UE903 course material

ABROUGUI Rim

We present a simple, robust generation system which performs content selection and surface realization in a unified, domain-independent framework. In our approach, we break up the end-to-end generation process into a sequence of local decisions, arranged hierarchically and each trained discriminatively. We deployed our system in three different domains—Robocup sportscasting, technical weather forecasts, and common weather forecasts, obtaining results comparable to state-ofthe-art domain-specific systems both in terms of BLEU scores and human evaluation.

AFARA Maria

This paper proposes a data-driven method for concept-to-text generation, the task of automatically producing textual output from non-linguistic input. A key insight in our approach is to reduce the tasks of content selection (“what to say”) and surface realization (“how to say”) into a common parsing problem. We define a probabilistic context-free grammar that describes the structure of the input (a corpus of database records and text describing some of them) and represent it compactly as a weighted hypergraph. The hypergraph structure encodes exponentially many derivations, which we rerank discriminatively using local and global features. We propose a novel decoding algorithm for finding the best scoring derivation and generating in this setting. Experimental evaluation on the ATIS domain shows that our model outperforms a competitive discriminative system both using BLEU and in a judgment elicitation study.

AKANI Aduenu

We present a simple, robust generation system which performs content selection and surface realization in a unified, domain-independent framework. In our approach, we break up the end-to-end generation process into a sequence of local decisions, arranged hierarchically and each trained discriminatively. We deployed our system in three different domains—Robocup sportscasting, technical weather forecasts, and common weather forecasts, obtaining results comparable to state-ofthe-art domain-specific systems both in terms of BLEU scores and human evaluation.

AKHMETOV Alisher

Most of the known stochastic sentence generators use syntactically annotated corpora, performing the projection to the surface in one stage. However, in full-fledged text generation, sentence realization usually starts from semantic (predicate-argument) structures. To be able to deal with semantic structures, stochastic generators require semantically annotated, or, even better, multilevel annotated corpora. Only then can they deal with such crucial generation issues as sentence planning, linearization and morphologization. Multilevel annotated corpora are increasingly available for multiple languages. We take advantage of them and propose a multilingual deep stochastic sentence realizer that mirrors the state-ofthe-art research in semantic parsing. The realizer uses an SVM learning algorithm. For each pair of adjacent levels of annotation, a separate decoder is defined. So far, we evaluated the realizer for Chinese, English, German, and Spanish.

BALARD Srilakshmi

We present a framework for text simplification based on applying transformation rules to a typed dependency representation produced by the Stanford parser. We test two approaches to regeneration from typed dependencies: (a) gen-light, where the transformed dependency graphs are linearised using the word order and morphology of the original sentence, with any changes coded into the transformation rules, and (b) gen-heavy, where the Stanford dependencies are reduced to a DSyntS representation and sentences are generating formally using the RealPro surface realiser. The main contribution of this paper is to compare the robustness of these approaches in the presence of parsing errors, using both a single parse and an n-best parse setting in an overgenerate and rank approach. We find that the gen-light approach is robust to parser error, particularly in the n-best parse setting. On the other hand, parsing errors cause the realiser in the genheavy approach to order words and phrases in ways that are disliked by our evaluators.

BOUZIGUES Aymeric

In this paper, we consider sentence simplification as a special form of translation with the complex sentence as the source and the simple sentence as the target. We propose a Tree-based Simplification Model (TSM), which, to our knowledge, is the first statistical simplification model covering splitting, dropping, reordering and substitution integrally. We also describe an efficient method to train our model with a large-scale parallel dataset obtained from the Wikipedia and Simple Wikipedia. The evaluation shows that our model achieves better readability scores than a set of baseline systems.

DIEUDONAT Lea

We consider the problem of producing a multi-document summary given a collection of documents. Since most successful methods of multi-document summarization are still largely extractive, in this paper, we explore just how well an extractive method can perform. We introduce an “oracle” score, based on the probability distribution of unigrams in human summaries. We then demonstrate that with the oracle score, we can generate extracts which score, on average, better than the human summaries, when evaluated with ROUGE. In addition, we introduce an approximation to the oracle score which produces a system with the best known performance for the 2005 Document Understanding Conference (DUC) evaluation.

GUILLAUME Maxime

Human evaluations of machine translation are extensive but expensive. Human evaluations can take months to finish and involve human labor that can not be reused. We propose a method of automatic machine translation evaluation that is quick, inexpensive, and language-independent, that correlates highly with human evaluation, and that has little marginal cost per run. We present this method as an automated understudy to skilled human judges which substitutes for them when there is need for quick or frequent evaluations.

HAN Kelvin

ROUGE stands for Recall-Oriented Understudy for Gisting Evaluation. It includes measures to automatically determine the quality of a summary by comparing it to other (ideal) summaries created by humans. The measures count the number of overlapping units such as n-gram, word sequences, and word pairs between the computer-generated summary to be evaluated and the ideal summaries created by humans. This paper introduces four different ROUGE measures: ROUGE-N, ROUGE-L, ROUGE-W, and ROUGE-S included in the ROUGE summarization evaluation package and their evaluations. Three of them have been used in the Document Understanding Conference (DUC) 2004, a large-scale summarization evaluation sponsored by NIST.

LEAVITT Phyllicia

We present a comparison of word-based and character-based sequence-to-sequence models for data-to-text natural language generation, which generate natural language descriptions for structured inputs. On the datasets of two recent generation challenges, our models achieve comparable or better automatic evaluation results than the best challenge submissions. Subsequent detailed statistical and human analyses shed light on the differences between the two input representations and the diversity of the generated texts. In a controlled experiment with synthetic training data generated from templates, we demonstrate the ability of neural models to learn novel combinations of the templates and thereby generalize beyond the linguistic structures they were trained on.

LY Sophea

A knowledge base is a large repository of facts that are mainly represented as RDF triples, each of which consists of a subject, a predicate (relationship), and an object. The RDF triple representation offers a simple interface for applications to access the facts. However, this representation is not in a natural language form, which is difficult for humans to understand. We address this problem by proposing a system to translate a set of RDF triples into natural sentences based on an encoder-decoder framework. To preserve as much information from RDF triples as possible, we propose a novel graph-based triple encoder. The proposed encoder encodes not only the elements of the triples but also the relationships both within a triple and between the triples. Experimental results show that the proposed encoder achieves a consistent improvement over the baseline models by up to 17.6%, 6.0%, and 16.4% in three common metrics BLEU, METEOR, and TER, respectively.

MARQUER Esteban

Data-to-text generation can be conceptually divided into two parts: ordering and structuring the information (planning), and generating fluent language describing the information (realization). Modern neural generation systems conflate these two steps into a single end-to-end differentiable system. We propose to split the generation process into a symbolic text-planning stage that is faithful to the input, followed by a neural generation stage that focuses only on realization. For training a plan-to-text generator, we present a method for matching reference texts to their corresponding text plans. For inference time, we describe a method for selecting high-quality text plans for new inputs. We implement and evaluate our approach on the WebNLG benchmark. Our results demonstrate that decoupling text planning from neural realization indeed improves the system’s reliability and adequacy while maintaining fluent output. We observe improvements both in BLEU scores and in manual evaluations. Another benefit of our approach is the ability to output diverse realizations of the same input, paving the way to explicit control over the generated text structure.

NADAR Fatima

Traditionally, Referring Expression Generation (REG) models first decide on the form and then on the content of references to discourse entities in text, typically relying on features such as salience and grammatical function. In this paper, we present a new approach (NeuralREG), relying on deep neural networks, which makes decisions about form and content in one go without explicit feature extraction. Using a delexicalized version of the WebNLG corpus, we show that the neural model substantially improves over two strong baselines. Data and models are publicly available1

The problem of AMR-to-text generation is to recover a text representing the same meaning as an input AMR graph. The current state-of-the-art method uses a sequence-to-sequence model, leveraging LSTM for encoding a linearized AMR structure. Although being able to model non-local semantic information, a sequence LSTM can lose information from the AMR graph structure, and thus facing challenges with large-graphs, which result in long sequences. We introduce a neural graph-to-sequence model, using a novel LSTM structure for directly encoding graph-level semantics. On a standard benchmark, our model shows superior results to existing methods in the literature.

RAZET Guilherme

Generating text from graph-based data, such as Abstract Meaning Representation (AMR), is a challenging task due to the inherent difficulty in how to properly encode the structure of a graph with labeled edges. To address this difficulty, we propose a novel graph-to-sequence model that encodes different but complementary perspectives of the structural information contained in the AMR graph. The model learns parallel top-down and bottom-up representations of nodes capturing contrasting views of the graph. We also investigate the use of different node message passing strategies, employing different state-of-the-art graph encoders to compute node representations based on incoming and outgoing perspectives. In our experiments, we demonstrate that the dual graph representation leads to improvements in AMR-to-text generation, achieving state-ofthe-art results on two AMR datasets.

SABDENOV Aidos

Neural sequence-to-sequence models have provided a viable new approach for abstractive text summarization (meaning they are not restricted to simply selecting and rearranging passages from the original text). However, these models have two shortcomings: they are liable to reproduce factual details inaccurately, and they tend to repeat themselves. In this work we propose a novel architecture that augments the standard sequence-to-sequence attentional model in two orthogonal ways. First, we use a hybrid pointer-generator network that can copy words from the source text via pointing, which aids accurate reproduction of information, while retaining the ability to produce novel words through the generator. Second, we use coverage to keep track of what has been summarized, which discourages repetition. We apply our model to the CNN / Daily Mail summarization task, outperforming the current abstractive state-of-the-art by at least 2 ROUGE points.

SRIVASTAVA Preprak

We propose a selective encoding model to extend the sequence-to-sequence framework for abstractive sentence summarization. It consists of a sentence encoder, a selective gate network, and an attention equipped decoder. The sentence encoder and decoder are built with recurrent neural networks. The selective gate network constructs a second level sentence representation by controlling the information flow from encoder to decoder. The second level representation is tailored for sentence summarization task, which leads to better performance. We evaluate our model on the English Gigaword, DUC 2004 and MSR abstractive sentence summarization datasets. The experimental results show that the proposed selective encoding model outperforms the state-of-the-art baseline models.

TSAI Yi-Ting

We present an LSTM approach to deletion-based sentence compression where the task is to translate a sentence into a sequence of zeros and ones, corresponding to token deletion decisions. We demonstrate that even the most basic version of the system, which is given no syntactic information (no PoS or NE tags, or dependencies) or desired compression length, performs surprisingly well: around 30% of the compressions from a large test set could be regenerated. We compare the LSTM system with a competitive baseline which is trained on the same amount of data but is additionally provided with all kinds of linguistic features. In an experiment with human raters the LSTMbased model outperforms the baseline achieving 4.5 in readability and 3.8 in informativeness.

YANG Ruoxiao (Lisa)

Single document summarization is the task of producing a shorter version of a document while preserving its principal information content. In this paper we conceptualize extractive summarization as a sentence ranking task and propose a novel training algorithm which globally optimizes the ROUGE evaluation metric through a reinforcement learning objective. We use our algorithm to train a neural summarization model on the CNN and DailyMail datasets and demonstrate experimentally that it outperforms state-of-the-art extractive and abstractive systems when evaluated automatically and by humans.