=Paper=
{{Paper
|id=Vol-2786/Paper23
|storemode=property
|title=An Ontology-based Sentiment Analysis Model towards Classification of Drug Reviews
|pdfUrl=https://ceur-ws.org/Vol-2786/Paper23.pdf
|volume=Vol-2786
|authors=Sridevi. U.K,Shanthi. P
|dblpUrl=https://dblp.org/rec/conf/isic2/KP21
}}
==An Ontology-based Sentiment Analysis Model towards Classification of Drug Reviews==
https://ceur-ws.org/Vol-2786/Paper23.pdf
An Ontology-based Sentiment Analysis Model towards
Classification of Drug Reviews
Sridevi. U.Ka, Shanthi. Pb
a
PSG College of Technology, Tamilnadu, India
b
Sri Krishna College of Engineering and Technology, Tamilnadu, India
Abstract
Due to the abundance of information that can be obtained through careful analysis of such data, the
need to analyze user-generated data on the internet has recently gained significance. Mining and
analyzing such data have become an important aspect for the companies to understand the opinion
of the people on specific drug information. The purpose of this research project is to use deep
learning models to analyze the feelings of patient reviews to determine the polarity of opinions
expressed in reviews that may be positive or negative. All positive relation instances are combined
Minto one class using a binary Support Vector Machine. Logistic regression and Long short-term
memory (LSTM) networks are used for sentiment classification of the drug review collected from
the Drug Reviews dataset. Logistic regression can be used for the prediction of a group membership.
The power of prediction with the given features can be identified using the LSTM classification. The
accuracy of 80% is achieved with LSTM classification method.
Keywords
Sentiment Analysis, Opinion Mining, Deep Learning
1. Introduction
Data from social media offer useful information Because of its wide scope, social media is a
on general health conditions. This applies valuable medium for exchanging knowledge
including in situations when clients of social relating to health. It makes it a strong choice for
networking sites are still not conscious of an tasks of controlling public safety, especially for
improvement in personal wellbeing. Depression co-vigilance in pharmaceuticals. The extracted
recently developed as the primary disorder for from social media about the adverse drug reaction
mental health concern among social scientists, study helps the people working in the healthcare
because it is a fairly common mental illness and sector [4]. The collection of technical knowledge
affects a variety of habits and trends. The advent from social platforms is difficult, primarily
and widespread acceptance of digital media make because of the brief and largely unstructured form
significant-time population growth-scale overall of the document especially in comparison with
sentiment estimation possible, an exceptional more detailed and structured medical documents.
ability that has important consequences for our The machine learning methods to determine the
understanding of the social activity. In automatic occurrence of mental illness and disorder among
emotion analysis, it is crucial to consider the participants on social media data [17]. The social
aspects of feeling dictionaries relate to their media data were collected, feature extracted, and
classification. The dictionary-based approaches classified effectively between depressed and
will keep playing a positive role-they are easy and healthy material, and comparing
suitable for online-scale data set [16].
___________ Depression is so prevalent in patients with a
ISICβ21:International Semantic Intelligence Conference, February mental disorder that it can be difficult to
25-27, 2021, Delhi, India determine whether a drug has directly induced
EMAIL: Srideviunni@gmail.com (A. 1); shanthi.slm@gmail.com
(A. 2)
depression, or whether the association is
ORCID: 0000-0003-2445-9193 (A. 1); 0000-0001-7721-8305 coincidental. We review the research on the
Β©οΈ 2020 Copyright for this paper by its authors. Use permitted under Creative
Commons License Attribution 4.0 International (CC BY 4.0). relationship between treatment and depressive
CEUR Workshop Proceedings (CEUR-WS.org) symptoms in this report. Mental illness is one of
the psychiatric disorders which is widespread and
enduring. This gives a real significant societal product sentiment ratings. The Internet such as
pressure, including spending on welfare and even social media platforms, blogs, online reviews, and
rates of suicide. The applications for medicines, or websites produces large amounts of information
drugs, can be a useful strategy for reducing the life in the context of user perspective. Common
cycle [6]. In this study we introduce a model of consumer opinions on the internet have a major
drug design by forecasting direction for a impact on authors, service suppliers, and decision-
particular disorder based on profiling of drug makers. It needs the unorganized kind of data.
expression, concentrating on psychological Digital medical forums offer a convenient way for
requirements. The sentiment analysis [14] patients to access health information and to
received a general overview of the deep learning communicate beyond clinical practice with
model. Sentiment analytics tools rely on doctors and users. Statistics show patients,
collections of words and phrases that do have specifically those recovering from serious
positively and negatively negative associations. illnesses, benefit greatly from the data provided in
The deep learning models that are focused on social media platforms and websites.
Sentiment Analysis with semantic ratings are
provided with them to decide the final polarization Throughout the study in [23], a classification
of a document. The effective tool introduced for system of attempts had been used to collect
consumer-generated content predictive analytics participant detail interests throughout online
on drug user reviews has not been thoroughly wellbeing platforms and also to uses the support
examined in comparison with other broader vector Machine (SVM) which is a multiclass
contexts, such as ratings and reviews [11]. A classifier to identify initial message articles as per
clause-level analysis methodology for opinion is their underlying purpose. Furthermore, large
established. The purpose of the analysis was to volumes of informal and consolidated content
make models forecasting the author's generated on those platforms make ingesting and
depressiveness based on his / her written text's retrieving relevant information challenging
statistical semantic indicators. Effective among participants. Ability to understand
identification of a patient at risk of depression is participant purposes will allow platforms to
advantageous for both patient and family in the recognize and propose necessary information to
beginning stages and a severe case. The study participants by posting off topics that don't
presents the original method of monitoring correspond to relevant intent and purpose. The
disorder focused on an evaluation of what the aim of the work seems to be to build software
participant uses the word to identify the possibility techniques to make wellness digitally. Developing
of mental illness. The objective of this to apply new applications for drug treatments, or
ontology to the drug reviews to analyze the repositioning drugs, can be a useful strategy for
sentiment analysis. This paper is organized as shortening the growth cycle. The drug review
follows: Section 2 includes the related work. system [19] developed has taken the reviews of
Section 3 describes the ontology-based sentiment the patient and opinion identification on the
analysis for drug reviews. Sentiment score review is extracted based on age and gender-wise
implementation is described in Section 4. It classification. Within this analysis the work dealt
further discusses the implementation of machine with a drug discovery approach by identifying
learning algorithms to improve the drug review significance for a disorder based on features of
classification. Concluding remarks are provided drug content, depending on medical uses.
in Section 5.
A tool [13] for polarity analysis of drug
interactions of consumers using contextual
2. Literature Study
information. Polarity identification is the key
subtask of emotion evaluation and opinion
Sentiment analysis on medical drugs in general
mining, excellently-known issues in the analysis
collecting user experience data in particular is a
of natural language which have received growing
complex research topic and the key challenge is
interest in recent times. Current techniques rely on
the absence of labeled data, which is important
the contextual portion of the document by which
for the classification of emotions. Opinion mining
feeling is conveyed directly by specific terms,
or sentiment analysis is used to handle cost-
called terms of emotion. Nevertheless, these
effective and detailed information related to
methods are also far from being successful in the
extracting large volumes of data to help determine
polarity classification of the observations of
patients. The word embedding method is individuals, illnesses, adverse drug reactions, and
commonly used in biomedical Natural Language diseases. Nevertheless, there was very little
Processing (NLP) technologies because it research on the medical environment and
provides vector representations of words numerous studies are provided to support the
capturing the semantic features of words and the recommended framework.
semantic association among terms [22]. Most Based on user feedback on different medications
biomedical applications use various textual tools together with associated disorders and an overall
to learn word embedding and extend this word patient experience score of 10 ratings [3]. The data
embedding to biomedical implementations. The was collected from drug analysis pages crawled
work [20] concentrated on forecasting the degree websites. The aim has been to study awareness
of drug interaction among the users who had examines of drug experience across various
encountered the drug influence already. Opinion dimensions, i.e. feelings experienced on particular
mining on the drug feedback is evaluated and aspects such as effectiveness. Many online retailer
classification approaches output on drug reviews sites will write comments about purchasing goods
is completed. The work exploring the effects of for customers due to the advancement of online
social networking on patients' studies show social marketing and internet technology. Consumer
counseling results in a positive impact on certain feedback shared opinions on services or goods
health problems. The opinion mining approach usually referred to as consumer input. Opinion
used in this study focused on forecasting the analysis through consumer feedback regarding
degree of drug satisfaction between the other users services has become an important area of study.
who have felt the impact of a medication
previously. Across certain disciplines, such as The work [18] suggested a great idea to
behavioral psychology, data sets derived from effectively consider opinions or sentences of each
social media are of importance. However, element from user reviews. The research focus
technological resources are anything but was on getting the processes of words sentiment
sufficient and unique solutions are desperately about the brand aspect analysis on text. The
needed. The research work of [21] explored features and opinions collected are beneficial for
applying data mining to the field of psychology to creating a concise description that provides a
identify distressed users in social media networks. valuable insightful strategy to help both the
Next, a data analysis approach is suggested using customer and marketers manage the most suitable
vocabularies and laws to measure each blog's consumer option [19]. To promote analysis and
anxiety tendency. Furthermore, a design for interaction among emotional research and mental
depression detection is founded on the suggested health diagnostics. The work [8] suggested and
approach and 10 characteristics of distressed users explored an ontological model for specifically
extracted from research studies. The author [15] describing the dynamic observable behaviors
suggested a module focused on sentiment analysis between emotional individuals and mental
to receive emotions and thoughts through texts illnesses. The framework was based on improved
linked to the medical field at the analysis and sentiment, impact, and behavior classification
entity level. within the predefined ontology, as well as
Contextual knowledge means the thoughts, psychological conditions in the Ontology of
values, emotions, and attitudes associated with Mental illness. To attempt to formulate the
people concerning various issues of concern. This linkages, this endeavor also builds on advances to
category of knowledge is of great value to conceptual ontology involving the relation
corporations, associations, or persons, as it helps between ordinary and abnormal. This conceptual
others to engage in acts that support them. analytical structure is important for requirements
Furthermore, sentiment analysis is the domain like classification of behavioral assessment
where contextual knowledge is studied from criteria, health data analytics as well as the
language processing, analytical cognitive science, incorporation and transmission of study findings
knowledge recovery, and machine learning among domains. Social networking has become
strategies. Research of emotions is very beneficial incredibly popular as a medium for exchanging
in different areas, including economics, information related to social safety. By the use of
advertising, hospitality, etc. Also, the healthcare natural language processing (NLP) methods, this
sector means a broad range of opportunities to knowledge can be used for health promotion
receive resources through data analysis, including surveillance objectives, notably for
collecting knowledge as to the disposition of pharmacovigilance. Nonetheless, social
networking sites terminology is extremely consideration. The reviews are categorized based
expressive and frequently non-technical, concise, on the different set of keywords related to mental
and difficult to obtain user-expressed health disorders.
terminology. Significant progress is made in An ontological model is capable of
overcoming such problems, and modern NLP expressing significant categories in the context of
methods focused on algorithms were underused. the entity or form of impact embodied by a
The study by [12] was developing a learning ontology structure. Also, the terms that could be
algorithm for extracting details of adverse drug annotated include not only those that are
reactions (ADRs) through largely unofficial social ontologically activating but also those that the
media text. entity action based on influencing condition. It
could also be that one factor may cause the
Starting from the DSM-5 definitions of formation of psychological disorders, while at the
several common mental illnesses has taken into same time getting an excitatory fundamental role
consideration the domain of mental illnesses [1]. of phobias. Conceptual structures are therefore
In this regard, the preference of a combined correlated with a structure of types of situations to
method to the interpretation of a behavioral study reflect these various kinds of influence. The
by integrating ontology method with a schematic Mental Disease Ontology describe and classify
interpretation of information based on semantic mental disease given by OBO foundry
frames. Analysis of sentiment is the method of organization. The main element of depression is
getting information from the thoughts, fear, anxiety disorder, and panic disorder, social
perceptions, and sentiments of the persons against phobia is taken into account
individuals, situations. The work [2] suggested a An Emotion Ontology (EM) [7]
study using the objective natural language constructed for the cognitive disorder's domain.
processing approach of the rule-based domain. The concepts it classifies and describes include
The approach suggested categorizes emotional sentiments, emotional states, and related to
and factual sentences from feedback and feedback various identities such as behavioral conduct,
on the forums. SentiWordNet extracts the body movements, personal beliefs, etc.,[9].
semantic rating of subjective phrases to measure Emotion Ontology and Hamilton Anxiety Rating
their sentiment analysis, depending on the Scale (HAM-A) are used to classify a collection
conceptual sentence structure. The framework of sentences describing those feelings that
[10] incorporating CNN with bi-directional long- individuals have [5]. Table .1 shows the words
term memory (Bi-LSTM) to detect harmful drug mapping with the emotions. Score patients by
effects based on user feedback through social seeking the response that better explains the
media and wellbeing-related blogging. degree to which they have these problems. The
key rating is given by the patient on a scale of 1 to
5. The low rating value is 1 which means that they
3. Ontology-based Sentiment Analysis
have a negative opinion on the drug. If the rating
is high then it's a positive rating and it means that
The drug feedback studies at
the drug has helped or cured the patient.
www.askapatient.com show that approximately
40 percent of texts are polarized details that reflect Table.1 Words mapping with emotions
the perceptions of people with good or negative
symptoms. The above suggests how conventional Concepts Words
Emotional mental Fear, anger, surprise
solutions to the analysis of polarity, which focuses
on qualitative assumptions, find only segments of
Emotional Crying, smiling, laughing
the information and overlook a significant amount Action
of useful knowledge. In this work set of reviews
about the drugs are taken and gender-wise Depressed/Dull Lonely, Loss of interest, lack of
comments and opinion words in each review are pleasure in hobbies, depression.
extracted. The opinion words help in deciding the
positive and negative sentences of the review. Fears Afraid, Abandon
Anxious uncertainty, uneasiness, misery,
Mental depression is taken into the category of suspense, panic, concern, doubt
study. There are several different types of mental
disorders and in this study e depression, anxiety Tension Not relaxing, Feelings of tension,
disorders, and drug abuse are taken into tears, shouting
.
The ontology-based sentiment analysis can be opinion terms and the analysis of the direction of
used to measure each post's average polarity and that word of view. The phrase Opinion inclination
emotion. By extracting the concepts related to the defines the review's positives and negatives. The
ontology class. The sentences were annotated task involved in polarity classification is to extract
with an opinion based on the characteristics and the opinion words from the review and to identify
with a polarity of opinion. A collection of word is positive or negative.
feedback and a set of opinion terms associated
with the feature in a sentence are mapped because Table .2
of the set of emotion labels. Examples of polar facts about drugs.
No Facts about drug Polarity
1 This antidepressant has Positive
helped me quite a bit. It
works particularly well for
my anxiety which has disap-
peared.
2 suicide thoughts perma- Negative
nently, worsening
depression
The semantic analysis can be inferred from the
wordnet dictionary. Based on opinion terms found
Figure.1: Ontology-based sentiment analysis on in the phrase the + 1 score is determined for
drug review positive words and -1 for negative words. The
final result is positive if the total score contains
Figure 1 shows the ontology-based sentiment more positive words. Determine each sentence's
analysis method based on drug reviews. The final values, and evaluate to determine sentiment
sentiment lexicons need to be adapted to cope analysis. Thus, the importance of opinions for
with the medical term due to different language both phrase and review is measured by allocating
usage clinical practices. Consider the "positive" the aggregate opinion weight to the phrase and
word. This term is often used differently in evaluating it using Eq.1 and 2.
clinical language from our normal use. A "positive
finding" often results in negative consequences βππ=1 πππππ(π) (1)
ππππ‘πππππππππ (π) =
for a patient. Polarity classification, which π
identifies texts into different categories such
as positive or negative is an important subtask of where, SentimentScore (S), is the sentence posi-
opinion mining. In recent times the concept of tive score or or negative score. Score (i) is the word
polarity has gained raising focus. Though much sentiment score of ith word in sentence S. n is the total
work had been done in this field, almost all of the no. of words in Sentence S.
current methods have focused on the contextual
βπ
π=1 πππππ(π) (2)
aspect of the document where the feeling is π
ππ£πππ€πππππ(π
) = π
conveyed directly by using specific terms called
emotion terms. The sentence gives a positive
opinion or negative opinion. Furthermore,
emotions are displayed not just in contextual 4. Experiment Results
declarations, but they can also be conveyed
in fact-based phrases that are harder to deal with. The collection of data obtained from
In certain phrases, a descriptive paragraph can www.askapatient.com for the drug analysis. The
have polarization without even being contextual. databases were created by gathering common
The patient opinion can be labeled as positive or drug feedback such as Cymbalta, Celexa, Effexor
negative without any sentences of feeling also. xr, Lexapro, Wellbutrin. The selected drugs are
Table 2 shows an example of polarity in drug used for treating people with depression. The
reviews. Ontology of Drug Adverse Events preference for such medicines was selected at
(ODAE) is obtained from the ONTOBEE website random from the review web site's list of its most
and provides a representation of adverse effects. highly ranked drugs. A total of 113,093 comments
The ODAE serves as a knowledge base. The are collected and 4773 comments are extracted
emotional category includes the classification of based on evaluations of depression and anxiety.
Figure 2 shows the distribution of men and represented as vectors is fed into the deep neural
women in the reviews. The drug reviews will layer. LSTM framework describes layer
contain the rating, side effects, duration, and parameter and level. The first layer is the
dosage information. Figure 3 shows the rating and embedded layer which represents each term using
side effects based on the reviews. The problem 32 length vectors. 100 units of memory units are
was a multi-classification problem for the included in the layer. The dense layer using the
sentiment classification of the drug reviews on the sigmoid activation function and the probability of
AskaPatient forum. The sample sentence score is each class output are given by SoftMax. For
given in Table 3 is used for sentiment classification Dense output layer is for making 0
classification. or 1 predictions using sigmoid activation function.
The log loss function is used in this binary
Long Short-Term Memory Network classification. The model is suitable for only 2
(LSTM) which is an RNN version solves the issue epochs since it overfits the problem quickly.
of categorizing opinions. Based on [8] Precision is the percentage of classified samples
contribution of the work is the integration of that are correct. The ratio of correctively classified
ontology information with a neural network to the total review is calculated for Recall. F-score
classification model. The LSTM architecture for is a measure that combines the score of precision
analysis of sentiments consists of a word and recall score. Table 4 and Table 5 show the F-
embedding layer as data. score accuracy of the model and comparison of
the models.
Table.3 Sentiment score of the reviews
ππ. ππ πΆππππππ‘ππ¦ ππππ π πππππ πππ£πππ€π
Sentence Sentence Average ππππππ πππ = (3)
πππ‘ππ ππ. ππ ππππ π πππππ πππ£πππ€π
ID Sentiment
1 Horrible -0.2046324
medicine for ππ.ππ πΆππππππ‘ππ¦ ππππ π πππππ πππ£πππ€π
π
πππππ = (4)
me. Switched πππ‘ππ ππ.ππ πππ£πππ€π
from lexapro
which crapped
out after 8 2 β (ππππππ πππ β π
πππππ)
ππππππ = (5)
years ππππππ πππ + π
πππππ)
2 Sweating. -0.1364216
Table.4 Accuracy of the model
Weird dreams
3 Mood swings, -0.6708204
Polarity Precision Recall F-Score Accuracy
insomnia,
depression,
Positive 0.63 0.63 0.63 0.63
irritability
Negative 0.56 0.64 0.60
4 Weight gain, -0.2834734
Neutral 0.67 0.62 0.65
brain zaps if
dose missed
5 suicide -1.118034 Table. 5 Comparison of the models
thoughts
permanently, Dataset (Predicted)
worsening Positive Negative Neutral
depression
6 Improved in 0.8660254 Logistic 79.8 79.0 81.5
confidence regression
7 Slept and no 0.375 (LR)
fear SVM 85.4 86.4 85.4
LSTM 92.3 88.0 85.4
A domain-specific ontology is the included in
the semantic embedding layer. The terms
Table.6 Classification error 5. Conclusion
Algorithm Mean Mean Online review platforms and forums on
emotional wellbeing are a huge pool of knowledge
Absolute Squared
that traditional psychology has not yet taped in.
Error Error
This paper demonstrates how natural text analysis
Logistic 0.332 0.315 of large databases can enable and speed up THE
Regression collection of information, analysis, and opinion
extraction by thousands of thousands of people
SVM 0.399 0.357 through millions of comments and posts.
Becoming able to present many of the results of
LSTM 0.171 0.178 this study with previous studies in psychiatric
research and even experience in a few cases shows
the importance of text analysis on large public
Table 6 shows the classification errors of a data is feasible. The effective method can make a
different model. The success of the prediction significant shift in emotional wellbeing research
model can be estimated using the classification and address the topic of counseling and care
table, 80% of the observations were correctly reform and administration. Data on drug analysis
classified to the appropriate group. 12 out of 15 in social networking sites and health forums will
observations were classified correctly. Table 7 provide us with useful resources. In future studies,
shows the sample new reviews classification our ongoing research will concentrate on
success and failure examining possible drug review interactions and
evaluating the effect of drug analysis technologies
Table .7 Actual and Predicted success and failure for adverse drug recognition, so that
on new reviews comprehensive online review data will better
serve individuals 'healthier lives.
Predicted Results
Success Failure 6. Reference
Actual Success 6 1
[1] Amoretti, Maria Cristina, Frixione, Marcello
Result Failure 2 6 Lieto, Antonio & Adamo, Greta. Ontologies,
8 7 Mental Disorders, and Prototypes. In Matteo
Vincenzo D'Alfonso & Don Berkich (eds.),
The observations that are correctly classified are On the Cognitive, Ethical, and Scientific
24.9%. A similar calculation of a failure gives Dimensions of Artificial Intelligence. Berlin,
24.9%. 25% of the reported data combining Germany: Springer Verlag. pp.189-204
success and failure is 62.2% which is obtained (2019)
using classification prediction. The observed 80% [2] Aurangzeb Khan, Baharum Baharudin, and
is higher than the classification probability of Khairullah Khan. Sentiment Classification
62.2%, which would support the usefulness of the Using Sentence-level Lexical Based
classifier. Our method achieved 80% of the Semantic Orientation of Online Reviews.
accuracy from the training set of 30% and appears Trends in Applied Sciences Research, 6:
consistent from the training set of 70%. The 1141-1157. (2011)
findings show that the size of the manually labeled [3] Felix GrΓ€Γer, Surya Kallumadi, Hagen
data didn't affect our model. The results of the data Malberg, and Sebastian Zaunseder. Aspect-
set for the classification are given in the table. To Based Sentiment Analysis of Drug Reviews
determine the accuracy of the method, the overall Applying Cross-Domain and Cross-Data
misclassification rate is determined as well. SVM Learning. In Proceedings of the 2018
is better than logistic classification The LSTM is International Conference on Digital Health
better than SVM. The improved LSTM (DH '18). ACM, New York, NY, USA, 121-
performance may be due to deep architectures 125. (2018)
with hidden layers that more effectively reflect [4] Gupta, S., Pawar, S., Ramrakhiyani, N. et
intelligent behavior than shallow architectures al. (2018). Semi-Supervised Recurrent
such as SVM. Neural Network for Adverse Drug Reaction
mention extraction. BMC Advanced Computer Science and
Bioinformatics 19, 212. (2018) Applications. 8. 6. (2017)
[5] Hamilton M. A rating scale for depression. [15] RamΓrez-Tinoco, Francisco &Alor-
(1960). J Neurol Neurosurg Psychiatry. 23. HernΓ‘ndez, Giner& SΓ‘nchez-Cervantes, JosΓ©
56β62. (1960) & Salas Zarate, MarΓa& Valencia-GarcΓa,
[6] Kai Zhao and Hon-Cheong So.. A machine Rafael. Use of Sentiment Analysis
learning approach to drug repositioning Techniques in Healthcare
based on drug expression profiles: Domain.10.1007/978-3-030-06149-4_8.
Applications to schizophrenia and (2019)
depression/anxiety disorders. IEEE Journal [16] Reagan, A.J., Danforth, C.M., Tivnan, B. et
Of Biomedical And Health Informatics. 23. al. Sentiment analysis methods for
3. (2019) understanding large-scale texts: a case for
[7] Hastings J, CeustersW, Smith B,Mulligan K. using continuum-scored words and word
The emotion ontology: enabling shift graphs. EPJ Data Sci. 6. 28. (2017)
interdisciplinary research in the affective [17] Reece, A.G., Reagan, A.J., Lix, K.L.M. et
sciences. In: International and al. Forecasting the onset and course of
Interdisciplinary Conference on Modeling mental illness with Twitter data. Sci
and Using Context. Berlin: Springer. p. 119β Rep. 7. 13006. (2017)
23. (2011) [18] Su SuHtay, KhinThidar Lynn, Extracting
[8] Lamurias, A., Sousa, D., Clarke, L. et al. BO- Product Features and Opinion Words Using
LSTM: classifying relations via long short- Pattern Knowledge in Customer Reviews,
term memory networks along biomedical The Scientific World Journal. Article
ontologies. BMC Bioinformatics 20, 10 ID 394758, 5. (2013)
(2019). [19] Vrushali Moon et al. Extraction of Drug
[9] Larsen RR and Hastings J. From Affective Reviews by Specific Aspects for Sentimental
Science to Psychiatric Disorder: Ontology as Analysis (IJCSIT) International Journal of
a Semantic Bridge. Front. Psychiatry. Computer Science and Information
9.487.(2018) Technologies. 7 .3. 1186-1189. (2016)
[10] Min Zhang and Guohua Geng. Adverse [20] VinodhiniGopalakrishnan,Chandrasekaran
Drug Event Detection Using a Weakly Ramaswamy. Patient opinion mining to
Supervised Convolutional Neural Network analyze drugs satisfaction using supervised
and Recurrent Neural Network Model, learning, Journal of Applied Research and
Information. 10, 276. (2019) Technology. 15. 311β319. (2017)
[11] Na Jin-Cheon and Kyaing, Wai. . Sentiment [21]Wang, Xinyu and Zhang, Chunhong & Ji,
Analysis of User-Generated Content on Drug Yang & Sun, Li & Wu, Leijia & Bao, Zhana.
Review Websites. Journal of Information A Depression Detection Model Based on
Science Theory and Practice. 3. 6-23. (2015) Sentiment Analysis in Micro-blog Social
[12] Nikfarjam A, et al. J Am Med Inform Assoc Network. 7867. 201-213. (2013)
.22:671β681. (2015) [22]Yanshan Wang, Sijia Liu, NaveedAfzal,
[13] Noferesti S, Shamsfard M. Using Linked MajidRastegar-Mojarad, Liwei Wang,
Data for polarity classification of patients' FeichenShen, Paul Kingsbury, Hongfang
experiences, J Biomed Inform. 57.6-19. Liu.. A Comparison of Word Embeddings for
(2015) the Biomedical Natural Language
[14] QuratTul Ain, Mubashir Ali, AmnaRiaz, Processing, J Biomed Inform. 87. 12-20.
AmnaNoureen, Muhammad Kamran, Babar (2018)
Hayat, and A. Rehman. Sentiment Analysis [23] Zhang T, Cho JH, Zhai C . Understanding
Using Deep Learning Techniques: A user intents in online health forums, IEEE J
Review, (IJACSA) International Journal of Biomed Health Inform. 19.4.1392-8. (2015)