Theses
Theses
Filip Šaina
Sažimanje sentimenta u studentskim upitnicima predmeta
Sentiment Summarization from Student Course Questionnaires
2017
Undergraduate
Jan Šnajder
FER
FER2
5329
42
EN
Sažimanje sentimenta jest zadatak obrade prirodnog jezika koji kombinira sažimanje teksta i analizu sentimenta. Osnovna ideja jest generirati sažetke subjektivnih tekstova koji u obzir uzimaju polaritet sentimenta prema pojedinim aspektima proizvoda ili usluge. Sustavi za zažimanje sentimenta omogućavaju analizu velikih količina subjektivnih tekstova, npr. korisničkih recenzija ili komentara na društvenim mrežama. Jedna zanimljiva i korisna primjena sažimanja sentimenta jest analiza studentskih upitnika o predmetima, kod kojih studenti iskazuju svoje mišljenje o pojedinačnim aspektima predmeta.
Tema završnoga rada jest sažimanje sentimenta iz tekstova studentskih upitnika na hrvatskome jeziku. U okviru završnoga rada proučeni su postupci za ekstraktivno sažimanje teksta, postupci za klasifikaciju polariteta sentimenta te postupci za sažimanje sentimenta temeljene na strojnom učenju. Osmišljen je i implementiran postupak za sažimanje studentskih upitnika koji može identificirati pojedine aspekte predmeta, grupirati komentare koji se odnose na slične aspekte te ih klasificirati prema sentimentu. Izrađen je prikladan skup podataka za treniranje i ispitivanje modela, temeljen na stvarnim studentskim upitnicima. Na prikladan i intuitivan način prikazani su rezultati sažimanja sentimenta. Provedeno je eksperimentalno vrednovanje postupka na ispitnim podatcima. Radu je priložen izvorni i izvršni kod razvijenog sustava, skup podataka i programska dokumentacija te citirana i korištena literatura.
Sentiment summarization is a natural language processing task that combines doc- ument summarization and sentiment analysis. The main idea is to generate summaries of subjective texts that account for sentiments toward the different aspects of a product or service. Sentiment summarization systems enable the analysis of large amount of subjective texts, e.g., user reviews or comments on social networks. One interesting and useful application of sentiment summarization is the analysis of student course questionnaires.
The topic of this thesis is sentiment summarization from student course questionnaires in Croatian language. Do a literature survey on extractive document summarization, sentiment polarity classification, and sentiment summarization. Devise and implement a method for sentiment summarization from student course questionnaires that can identify the various aspects of a course, group comments pertaining to similar aspects, and classify their sentiment polarity. Compile a suitable dataset for training and evaluating the model, derived from real student questionnaires. Devise a suitable and intuitive way for visualizing the sentiment summarization results. Carry out an experimental evaluation of the method on test data. All references must be cited, and all source code, documentation, executables, and datasets must be provided with the thesis.
obrada prirodnog jezika, strojno učenje, analiza sentimenta, analiza sentimenta na osnovu aspekata, hrvatski jezik, upitnici, recenzije
natural language processing, machine learning, sentiment analysis, aspect-based sentiment analysis, Croatian language, student questionnaires
5.7.2017.
Sentiment summarization is a natural language processing task that combines document summarization and sentiment analysis. The main idea is to generate summaries of subjective texts that account for sentiment toward the different aspects of a product or service. Sentiment summarization systems enable the analysis of large amount of subjective texts, e.g., user reviews or comments on social networks. One interesting and useful application of sentiment summarization is the analysis of student course questionnaires.
The topic of this thesis is sentiment summarization from student course questionnaires in Croatian language. Do a literature survey on extractive document summarization, sentiment polarity classification, and sentiment summarization based on machine learning. Devise and implement a method for sentiment summarization from student course questionnaires that can identify the various aspects of a course, group comments pertaining to similar aspects, and classify their sentiment polarity. Compile a suitable dataset for training and evaluating the model, derived from real student questionnaires. Devise a suitable and intuitive way for visualizing the sentiment summarization results. Carry out an experimental evaluation of the method on test data. All references must be cited, and all source code, documentation, executables, and datasets must be provided with the thesis.