隨著網路世界的興起，在面臨選擇難題時，民眾不僅會接收口耳相傳的資訊，也會以關鍵字上網搜尋目標資訊，但是在海量資料的浪潮中，如何快速的整合資料是一大挑戰。電影影評文章摘要可以幫助民眾進電影院前了解電影的資訊，透過這樣的方式確認電影是自身有興趣的電影。 本研究以電影：復仇者聯盟2影評66篇4616句、蝙蝠俠對超人：正義曙光60篇9345句、動物方城市60篇5545句、星際效應50篇4616句、高年級實習生62篇5622句為資料來源，以分群概念結合摘句之方法生成影評摘要。其中，利用K-Means演算法將五部電影的多篇影評特徵詞、句子進行分群後，使用TFIDF評比各分群語句的重要性來選取高權重語句，再以WWA方法挑選分群中不同面向的語句，最後以相似度計算最佳範本與各分群內容的相似度來決定每一群聚的排序順序，產生一篇具有相似內容段落和段落順序的影評多篇摘要。 研究結果顯示，原本五部電影影評對最佳範本之相似度為15.87%，經由本研究方法產生之摘要對最佳範本單篇摘要之相似度為21.19%。另外，因為影評中各分群的順序是比對最佳範本相似度而產生的排序，整篇摘要會具有與最佳範本相似段落排序的摘要內容，其中內容包含了電影影評中廣泛提到的相似內容，不同的相似段落讓文章摘要的呈現更具廣泛性。藉由此摘要方法，可以幫助民眾藉由自動化彙整、萃取的摘要快速了解相關電影資訊內容和協助決策。 / Abstract As Facing the Big Data issue, there are too many information on the website for reader to understand. How to perform and summarize essential information quickly is a challenge. People who want to go to a movie will also face this situation. Before choosing movies, they will search relative information of the movies. However, there are many film reviews all over the websites. Automatic text summarization can efficiently extract important information for readers, and conclude concepts of reviews on the websites. Through this method, readers can easily comprehend the best idea of all the reviews and save their time. The research presents a multi-concept and extractive film review summary for readers. It generates film review summary from the most popular blog platform, PIXNET, with extract-based method and clustering concept. The method using K-Means algorism let the film review summary focus on specific film to cluster the sentences by features, and having statistical sense and WWA method to measure the weight of sentences in order to choose the representative sentences. On the last step, it will compare to templates to decide the sequence of classified sentences and summary all represent sentences from each cluster. The research provides a multi-concept and extractive film review summary for people. From the result, there are five movies, which are used summary method increase the average similarity to 21.19% that comparing between the film reviews summary and templates summary. It shows that the automatic film reviews summarization can extract the important sentences from the reviews. Also, with comparing template method to order the cluster, it can sequentially list the cluster of the sentences to generate a movie review, which saves readers’ time and easily comprehend.

Σκοπός της Μεταπτυχιακής Εργασίας είναι η επέκταση και αναβάθμιση του μηχανισμού που είχε δημιουργηθεί στα πλαίσια της Διπλωματικής Εργασίας που εκπόνησα με τίτλο «Δημιουργία Πύλης Προσωποποιημένης Πρόσβασης σε Περιεχόμενο του WWW». Η παραπάνω Διπλωματική εργασία περιλάμβανε τη δημιουργία ενός μηχανισμού που ξεκινούσε με ανάκτηση πληροφορίας από το Διαδίκτυο (HTML σελίδες από news portals), εξαγωγή χρήσιμου κειμένου και προεπεξεργασία της πληροφορίας, αυτόματη κατηγοριοποίηση της πληροφορίας και τέλος παρουσίαση στον τελικό χρήστη με προσωποποίηση με στοιχεία που εντοπίζονταν στις επιλογές του χρήστη. Στην παραπάνω εργασία εξετάστηκαν διεξοδικά θέματα που είχαν να κάνουν με τον τρόπο προεπεξεργασίας της πληροφορίας καθώς και με τον τρόπο αυτόματης κατηγοριοποίησης ενώ υλοποιήθηκαν αλγόριθμοι προεπεξεργασίας πληροφορίας τεσσάρων σταδίων και αλγόριθμος αυτόματης κατηγοριοποίησης βασισμένος σε πρότυπες κατηγορίες. Τέλος υλοποιήθηκε portal το οποίο εκμεταλλευόμενο την επεξεργασία που έχει πραγματοποιηθεί στην πληροφορία παρουσιάζει το περιεχόμενο στους χρήστες προσωποποιημένο βάσει των επιλογών που αυτοί πραγματοποιούν. Σκοπός της μεταπτυχιακής εργασίας είναι η εξέταση περισσοτέρων αλγορίθμων για την πραγματοποίηση της παραπάνω διαδικασίας αλλά και η υλοποίησή τους προκειμένου να γίνει σύγκριση αλγορίθμων και παραγωγή ποιοτικότερου αποτελέσματος. Πιο συγκεκριμένα αναβαθμίζονται όλα τα στάδια λειτουργίας του μηχανισμού. Έτσι, το στάδιο λήψης πληροφορίας βασίζεται σε έναν απλό crawler λήψης HTML σελίδων από αγγλόφωνα news portals. Η διαδικασία βασίζεται στο γεγονός πως για κάθε σελίδα υπάρχουν RSS feeds. Διαβάζοντας τα τελευταία νέα που προκύπτουν από τις εγγραφές στα RSS feeds μπορούμε να εντοπίσουμε όλα τα URL που περιέχουν HTML σελίδες με τα άρθρα. Οι HTML σελίδες φιλτράρονται προκειμένου από αυτές να γίνει εξαγωγή μόνο του κειμένου και πιο αναλυτικά του χρήσιμου κειμένου ούτως ώστε το κείμενο που εξάγεται να αφορά αποκλειστικά άρθρα. Η τεχνική εξαγωγής χρήσιμου κειμένου βασίζεται στην τεχνική web clipping. Ένας parser, ελέγχει την HTML δομή προκειμένου να εντοπίσει τους κόμβους που περιέχουν μεγάλη ποσότητα κειμένου και βρίσκονται κοντά σε άλλους κόμβους που επίσης περιέχουν μεγάλες ποσότητες κειμένου. Στα εξαγόμενα άρθρα πραγματοποιείται προεπεξεργασία πέντε σταδίων με σκοπό να προκύψουν οι λέξεις κλειδιά που είναι αντιπροσωπευτικές του άρθρου. Πιο αναλυτικά, αφαιρούνται όλα τα σημεία στίξης, όλοι οι αριθμοί, μετατρέπονται όλα τα γράμματα σε πεζά, αφαιρούνται όλες οι λέξεις που έχουν λιγότερους από 4 χαρακτήρες, αφαιρούνται όλες οι κοινότυπες λέξεις και τέλος εφαρμόζονται αλγόριθμοι εύρεσης της ρίζας μίας λέξεις. Οι λέξεις κλειδιά που απομένουν είναι stemmed το οποίο σημαίνει πως από τις λέξεις διατηρείται μόνο η ρίζα. Από τις λέξεις κλειδιά ο μηχανισμός οδηγείται σε δύο διαφορετικά στάδια ανάλυσης. Στο πρώτο στάδιο υπάρχει μηχανισμός ο οποίος αναλαμβάνει να δημιουργήσει μία αντιπροσωπευτική περίληψη του κειμένου ενώ στο δεύτερο στάδιο πραγματοποιείται αυτόματη κατηγοριοποίηση του κειμένου βασισμένη σε πρότυπες κατηγορίες που έχουν δημιουργηθεί από επιλεγμένα άρθρα που συλλέγονται καθ’ όλη τη διάρκεια υλοποίησης του μηχανισμού. Η εξαγωγή περίληψης βασίζεται σε ευρεστικούς αλγορίθμους. Πιο συγκεκριμένα προσπαθούμε χρησιμοποιώντας λεξικολογική ανάλυση του κειμένου αλλά και γεγονότα για τις λέξεις του κειμένου αν δημιουργήσουμε βάρη για τις προτάσεις του κειμένου. Οι προτάσεις με τα μεγαλύτερη βάρη μετά το πέρας της διαδικασίας είναι αυτές που επιλέγονται για να διαμορφώσουν την περίληψη. Όπως θα δούμε και στη συνέχεια για κάθε άρθρο υπάρχει μία γενική περίληψη αλλά το σύστημα είναι σε θέση να δημιουργήσει προσωποποιημένες περιλήψεις για κάθε χρήστη. Η διαδικασία κατηγοριοποίησης βασίζεται στη συσχέτιση συνημίτονου συγκριτικά με τις πρότυπες κατηγορίες. Η κατηγοριοποίηση δεν τοποθετεί μία ταμπέλα σε κάθε άρθρο αλλά μας δίνει τα αποτελέσματα συσχέτισης του άρθρου με κάθε κατηγορία. Ο συνδυασμός των δύο παραπάνω σταδίων δίνει την πληροφορία που εμφανίζεται σε πρώτη φάση στο χρήστη που επισκέπτεται το προσωποποιημένο portal. Η προσωποποίηση στο portal βασίζεται στις επιλογές που κάνουν οι χρήστες, στο χρόνο που παραμένουν σε μία σελίδα αλλά και στις επιλογές που δεν πραγματοποιούν προκειμένου να δημιουργηθεί προφίλ χρήστη και να είναι εφικτό με την πάροδο του χρόνου να παρουσιάζεται στους χρήστες μόνο πληροφορία που μπορεί να τους ενδιαφέρει. / The scope of this MsC thesis is the extension and upgrade of the mechanism that was constructed during my undergraduate studies under my undergraduate thesis entitled “Construction of a Web Portal with Personalized Access to WWW content”. The aforementioned thesis included the construction of a mechanism that would begin with information retrieval from the WWW and would conclude to representation of information through a portal after applying useful text extraction, text pre-processing and text categorization techniques. The scope of the MsC thesis is to locate the problematic parts of the system and correct them with better algorithms and also include more modules on the complete mechanism. More precisely, all the modules are upgraded while more of them are constructed in every aspect of the mechanism. The information retrieval module is based on a simple crawler. The procedure is based on the fact that all the major news portals include RSS feeds. By locating the latest articles that are added to the RSS feeds we are able to locate all the URLs of the HTML pages that include articles. The crawler then visits every simple URL and downloads the HTML page. These pages are filtered by the useful text extraction mechanism in order to extract only the body of the article from the HTML page. This procedure is based on the web-clipping technique. An HTML parser analyzes the DOM model of HTML and locates the nodes (leafs) that include large amounts of text and are close to nodes with large amounts of text. These nodes are considered to include the useful text. In the extracted useful text we apply a 5 level preprocessing technique in order to extract the keywords of the article. More analytically, we remove the punctuation, the numbers, the words that are smaller than 4 letters, the stopwords and finally we apply a stemming algorithm in order to produce the root of the word. The keywords are utilized into two different interconnected levels. The first is the categorization subsystem and the second is the summarization subsystem. During the summarization stage the system constructs a summary of the article while the second stage tries to label the article. The labeling is not unique but the categorization applies multi-labeling techniques in order to detect the relation with each of the standard categories of the system. The summarization technique is based on heuristics. More specifically, we try, by utilizing language processing and facts that concern the keywords, to create a score for each of the sentences of the article. The more the score of a sentence, the more the probability of it to be included to the summary which consists of sentences of the text. The combination of the categorization and summarization provides the information that is shown to our web portal called perssonal. The personalization issue of the portal is based on the selections of the user, on the non-selections of the user, on the time that the user remains on an article, on the time that spends reading similar or identical articles. After a short period of time, the system is able to adopt on the user’s needs and is able to present articles that match the preferences of the user only.

The advancements in abstractive summarization using Large Language Models (LLMs) have brought with it new challenges in evaluating the quality and faithfulness of generated summaries. This thesis explores a human-like automated method for evaluating news article summaries. By leveraging two LLMs with instruction-following capabilities (GPT-4 and Claude), the aim is to examine to what extent the quality of summaries can be measured by predictions of an LLM. The proposed framework involves defining specific attributes of desired summaries, which are used to design generation prompts and evaluation questions. These questions are presented to the LLMs in natural language during evaluation to assess of various summary qualities. To validate the effectiveness of the evaluation method, an adversarial approach is employed, in which a dataset comprising summaries with distortions related to various summary attributes is generated. In an experiment, the two LLMs evaluate the adversarial dataset, and their ability to detect known distortions is measured and analyzed. The findings suggest that the LLM-based evaluations demonstrate promise in detecting binary qualitative issues, such as incorrect facts. However, the reliability of the zero-shot evaluation varies depending on the evaluating LLM and the specific questions used. Further research is required to validate the accuracy and generalizability of the results, particularly in subjective dimensions where the results of this thesis are inconclusive. Nonetheless, this thesis provides insights that can serve as a foundation for future advancements in the field of automatic text evaluation. / De framsteg som gjorts inom abstrakt sammanfattning med hjälp av stora språkmodeller (LLM) har medfört nya utmaningar när det gäller att utvärdera kvaliteten och sanningshalten hos genererade sammanfattningar. Detta examensarbete utforskar en mänskligt inspirerad automatiserad metod för att utvärdera sammanfattningar av nyhetsartiklar. Genom att dra nytta av två LLM:er med instruktionsföljande förmågor (GPT-4 och Claude) är målet att undersöka i vilken utsträckning kvaliteten av sammanfattningar kan bestämmas med hjälp av språkmodeller som utvärderare. Det föreslagna ramverket innefattar att definiera specifika egenskaper hos önskade sammanfattningar, vilka används för att utforma genereringsuppmaningar (prompts) och utvärderingsfrågor. Dessa frågor presenteras för språkmodellerna i naturligt språk under utvärderingen för att bedöma olika kvaliteter hos sammanfattningar. För att validera utvärderingsmetoden används ett kontradiktoriskt tillvägagångssätt där ett dataset som innefattar sammanfattningar med förvrängningar relaterade till olika sammanfattningsattribut genereras. I ett experiment utvärderar de två språkmodellerna de motstridiga sammanfattningar, och deras förmåga att upptäcka kända förvrängningar mäts och analyseras. Resultaten tyder på att språkmodellerna visar lovande resultat vid upptäckt av binära kvalitativa problem, såsom faktafel. Dock varierar tillförlitligheten hos utvärderingen beroende på vilken språkmodell som används och de specifika frågorna som ställs. Ytterligare forskning krävs för att validera tillförlitligheten och generaliserbarheten hos resultaten, särskilt när det gäller subjektiva dimensioner där resultaten är osäkra. Trots detta ger detta arbete insikter som kan utgöra en grund för framtida framsteg inom området för automatisk textutvärdering.

This thesis aims to evaluate different abstractive text summarization models and techniques for summarizing resumes. It has two main objectives: investigate the models’ performance on resume summarization and assess the suitability of the generated summaries as resume introductions. Although automatic abstractive text summarization has gained traction in various areas, its application in the resume domain has not yet been explored. Resumes present a unique challenge for abstractive summarization due to their diverse style, content, and length. To address these challenges, three state-of-the-art pre-trained text generation models: BART, T5, and ProphetNet, were selected. Additionally, two approaches that can handle longer resumes were investigated. The first approach, named LongBART, modified the BART architecture by incorporating the Longformer’s self-attention into the encoder. The second approach, named HybridBART, used an extractive-then-abstractive summarization strategy. The models were fine-tuned on a dataset of 653 resume-introduction pairs and were evaluated using automatic metrics as well as two types of human evaluations: a survey and expert interviews. None of the models demonstrated superiority across all criteria and evaluation metrics. However, the survey responses indicated that LongBART showed promising results, receiving the highest scores in three out of five criteria. On the other hand, ProphetNet consistently received the lowest scores across all criteria in the survey, and across all automatic metrics. Expert interviews emphasized that the generated summaries cannot be considered correct summaries due to the presence of hallucinated personal attributes. However, there is potential for using the generated texts as resume introductions, given that measures are taken to ensure the hallucinated personal attributes are sufficiently generic. / Denna avhandling utvärderar olika modeller och tekniker för automatisk textsammanfattning för sammanfattning av CV:n. Avhandlingen har två mål: att undersöka modellernas prestanda på sammanfattning av CV:n och bedöma lämpligheten att använda de genererade sammanfattningar som CV-introduktioner. Även om automatisk abstrakt textsummering har fått fotfäste inom olika sammanhang är dess tillämpning inom CV-domänen ännu outforskad. CV:n utgör en unik utmaning för abstrakt textsammanfattning på grund av deras varierande stil, innehåll och längd. För att hantera dessa utmaningar valdes tre av de främsta förtränade modellerna inom textgenerering: BART, T5 och ProphetNet. Dessutom undersöktes två extra metoder som kan hantera längre CV:n. Det första tillvägagångssättet, kallat LongBART, modifierade BART-arkitekturen genom att inkludera självuppmärksamhet från Longformer-arkitekturen i kodaren. Det andra tillvägagångssättet, kallat HybridBART, använde en extraktiv-sen-abstraktiv sammanfattningsstrategi. Modellerna finjusterades med ett dataset med 653 CV-introduktionspar och utvärderades med hjälp av automatiska mått, samt två typer av mänsklig utvärdering: en enkätundersökning och intervjuer med experter. Ingen av modellerna visade överlägsenhet på alla kriterier och utvärderingsmått. Dock indikerade enkätsvaren att LongBART visade lovande resultat, genom att få högst poäng i tre av fem utvärderingskategorier. Å andra sidan fick ProphetNet lägst poäng i samtliga utvärderingskategorier, samt lägst poäng i alla automatiska mätningar. Expertintervjuer framhävde att de genererade sammanfattningarna inte kan anses vara pålitliga som fristående sammanfattningar på grund av förekomsten av hallucinerade personliga egenskaper. Trots detta finns det potential att använda dessa sammanfattningar som introduktioner, under förutsättningen att åtgärder vidtas för att säkerställa att hallucinerade personliga attribut är tillräckligt generiska.

No description available.

Information gathering and information overload is an ever growing topic of concernfor Information Technology (IT) experts. The amount of information dealt withon an everyday basis is large enough to take up valuable time having to scatterthrough it all to find the relevant information. As for the application area of IT,time is directly related to money as having to waste valuable production time ininformation gathering and allocation of human resources is a direct loss of profitsfor any given company. Two issues are mainly addressed through this thesis: textsare too lengthy and the difficulty of finding relevant information. Through the useof Natural Language Processes (NLP) methods such as topic modelling and textsummarization, a proposed solution is constructed in the form of a technical basiswhich can be implemented in most business areas. An experiment along with anevaluation session is setup in order to evaluate the performance of the technical basisand enforce the focus of this paper, namely ”How effective is text summarizationcombined with individualized information recommendation in improving informationgathering of IT experts?”. Furthermore, the solution includes a construction of userprofiles in an attempt to individualize content and theoretically present more relevantinformation. The results for this project are affected by the substandard quality andmagnitude of data points, however positive trends are discovered. It is stated thatthe use of user profiles further enhances the amount of relevant articles presentedby the model along with the increasing recall and precision values per iteration andaccuracy per number of updates made per user. Not enough time is spent as for theextent of the evaluation process to confidently state the validity of the results morethan them being inconsistent and insufficient in magnitude. However, the positivetrends discovered creates further speculations on if the project is given enough timeand resources to reach its full potential. Essentially, one can theoretically improveinformation gathering by summarizing texts combined with individualization.

Automatic text summarization techniques, which can reduce a source text to a summary text by content generalization or selection have assumed signifi- cance in recent times due to the ever expanding information explosion created by the World Wide Web. Summaries generated by generalization of information are called abstracts and those generated by selection of portions of text (sentences, phrases etc.) are called extracts. Further, summaries could for each document separately or multiple documents could be summarized together to produce a single summary. The challenges in making machines generate extracts or abstracts are primarily due to the lack of understanding of human cognitive processes. Summary generated by humans seems to be influenced by their moral, emotional and ethical stance on the subject and their background knowledge of the content being summarized.These characteristics are hardly understood and difficult to model mathematically. Further automatic summarization is very much handicapped by limitations of existing computing resources and lack of good mathematical models of cognition. In view of these, the role of rigorous mathematical theory in summarization has been limited hitherto. The research reported in this thesis is a contribution towards bringing in the awesome power of well-established concepts information theory to the field of summarization. Contributions of the Thesis The specific focus of this thesis is on extractive summarization. Its domain spans multi-document summarization as well as single document summarization. In the whole thesis the word "summarization" and "summary", imply extract generation and sentence extracts respectively. In this thesis, two new and novel summarizers referred to as ESCI (Extractive Summarization using Collocation Information) and De-ESCI (Dictionary enhanced ESCI) have been proposed. In addition, an automatic summary evaluation technique called DeFuSE (Dictionary enhanced Fuzzy Summary Evaluator) has also been introduced.The mathematical basis for the evolution of the scoring scheme proposed in this thesis and its relationship with other well-known summarization algorithms such as latent Semantic Indexing (LSI) is also derived. The work detailed in this thesis is specific to the domain of extractive summarization of unstructured text without taking into account the data set characteristics such as the positional importance of sentences. This is to ensure that the summarizer works well for a broad class of documents, and to keep the proposed models as generic as possible. Central to the proposed work is the concept of "Collocation Information of a word", its quantification and application to summarization. "Collocation Information" (CI) is the amount of information (Shannon’s measure) that a word and its collocations together contribute to the total information in the document(s) being summarized.The CI of a word has been computed using Shannon’s measure for information using a joint probability distribution. Further, a base value of CI called "Discrimination Threshold" (DT) has also been derived. To determine DT, sentences from a large collection of documents covering various topics including the topic covered by the document(s) being summarized were broken down into sequences of word collocations.The number of possible neighbors for a word within a specified collocation window was determined. This number has been called the "cardinality of the collocating set" and is represented as |ℵ (w)|. It is proved that if |ℵ (w)| determined from this large document collection for any word w is fixed, then the maximum value of the CI for a word w is proportional to |ℵ (w)|. This constrained maximum is the "Discrimination Threshold" and is used as the base value of CI. Experimental evidence detailed in this thesis shows that sentences containing words with CI greater than DT are most likely to be useful in an extract. Words in every sentence of the document(s) being summarized have been assigned scores based on the difference between their current value of CI and their respective DT. Individual word scores have been summed to derive a score for every sentence. Sentences are ranked according to their scores and the first few sentences in the rank order have been selected as the extract summary. Redundant and semantically similar sentences have been excluded from the selection process using a simple similarity detection algorithm. This novel method for extraction has been called ESCI in this thesis. In the second part of the thesis, the advantages of tagging words as nouns, verbs, adjectives and adverbs without the use of sense disambiguation has been explored. A hierarchical model for abstraction of knowledge has been proposed, and those cases where such a model can improve summarization accuracy have been explained. Knowledge abstraction has been achieved by converting collocations into their hypernymous versions. In the second part of the thesis, the advantages of tagging words as nouns, verbs, adjectives and adverbs without the use of sense disambiguation has been explored. A hierarchical model for abstraction of knowledge has been proposed, and those cases where such a model can improve summarization accuracy have been explained. Knowledge abstraction has been achieved by converting collocations into their hypernymous versions. The number of levels of abstraction varies based on the sense tag given to each word in the collocation being abstracted. Once abstractions have been determined, Expectation- Maximization algorithm is used to determine the probability value of each collocation at every level of abstraction. A combination of abstracted collocations from various levels is then chosen and sentences are assigned scores based on collocation information of these abstractions.This summarization scheme has been referred to as De-ESCI (Dictionary enhanced ESCI). It had been observed in many human summary data sets that the factual attribute of the human determines the choice of noun and verb pairs. Similarly, the emotional attribute of the human determines the choice of the number of noun and adjective pairs. In order to bring these attributes into the machine generated summaries, two variants of DeESCI have been proposed. The summarizer with the factual attribute has been called as De-ESCI-F, while the summarizer with the emotional attribute has been called De-ESCI-E in this thesis. Both create summaries having two parts. First part of the summary created by De-ESCI-F is obtained by scoring and selecting only those sentences where a fixed number of nouns and verbs occur.The second part of De-ESCI-F is obtained by ranking and selecting those sentences which do not qualify for the selection process in the first part. Assigning sentence scores and selecting sentences for the second part of the summary is exactly like in ESCI. Similarly, the first part of De-ESCI-E is generated by scoring and selecting only those sentences where fixed number of nouns and adjectives occur. The second part of the summary produced by De-ESCI-E is exactly like the second part in De-ESCI-F. As the model summary generated by human summarizers may or may not contain sentences with preference given to qualifiers (adjectives), the automatic summarizer does not know apriori whether to choose sentences with qualifiers over those without qualifiers. As there are two versions of the summary produced by De-ESCI-F and De-ESCIE, one of them should be closer to the human summarizer’s point of view (in terms of giving importance to qualifiers). This technique of choosing the best candidate summary has been referred to as De-ESCI-F/E. Performance Metrics The focus of this thesis is to propose new models and sentence ranking techniques aimed at improving the accuracy of the extract in terms of sentences selected, rather than on the readability of the summary. As a result, the order of sentences in the summary is not given importance during evaluation. Automatic evaluation metrics have been used and the performance of the automatic summarizer has been evaluated in terms of precision, recall and f-scores obtained by comparing its output with model human generated extract summaries. A novel summary evaluator called DeFuSE has been proposed in this thesis, and its scores are used along with the scores given by a standard evaluator called ROUGE. DeFuSE evaluates an extract in terms of precision, recall and f-score relying on The use of WordNet hypernymy structure to identify semantically similar sentences in a document. It also uses fuzzy set theory to compute the extent to which a sentence from the machine generated extract belongs to the model summary. Performance of candidate summarizers has been discussed in terms of percentage improvement in fscore relative to the baselines. Average of ROUGE and DeFuSE f-score for every summary is computed, and the mean value of these scores is used to compare performance improvement. Performance For illustrative purposes, DUC 2002 and DUC 2003 multi-document data sets have been used. From these data sets only the 400 word summaries of DUC 2002 and track-4 (novelty track) summaries of DUC 2003 are useful for evaluation of sentence extracts and hence only these have been used. f-score has been chosen as a measure of performance. Standard baselines such as coverage, size and lead and also probabilistic baselines have been used to measure percentage improvement in f-score of candidate summarizers relative to these baselines. Further, summaries generated by MEAD using centroid and length as features for ranking (MEAD-CL), MEAD using positional, centroid and length as features for ranking (MEAD-CLP), Microsoft Word automatic summarizer (MS-Word) and Latent Semantic Indexing (LSI) based summarizer were used to compare the performance of the proposed summarization schemes.

Very often, people buy goods on the Internet that they can not see and try. They therefore rely on reviews of other customers. However, there may be too many reviews for a human to handle them quickly and comfortably. The aim of this work is to offer an application that can recognize in Czech reviews what features of a product are most commented and whether the commentary is positive or negative. The results can save a lot of time for e-shop customers and provide interesting feedback to the manufacturers of the products.

The Linz Center of Mechatronics’ SymSpace platform is designed to provide intelligent design and training for the traditional engineer training and industrial design approach in the field of motor design, which relies on the engineer’s own experience and manual work. This paper first analyzes and explores the usage patterns and possible improvement perspectives of motor design components using SymSpace user data. Then an attempt is made to summarize the motor design manual provided by LCM using a text summary model and use it for training engineers. Next, a question-and-answer system model was used to try to provide an aid system for engineers in design. The evaluation of text summaries and question and answer systems is difficult in the motor design domain because the amount of redundant textual information in this domain is small and key information is often presented in detail rather than in the main stem of the sentence. In this case, instead of evaluating the model using traditional machine scores, this paper refers to the feedback from LCM experts as future users. The final results show that, despite the problems of difficulty in explaining the reasons; the possibility of being misleading; and the loss of information details, both attempts are generally positive and the exploration in this direction is worthwhile. / Symspace från Linz Center of Mechatronics är utformad för att tillhandahålla intelligent design och utbildning för den traditionella ingenjörsutbildningen och den industriella designmetoden inom motorkonstruktion, som bygger på ingenjörens egen erfarenhet och manuellt arbete. I den här artikeln analyseras och utforskas först användningsmönster och möjliga förbättringsperspektiv för komponenter för motorkonstruktion med hjälp av användaruppgifter från Symspace. Därefter görs ett försök att sammanfatta den motorkonstruktionshandbok som tillhandahålls av LCM med hjälp av en modell för textsammanfattningar och använda den för att utbilda ingenjörer. Därefter användes en modell för ett system med frågor och svar för att försöka tillhandahålla ett hjälpsystem för ingenjörer vid konstruktion. Utvärderingen av textsammanfattningar och fråga-och-svar-system är svår inom motorkonstruktionsområdet eftersom mängden överflödig textinformation inom detta område är liten och nyckelinformation ofta presenteras i detalj snarare än i huvudstammen av meningen. I det här fallet hänvisar den här artikeln i stället för att utvärdera modellen med hjälp av traditionella maskinpoäng till feedback från LCM-experter som framtida användare. De slutliga resultaten visar att trots problemen med svårigheten att förklara orsakerna, möjligheten att vara vilseledande och förlusten av informationsdetaljer är båda försöken generellt sett positiva och att utforskningen i denna riktning är värd att fortsätta.

During an investigation carried out by a police officer or a gendarme, audition reports are written, the length of which can be up to several pages. The high-level goal of this thesis is to study various automatic and reliable text summarization methods to help with this time-consuming task. One challenge comes from the specific, French and judicial data that we wish to summarize; and another challenge comes from the need for reliable and factual models. First, this thesis focuses on automatic summarization evaluation, in terms of both content (how well the summary captures essential information of the source text) and factuality (to what extent the summary only includes information from or coherent with the source text). Factuality evaluation, in particular, is of crucial interest when using LLMs for judicial purposes, because of their hallucination risks. Notably, we propose a light variation of SelfCheckGPT, which has a stronger correlation with human judgment (0.743) than the wide-spread BARTScore (0.542), or our study dataset. Other paradigms, such as Question-Answering, are studied in this thesis, which however underperform compared to these. Then, extractive summarization methods are explored and compared, including one based on graphs via the TextRank algorithm, and one based on greedy optimization. The latter (overlap rate: 0.190, semantic similarity: 0.513) clearly outperforms the base TextRank (overlap rate: 0.172, semantic similarity: 0.506). An improvement of the TextRank with a threshold mechanism is also proposed, leading to a non-negligible improvement (overlap rate: 0.180, semantic similarity: 0.513). Finally, abstractive summarization, with pre-trained LLMs based on a Transformer architecture, is studied. In particular, several general-purpose and multilingual models (Llama-2, Mistral and Mixtral) were objectively compared on a summarization dataset of judicial procedures from the French police. Results show that the performances of these models are highly related to their size: Llama-2 7B struggles to adapt to uncommon data (overlap rate: 0.083, BARTScore: -3.099), while Llama-2 13B (overlap rate: 0.159, BARTScore: -2.718) and Llama-2 70B (overlap rate: 0.191, BARTScore: -2.479) have proven quite versatile and efficient. To improve the performances of the smallest models, empirical prompt-engineering and parameter-efficient fine-tuning are explored. Notably, our fine-tuned version of Mistral 7B reaches performances comparable to those of much larger models (overlap rate: 0.185, BARTScore: -2.060), without the need for empirical prompt-engineering, and with a linguistic style closer to what is expected. / Under en utredning som görs av en polis eller en gendarm skrivs förhörsprotokoll vars längd kan vara upp till flera sidor. Målet på hög nivå med denna rapport är att studera olika automatiska och tillförlitliga textsammanfattningsmetoder för att hjälpa till med denna tidskrävande uppgift. En utmaning kommer från de specifika franska och rättsliga uppgifter som vi vill sammanfatta; och en annan utmaning kommer från behovet av pålitliga, sakliga och uppfinningsfria modeller. För det första fokuserar denna rapport på automatisk sammanfattningsutvärdering, både vad gäller innehåll (hur väl sammanfattningen fångar väsentlig information i källtexten) och fakta (i vilken utsträckning sammanfattningen endast innehåller information från eller överensstämmer med källtexten). Faktautvärdering, i synnerhet, är av avgörande intresse när man använder LLM för rättsliga ändamål, på grund av deras hallucinationsrisker. Vi föreslår särskilt en lätt variant av SelfCheckGPT, som har en starkare korrelation med mänskligt omdöme (0,743) än den utbredda BARTScore (0,542), eller vår studiedatauppsättning. Andra paradigm, såsom Question-Answering, studeras i denna rapport, som dock underpresterar jämfört med dessa. Sedan utforskas och jämförs extraktiva sammanfattningsmetoder, inklusive en baserad på grafer via TextRank-algoritmen och en baserad på girig optimering. Den senare (överlappning: 0,190, semantisk likhet: 0,513) överträffar klart basen TextRank (överlappning: 0,172, semantisk likhet: 0,506). En förbättring av TextRank med en tröskelmekanism föreslås också, vilket leder till en icke försumbar förbättring (överlappning: 0,180, semantisk likhet: 0,513). Slutligen studeras abstrakt sammanfattning, med förutbildade LLM baserade på en transformatorarkitektur. I synnerhet jämfördes flera allmänna och flerspråkiga modeller (Llama-2, Mistral och Mixtral) objektivt på en sammanfattningsdatauppsättning av rättsliga förfaranden från den franska polisen. Resultaten visar att prestandan för dessa modeller är starkt relaterade till deras storlek: Llama-2 7B kämpar för att anpassa sig till ovanliga data (överlappning: 0,083, BARTScore: -3,099), medan Llama-2 13B (överlappning: 0,159, BARTScore: -2,718) och Llama-2 70B (överlappning: 0,191, BARTScore: -2,479) har visat sig vara ganska mångsidiga och effektiva. För att förbättra prestandan för de minsta modellerna utforskas empirisk prompt-teknik och parametereffektiv finjustering. Noterbart är att vår finjusterade version av Mistral 7B når prestanda som är jämförbara med de för mycket större modeller (överlappning: 0,185, BARTScore: -2,060), utan behov av empirisk prompt-teknik och med en språklig stil som ligger närmare vad som förväntas.