Towards Compositional Distributional Discourse Analysis

Abstract

Categorical compositional distributional semantics provide a method to derive the meaning of a sentence from the meaning of its individual words: the grammatical reduction of a sentence automatically induces a linear map for composing the word vectors obtained from distributional semantics. In this paper, we extend this passage from word-to-sentence to sentence-to-discourse composition. To achieve this we introduce a notion of basic anaphoric discourses as a mid-level representation between natural language discourse formalised in terms of basic discourse representation structures (DRS); and knowledge base queries over the Semantic Web as described by basic graph patterns in the Resource Description Framework (RDF). This provides a high-level specification for compositional algorithms for question answering and anaphora resolution, and allows us to give a picture of natural language understanding as a process involving both statistical and logical resources.

Functorial Question Answering

“…As a corollary, the computational problems for RelCoCat Semantics and Entailment reduce to conjunctive query evaluation and containment respectively. Building on previous work [4], we then show that the QuestionAnswering problem is NP − complete. Logical semantics for pregroups has been developped in a line of work by Preller [5,6,7], however the corresponding reasoning problems were undecidable.…”

“…Membership follows immediately by reduction to Evaluation. Hardness follows by reduction from graph homomorphism, we only give a sketch of proof and refer to [4] where EL is called matching. Any graph can be encoded in a corpus given by a set of subject-verb-object sentences, where EL maps nouns to their corresponding node.…”

Pregroup Grammars, Their Syntax and Semantics

“…There is some other relevant work in this area, but not directly via pregroups or vectors, e.g. for knowledge bases [8] and using Dynamic Syntax [39].…”

Functorial Language Games for Question Answering

“…The image of the noun type n ∈ B is a vector space where the inner product computes noun-phrase similarity [SCC13]. When applied to question answering tasks, distributional models can be used to compute the distance between a question and its answer [CdMT18].…”