Recent Advances in Natural Language Generation: A Survey and Classification of the Empirical Literature

Abstract

Abstract. Natural Language Generation (NLG) is defined as the systematic approach for producing human understandable natural language text based on nontextual data or from meaning representations. This is a significant area which empowers human-computer interaction. It has also given rise to a variety of theoretical as well as empirical approaches. This paper intends to provide a detailed overview and a classification of the state-of-the-art approaches in Natural Language Generation. The paper explores NLG architectures and tasks classed under document planning, micro-planning and surface realization modules. Additionally, this paper also identifies the gaps existing in the NLG research which require further work in order to make NLG a widely usable technology.

Automatically Generating Interesting Facts from Wikipedia Tables

“…Natural Language Generation (NLG) is the problem of transforming structured data, such as RDF triples, into humancomprehensible text [15], as opposed to Information Extraction (IE), which extracts structured relations from natural language (cf., [20]). Within this thread, [18] and [8] employ recurrent neural networks to transduce a Freebase triple into natural language questions for which the triple ("fact") would provide an answer to its corresponding generated question.…”

“…For a fixed table at the previous epoch in time, we measure the similarity to all tables from the same page at the current epoch to determine which one, if any, is its counterpart. 15 For cases where multiple tables have the same (highest) similarity score, we break ties by considering table schema. However, if none of the tying tables has the same schema, or if more than one of them does, ties are then further broken by considering the section name hierarchy from the page's table of contents, for pages containing one, and choosing the table having longest subpath match with regards to the section hierarchy.…”

A Context-Aware Conversational Agent in the Rehabilitation Domain

“…Nowadays, state-of-the-art speech recognition Application Programming Interfaces (APIs) are commonplace and are being offered by global industry conglomerates (e.g., Microsoft, Amazon, and Google), integrating easily into custom frameworks. Language generation is the process of producing meaningful text from nonlinguistic knowledge representations [13] and has preoccupied researchers since the 2000s, with most contemporary studies (References [14,15]) presenting promising results by utilising Deep Neural Network (DNN)-based systems. DNNs [16] are also being exploited in speech-generation systems, which have attracted much attention as well, with other methods varying between Hidden Markov Model (HMM)-based models [17] and incremental dialogue systems [18].…”

Make data sing: The automation of storytelling

“…The third and final level is Realization, which is also often referred to as surface realization (again emphasizing the depth metaphor). This describes the stage at which the actual text that will be read is generated, involving a linguistic and a structure realization, abiding to the rules of syntax, morphology and orthography (Reiter, 2012;Reiter and Dale, 2006;Perera and Nand, 2017). When going through the different stages of decisionmaking in a table like this, it becomes apparent, firstly, how the programming of NLG relies upon a series of decisions executed in accordance with a set of specific rules and, secondly how there are similarities in the ways of thinking in NLG and in structuralism and semiotics, which also operate from the idea of a more-or-less universal pattern of discernible codes.…”