Sentiment Lexicon Research Articles

Recognizing emotions in restaurant online reviews: a hybrid model integrating deep learning and a sentiment lexicon

PurposeThis study aims to establish a model for rapid and accurate emotion recognition in restaurant online reviews, thus advancing the literature and providing practical insights into electronic word-of-mouth management for the industry.Design/methodology/approachThis study elaborates a hybrid model that integrates deep learning (DL) and a sentiment lexicon (SL) and compares it to five other models, including SL, random forest (RF), naïve Bayes, support vector machine (SVM) and a DL model, for the task of emotion recognition in restaurant online reviews. These models are trained and tested using 652,348 online reviews from 548 restaurants.FindingsThe hybrid approach performs well for valence-based emotion and discrete emotion recognition and is highly applicable for mining online reviews in a restaurant setting. The performances of SL and RF are inferior when it comes to recognizing discrete emotions. The DL method and SVM can perform satisfactorily in the valence-based emotion recognition.Research limitations/implicationsThese findings provide methodological and theoretical implications; thus, they advance the current state of knowledge on emotion recognition in restaurant online reviews. The results also provide practical insights into intelligent service quality monitoring and electronic word-of-mouth management for the industry.Originality/valueThis study proposes a superior model for emotion recognition in restaurant online reviews. The methodological framework and steps are elucidated in detail for future research and practical application. This study also details the performances of other commonly used models to support the selection of methods in research and practical applications.

A large-scale sentiment analysis using political tweets

<p><span lang="EN-US">Twitter has become a key element of political discourse in candidates’ campaigns. The political polarization on Twitter is vital to politicians as it is a popular public medium to analyze and predict public opinion concerning political events. The analysis of the sentiment of political tweet contents mainly depends on the quality of sentiment lexicons. Therefore, it is crucial to create sentiment lexicons of the highest quality. In the proposed system, the domain-specific of the political lexicon is constructed by using the supervised approach to extract extreme political opinions words, and features in tweets. Political multi-class sentiment analysis (PMSA) system on the big data platform is developed to predict the inclination of tweets to infer the results of the elections by conducting the analysis on different political datasets: including the Trump election dataset and the BBC News politics. The comparative analysis is the experimental results which are better political text classification by using the three different models (multinomial naïve Bayes (MNB), decision tree (DT), linear support vector classification (SVC)). In the comparison of three different models, linear SVC has the better performance than the other two techniques. The analytical evaluation results show that the proposed system can be performed with 98% accuracy in linear SVC.</span></p>

A novel meta-embedding technique for drug reviews sentiment analysis

<p>Traditional word embedding models have been used in the feature extraction process of deep learning models for sentiment analysis. However, these models ignore the sentiment properties of words while maintaining the contextual relationships and have inadequate representation for domainspecific words. This paper proposes a method to develop a meta embedding model by exploiting domain sentiment polarity and adverse drug reaction (ADR) features to render word embedding models more suitable for medical sentiment analysis. The proposed lexicon is developed from the medical blogs corpus. The polarity scores of the existing lexicons are adjusted to assign new polarity score to each word. The neural network model utilizes sentiment lexicons and ADR in learning refined word embedding. The refined embedding obtained from the proposed approach is concatenated with original word vectors, lexicon vectors, and ADR feature to form a meta-embedding model which maintains both contextual and sentimental properties. The final meta-embedding acts as a feature extractor to assess the effectiveness of the model in drug reviews sentiment analysis. The experiments are conducted on global vectors (GloVE) and skip-gram word2vector (Word2Vec) models. The empirical results demonstrate the proposed meta-embedding model outperforms traditional word embedding in different performance measures.</p>

Analisis Sentimen Brand Ambassador Artis Korea Selatan pada Produk Indonesia dengan Lexicon

The use of South Korean artists as brand ambassadors for Indonesian products is one way that e-commerce and local companies do to attract consumers. However, it reaped positive and negative responses from the public. The social media used to issue opinions regarding this matter is Twitter. This study conducts sentiment analysis from tweets on Twitter regarding this topic. The method used is Lexicon sentiment. The sentiment Lexicon works by first making a dictionary of opinion words and matching them with the sentiment words contained in the Lexicon dictionary. The Lexicon dictionary used is InSet Lexicon. Sentiment classification is divided into 2 namely, positive and negative. Data collection uses the Twitter API with the keyword "BA Artis Korea" as many as 366 data tweets. The research phase consists of data collection, text preprocessing, Lexicon sentiment, visualization, and testing. The results of the classification using the Lexicon method obtained 67.1% positive and 32.9% negative. It was found that the use of South Korean artists to become brand ambassadors for Indonesian products received positive sentiment. Accuracy, precision, and recall values ​​obtained by looking at the confusion matrix table are 78% accuracy, 68% precision, and 61% recall.

The Use of Natural Language Processing for Computer-Aided Diagnostics and Monitoring of Body Image Perception in Patients with Cancers.

Head and neck cancers (H&NCs) constitute a significant part of all cancer cases. H&NC patients experience unintentional weight loss, poor nutritional status, or speech disorders. Medical interventions affect appearance and interfere with patients' self-perception of their bodies. Psychological consultations are not affordable due to limited time. We used NLP to analyze the basic emotion intensity, sentiment about one's body, characteristic vocabulary, and potential areas of difficulty in free notes. The emotion intensity research uses the extended NAWL dictionary developed using word embedding. The sentiment analysis used a hybrid approach: a sentiment dictionary and a deep recursive network. The part-of-speech tagging and domain rules defined by a psycho-oncologist determine the distinct language traits. Potential areas of difficulty were analyzed using the dictionaries method with word polarity to define a given area and the presentation of a note using bag-of-words. Here, we applied the LSA method using SVD to reduce dimensionality. A total of 50 cancer patients requiring enteral nutrition participated in the study. The results confirmed the complexity of emotions in patients with H&NC in relation to their body image. A negative attitude towards body image was detected in most of the patients. The method presented in the study appeared to be effective in assessing body image perception disturbances, but it cannot be used as the sole indicator of body image perception issues. The main problem in the research was the fairly wide age range of participants, which explains the potential diversity of vocabulary. The combination of the attributes of a patient's condition, possible to determine using the method for a specific patient, can indicate the direction of support for the patient, relatives, direct medical personnel, and psycho-oncologists.

20 Using Automated Sentiment Analysis to Examine Self-Evaluation in Youth with Autism Spectrum Disorder

Objective:Individuals diagnosed with autism spectrum disorder (ASD) experience negative self-evaluation, indicated by low levels of self-esteem and describing themselves more negatively to others. Variations in reading comprehension, difficulty identifying emotions, and masking (camouflaging of autistic traits) make it difficult to accurately measure self-evaluation of individuals with ASD using subjective self-report scales such as the Rosenburg Self-Esteem Scale. Therefore, it is important to explore more objective methods of measuring self-evaluation in ASD. Sentiment analysis is a popular Natural Language Processing (NLP) technique used to quantify the emotional content of language programmatically by automatically transforming text into a data frame of words represented as individual values or tokens. Each token can then be categorized as positive or negative with a sentiment dictionary. The current study aims to investigate an automated sentiment analysis approach to evaluate self-evaluation by quantifying implicit linguistic affective valence of ASD participants' verbal self-describing statements in a naturalistic setting. Specifically, we evaluated the frequency of positive or negative words used during a mock job interview in which individuals with ASD were asked to describe themselves. We then examined the relationship between positive and negative word usage and standard self-report measures of self-evaluation.Participants and Methods:Twenty-four young adults with ASD were included in this study with an age range of 15-24 and a mean age of 19.2 years. Participants completed a battery of assessments including a mock job interview in which they were asked to describe themselves as a measure of implicit self-esteem. Self-esteem and knowledge of personal strengths was assessed using the Rosenberg Self-Esteem Scale and Strengths Knowledge Scale, respectively. Interview transcripts were automatically transformed into word token data frames using the tidytext package in Rstudio. Frequencies of positive and negative words were calculated and their ratio to total word count was used to measure the implicit positivity and negativity of transcripts.Results:There was a significant negative correlation between the frequency of negative sentiment in transcripts and measure on the Rosenburg Self-Esteem Scale (r = -.376, p = .035) and the Strengths Knowledge Scale (r = -.387, p = .031) indicating that individuals with higher self-esteem and knowledge of their strengths used fewer negative words when talking during a mock interview.Conclusions:While our results are preliminary, this pilot study represents the first to use automated sentiment analysis to study self-evaluation in individuals with ASD. The use of this technique on natural linguistic data collected through a mock job interview allows researchers to quantitatively analyze the emotionality of transcriptions and create insights that would otherwise be unavailable using more subjective qualitative techniques. Limited research into self-evaluation in this population has yielded inconsistent results, relying too heavily upon qualitative or self-report measures. The ability to programmatically quantify affective valence in transcripts is a time and cost-effective technique for improving validity of future measures of self-evaluation.

텍스트 마이닝을 활용한 감정 비율 단어 그래프

Unstructured data consisting of natural language such as SNS, papers, and questionnaire subjective question answers are generally compared or visualized using text mining. When analyzing text data by setting a random interval representing a time such as year, quarter, month, and day, it may be necessary to determine which interval has the most data, which sentiment words are used a lot for each interval, and how much text data in a particular interval is written negatively than positive. As a way to solve this problem, this study proposes an sentiment ratio word graph that can deliver the above three information at once. An sentiment ratio word graph is a graph created by dividing text data by a random interval standard representing time such as year, quarter, month, and day of the week and then assigning the sentiment score in the sentiment dictionary to the text data. When visualizing an sentiment ratio word graph, if you also use the pie coefficient to display words, you can further convey information about which words are most relevant to the sentiment word in a particular interval.

A text analytics model for agricultural knowledge discovery and sustainable food production: A case study from Oklahoma Panhandle

With recent increases in the use of social media in agricultural communities, many farmers are showing more and more interest in participating in social media and sharing different aspects of their profession with peers and policymakers. However, researchers have not explored this valuable data source well enough to help improve agribusiness decision-making. This study aims to investigate the potential capability and richness of social media for agricultural knowledge discovery, which can help monitor, detect, and predict critical agricultural events and activities and develop more sustainable food production and agricultural economy. This research utilizes text-mining tools and techniques to collect, process, and mine unstructured textual data from Twitter. Then, it examines the agreement between the retrieved information from tweets and that from the current monitoring systems and common cultural practices. Our findings illustrate that social media data can effectively provide information regarding the commencement and duration of significant cultural activities This study also examines the classification performance of several popular sentiment analysis tools on the farmer’s tweets and provides suggestions for future research on domain-specific sentiment lexicons for agricultural purposes.

Identifying equipment health status from maintenance records using Lexicon based Unsupervised Sentiment Analysis Adjusted for Negation (LUSAA-N)

During periodic inspection and repair of industrial equipment, technicians create maintenance records in the form of unstructured text to document valuable information regarding the equipment condition, operating environment, failed components, failure mechanism, etc. Inferring the equipment health status from these records can help in planning future maintenance actions to avoid expensive machine failures. However, automatic inference of the equipment health status is non-trivial due to various challenges inherent in analyzing unstructured text such as domain-specific vocabulary and negation words. Further, manual labeling of the records for supervised classification is arduous and impractical. Thus, in this paper, we present a model to automatically extract the equipment health status from maintenance records in a completely unsupervised manner. We show that using general English sentiment lexicons is highly inefficient for analyzing maintenance records, and thus propose a method that leverages sentiment lexicons curated for the maintenance domain and addresses the central issue of negation. We demonstrate the effectiveness of our proposed model over extant research using data from real-life maintenance records from oil rig equipment.

Multivariate SVR Demand Forecasting for Beauty Products Based on Online Reviews

Owing to changes in consumer attitudes, the beauty consumer population is growing rapidly and the demands of beauty consumers are variable. With a wide range of beauty products and exaggerated product promotions, consumers rely more on online reviews to perceive product information. In this paper, we propose a demand forecasting model that takes into account both product features and product emotional needs based on online reviews to help companies better develop production and sales plans. Firstly, a Word2vec model and sentiment analysis method based on a sentiment dictionary are used to extract product features and factors influencing product sentiment; secondly, a multivariate Support Vector Regression (SVR) demand prediction model is constructed and the model parameters are optimized using particle swarm optimization; and finally, an example analysis is conducted with beauty product Z. The results show that compared with the univariate SVR model and the multivariate SVR model with only product feature demand as the influencing factor, the multivariate SVR model with both product feature and product sentiment demand as influencing factors has a smaller prediction error, which can enable beauty retail enterprises to better grasp consumer demand dynamics, make flexible production and sales plans, and effectively reduce production costs.

SKEDS — An external knowledge supported logistic regression approach for document-level sentiment classification

Due to the enormous amount of user-generated content being generated on the web, labeling such data is a time-consuming and expensive endeavor. As a result, we have limited annotated data and the vast majority of data are unlabeled. Analysis reveals that extracting (external) knowledge from unlabeled data and integrating it with knowledge extracted from labeled data is a beneficial task for text information processing, in particular text classification. In this paper, we present a hybrid approach for classifying sentiments that employs external knowledge, which is categorized as either general-purpose sentiment knowledge or domain-related knowledge. General-purpose sentiment knowledge is extracted from sentiment lexicons, whereas domain-related knowledge is extracted from unlabeled data from the same or related domains. Similar domains for a given domain are identified based on their similarity score in terms of overlapping features. The proposed approach utilizes both forms of external knowledge and combines them with logistic regression to train an improved classification model. The classification model uses the conventional gradient descent algorithm for optimization, and its convergence analysis indicates that it is convex and converges to the global optimum. The proposed approach is empirically evaluated and compared to three baselines and one state-of-the-art method using standard performance evaluation metrics on a multi-domain sentiment dataset. The experiment results are encouraging, demonstrating that the proposed approach considerably outperforms the baseline approaches and outperforms the state-of-the-art approach by up to 2% in terms of both f-score and accuracy.

Incorporating Multiple Textual Factors into Unbalanced Financial Distress Prediction: A Feature Selection Methods and Ensemble Classifiers Combined Approach

Textual-based factors have been widely regarded as a promising feature that can be applied to financial issues. This study focuses on extracting both basic and semantic textual features to supplement the traditionally used financial indicators. The main is to improve Chinese listed companies’ financial distress prediction (FDP). A unique paradigm is proposed in this study that combines financial and multi-type textual predictive factors, feature selection methods, classifiers, and time spans to achieve the optimal FDP. The frequency counts, TF-IDF, TextRank, and word embedding approaches are employed to extract frequency count-based, keyword-based, sentiment, and readability indicators. The experimental results prove that financial domain sentiment lexicons, word embedding-based readability analysis approaches, and the basic textual features of Management Discussion and Analysis can be important elements of FDP. Moreover, the finding highlights the fact that incorporating financial and textual features can achieve optimal performance 4 or 5 years before the expected baseline year; applying the RF-GBDT combined model can also outperform other classifiers. This study makes an innovative contribution, since it expands the multiple text analysis method in the financial text mining field and provides new findings on how to provide early warning signs related to financial risk. The approaches developed in this research can serve as a template that can be used to resolve other financial issues.

Sentiment Classification of Fuel Price Rise in Economic Aspects Using Lexicon and SVM Method

After being hit by COVID-19 for a long time around the world which resulted in the paralysis of all countries, especially the economic aspects of all countries that dropped dramatically, the world was again shocked by the conflict between Russia and Ukraine which resulted in an increase in world oil prices including in Indonesia, many people complained and opposed the government's policy of increasing fuel prices because fuel affects various aspects, including economic aspects. Based on these problems, researchers use sentiment analysis methods that aim to find out people's opinions on issues that are being discussed throughout Indonesia and this research focuses on comparing the SVM algorithm with TF-IDF feature extraction then using K-Fold Cross Validation after that it is compared with the Lexicon Inset dictionary, in this case the model with Lexicon Inset which contains weighting on each word. In this study, it was found that the dataset model using the SVM algorithm with TF-IDF feature extraction and then using K-Fold Cross Validation obtained an average accuracy of 0.85 using the SVM algorithm. While the model using the automatic labeling dataset using the Indonesian sentiment Lexicon (Lexicon Inset) obtained an average accuracy of 0.68. Classification using SVM with TF-IDF feature extraction is superior to using Lexicon Inset.

Research on a Hotel Collaborative Filtering Recommendation Algorithm Based on the Probabilistic Language Term Set

In the face of problems such as information overload and the information cocoon resulting from big data, it is a key point of current research to solve the problem of semantic fuzziness of online reviews and improve the accuracy of personalized recommendation algorithms by using online reviews. Based on the advantage of the probabilistic language term set to deal with fuzzy information and the historical data of online hotel reviews, this paper proposes a collaborative filtering recommendation algorithm for hotels. Firstly, the text data of hotel online reviews are crawled by a crawler and processed by jieba and TF-IDF tools. Secondly, the hotel evaluation attribute set is constructed, and the sentiment analysis of the review statements is carried out with the help of the HowNet sentiment dictionary and manual annotation method. The probabilistic language term set is used to classify the data and derive statistics, and the maximum deviation method is used to determine the weight of each attribute. Then, the cosine similarity formula is fused with the modified cosine similarity formula to calculate the similarity and construct the decision matrix. Finally, combined with the historical data of the user’s hotel selection, the hotel recommendation results are generated. This paper collected review data from 10 hotels in Macau from the official “Ctrip” website. The proposed recommendation algorithm model was then applied to process and analyze the data, resulting in the generation of a ranked list of hotel recommendations. To validate the accuracy and effectiveness of this research, the recommendation results were compared with those produced by other algorithms.

Negativity and Positivity in the ICU: Exploratory Development of Automated Sentiment Capture in the Electronic Health Record.

To develop proof-of-concept algorithms using alternative approaches to capture provider sentiment in ICU notes. Retrospective observational cohort study. The Multiparameter Intelligent Monitoring of Intensive Care III (MIMIC-III) and the University of California, San Francisco (UCSF) deidentified notes databases. Adult (≥18 yr old) patients admitted to the ICU. We developed two sentiment models: 1) a keywords-based approach using a consensus-based clinical sentiment lexicon comprised of 72 positive and 103 negative phrases, including negations and 2) a Decoding-enhanced Bidirectional Encoder Representations from Transformers with disentangled attention-v3-based deep learning model (keywords-independent) trained on clinical sentiment labels. We applied the models to 198,944 notes across 52,997 ICU admissions in the MIMIC-III database. Analyses were replicated on an external sample of patients admitted to a UCSF ICU from 2018 to 2019. We also labeled sentiment in 1,493 note fragments and compared the predictive accuracy of our tools to three popular sentiment classifiers. Clinical sentiment terms were found in 99% of patient visits across 88% of notes. Our two sentiment tools were substantially more predictive (Spearman correlations of 0.62-0.84, p values < 0.00001) of labeled sentiment compared with general language algorithms (0.28-0.46). Our exploratory healthcare-specific sentiment models can more accurately detect positivity and negativity in clinical notes compared with general sentiment tools not designed for clinical usage.

Construction of an Emotional Lexicon of Patients With Breast Cancer: Development and Sentiment Analysis.

The innovative method of sentiment analysis based on an emotional lexicon shows prominent advantages in capturing emotional information, such as individual attitudes, experiences, and needs, which provides a new perspective and method for emotion recognition and management for patients with breast cancer (BC). However, at present, sentiment analysis in the field of BC is limited, and there is no emotional lexicon for this field. Therefore, it is necessary to construct an emotional lexicon that conforms to the characteristics of patients with BC so as to provide a new tool for accurate identification and analysis of the patients' emotions and a new method for their personalized emotion management. This study aimed to construct an emotional lexicon of patients with BC. Emotional words were obtained by merging the words in 2 general sentiment lexicons, the Chinese Linguistic Inquiry and Word Count (C-LIWC) and HowNet, and the words in text corpora acquired from patients with BC via Weibo, semistructured interviews, and expressive writing. The lexicon was constructed using manual annotation and classification under the guidance of Russell's valence-arousal space. Ekman's basic emotional categories, Lazarus' cognitive appraisal theory of emotion, and a qualitative text analysis based on the text corpora of patients with BC were combined to determine the fine-grained emotional categories of the lexicon we constructed. Precision, recall, and the F1-score were used to evaluate the lexicon's performance. The text corpora collected from patients in different stages of BC included 150 written materials, 17 interviews, and 6689 original posts and comments from Weibo, with a total of 1,923,593 Chinese characters. The emotional lexicon of patients with BC contained 9357 words and covered 8 fine-grained emotional categories: joy, anger, sadness, fear, disgust, surprise, somatic symptoms, and BC terminology. Experimental results showed that precision, recall, and the F1-score of positive emotional words were 98.42%, 99.73%, and 99.07%, respectively, and those of negative emotional words were 99.73%, 98.38%, and 99.05%, respectively, which all significantly outperformed the C-LIWC and HowNet. The emotional lexicon with fine-grained emotional categories conforms to the characteristics of patients with BC. Its performance related to identifying and classifying domain-specific emotional words in BC is better compared to the C-LIWC and HowNet. This lexicon not only provides a new tool for sentiment analysis in the field of BC but also provides a new perspective for recognizing the specific emotional state and needs of patients with BC and formulating tailored emotional management plans.

Design of Sentiment Analysis Framework of Digital Media Short Text Based on Multi-pattern Sentiment Lexicon

Along the continuous advancement of the network and the rise of digital media, the amount of data produced by the exponential explosion. And how to use these data to provide personalized services for users is one of the current research focuses. To address the issue of insufficient coverage in the current sentiment lexicon and the difficulty of constructing sentiment lexicon in specific fields, this study proposes a multi-modal emotional thesaurus. Semi-supervised learning is used to solve the problem of insufficient coverage of emotional thesaurus, and a semi-supervised classification algorithm is realized by using a large number of unlabeled sample data combined with a small number of labeled sample data. Optimized learning is used to solve the problem of difficult construction of emotional thesaurus in specific fields, the corresponding specific emotional thesaurus is constructed by adaptive adjustment of emotional word score, and finally the improved emotional thesaurus is used to build a digital media short text sentiment analysis framework. For testing, the NLPCC dataset was used in this study, Experiments show that the framework constructed in this study requires 87 iterations, a Recall value of 0.912, a F1 value of 0.753, and an average accuracy of 83.39%, all of which are better than the sentiment analysis framework without the use of multi-pattern sentiment lexicon. In the simulation experiment, the recognition accuracy reached 85.88%, which was 16.85%, 11.57% and 6.72% higher than the test scenarios using a single emotion thesaurus selected in this study. The above results show that the digital media short-text sentiment analysis framework built in this research based on multi-pattern sentiment lexicon can carry out short-text sentiment analysis more accurately and efficiently, so as to accurately analyze users’ needs and provide customized services precisely.

Sentiment Analysis in Online Product Reviews: Mining Customer Opinions for Sentiment Classification

Online product reviews have become a valuable resource for consumers seeking detailed information and making informed choices. The process of automatically extracting sentiment or opinions from these reviews heavily relies on sentiment analysis, a branch of Natural Language Processing (NLP). This research article focuses on sentiment categorization in online product evaluations, utilizing innovative techniques for mining consumer opinions. The project aims to establish a robust framework for sentiment analysis that accurately classifies emotions expressed in these reviews. The proposed system incorporates advanced deep learning and machine learning methods to enhance data classification and extract fine-grained sentiment information. The study addresses the unique challenges of sentiment analysis in the context of online product evaluations, including polarity changes, sarcasm, and domain-specific sentiment expressions, which often pose significant obstacles to precise sentiment classification. The approach combines feature engineering and deep learning techniques, extracting lexical, syntactic, and semantic features such as part-of-speech tags, n-grams, sentiment lexicons, and word embeddings from the review texts. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are employed as sophisticated neural network architectures to leverage these features, creating robust representations and capturing contextual information. The suggested architecture is extensively evaluated on a large dataset of online product reviews, demonstrating superior performance in sentiment categorization compared to existing approaches. The evaluation encompasses various sentiment classes, measuring metrics like accuracy, recall, and F1-score, and assessing the framework's adaptability to different product domains. The study showcases the effectiveness of advanced machine learning and deep learning algorithms in sentiment categorization, advancing the field of sentiment analysis for online product evaluations. Businesses can gain valuable insights into customer sentiment and make well-informed decisions regarding product enhancements and marketing strategies by leveraging the proposed framework

Automatically generate sentiment lexicon for the Persian stock market

Automatically generate sentiment lexicon for the Persian stock market

Identifying causal effects of the clinical sentiment of patients' nursing notes on anticipated fall risk stratification

Stratifying patients with high risks of falling based on assessment with nursing notes is essential for tailoring anticipated fall prevention strategies. However, the average exposure effects of the clinical sentiment polarity (CSP) of nursing notes on the fall risks of patients are still not well understood. Thus, this study proposes a causal inference framework to identify the exposure effect on patients' anticipated fall risks using electronic nursing notes. The augmented inverse probability weighting model is leveraged to estimate the average exposure effect in a quasi-experiment. The experimental data contains 334,000 words of 2,434 nursing notes from the MIMIC dataset. The results show that the fall risk of exposure to the positive CSP is 0.0054 lower than the control group. The contribution of this paper is three-fold. First, the exposure effect of the CSP in the clinical nursing notes is identified with the causal inference method, which augments the Morse fall scale by analyzing the nursing notes. Second, the Pearson correlations between the exposure and the MFS are sensitive to the cut-off of the clinical sentiment score. By contrast, our results show that their impact on the fall risks in the weighting model can be identified (negative effect) when reducing the bias of covariate imbalance. Compared to the risk factors in the Morse fall scale, the CSP takes a scale of 10 for predicting anticipated fall risks. Third, the CSP factor can be extracted from the sentiment lexicons of nursing notes on the patient's fall risks with information processing methods, which contributes to the clinical decision support of anticipated fall prevention.