An Opinion Mining Approach for Drug Reviews in Spanish

Introduction

Opinion mining is an area of great importance for the coarse application that has, focuses on analyzing opinions, sentiments, evaluations, assessments, attitudes, and emotions of people towards entities such as products, services, organizations, individuals, problems, and events (Liu, 2012; Jiménez et al., 2018). This technique, emerged thanks to the accelerated growth of resources available on the web, a representative work of the application, in its beginnings, of opinion mining is the study of Das and Chen (2004) in which they found that, in the case of Amazon Inc., there were cumulatively 70,000 messages by the end of 1998 on Yahoo’s message board, and this had grown to about 750,000 messages by early 2004. The authors found that many of the messages from Amazon’s board offered favorable, pessimistic, confusing, and even spamming opinions, so in their study, they demonstrated the possibility of capturing sentiment by applying statistical language and natural language processing techniques. Thus, sentiment analysis, also known as opinion mining, began to take on relevance. Also, the availability of potential resources for analysis continued to grow exponentially, now through online review sites, shopping sites, and blogs, which increased the challenge of understanding people’s opinions, those opinions important to the decision-making process. Pang and Lee (2008) analyzed surveys of American adults, where they found that consumers reported being willing to pay 20% to 99% more for a 5-star item than a 4-star item in an online store, clearly identifying the importance of knowing other people’s opinions and feelings about a product.

It is noteworthy that opinion mining is not only applied for analysis in the consumption of goods and services; it has also been applied in the field of politics; generally, it recognized by the full range of applications it has at present. On the other hand, to implement opinion mining, it is necessary to use lexical resources that help to carry out sentiment classification. A widely known resource is SENTIWORDNET 3.0 (Baccianella et al., 2014) created for research purposes, which provides automatic annotation of all WORDNET synsets according to their degrees of positivity, negativity, and neutrality. The process presented by the authors consists of two steps,

• 1)

  weak-supervision, semi-supervised learning step;

• 2)

  a random-walk step, used to support sentiment classification in opinion mining.

Thus, the industry surrounding sentiment has grown due to the proliferation of analytics for commercial applications, as well as the exponential increase, in recent years, of social networks and video blogs accessed by millions of users, which generate large amounts of unstructured data. Given this fact, manual revision of data is not very feasible, as a consequence of the number of data continuously generated; therefore, it highlights the use and implementation of opinion mining in current systems (Liu, 2012).

Notwithstanding, opinion mining in the area of health care has increased because of the benefits provided for decision-making, one of the analyses that can be performed using this technique is pharmacovigilance, defined as the science and activities related to the detection, assessment, understanding, and prevention of adverse effects or any other drug-related problem (World Health Organization, 2015).