Comparing automated text classification methods

Online social media drive the growth of unstructured text data. Many marketing applications require structuring this data at scales non-accessible to human coding, e.g., to detect communication shifts in sentiment or other researcher-defined content categories. Several methods have been proposed to automatically classify unstructured text. This paper compares the performance of ten such approaches (five lexicon-based, five machine learning algorithms) across 41 social media datasets covering major social media platforms, various sample sizes, and languages. So far, marketing research relies predominantly on support vector machines (SVM) and Linguistic Inquiry and Word Count (LIWC). Across all tasks we study, either random forest (RF) or naive Bayes (NB) performs best in terms of correctly uncovering human intuition. In particular, RF exhibits consistently high performance for three-class sentiment, NB for small samples sizes. SVM never outperform the remaining methods. All lexicon-based approaches, LIWC in particular, perform poorly compared with machine learning. In some applications, accuracies only slightly exceed chance. Since additional considerations of text classification choice are also in favor of NB and RF, our results suggest that marketing research can benefit from considering these alternatives.     

Discovering bank risk factors from financial statements based on a new semi‐supervised text mining algorithm

This paper aims to comprehensively uncover bank risk factors from qualitative textual risk disclosures reported in financial statements, which contain a huge amount of information on bank risks. We propose a new semi‐supervised text mining approach named naive collision algorithm to analyse the textual risk disclosures, which can more accurately identify bank risk factors compared with the typical unsupervised text mining approach. We identified 21 bank risk factors in total, which is far more than identified in previous studies. We further analyse the importance of each bank risk factor and how the importance of each risk factor changes over time.     

Employing structural topic modelling to explore perceived service quality attributes in Airbnb accommodation

This study employs the structural topic model to extract service quality attributes from 242,020 Airbnb reviews in Malaysia. 22 service related topics were extracted from the corpus and four topics have not appeared in previous Airbnb studies. A widely used modified SERVQUAL questionnaire (MSQ) is cross-validated in this study by comparing its service quality attributes with the results of the topic modelling, which indicates that this MSQ can cover general Airbnb service quality attributes. This study also examines the different preferences of Malaysian and international Airbnb users and the changing patterns of the top six service quality attributes during a five-year period. The findings reveal that Malaysian Airbnb users care more about the appearance and location of the property, and international Airbnb users pay more attention to whether the property can accommodate a group of people. In addition, communication with the host is found to play an increasingly important role in Airbnb users’ lodging experiences.     

Copper trade and production of copper,brass and bronze goods in the Oldenburg monarchy: copperworks and copper users in the eighteenth century

This paper explores trade connections – or the lack of such – between copperworks and copper processing plants in the Oldenburg Monarchy in the eighteenth century. Domestic customs areas, high tariffs on raw material export and import bans sought to encourage domestic copper and brass goods production of Norwegian copper raw material, however this was only realised halfway. The raw material from Norway was largely exported, and copper and brass materials used to produce copper-, brass and bronze goods were imported from all over the world. The copperworks and processing plants in the Monarchy never became strongly integrated due to several reasons. First, shareholders of copperworks acquired favourable credit deals abroad, and preferred to export the copper, and second, copper materials had different features and processing plants used all sorts of copper inputs in the making of goods, not only copper raw material. Norway produced mostly gar copper, so copper plants and coppersmiths had to turn elsewhere for other types of copper. Production of copper and brass goods increased, but did not meet the domestic demand partly due to a strong foreign competition. The optimal goal of ‘mercantilist theory’ regarding copper and brass import substitution was not reached.     

Mapping hotel brand positioning and competitive landscapes by text-mining user-generated content

This study uncovers hotel brand positioning and competitive landscape mapping by text-mining user-generated content (UGC). Rather than relying on a single dimension of consumer evaluation, the current study detects brand attributes by using both customer preferences as well as perceptual performance to develop meaningful insights. For this, the study combines content analysis and repertory grid analysis (RGA) to answer three key research issues. 111,986 hotel reviews from two biggest Chinese cities are used to explore and visualize the competitive landscape of six selected hotel brands across three hotel categories. Findings from the study will not only advance the existing literature on brand positioning and competitive landscape mapping but also help practitioners in developing brand positioning strategies to fight competitors within and across hotel categories.     

Evaluation of forestland use in mining operation activities in Turkey in terms of sustainable natural resources

Mining has been necessary for human activities and is conducted in line with this need. The location of mines must sometimes be where land use overlaps with other activities because the location of the mined substance cannot be changed. In Turkey, forestland are the most common of these overlapping areas. Therefore, mining has frequently occurred on forestland in Turkey—and worldwide. After the mining operation activities are conducted, the forestland are rehabilitated and returned to the forest administration. The examination of used and returned areas provides an opportunity to create an optimal situation between “mining for sustainable development” and “protection of forestland.”Accordingly, several questions, such as mining production amounts, degrees of social and economic development of the cities in which enterprises are conducting mining, the quantity of the areas they used for mining activities in forestland, the areas which were returned to the forest administration, operating license areas and operation permit areas, and the life of mining operation, were asked to the mining enterprises in Turkey through the “Survey Monkey” program in 2018. Thus, according to mineral groups, different land use rates were compared with the operating license areas, and the land uses for each mineral group were analyzed by considering the operation activity periods. The results indicate that the sustainability of the use of forestry in mining activities in Turkey has changed in a positive direction, particularly because of changes in mining and environmental legislation in Turkey over the last decade.     

Incorporating textual information in customer churn prediction models based on a convolutional neural network

This study investigates the value added by incorporating textual data into customer churn prediction (CCP) models. It extends the previous literature by benchmarking convolutional neural networks (CNNs) against current best practices for analyzing textual data in CCP, and, using real life data from a European financial services provider, validates a framework that explains how textual data can be incorporated in a predictive model. First, the results confirm previous research showing that the inclusion of textual data in a CCP model improves its predictive performance. Second, CNNs outperform current best practices for text mining in CCP. Third, textual data are an important source of data for CCP, but unstructured textual data alone cannot create churn prediction models that are competitive with models that use traditional structured data. A calculation of the additional profit obtained from a customer retention campaign through the inclusion of textual information can be used by practitioners directly to help them make more informed decisions on whether to invest in text mining.     

The role of text-extracted investor sentiment in Chinese stock price prediction with the enhancement of deep learning

Whether investor sentiment affects stock prices is an issue of long-standing interest for economists. We conduct a comprehensive study of the predictability of investor sentiment, which is measured directly by extracting expectations from online user-generated content (UGC) on the stock message board of Eastmoney.com in the Chinese stock market. We consider the influential factors in prediction, including the selections of different text classification algorithms, price forecasting models, time horizons, and information update schemes. Using comparisons of the long short-term memory (LSTM) model, logistic regression, support vector machine, and Naïve Bayes model, the results show that daily investor sentiment contains predictive information only for open prices, while the hourly sentiment has two hours of leading predictability for closing prices. Investors do update their expectations during trading hours. Moreover, our results reveal that advanced models, such as LSTM, can provide more predictive power with investor sentiment only if the inputs of a model contain predictive information.     

环境规制科技政策对科技进步与经济增长的影响

基于事件系统理论构建系统性、多层次性和综合性的政策量化分析框架，从政策属性、政策目标和政策工具3个维度对我国1985—2015年制定的198条环境规制科技政策进行量化研究。基于政策评估中的工具理性和价值理性两个层面，从环保科技进步效果和经济增长效应两个维度对环境规制科技政策有效性进行评估。研究发现，我国环境规制科技政策工具及政策工具协同对环保科技进步和经济增长的影响存在显著方向性差异。为此，进一步讨论了我国环境规制科技政策工具及工具协同的政策有效性评估结果，可为我国环境规制科技政策的完善和有效实施提供决策依据。     

Motivation and satisfaction of Chinese and U.S. tourists in restaurants: A cross-cultural text mining of online reviews

Tourists with dissimilar cultural backgrounds think and behave differently. Precisely capturing and correctly understanding the cultural difference will help tourism managers generate greater customer satisfaction and increased business revenue. To this end, this paper uncovers and compares the motivation and satisfaction of restaurant tourist customers coming from China and U.S. by investigating their online ratings and reviews. From two major online review communities, customer ratings and reviews have been retrieved, quantified, text-mined, compared, and interpreted using statistics, latent Dirichlet allocation, and frequency analysis. Results suggest that Chinese tourists are less inclined to assign lower ratings to restaurants, and are more strongly fascinated by the food offered, whereas U.S. tourists are more apt to be fun-seeking, and are less uncomfortable with crowdedness.