No description available.

Automatic movie analysis is the task of employing Machine Learning methods to the field of screenplays, movie scripts, and motion pictures to facilitate or enable various tasks throughout the entirety of a movie’s life-cycle. From helping with making informed decisions about a new movie script with respect to aspects such as its originality, similarity to other movies, or even commercial viability, all the way to offering consumers new and interesting ways of viewing the final movie, many stages in the life-cycle of a movie stand to benefit from Machine Learning techniques that promise to reduce human effort, time, or both. Within this field of automatic movie analysis, this thesis addresses the task of summarising the content of screenplays, enabling users at any stage to gain a broad understanding of a movie from greatly reduced data. The contributions of this thesis are four-fold: (i)We introduce ScriptBase, a new large-scale data set of original movie scripts, annotated with additional meta-information such as genre and plot tags, cast information, and log- and tag-lines. To our knowledge, Script- Base is the largest data set of its kind, containing scripts and information for almost 1,000 Hollywood movies. (ii) We present a dynamic summarisation model for the screenplay domain, which allows for extraction of highly informative and important scenes from movie scripts. The extracted summaries allow for the content of the original script to stay largely intact and provide the user with its important parts, while greatly reducing the script-reading time. (iii) We extend our summarisation model to capture additional modalities beyond the screenplay text. The model is rendered multi-modal by introducing visual information obtained from the actual movie and by extracting scenes from the movie, allowing users to generate visual summaries of motion pictures. (iv) We devise a novel end-to-end neural network model for generating natural language screenplay overviews. This model enables the user to generate short descriptive and informative texts that capture certain aspects of a movie script, such as its genres, approximate content, or style, allowing them to gain a fast, high-level understanding of the screenplay. Multiple automatic and human evaluations were carried out to assess the performance of our models, demonstrating that they are well-suited for the tasks set out in this thesis, outperforming strong baselines. Furthermore, the ScriptBase data set has started to gain traction, and is currently used by a number of other researchers in the field to tackle various tasks relating to screenplays and their analysis.

This thesis focuses on two important aspects of non-native reading comprehension: text readability assessment, which estimates the reading difficulty of a given text for L2 learners, and learner summarisation assessment, which evaluates the quality of learner summaries to assess their reading comprehension. We approach both tasks as supervised machine learning problems and present automated assessment systems that achieve state-of-the-art performance. We first address the task of text readability assessment for L2 learners. One of the major challenges for a data-driven approach to text readability assessment is the lack of significantly-sized level-annotated data aimed at L2 learners. We present a dataset of CEFR-graded texts tailored for L2 learners and look into a range of linguistic features affecting text readability. We compare the text readability measures for native and L2 learners and explore methods that make use of the more plentiful data aimed at native readers to help improve L2 readability assessment. We then present a summarisation task for evaluating non-native reading comprehension and demonstrate an automated summarisation assessment system aimed at evaluating the quality of learner summaries. We propose three novel machine learning approaches to assessing learner summaries. In the first approach, we examine using several NLP techniques to extract features to measure the content similarity between the reading passage and the summary. In the second approach, we calculate a similarity matrix and apply a convolutional neural network (CNN) model to assess the summary quality using the similarity matrix. In the third approach, we build an end-to-end summarisation assessment model using recurrent neural networks (RNNs). Further, we combine the three approaches to a single system using a parallel ensemble modelling technique. We show that our models outperform traditional approaches that rely on exact word match on the task and that our best model produces quality assessments close to professional examiners.

The rapid growth of digital video content in recent years has imposed the need for the development of technologies with the capability to produce condensed but semantically rich versions of the input video stream in an effective manner. Consequently, the topic of Video Summarisation is becoming increasingly popular in multimedia community and numerous video abstraction approaches have been proposed accordingly. These recommended techniques can be divided into two major categories of automatic and semi-automatic in accordance with the required level of human intervention in summarisation process. The fully-automated methods mainly adopt the low-level visual, aural and textual features alongside the mathematical and statistical algorithms in furtherance to extract the most significant segments of original video. However, the effectiveness of this type of techniques is restricted by a number of factors such as domain-dependency, computational expenses and the inability to understand the semantics of videos from low-level features. The second category of techniques however, attempts to alleviate the quality of summaries by involving humans in the abstraction process to bridge the semantic gap. Nonetheless, a single user’s subjectivity and other external contributing factors such as distraction will potentially deteriorate the performance of this group of approaches. Accordingly, in this thesis we have focused on the development of three user-centred effective video summarisation techniques that could be applied to different video categories and generate satisfactory results. According to our first proposed approach, a novel mechanism for a user-centred video summarisation has been presented for the scenarios in which multiple actors are employed in the video summarisation process in order to minimise the negative effects of sole user adoption. Based on our recommended algorithm, the video frames were initially scored by a group of video annotators ‘on the fly’. This was followed by averaging these assigned scores in order to generate a singular saliency score for each video frame and, finally, the highest scored video frames alongside the corresponding audio and textual contents were extracted to be included into the final summary. The effectiveness of our approach has been assessed by comparing the video summaries generated based on our approach against the results obtained from three existing automatic summarisation tools that adopt different modalities for abstraction purposes. The experimental results indicated that our proposed method is capable of delivering remarkable outcomes in terms of Overall Satisfaction and Precision with an acceptable Recall rate, indicating the usefulness of involving user input in the video summarisation process. In an attempt to provide a better user experience, we have proposed our personalised video summarisation method with an ability to customise the generated summaries in accordance with the viewers’ preferences. Accordingly, the end-user’s priority levels towards different video scenes were captured and utilised for updating the average scores previously assigned by the video annotators. Finally, our earlier proposed summarisation method was adopted to extract the most significant audio-visual content of the video. Experimental results indicated the capability of this approach to deliver superior outcomes compared with our previously proposed method and the three other automatic summarisation tools. Finally, we have attempted to reduce the required level of audience involvement for personalisation purposes by proposing a new method for producing personalised video summaries. Accordingly, SIFT visual features were adopted to identify the video scenes’ semantic categories. Fusing this retrieved data with pre-built users’ profiles, personalised video abstracts can be created. Experimental results showed the effectiveness of this method in delivering superior outcomes comparing to our previously recommended algorithm and the three other automatic summarisation techniques.

Yes / The online shopping is increasing rapidly because of its convenience to buy from home and comparing products from their reviews written by other purchasers. When people buy a product, they express their emotions about that product in the form of review. In Indian context, it is found that the reviews contain Hindi text along with English. It is also found that most of the Hindi text contains opinionated words like bahut achha, bakbas, pesa wasool etc. We have tried to find out different Hindi texts appearing in product reviews written on Indian E-commerce portals. We have also developed a system which takes all those reviews containing Hindi as well as English texts and find out the sentiment expressed in that review for each attribute of the product as well as a final review of the product.

Yes / Welcome to this special issue of the Semantic Web (SWJ) journal. The special issue compiles three technical contributions that significantly advance the state-of-the-art in exploration of semantic data using semantic web techniques and technologies.

Summarization models which operate on meaning representations of documents have been neglected in the past, although they are a very promising and interesting class of methods for summarization and text understanding. In this thesis, I present one such summarizer, which uses the proposition as its meaning representation. My summarizer is an implementation of Kintsch and van Dijk's model of comprehension, which uses a tree of propositions to represent the working memory. The input document is processed incrementally in iterations. In each iteration, new propositions are connected to the tree under the principle of local coherence, and then a forgetting mechanism is applied so that only a few important propositions are retained in the tree for the next iteration. A summary can be generated using the propositions which are frequently retained. Originally, this model was only played through by hand by its inventors using human-created propositions. In this work, I turned it into a fully automatic model using current NLP technologies. First, I create propositions by obtaining and then transforming a syntactic parse. Second, I have devised algorithms to numerically evaluate alternative ways of adding a new proposition, as well as to predict necessary changes in the tree. Third, I compared different methods of modelling local coherence, including coreference resolution, distributional similarity, and lexical chains. In the first group of experiments, my summarizer realizes summary propositions by sentence extraction. These experiments show that my summarizer outperforms several state-of-the-art summarizers. The second group of experiments concerns abstractive generation from propositions, which is a collaborative project. I have investigated the option of compressing extracted sentences, but generation from propositions has been shown to provide better information packaging.

Cognitive limitations such as those described in Miller's (1956) work on channel capacity and Cowen's (2001) on short-term memory are factors in determining user cognitive load and in turn task performance. Inappropriate user cognitive load can reduce user efficiency in goal realization. For instance, if the user's attentional capacity is not appropriately applied to the task, distractor processing can tend to appropriate capacity from it. Conversely, if a task drives users beyond their short-term memory envelope, information loss may be realized in its translation to long-term memory and subsequent retrieval for task base processing. To manage user cognitive capacity in the task of text search the interface should allow users to draw on their powerful and innate pattern recognition abilities. This harmonizes with Johnson-Laird's (1983) proposal that propositional representation is tied to mental models. Combined with the theory that knowledge is highly organized when stored in memory an appropriate approach for cognitive load optimization would be to graphically present single documents, or clusters thereof, with an appropriate number and type of descriptors. These descriptors are commonly words and/or phrases. Information theory research suggests that words have different levels of importance in document topic differentiation. Although key word identification is well researched, there is a lack of basic research into human preference regarding query formation and the heuristics users employ in search. This lack extends to features as elementary as the number of words preferred to describe and/or search for a document. Contrastive understanding these preferences will help balance processing overheads of tasks like clustering against user cognitive load to realize a more efficient document retrieval process. Common approaches such as search engine log analysis cannot provide this degree of understanding and do not allow clear identification of the intended set of target documents. This research endeavours to improve the manner in which text search returns are presented so that user performance under real world situations is enhanced. To this end we explore both how to appropriately present search information and results graphically to facilitate optimal cognitive and perceptual load/utilization, as well as how people use textual information in describing documents or constructing queries.

This thesis investigates the problem of content selection for timeline generation from single history articles. While the task of timeline generation has been addressed before, most previous approaches assume the existence of a large corpus of history articles from the same era. They exploit the fact that salient information is likely to be mentioned multiple times in such corpora. However, large resources of this kind are only available for historical events that happened in the most recent decades. In this thesis, I present approaches which can be used to create history timelines for any historical period, even for eras such as the Middle Ages, for which no large corpora of supplementary text exist. The thesis first presents a system that selects relevant historical figures in a given article, a task which is substantially easier than full timeline generation. I show that a supervised approach which uses linguistic, structural and semantic features outperforms a competitive baseline on this task. Based on the observations made in this initial study, I then develop approaches for timeline generation. I find that an unsupervised approach that takes into account the article's subject area outperforms several supervised and unsupervised baselines. A main focus of this thesis is the development of evaluation methodologies and resources, as no suitable corpora existed when work began. For the initial experiment on important historical figures, I construct a corpus of existing timelines and textual articles, and devise a method for evaluating algorithms based on this resource. For timeline generation, I present a comprehensive evaluation methodology which is based on the interpretation of the task as a special form of single-document summarisation. This methodology scores algorithms based on meaning units rather than surface similarity. Unlike previous semantic-units-based evaluation methods for summarisation, my evaluation method does not require any manual annotation of system timelines. Once an evaluation resource has been created, which involves only annotation of the input texts, new timeline generation algorithms can be tested at no cost. This crucial advantage should make my new evaluation methodology attractive for the evaluation of general single-document summaries beyond timelines. I also present an evaluation resource which is based on this methodology. It was constructed using gold-standard timelines elicited from 30 human timeline writers, and has been made publicly available. This thesis concentrates on the content selection stage of timeline generation, and leaves the surface realisation step for future work. However, my evaluation methodology is designed in such a way that it can in principle also quantify the degree to which surface realisation is successful.

In this thesis, we work towards bridging the gap between two distinct areas: noisy text handling and text summarisation. The overall goal of the paper is to examine the effects of noise in the training of convolutional neural networks for text summarisation, with a view to understanding how to effectively create a noise-robust text-summarisation system. We look specifically at the problem of abstractive text summarisation of noisy data in the context of summarising error-containing documents from automatic speech recognition (ASR) output. We experiment with adding varying levels of noise (errors) to the 4 million-article Gigaword corpus and training an encoder-decoder CNN on it with the aim of producing a noise-robust text summarisation system. A total of six text summarisation models are trained, each with a different level of noise. We discover that the models with a high level of noise are indeed able to aptly summarise noisy data into clean summaries, despite a tendency for all models to overfit to the level of noise on which they were trained. Directions are given for future steps in order to create an even more noise-robust and flexible text summarisation system.