Natural Language Sentences Research Articles

Text Generation From Tables

This paper proposes a neural generative model, namely Table2Seq, to generate a natural language sentence based on a table. Specifically, the model maps a table to continuous vectors and then generates a natural language sentence by leveraging the semantics of a table. Since rare words, e.g., entities and values, usually appear in a table, we develop a flexible copying mechanism that selectively replicates contents from the table to the output sequence. We conduct extensive experiments to demonstrate the effectiveness of our Table2Seq model and the utility of the designed copying mechanism. On the WIKIBIO and SIMPLEQUESTIONS datasets, the Table2Seq model improves the state-of-the-art results from 34.70 to 40.26 and from 33.32 to 39.12 in terms of BLEU-4 scores, respectively. Moreover, we construct an open-domain dataset WIKITABLETEXT that includes 13 318 descriptive sentences for 4962 tables. Our Table2Seq model achieves a BLEU-4 score of 38.23 on WIKITABLETEXT outperforming template-based and language model based approaches. Furthermore, through experiments on 1 M table-query pairs from a search engine, our Table2Seq model considering the structured part of a table, i.e., table attributes and table cells, as additional information outperforms a sequence-to-sequence model considering only the sequential part of a table, i.e., table caption.

Relation Extraction for Massive News Texts

With the development of information technology including Internet technologies, the amount of textual information that people need to process daily is increasing. In order to automatically obtain valuable and user-informed information from massive amounts of textual data, many researchers have conducted in-depth research in the area of entity relation extraction. Based on the existing research of word vector and the method of entity relation extraction, this paper designs and implements an method based on support vector machine (SVM) for extracting English entity relationships from massive news texts. The method converts sentences in natural language into a form of numerical matrix that can be understood and processed by computers through word embedding and position embedding. Then the key features are extracted, and feature vectors are constructed and sent to the SVM classifiers for relation classification. In the process of feature extraction, we had two different models to finish the job, one by Principal Component Analysis (PCA) and the other by Convolutional Neural Networks (CNN). We designed experiments to evaluate the algorithm.

Разработка автоматизированных методов представления знаний о действиях и ситуациях

The article is devoted to the development of semi-automated methods for integrating knowledge extracted from natural language texts. To solve this problem, we use methods for converting natural language sentences into fragments of atomic diagram of algebraic systems. The knowledge extracted from texts is formalized with atomic sentences of signature consisting of constant symbols, binary predicates and additional situation-constants. We developed methods for integrating knowledge contained in several sentences of natural language, allowing to take their semantic contexts into account.

A Natural-language-based Visual Query Approach of Uncertain Human Trajectories.

Visual querying is essential for interactively exploring massive trajectory data. However, the data uncertainty imposes profound challenges to fulfill advanced analytics requirements. On the one hand, many underlying data does not contain accurate geographic coordinates, e.g., positions of a mobile phone only refer to the regions (i.e., mobile cell stations) in which it resides, instead of accurate GPS coordinates. On the other hand, domain experts and general users prefer a natural way, such as using a natural language sentence, to access and analyze massive movement data. In this paper, we propose a visual analytics approach that can extract spatial-temporal constraints from a textual sentence and support an effective query method over uncertain mobile trajectory data. It is built up on encoding massive, spatially uncertain trajectories by the semantic information of the POls and regions covered by them, and then storing the trajectory documents in text database with an effective indexing scheme. The visual interface facilitates query condition specification, situation-aware visualization, and semantic exploration of large trajectory data. Usage scenarios on real-world human mobility datasets demonstrate the effectiveness of our approach.

FEATURES OF DIAGNOSTIC ARTIFICIAL NEURAL NETWORKS FOR HYBRID EXPERT SYSTEMS

In the proposed article, various methods of constructing an artificial neural network as one of the components of a hybrid expert system for diagnosis were investigated. A review of foreign literature in recent years was conducted, where hybrid expert systems were considered as an integral part of complex technical systems in the field of security. The advantages and disadvantages of artificial neural networks are listed, and the main problems in creating hybrid expert systems for diagnostics are indicated, proving the relevance of further development of artificial neural networks for hybrid expert systems. The approaches to the analysis of natural language sentences, which are used for the work of hybrid expert systems with artificial neural networks, are considered. A bulletin board is shown, its structure and principle of operation are described. The structure of the bulletin board is divided into levels and sublevels. At sublevels, a confidence factor is applied. The dependence of the values of the confidence factor on the fulfillment of a particular condition is shown. The links between the levels and sublevels of the bulletin board are also described. As an artificial neural network architecture, the «key-threshold» model is used, the rule of neuron operation is shown. In addition, an artificial neural network has the property of training, based on the application of the penalty property, which is able to calculate depending on the accident situation. The behavior of a complex technical system, as well as its faulty states, are modeled using a model that describes the structure and behavior of a given system. To optimize the data of a complex technical system, an evolutionary algorithm is used to minimize the objective function. Solutions to the optimization problem consist of Pareto solution vectors. Optimization and training tasks are solved by using the Hopfield network. In general, a hybrid expert system is described using semantic networks, which consist of vertices and edges. The reference model of a complex technical system is stored in the knowledge base and updated during the acquisition of new knowledge. In an emergency, or about its premise, with the help of neural networks, a search is made for the cause and the control action necessary to eliminate the accident. The considered approaches, interacting with each other, can improve the operation of diagnostic artificial neural networks in the case of emergency management, showing more accurate data in a short time. In addition, the use of such a network for analyzing the state of health, as well as forecasting based on diagnostic data using the example of a complex technical system, is presented.

Deep Audio-Visual Speech Recognition.

The goal of this work is to recognise phrases and sentences being spoken by a talking face, with or without the audio. Unlike previous works that have focussed on recognising a limited number of words or phrases, we tackle lip reading as an open-world problem - unconstrained natural language sentences, and in the wild videos. Our key contributions are: (1) we compare two models for lip reading, one using a CTC loss, and the other using a sequence-to-sequence loss. Both models are built on top of the transformer self-attention architecture; (2) we investigate to what extent lip reading is complementary to audio speech recognition, especially when the audio signal is noisy; (3) we introduce and publicly release a new dataset for audio-visual speech recognition, LRS2-BBC, consisting of thousands of natural sentences from British television. The models that we train surpass the performance of all previous work on a lip reading benchmark dataset by a significant margin.

Augmenting Statistical Data Dissemination by Short Quantified Sentences of Natural Language

AbstractData from National Statistical Institutes is generally considered an important source of credible evidence for a variety of users. Summarization and dissemination via traditional methods is a convenient approach for providing this evidence. However, this is usually comprehensible only for users with a considerable level of statistical literacy. A promising alternative lies in augmenting the summarization linguistically. Less statistically literate users (e.g., domain experts and the general public), as well as disabled people can benefit from such a summarization. This article studies the potential of summaries expressed in short quantified sentences. Summaries including, for example, “most visits from remote countries are of a short duration” can be immediately understood by diverse users. Linguistic summaries are not intended to replace existing dissemination approaches, but can augment them by providing alternatives for the benefit of diverse users of official statistics. Linguistic summarization can be achieved via mathematical formalization of linguistic terms and relative quantifiers by fuzzy sets. To avoid summaries based on outliers or data with low coverage, a quality criterion is applied. The concept based on linguistic summaries is demonstrated on test interfaces, interpreting summaries from real municipal statistical data. The article identifies a number of further research opportunities, and demonstrates ways to explore those.

De Morgan’s laws and NEG-raising: a syntactic view

Abstract In this paper, we will motivate the application of specific rules of inference from the propositional calculus to natural language sentences. Specifically, we will analyse De Morgan’s laws, which pertain to the interaction of two central topics in syntactic research: negation and coordination. We will argue that the applicability of De Morgan’s laws to natural language structures can be derived from independently motivated operations of grammar and principles restricting the application of these operations. This has direct empirical consequences for the hypothesised relations between natural language and logic.

3G structure for image caption generation

It is a big challenge of computer vision to make machine automatically describe the content of an image with a natural language sentence. Previous works have made great progress on this task, but they only use the global or local image feature, which may lose some important subtle or global information of an image. In this paper, we propose a model with 3-gated model which fuses the global and local image features together for the task of image caption generation. The model mainly has three gated structures. (1) Gate for the global image feature, which can adaptively decide when and how much the global image feature should be imported into the sentence generator. (2) The gated recurrent neural network (RNN) is used as the sentence generator. (3) The gated feedback method for stacking RNN is employed to increase the capability of nonlinearity fitting. More specially, the global and local image features are combined together in this paper, which makes full use of the image information. The global image feature is controlled by the first gate and the local image feature is selected by the attention mechanism. With the latter two gates, the relationship between image and text can be well explored, which improves the performance of the language part as well as the multi-modal embedding part. Experimental results show that our proposed method outperforms the state-of-the-art for image caption generation.

МЕТОД АНАЛІЗУ ТА ОБРОБКИ ІНФОРМАЦІЇ В СППР АСУ

Предмет: У статті розглянуті питання семантичного та синтаксичного аналізу текстів, процесів декомпозиції речення на складові, побудови синтаксичної структури речення; семантичної інтерпретації фрагментів речення природної мови у фрагменти М-мови; композиції фрагментів М-мови в структуру реферату, що описує проблемне середовище. Мета: розробка методу аналізу та обробки інформації в СППР АСУ. Завдання: Провести аналіз досліджень останніх років пов'язаних з вирішенням проблеми аналізу змісту мовних структур з погляду на створення систем діалогу з програмним забезпеченням, зроблено висновки щодо доцільності напрямку досліджень. Виконати детальний опис етапів виконання семантичного та синтаксичного аналізу з описом вхідних, проміжних та вихідних даних. Визначити рекомендації щодо варіативності вхідних даних в залежності від структури речень предметної області (тексту викладеного природною мовою). Запропонувати правила формування предикатів для різних частин української мови. Визначити типовий набір службових слів для наповнення еталонного словника. Обґрунтувати виконання наведених операцій щодо об’єднання в одну обчислювальну структуру, яка виконується за визначеними етапами. Розробити процедури проходження кожного з етапів. Запропонувати правила формування синтаксичних груп з визначених пар лексем. Висновки: Застосування розробленого методу на основі визначених правил до предикатів дозволить виконати вибірку інформативно суттєвої інформації із набору текстової інформації та представити її у вигляді анотації (реферату).

VuWaDB

This paper introduces VuWaDB, which is a database for vulnerability workarounds. When faced with a newly discovered vulnerability, experts tend to use workarounds in order to mitigate the risks until a patch is issued by the vendor. Currently, the available vulnerability databases suffer from the lack of comprehensive workaround solutions. Furthermore, the presented workarounds are only limited to a few vulnerabilities but also poorly recorded in natural language sentences and plain text format. In order to construct VuWaDB, first, the related information was gathered from a number of well-known vulnerability databases and then extracted, analyzed, and labeled. In this regard, VuWaDB organizes the workarounds in six categories: configuration, modify code, redirect, remove, restrict access and, utility tool. Lastly, it was analyzed from the statistical point of view.

A 5W1H Based Annotation Scheme for Semantic Role Labeling of English Tweets

Semantic Role Labeling (SRL) is a well researched area of Natural Language Processing. State-of-the-art lexical resources have been developed for SRL on formal texts that involve a tedious annotation scheme and require linguistic expertise. The difficulties increase manifold when such complex annotation scheme is applied on tweets for identifying predicates and role arguments. In this paper, we present asimplified approach for annotation of English tweets for identification of predicates and corresponding semantic roles. For annotation purpose, we adopted the 5W1H (Who, What, When, Where, Why and How) concept which is widely used in journalism. The 5W1H task seeks to extract the semantic information in a natural language sentence by distilling it into the answers to the 5W1H questions: Who, What, When, Where, Why and How. The 5W1H approach is comparatively simple and convenient with respect to the ProbBank Semantic Role Labeling task. We report an the performance of our annotation scheme for SRL on tweets and show that non-expert annotators can produce quality SRL datafor tweets. This paper also reports the difficulties and challenges involved with semantic role labeling on twitter data and propose solutions to them.

Building Compact Entity Embeddings Using Wikidata

Representing natural language sentences has always been a challenge in statistical language modelling. Atomic discrete representations of words make it difficult to represent semantically related sentences. Other sentence components such as phrases and named-entities should be recognized and given representations as units instead of individual words. Different entity senses should be assigned different representations regardless the fact that they share identical words. In this paper, we focus on building the vector representations (embeddings) of named-entities from their contexts to facilitate the task of ontology population where named-entities need to be recognized and disambiguated in natural language text. Given a list of target named-entities, Wikidata is used to compensate for the lack of a labelled corpus to build the contexts of all target named-entities as well as all their senses. Description text and semantic relations with other named-entities are considered when building the contexts from Wikidata. To avoid noisy and uninformative features in the embeddings generated from artificially built contexts, we propose a method to build compact entity representations to sharpen entity embeddings by removing irrelevant features and emphasizing the most descriptive ones. An extended version of the Continuous Bag-of-Words model (CBOW) is used to build the joint vector representations of words and named-entities using Wikidata contexts. Each entity context is then represented by a subset of elements that maximizes the chances of keeping the most descriptive features about the target entity. The final entity representations are built by compressing the embeddings of the chosen subset using a deep stacked autoencoders model. Cosine similarity and t-SNE visualisation technique are used to evaluate the final entity vectors. Results show that semantically related entities are clustered near each other in the vector space. Entities that appear in similar contexts are assigned similar compact vector representations based on their contexts.

Augmenting Visualizations with Interactive Data Facts to Facilitate Interpretation and Communication.

Recently, an increasing number of visualization systems have begun to incorporate natural language generation (NLG) capabilities into their interfaces. NLG-based visualization systems typically leverage a suite of statistical functions to automatically extract key facts about the underlying data and surface them as natural language sentences alongside visualizations. With current systems, users are typically required to read the system-generated sentences and mentally map them back to the accompanying visualization. However, depending on the features of the visualization (e.g., visualization type, data density) and the complexity of the data fact, mentally mapping facts to visualizations can be a challenging task. Furthermore, more than one visualization could be used to illustrate a single data fact. Unfortunately, current tools provide little or no support for users to explore such alternatives. In this paper, we explore how system-generated data facts can be treated as interactive widgets to help users interpret visualizations and communicate their findings. We present Voder, a system that lets users interact with automatically-generated data facts to explore both alternative visualizations to convey a data fact as well as a set of embellishments to highlight a fact within a visualization. Leveraging data facts as interactive widgets, Voder also facilitates data fact-based visualization search. To assess Voder's design and features, we conducted a preliminary user study with 12 participants having varying levels of experience with visualization tools. Participant feedback suggested that interactive data facts aided them in interpreting visualizations. Participants also stated that the suggestions surfaced through the facts helped them explore alternative visualizations and embellishments to communicate individual data facts.

Scene understanding using natural language description based on 3D semantic graph map

A natural language description for working environment understanding is an important component in human---robot communication. Although 3D semantic graph mappings are widely studied for perceptual aspects of the environment, these approaches hardly apply to the communication issues such as natural language descriptions for a semantic graph map. There are many researches on workspace understanding over images in the field of computer vision, which automatically generate sentences while they usually never utilize multiple scenes and 3D information. In this paper, we introduce a novel natural language description method using 3D semantic graph map. An object-oriented semantic graph map is first constructed using 3D information. A graph convolutional neural network and a recurrent neural network are then used to generate a description of the map. A natural language sentence focusing on objects over 3D semantic graph map can be eventually generated consisting of a single scene or multiple scenes. We validate the proposed method using publicly available dataset and compare it with conventional methods.

Maverick

This paper presents Maverick, a system for discovering exceptional facts about entities in knowledge graphs. Maverick is built upon a beam-search based algorithmic framework which we proposed in a research paper that is published in SIGMOD 2018. In this demonstration proposal, we showcase an end-to-end system that includes a user-facing portal and a cache server. In Maverick, an exceptional fact about an entity of interest is modeled as a context-subspace pair, in which the subspace is a set of attributes and the context is defined by a graph query pattern of which the entity is a match, together with other matching entities. The entity is exceptional among the entities in the context, with regard to the sub-space. The portal allows users to search entities in a knowledge graph and explores exceptional facts about the entities of interest. It presents exceptional facts to users in forms of natural language sentences and illustration charts, for better interpretability of the discovered exceptional facts. The cache server stores intermediate computation results, such as pattern evaluations, exceptionality calculations, and candidate patterns. It is built for sharing computation across entities, such that repetitive computation across entities can be avoided.

A system for transformation of sentences from the enriched formalized Node of Knowledge record into relational database

Verbalized text contains knowledge necessary and sufficient for transfer of numerous human cognitions. The question is how this knowledge was saved into text. Authors believe that they found an important idea how to assemble knowledge into nodes of knowledge. Similar methods in the field of knowledge networks exist, but none of them does it in the same way as the Node of Knowledge (NOK) method. Basic terms and concepts in a language are represented in words whose meaning is final and cannot be divided in subterms. More complex meaning can be achieved by combining words in sentences. According to authors’ opinion, each sentence includes connective medium for words in the sentence (related to semantic reasons), which is inwrought in the Node of Knowledge method. Using the Node of Knowledge method each sentence can be presented as a network of connected words. This network is enriched with links between words so a computer can interpret meaning and knowledge of the sentence in the same way an intelligent person does. A formalized and semantically enriched record of sentences (called Formalized Node of Knowledge – FNOK) is developed. Authors find that in this way, even without statistical text analysis, an algorithm can give correct answer to a question set based on written text. This paper presents the system for transformation of textual knowledge expressed in natural language sentences into a relational database. The system is a part of a larger knowledge-based system based on the Node of Knowledge (NOK) conceptual framework for knowledge-based system development. This paper starts with sentences written in the formalized and enriched form for which a logical transformation into the structure of a relational database is proposed. The system and algorithms for the transformation of formalized sentences into n-tuples of the relational database are distributed and represented in several steps. The research has shown that it is possible to store semantically enriched sentences in relational databases. The solution presented in this paper is important for further development of the system for receiving questions from users and providing answers (i.e. question-answering system), with the ability to use well-developed relational SQL languages. Relational database of texts enables numerous applications in the field of expert and intelligent systems.

Aggregation with Recombination Patterns

In this paper, we show the commonalities between aggregation processes in Natural Language Generation and recombination patterns, a framework introduced recently as a way of generating complex sentences in natural languages using very simple recombination –and therefore biological– rules. By showing similarities between these two mechanisms, we suggest the possibility of carrying out aggregation by means of recombination patterns. We also refer to the possibility of using such a biological-motivated framework in the design of efficient and simple natural language generation devices.

Contentual approach to negation1

Abstract Interpretations of logics with only truth-functional connectives create a number of problems regarding the understanding of interpreted sentences. A particular problem is caused by the understanding of a sentence that is the negation of another. What is the meaning of sentence ¬p, for a particular sentence p? Even when we know what the semantic correlate of the sentence p is, we still do not know how to understand the semantic correlate of the sentence ¬p. The standard algebraic approach does not explain much. The problem is still open, because it is still unclear how to understand the element of algebra −v(p). The propositional logic with the connective of the content implication, as it is a non-Fregean logic, proposes a simple solution to this problem. The key to understanding the content of the sentence ¬p is to understand the content of the sentence p. Because, the content of the natural language sentence usually has a partial understanding, so it can be understood differently. Therefore, the negation ¬p is here understood adequately to the current and partial understanding of the sentence p. It seems that the proposed approach is consistent with our daily thinking.

An Efficient Framework for Development of Task-Oriented Dialog Systems in a Smart Home Environment.

In recent times, with the increasing interest in conversational agents for smart homes, task-oriented dialog systems are being actively researched. However, most of these studies are focused on the individual modules of such a system, and there is an evident lack of research on a dialog framework that can integrate and manage the entire dialog system. Therefore, in this study, we propose a framework that enables the user to effectively develop an intelligent dialog system. The proposed framework ontologically expresses the knowledge required for the task-oriented dialog system’s process and can build a dialog system by editing the dialog knowledge. In addition, the framework provides a module router that can indirectly run externally developed modules. Further, it enables a more intelligent conversation by providing a hierarchical argument structure (HAS) to manage the various argument representations included in natural language sentences. To verify the practicality of the framework, an experiment was conducted in which developers without any previous experience in developing a dialog system developed task-oriented dialog systems using the proposed framework. The experimental results show that even beginner dialog system developers can develop a high-level task-oriented dialog system.