Natural Language Understanding Wiki

A very important reference: Hobbs (1979)[1]. Coherence is seen as a device a speaker uses to make sure the hearer understands what she has to say. It aims at eliminating unwanted interpretations and reducing the processing overhead of the hearer.

From Barzilay and Lapata (2008)[2]:

"McKoon and Ratcliff (1992) argue that local coherence is the primary source of inference-making during reading"

"the distribution of entities in locally coherent texts exhibits certain regularities. This assumption is not arbitrary—some of these regularities have been recognized in Centering Theory (Grosz, Joshi, and Weinstein 1995) and other entity-based theories of discourse (e.g., Givon 1987; Prince 1981)"

From Roth and Frank (2015)[3]:

The most prominent approach to entity-based coherence modeling nowadays is the entity grid model by Barzilay and Lapata (2005)[4]. It has originally been proposed for automatic sentence ordering but has also been applied in coherence evaluation and read-ability assessment (Barzilay and Lapata, 2008; Pitler and Nenkova, 2008), and story generation (McIntyre and Lapata, 2009). Based on the original model, a few extensions have been proposed: for exam- ple, Filippova and Strube (2007) and Elsner and Charniak (2011b) suggested additional features to characterize semantic relatedness between entities and features specific to single entities, respectively. Other entity-based approaches to coherence modeling include the pronoun model by Charniak and Elsner (2009) and the discourse-new model by Elsner and Charniak (2008). All of these approaches are, however, based on explicitly realized entity mentions only, ignoring references that are inferrable.

From Barzilay and Lapata (2005)[4]:

In the discourse literature, entity-based theories are primarily applied at the level of local coherence, while relational models, such as Rhetorical Structure Theory (Mann and Thomson, 1988; Marcu, 2000), are used to model the global structure of discourse.

Modeling coherence[]

Modeling coherence is the task of judging the level of coherence of a text.


Sentence ordering dataset of Barzilay and Lapata (2005)[4] available here:


  1. Hobbs, J. R. (1979). Coherence and coreference. Cognitive Science, 3, 67–90.
  2. Barzilay, Regina, and Mirella Lapata. "Modeling local coherence: An entity-based approach." Computational Linguistics 34, no. 1 (2008): 1-34.
  3. Roth, M., & Frank, A. (2013). Automatically Identifying Implicit Arguments to Improve Argument Linking and Coherence Modeling. Proceedings of the Second Joint Conference on Lexical and Computational Semantics (*SEM), 1, 306–316. Retrieved from
  4. 4.0 4.1 4.2 Regina Barzilay and Mirella Lapata. 2005. Modeling local coherence: An entity-based approach. In Pro- ceedings of the 43rd Annual Meeting of the Associa- tion for Computational Linguistics, Ann Arbor, Michi- gan, USA, 25–30 June 2005, pages 141–148.