Generation that Exploits Corpus-Based Statistical Knowledge

By Irene Langkilde-Geary, Kevin Knight. 1998

We describe novel aspects of a new natural language generator called Nitrogen. This generator has a highly flexible input representation that allows a spectrum of input from syntactic to semantic depth, and shifts the burden of many linguistic decisions to the statistical post-processor. The generation algorithm is compositional, making it efficient, yet it also handles non-compositional aspects of language. Nitrogen's design makes it robust and scalable, operating with lexicons and knowledge bases of one hundred thousand entities.

Subfields: Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

Citations in the corpus

Repairing the Cracked Foundation: A Survey of Obstacles in Evaluation Practices for Generated Text (2022, methodology): “rics which have gained popularity with the advent of statistical NLG (Langkilde and Knight, 1998), which led to the standardization of tasks.”
Neural Approaches to Conversational AI (2018, neutral): “rts design a set of templates or rules, and hand-craft heuristics to select a proper candidate to generate sentences. Even though machine learning can be used to train certain parts of these systems (Langkilde and Knight, 1998; Stent et al., 2004; Walker et al., 2007), the cost to write and maintain”
Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems (2015, methodology): “The trainable generator approach exempliﬁed by the HALOGEN (Langkilde and Knight, 1998) and SPaRKy system (Stent et al., 2004) provides a possible way forward.”
Neural Text Generation from Structured Data with Application to the Biography Domain (2016, methodology): “At the intersection of rule-based and statistical methods, hybrid systems aim at leveraging human contributed rules and corpus statistics (Langkilde and Knight, 1998; Soricut and Marcu, 2006; Mairesse and Walker, 2011).”
A Survey of Available Corpora For Building Data-Driven Dialogue Systems: The Journal Version (2015, background): “…2013; Ren et al., 2013; Henderson et al., 2013; Williams et al., 2013; Henderson et al., 2014c; Kim et al., 2015), natural language generation (Langkilde and Knight, 1998; Oh and Rudnicky, 2000; Walker et al., 2002; Ratnaparkhi, 2002; Stent et al., 2004; Rieser and Lemon, 2010; Mairesse et…”
Multi-domain Neural Network Language Generation for Spoken Dialogue Systems (2016, background): “In subjective testing, human judges confirm that the procedure greatly improves generator performance when only a small amount of data is available in the domain.”
Stochastic Language Generation in Dialogue using Recurrent Neural Networks with Convolutional Sentence Reranking (2015, methodology): “An early example of a statistical NLG system is HALOGEN by Langkilde and Knight (1998) which uses an n-gram language model (LM) to rerank a set of candidates generated by a handcrafted generator.”
Joint Video and Text Parsing for Understanding Events and Answering Queries (2013, methodology): “During this process, the information from the parse graph is converted to a functional description [66], which speciﬁes elements of a sentence such as the event, subjects, objects, and their functional properties.”
Few-shot Natural Language Generation for Task-Oriented Dialog (2020, methodology): “( i ) Template-based methods require domain experts to handcraft templates for each domain, and the system ﬁlls in slot-values afterward (Cheyer and Guzzoni, 2014; Langkilde and Knight, 1998).”
Modeling Local Coherence: An Entity-Based Approach (2005, background): “Local coherence has been extensively studied within the modeling framework put forward by Centering Theory (Grosz et al., 1995; Walker et al., 1998; Strube and Hahn, 1999; Poesio et al., 2004; Kibble and Power, 2004).”
Controlling User Perceptions of Linguistic Style: Trainable Generation of Personality Traits (2011, background): “The seminal work of Langkilde and Knight (1998) in this area showed that high quality paraphrases can be generated from an underspeciﬁed representation of meaning, by ﬁrst applying a very underconstrained, rule-based overgeneration phase, whose outputs are then ranked by an SLM scoring phase.”
Individual and Domain Adaptation in Sentence Planning for Dialogue (2007, methodology): “…as we explain below, with techniques that overgenerate using user or domain-independent rules, and then filter or rank the possibilities using domain or userspecific corpora or feedback (Langkilde & Knight, 1998; Langkilde-Geary, 2002; Bangalore & Rambow, 2000; Rambow, Rogati, & Walker, 2001).”
Broad Coverage Multilingual Deep Sentence Generation with a Stochastic Multi-Level Realizer (2010, methodology): “…(Bangalore and Ram-bow, 2000; Langkilde-Geary, 2002; Filippova and Strube, 2008), or draw upon hybrid models that imply a symbolic submodule which derives the syntactic representation that is then used by the stochastic submodule (Knight and Hatzivas-siloglou, 1995; Langkilde and Knight, 1998).”
Generation from Abstract Meaning Representation using Tree Transducers (2016, methodology): “The Nitrogen and Halogen systems (Langkilde and Knight, 1998; Langkilde, 2000) used an input representation that was a precursor to the modern version of AMR, which was also called AMR, although it was not the same representation as Ba-narescu et al. (2013).”
Hierarchical Reinforcement Learning for Adaptive Text Generation (2010, background): “Approaches to NLG in the past have been either rule-based (Reiter and Dale, 1997) or statistical (Langkilde and Knight, 1998).”
Extracting Paraphrases from a Parallel Corpus (2001, background): “In some applications, only synonyms are considered as para-phrases (Langkilde and Knight, 1998); in others, looser deﬁnitions are used (Barzilay and Elhadad, 1997).”
Phrase-Based Statistical Language Generation Using Graphical Models and Active Learning (2010, background): “The ﬁrst one, pioneered by Langkilde and Knight (1998), introduces statistics in the generation process by training a model which reranks candidate outputs of a handcrafted generator.”
Natural Language Generation (2004, methodology): “A related concept of salience as a quantity composed of several factors (including extra-linguistic ones) has been studied in the psycholinguistics literature; cf. [11] for a detailed discussion and perspectives on its application in NLG. Finally, Reed [13] modelled “salience” of a proposition as an”