlorine dosage and select on profitable sites in the network map.

• Journal of Intelligence and Information Systems
• /
• v.23 no.2
• /
• pp.71-88
• /
• 2017

Language models were originally developed for speech recognition and language processing. Using a set of example sentences, a language model predicts the next word or character based on sequential input data. N-gram models have been widely used but this model cannot model the correlation between the input units efficiently since it is a probabilistic model which are based on the frequency of each unit in the training set. Recently, as the deep learning algorithm has been developed, a recurrent neural network (RNN) model and a long short-term memory (LSTM) model have been widely used for the neural language model (Ahn, 2016; Kim et al., 2016; Lee et al., 2016). These models can reflect dependency between the objects that are entered sequentially into the model (Gers and Schmidhuber, 2001; Mikolov et al., 2010; Sundermeyer et al., 2012). In order to learning the neural language model, texts need to be decomposed into words or morphemes. Since, however, a training set of sentences includes a huge number of words or morphemes in general, the size of dictionary is very large and so it increases model complexity. In addition, word-level or morpheme-level models are able to generate vocabularies only which are contained in the training set. Furthermore, with highly morphological languages such as Turkish, Hungarian, Russian, Finnish or Korean, morpheme analyzers have more chance to cause errors in decomposition process (Lankinen et al., 2016). Therefore, this paper proposes a phoneme-level language model for Korean language based on LSTM models. A phoneme such as a vowel or a consonant is the smallest unit that comprises Korean texts. We construct the language model using three or four LSTM layers. Each model was trained using Stochastic Gradient Algorithm and more advanced optimization algorithms such as Adagrad, RMSprop, Adadelta, Adam, Adamax, and Nadam. Simulation study was done with Old Testament texts using a deep learning package Keras based the Theano. After pre-processing the texts, the dataset included 74 of unique characters including vowels, consonants, and punctuation marks. Then we constructed an input vector with 20 consecutive characters and an output with a following 21st character. Finally, total 1,023,411 sets of input-output vectors were included in the dataset and we divided them into training, validation, testsets with proportion 70:15:15. All the simulation were conducted on a system equipped with an Intel Xeon CPU (16 cores) and a NVIDIA GeForce GTX 1080 GPU. We compared the loss function evaluated for the validation set, the perplexity evaluated for the test set, and the time to be taken for training each model. As a result, all the optimization algorithms but the stochastic gradient algorithm showed similar validation loss and perplexity, which are clearly superior to those of the stochastic gradient algorithm. The stochastic gradient algorithm took the longest time to be trained for both 3- and 4-LSTM models. On average, the 4-LSTM layer model took 69% longer training time than the 3-LSTM layer model. However, the validation loss and perplexity were not improved significantly or became even worse for specific conditions. On the other hand, when comparing the automatically generated sentences, the 4-LSTM layer model tended to generate the sentences which are closer to the natural language than the 3-LSTM model. Although there were slight differences in the completeness of the generated sentences between the models, the sentence generation performance was quite satisfactory in any simulation conditions: they generated only legitimate Korean letters and the use of postposition and the conjugation of verbs were almost perfect in the sense of grammar. The results of this study are expected to be widely used for the processing of Korean language in the field of language processing and speech recognition, which are the basis of artificial intelligence systems.

• Journal of Intelligence and Information Systems
• /
• v.24 no.2
• /
• pp.59-83
• /
• 2018

With the increasing importance of sentiment analysis to grasp the needs of customers and the public, various types of deep learning models have been actively applied to English texts. In the sentiment analysis of English texts by deep learning, natural language sentences included in training and test datasets are usually converted into sequences of word vectors before being entered into the deep learning models. In this case, word vectors generally refer to vector representations of words obtained through splitting a sentence by space characters. There are several ways to derive word vectors, one of which is Word2Vec used for producing the 300 dimensional Google word vectors from about 100 billion words of Google News data. They have been widely used in the studies of sentiment analysis of reviews from various fields such as restaurants, movies, laptops, cameras, etc. Unlike English, morpheme plays an essential role in sentiment analysis and sentence structure analysis in Korean, which is a typical agglutinative language with developed postpositions and endings. A morpheme can be defined as the smallest meaningful unit of a language, and a word consists of one or more morphemes. For example, for a word '예쁘고', the morphemes are '예쁘(= adjective)' and '고(=connective ending)'. Reflecting the significance of Korean morphemes, it seems reasonable to adopt the morphemes as a basic unit in Korean sentiment analysis. Therefore, in this study, we use 'morpheme vector' as an input to a deep learning model rather than 'word vector' which is mainly used in English text. The morpheme vector refers to a vector representation for the morpheme and can be derived by applying an existent word vector derivation mechanism to the sentences divided into constituent morphemes. By the way, here come some questions as follows. What is the desirable range of POS(Part-Of-Speech) tags when deriving morpheme vectors for improving the classification accuracy of a deep learning model? Is it proper to apply a typical word vector model which primarily relies on the form of words to Korean with a high homonym ratio? Will the text preprocessing such as correcting spelling or spacing errors affect the classification accuracy, especially when drawing morpheme vectors from Korean product reviews with a lot of grammatical mistakes and variations? We seek to find empirical answers to these fundamental issues, which may be encountered first when applying various deep learning models to Korean texts. As a starting point, we summarized these issues as three central research questions as follows. First, which is better effective, to use morpheme vectors from grammatically correct texts of other domain than the analysis target, or to use morpheme vectors from considerably ungrammatical texts of the same domain, as the initial input of a deep learning model? Second, what is an appropriate morpheme vector derivation method for Korean regarding the range of POS tags, homonym, text preprocessing, minimum frequency? Third, can we get a satisfactory level of classification accuracy when applying deep learning to Korean sentiment analysis? As an approach to these research questions, we generate various types of morpheme vectors reflecting the research questions and then compare the classification accuracy through a non-static CNN(Convolutional Neural Network) model taking in the morpheme vectors. As for training and test datasets, Naver Shopping's 17,260 cosmetics product reviews are used. To derive morpheme vectors, we use data from the same domain as the target one and data from other domain; Naver shopping's about 2 million cosmetics product reviews and 520,000 Naver News data arguably corresponding to Google's News data. The six primary sets of morpheme vectors constructed in this study differ in terms of the following three criteria. First, they come from two types of data source; Naver news of high grammatical correctness and Naver shopping's cosmetics product reviews of low grammatical correctness. Second, they are distinguished in the degree of data preprocessing, namely, only splitting sentences or up to additional spelling and spacing corrections after sentence separation. Third, they vary concerning the form of input fed into a word vector model; whether the morphemes themselves are entered into a word vector model or with their POS tags attached. The morpheme vectors further vary depending on the consideration range of POS tags, the minimum frequency of morphemes included, and the random initialization range. All morpheme vectors are derived through CBOW(Continuous Bag-Of-Words) model with the context window 5 and the vector dimension 300. It seems that utilizing the same domain text even with a lower degree of grammatical correctness, performing spelling and spacing corrections as well as sentence splitting, and incorporating morphemes of any POS tags including incomprehensible category lead to the better classification accuracy. The POS tag attachment, which is devised for the high proportion of homonyms in Korean, and the minimum frequency standard for the morpheme to be included seem not to have any definite influence on the classification accuracy.

• Asia pacific journal of information systems
• /
• v.23 no.4
• /
• pp.53-84
• /
• 2013

With rapid development of ICT (Information and Communications Technology), new services by the convergence of mobile network and application technology began to appear. Today, smart phone with new ICT convergence network capabilities is exceedingly popular and very useful as a new tool for the development of business opportunities. Previous studies based on Technology Acceptance Model (TAM) suggested critical factors, which should be considered for acquiring new customers and maintaining existing users in smart phone market. However, they had a limitation to focus on technology acceptance, not value based approach. Prior studies on customer's adoption of electronic utilities like smart phone product showed that the antecedents such as the perceived benefit and the perceived sacrifice could explain the causality between what is perceived and what is acquired over diverse contexts. So, this research conceptualizes perceived value as a trade-off between perceived benefit and perceived sacrifice, and we need to research the perceived value to grasp user's continuous intention to use of smart phone. The purpose of this study is to investigate the structured relationship between benefit (quality, usefulness, playfulness) and sacrifice (technicality, cost, security risk) of smart phone users, perceived value, and continuous intention to use. In addition, this study intends to analyze the differences between two subgroups of smart phone users by the degree of personal innovativeness. Personal innovativeness could help us to understand the moderating effect between how perceptions are formed and continuous intention to use smart phone. This study conducted survey through e-mail, direct mail, and interview with smart phone users. Empirical analysis based on 330 respondents was conducted in order to test the hypotheses. First, the result of hypotheses testing showed that perceived usefulness among three factors of perceived benefit has the highest positive impact on perceived value, and then followed by perceived playfulness and perceived quality. Second, the result of hypotheses testing showed that perceived cost among three factors of perceived sacrifice has significantly negative impact on perceived value, however, technicality and security risk have no significant impact on perceived value. Also, the result of hypotheses testing showed that perceived value has significant direct impact on continuous intention to use of smart phone. In this regard, marketing managers of smart phone company should pay more attention to improve task efficiency and performance of smart phone, including rate systems of smart phone. Additionally, to test the moderating effect of personal innovativeness, this research conducted multi-group analysis by the degree of personal innovativeness of smart phone users. In a group with high level of innovativeness, perceived usefulness has the highest positive influence on perceived value than other factors. Instead, the analysis for a group with low level of innovativeness showed that perceived playfulness was the highest positive factor to influence perceived value than others. This result of the group with high level of innovativeness explains that innovators and early adopters are able to cope with higher level of cost and risk, and they expect to develop more positive intentions toward higher performance through the use of an innovation. Also, hedonic behavior in the case of the group with low level of innovativeness aims to provide self-fulfilling value to the users, in contrast to utilitarian perspective, which aims to provide instrumental value to the users. However, with regard to perceived sacrifice, both groups in general showed negative impact on perceived value. Also, the group with high level of innovativeness had less overall negative impact on perceived value compared to the group with low level of innovativeness across all factors. In both group with high level of innovativeness and with low level of innovativeness, perceived cost has the highest negative influence on perceived value than other factors. Instead, the analysis for a group with high level of innovativeness showed that perceived technicality was the positive factor to influence perceived value than others. However, the analysis for a group with low level of innovativeness showed that perceived security risk was the second high negative factor to influence perceived value than others. Unlike previous studies, this study focuses on influencing factors on continuous intention to use of smart phone, rather than considering initial purchase and adoption of smart phone. First, perceived value, which was used to identify user's adoption behavior, has a mediating effect among perceived benefit, perceived sacrifice, and continuous intention to use smart phone. Second, perceived usefulness has the highest positive influence on perceived value, while perceived cost has significant negative influence on perceived value. Third, perceived value, like prior studies, has high level of positive influence on continuous intention to use smart phone. Fourth, in multi-group analysis by the degree of personal innovativeness of smart phone users, perceived usefulness, in a group with high level of innovativeness, has the highest positive influence on perceived value than other factors. Instead, perceived playfulness, in a group with low level of innovativeness, has the highest positive factor to influence perceived value than others. This result shows that early adopters intend to adopt smart phone as a tool to make their job useful, instead market followers intend to adopt smart phone as a tool to make their time enjoyable. In terms of marketing strategy for smart phone company, marketing managers should pay more attention to identify their customers' lifetime value by the phase of smart phone adoption, as well as to understand their behavior intention to accept the risk and uncertainty positively. The academic contribution of this study primarily is to employ the VAM (Value-based Adoption Model) as a conceptual foundation, compared to TAM (Technology Acceptance Model) used widely by previous studies. VAM is useful for understanding continuous intention to use smart phone in comparison with TAM as a new IT utility by individual adoption. Perceived value dominantly influences continuous intention to use smart phone. The results of this study justify our research model adoption on each antecedent of perceived value as a benefit and a sacrifice component. While TAM could be widely used in user acceptance of new technology, it has a limitation to explain the new IT adoption like smart phone, because of customer behavior intention to choose the value of the object. In terms of theoretical approach, this study provides theoretical contribution to the development, design, and marketing of smart phone. The practical contribution of this study is to suggest useful decision alternatives concerned to marketing strategy formulation for acquiring and retaining long-term customers related to smart phone business. Since potential customers are interested in both benefit and sacrifice when evaluating the value of smart phone, marketing managers in smart phone company has to put more effort into creating customer's value of low sacrifice and high benefit so that customers will continuously have higher adoption on smart phone. Especially, this study shows that innovators and early adopters with high level of innovativeness have higher adoption than market followers with low level of innovativeness, in terms of perceived usefulness and perceived cost. To formulate marketing strategy for smart phone diffusion, marketing managers have to pay more attention to identify not only their customers' benefit and sacrifice components but also their customers' lifetime value to adopt smart phone.

• Journal of Intelligence and Information Systems
• /
• v.24 no.3
• /
• pp.1-19
• /
• 2018

Large amount of data is now available for research and business sectors to extract knowledge from it. This data can be in the form of unstructured data such as audio, text, and image data and can be analyzed by deep learning methodology. Deep learning is now widely used for various estimation, classification, and prediction problems. Especially, fashion business adopts deep learning techniques for apparel recognition, apparel search and retrieval engine, and automatic product recommendation. The core model of these applications is the image classification using Convolutional Neural Networks (CNN). CNN is made up of neurons which learn parameters such as weights while inputs come through and reach outputs. CNN has layer structure which is best suited for image classification as it is comprised of convolutional layer for generating feature maps, pooling layer for reducing the dimensionality of feature maps, and fully-connected layer for classifying the extracted features. However, most of the classification models have been trained using online product image, which is taken under controlled situation such as apparel image itself or professional model wearing apparel. This image may not be an effective way to train the classification model considering the situation when one might want to classify street fashion image or walking image, which is taken in uncontrolled situation and involves people's movement and unexpected pose. Therefore, we propose to train the model with runway apparel image dataset which captures mobility. This will allow the classification model to be trained with far more variable data and enhance the adaptation with diverse query image. To achieve both convergence and generalization of the model, we apply Transfer Learning on our training network. As Transfer Learning in CNN is composed of pre-training and fine-tuning stages, we divide the training step into two. First, we pre-train our architecture with large-scale dataset, ImageNet dataset, which consists of 1.2 million images with 1000 categories including animals, plants, activities, materials, instrumentations, scenes, and foods. We use GoogLeNet for our main architecture as it has achieved great accuracy with efficiency in ImageNet Large Scale Visual Recognition Challenge (ILSVRC). Second, we fine-tune the network with our own runway image dataset. For the runway image dataset, we could not find any previously and publicly made dataset, so we collect the dataset from Google Image Search attaining 2426 images of 32 major fashion brands including Anna Molinari, Balenciaga, Balmain, Brioni, Burberry, Celine, Chanel, Chloe, Christian Dior, Cividini, Dolce and Gabbana, Emilio Pucci, Ermenegildo, Fendi, Giuliana Teso, Gucci, Issey Miyake, Kenzo, Leonard, Louis Vuitton, Marc Jacobs, Marni, Max Mara, Missoni, Moschino, Ralph Lauren, Roberto Cavalli, Sonia Rykiel, Stella McCartney, Valentino, Versace, and Yve Saint Laurent. We perform 10-folded experiments to consider the random generation of training data, and our proposed model has achieved accuracy of 67.2% on final test. Our research suggests several advantages over previous related studies as to our best knowledge, there haven't been any previous studies which trained the network for apparel image classification based on runway image dataset. We suggest the idea of training model with image capturing all the possible postures, which is denoted as mobility, by using our own runway apparel image dataset. Moreover, by applying Transfer Learning and using checkpoint and parameters provided by Tensorflow Slim, we could save time spent on training the classification model as taking 6 minutes per experiment to train the classifier. This model can be used in many business applications where the query image can be runway image, product image, or street fashion image. To be specific, runway query image can be used for mobile application service during fashion week to facilitate brand search, street style query image can be classified during fashion editorial task to classify and label the brand or style, and website query image can be processed by e-commerce multi-complex service providing item information or recommending similar item.

• Journal of Intelligence and Information Systems
• /
• v.22 no.3
• /
• pp.71-89
• /
• 2016

Sentiment analysis is used for identifying emotions or sentiments embedded in the user generated data such as customer reviews from blogs, social network services, and so on. Various research fields such as computer science and business management can take advantage of this feature to analyze customer-generated opinions. In previous studies, the star rating of a review is regarded as the same as sentiment embedded in the text. However, it does not always correspond to the sentiment polarity. Due to this supposition, previous studies have some limitations in their accuracy. To solve this issue, the present study uses a supervised sentiment classification model to measure a more accurate sentiment polarity. This study aims to propose an advanced sentiment classifier and to discover the correlation between movie reviews and box-office success. The advanced sentiment classifier is based on two supervised machine learning techniques, the Support Vector Machines (SVM) and Feedforward Neural Network (FNN). The sentiment scores of the movie reviews are measured by the sentiment classifier and are analyzed by statistical correlations between movie reviews and box-office success. Movie reviews are collected along with a star-rate. The dataset used in this study consists of 1,258,538 reviews from 175 films gathered from Naver Movie website (movie.naver.com). The results show that the proposed sentiment classifier outperforms Naive Bayes (NB) classifier as its accuracy is about 6% higher than NB. Furthermore, the results indicate that there are positive correlations between the star-rate and the number of audiences, which can be regarded as the box-office success of a movie. The study also shows that there is the mild, positive correlation between the sentiment scores estimated by the classifier and the number of audiences. To verify the applicability of the sentiment scores, an independent sample t-test was conducted. For this, the movies were divided into two groups using the average of sentiment scores. The two groups are significantly different in terms of the star-rated scores.