Language Use And Sentiment In Psychotherapy:Predicting The Process And Outcome

Language use and sentiment in psychotherapy can provide important information about the process and outcome of counseling.Qualitative coding which is an usual method for process and outcome research costs much.Language Inquiry and Word Count(LIWC)is another method to explore the relationship between language use and psychology.Sentiment analysis can compute the sentiment of psychotherapy automatically.The present study applied LIWC and sentiment analysis to analysis the relationship between language use/sentiment and the symptoms of clients as well as the evaluation of the psychotherapy process.28 clients' data was collected including the recordings of psychotherapy and the process and outcome questionnaires.All 154 recordings were translated into texts.Data cleaning was conducted including cutting words and cutting talk turns.The first study focused on LIWC features.Principal component analysis was conducted to find out the main factors of LIWC features.Results revealed that 8 factors interpret 75.11%of all the variance.Among the factors,biological feelings,function words,emotion,eognition and filler word can predict the symptom of clients.K-means clustering was used to explore the classifications of transcripts.The accuracy of clustering for speakers attend 99%.The clients' transcripts have 3 categories.And the outcome of psychotherapy have significant difference between 3 categories.In a multilevel model,biological feelings and emotion words as well as the clusters can be used to predict the outcome of psychotherapy.The second study used baiduSenta and sentiment analysis based on dependent sentence analysis and sentiment dictionary to compute the sentiment of psychotherapy transcripts.The Kappa between human coding and baiduSenta is 0.51 and Kappa between human coding and sentiment analysis based on dependent sentence analysis and sentiment dictionary is 0.62.The second method of sentiment analysis is more accurate.In a regression model,the negative sentiment of clients predicts their symptom(?=0.26,p<0.01).The third study computed the language style match(LSM)and sentiment similarity between therapists and clients.Results revealed that positive sentiment similarity can predict the the competence of therapists(?=0.60,p<0.05).The negative sentiment similarity predict the therapists' competence as well(?=0.59,p<0.05).Besides,the positive sentiment similarity predicts the evaluation of psychotherapy(??0.56,p<0.05).However,LSM is not a significant predictor.Finally,the study discussed the values and the limitations of the designs and methods.Future directions were reviewed.