Double R

A Computational Cognitive Construction Grammar of English
April 2011

Jerry T. Ball (jerry.ball@mesa.afmc.af.mil)

What could he have been given?

Page Links:

• Ditransitive Verb
• Declarative Clause
• Ditransitive Verb + Indirect Object
• Ditransitive Verb + Recipient
• Declarative Clause + Grammatical Features
• Declarative Clause + Grammatical Features + Incremental Processing
• Imperative Clause
• Imperative Clause + Grammatical Features
• Yes-No Question
• Yes-No Question + Grammatical Features + Incremental Processing
• Wh Question
• Wh Question + Passive
• Wh Question + Passive + Grammatical Features
• Raising Verb
• Control Verb
• Subject Control Verb
• Object Control Verb
• Control Adjective
• Subject Control Adjective
• Object Control Adjective
• Predicate Preposition
• Locative Argument Verb

This document demonstrates the capabilities of Double R, a computational cognitive construction grammar of English implemented in the  ACT-R Cognitive Architecture.  Input to the computational model is in written form.  Although most of the examples in this document demonstrate the processing of isolated sentences, the model accepts input  from single words up to an entire document of text.  Current capabilities for resolving cross-sentential dependencies (e.g. anaphora) are limited, as are discourse capabilities more generally.  During the processing of an input text,  the model creates a collection of nested ACT-R chunks  (i.e. frame-like representations consisting of a collection of slot-value pairs where the value of a slot may be chunk).  At the end of processing these ACT-R chunks are converted into tree diagrams for visualization.  The diagram creation capability consists of Lisp code that generates bracketed structures from the ACT-R chunks and phpSyntaxTree which generates diagrams from the bracketed structures.  The bracketing code was developed by Andrea Heiberg.  In linearizing the slot values of the nested ACT-R chunks, the bracketing code provides the rough equivalent of a language generation capability.  phpSyntaxTree is a product of Mei and André Eisenbach.  Andrea worked with Jack Harris to interface the bracketing code to phpSyntaxTree.  The tree diagrams are customizable to some extent.  Grammatical features may or may not be displayed, and when they are displayed they are only displayed at the level of the clause or nominal.  Some elements of the underlying chunk representation are not displayed in the diagrams--especially slots lacking a value.

We refer to these representations as grammatical or linguistic (but not syntactic) representations.  By this we mean that they include information about the grammatical function as well as the form of the linguistic elements in the input.  We assume that grammatical functions like subject, object, specifier, head, and modifier,  and grammatical features like animacy, gender and number are semantically motivated.  We also assume, as in Langacker's Cognitive Grammar, that parts of speech like noun, verb and adjective are semantically motivated.  We use the term "grammatical" as it is traditionally used, to reflect these assumptions.  Although not considered in this document, these grammatical representations map into non-linguistic representations of the situations and objects  that they describe. These non-linguistic representations are in the spirit of Jackendoff's Conceptual Semantics and are under development by Stu Rodgers.

The grammatical representations encode two key dimensions of meaning: referential and relational--hence the name Double R.  Double R identifies the referring expressions in the input (e.g. object referring expression or nominal, situation referring expression or clause)  and the relationships between these referring expressions.  The key relational elements include verbs, adjectives, adverbs, prepositions and conjunctions (but not nouns).  The processing of relational elements leads to projection of constructions which predict the occurrence of the elements they relate.

Double R is most closely aligned with Langacker's Cognitive Grammar and collectively Construction Grammar  (most recently Sag's shift from HPSG to Sign Based Construction Grammar and Jackendoff's shift in this direction as well).  Double R can best be viewed as a formalization and computational implementation of ideas from Cognitive and Construction Grammar.

From the perspective of English grammar, Double R aligns with Huddleston and Pullum's "The Cambridge Grammar of the English Language" and to lesser extent  Quirk, Greenbaum, Leech and Svartvik's "A Comprehensive Grammar of the English Language",  and Biber, Johansson, Leech, Conrad & Finegan's "Longman Grammar of Spoken and Written English".

From the perspective of formal linguistics, Double R aligns with the "Simpler Syntax" of Culicover and Jackendoff.

Double R adheres to well-established cognitive constraints on Human Language Processing (HLP), including the incremental and interactive nature of HLP.  Double R processes the input incrementally, one word or multi-word unit at a time, using a perceptual span of 12 characters to perceive the current input.  No preprocessing of the input is required.  There is no separate tokenizing, part of speech tagging or syntactic parsing.  All processing occurs interactively within Double R.

The word recognition subcomponent, which is fully integrated with the rest of the system, is being developed by Mary Freiman.  The contents of the perceptual span activate words and multi-word units in declarative memory.  Activation spreads from the letters, trigrams and space delimited units in the preceptual span.  The most highly activated declarative memory element, which need not be an exact match, is retrieved from memory and used in subsequent processing.

The current computational implementation comprises ~700 productions and ~60,000 lexical items, and is capable of processing a broad range of English language constructions.  The mental lexicon recently increased from ~8000 to ~60,000 lexical items.  We are working on stabilizing the behavior of the model with this comprehensive lexicon  and on extending the grammatical coverage to encompass the expanded lexicon.

On a 64-bit quad core computer with 8 Gig RAM, the model is capable of processing ~285 words per minute (wpm) in real time  and ~140 wpm in ACT-R cognitive processing time.  By comparison, fluent adult reading rates are in the range of 200-300 wpm.  The key to achieving adult reading rates is the ability to process multi-word units, including units that are larger than a single perceptual span.  Such units not only speed up processing, but they are less ambiguous than individual words and facilitate determination of meaning.  We are working on extending the number of multi-word units in the mental lexicon and adding the capability to recognize units larger than a single perceptual span.

For detailed description of Double R, see the Double R web site at:

Double R Theory

Double R is currently a key component of the larger

Synthetic Teammate project

Regression Test Folder: demo-sentences: basic construction types

Ditransitive Verb "give"

Example 1: he gave me it

• [subj + spec_optional + head]

• [head_ditrans-verb + iobj_x-or + obj + recip_x-or]

• [spec_optional + head_pronoun]

• x-or = exclusive or -- only an iobj or a recipient is allowed -- not both
• n/a = not applicable (in this example)

• definite (def) = definite (*def*)
• number = singular (*sing*)
• animacy (animate) = human (*human*)
• gender = male (*male*)
• person = third (*third*)
• case = subject (*subj*)

• tense (tense-1) = finite (*fin*)
• tense (tense) = past (*past*)
• voice = active (*act*)

Example 2: he gave it to me

• [subj + spec_optional + head]

• [head_ditrans-verb + iobj_x-or + obj + recip_x-or]

• [spec_optional + head_pronoun]

• [head + object]

Example 3: did he give me it

• [oper + subj + spec_optional + head]

• [head_ditrans-verb + iobj_x-or + obj + recip_x-or]

Example 4: did he give it to me

Example 5: Who gave it to me

Example 6: What did he give you

Example 7: Give me it

Example 8: Give it to me

Example 9: Give it me

Example 10: It was given to me

Example 11: Who did he give the ball to?

Example 12: What could he have been given?

Example 13: Who could he be giving the ball

Example 13: Who could he be giving the ball to

Raising Verb "seem"

• they function as the specifier, not the head (i.e. main predicate) of the clause
• the subject is an external argument of the main predicate, not the raising verb

Example 14: He seems happy

Example 15: He does not seem happy

Example 16: He seems to be happy

Example 17: He seems to be very happy

Example 18: He seems like a happy man

Example 19: He could not have seemed to have been happy

Control Verbs "begin", "start", "try"

Example 20a: He began to read the book

Example 20b: He started reading the book

Example 21: Try reading the book

Example 22: He tried the book

Control Verb "allow"

Example 23: They allow smoking

Example 24: They allow smokers

Example 25: They allow us to smoke

Example 26: They allow us to be smokers

Example 27: They allow our smoking cigarettes

Example 28: They allow that we smoke

The subject and (indirect) object control verbs "promise" and "persuade"

Example 29: He promised to go

Example 30: He promised me to go

Example 31: He persuaded me to go

The object control verb "made"

Example 32: He made me go

Example 33: He made me sad

Example 34: He made me very sad

Example 35: He made me a drink

Example 36: It made me a man

The linguistically special adjectives "easy" and "eager"

Example 37: He is easy

Example 38: He is easy to please

Example 39: He is eager to please you

Predicate Prepositions

Example 40: The book is on the table

Example 41: Isn't the book on the table?

Example 42: What is the book on?

The locative argument verb "put"

Example 43: He put the book on the table

Example 44: Where did he put the book?