Generation of Referring Expressions GRE Reading Dale Reiter

Generation of Referring Expressions (GRE) Reading: Dale & Reiter (1995) (key paper in this area)

The task: GRE • NLG can have different kinds of inputs: – ‘Flat’ data (collections of atoms, e. g. , in the tables of a database) – Logically complex data • In both cases, unfamiliar constants may be used, and this is sometimes unavoidable

No familiar constant available: 1. The referent has a familiar name, but it’s not unique, e. g. , ‘John Smith’ 2. The referent has no familiar name: trains, furniture, trees, atomic particles, … ( In such cases, databases use database keys, e. g. , ‘Smith$73527$’, ‘TRAIN-3821’ ) 3. Similar: sets of objects (lecture 4).

• Natural Languages are too economic to have a proper name for everything • Names may not even be most appropriate • So, speakers/NLG systems have to invent ways of referring to things. E. g. , ‘the 7: 38 Trenton express’ • Note: the problem arises whether the referent is a token or a type

• GRE tries to find the best description • GRE is microcosm of NLG: e. g. , determines – which properties to express (Content Determination) – which syntactic configuration to use (Syntactic Realization) – which words to choose (Lexical Choice)

This lecture: • Simplification 1: Content Determination only (until lecture 5). • Simplification 2: Definite descriptions only (Pronouns, demonstratives, etc. , are disregarded; until tomorrow)

Dale & Reiter (1995): best description fulfills the Gricean maxims. E. g. , • (Quality: ) list properties truthfully • (Quantity: ) list sufficient properties to allow hearer to identify referent – but not more • (Relevance: ) use properties that are of interest in themselves * • (Manner: ) be brief * Slightly different from D&R 1995

D&R’s expectation: • Violation of a maxim leads to implicatures. • For example, – [Quantity] ‘the pitbull’ (when there is only one dog). – [Manner] ‘Get the cordless drill that’s in the toolbox’ (Appelt). • There’s just one problem: …

…people don’t speak this way For example, – [Manner] ‘the red chair’ (when there is only one red object in the domain). – [Manner/Quantity] ‘I broke my arm’ (when I have two). General: empirical work shows much redundancy Similar for other maxims, e. g. , – [Quality] ‘the man with the martini’ (Donellan)

Example Situation c, £ 100 Swedish a, £ 100 d, £ 150 Italian b, £ 150 e, £?

Formalized in a KB • • • Type: furniture (abcde), desk (ab), chair (cde) Origin: Sweden (ac), Italy (bde) Colours: dark (ade), light (bc), grey (a) Price: 100 (ac), 150 (bd) , 250 ({}) Contains: wood ({}), metal ({abcde}), cotton(d) Assumption: all this is shared knowledge.

Violations of … • Manner: * ‘The £ 100 grey Swedish desk which is made of metal’ (Description of a) • Relevance: ‘The cotton chair is a fire hazard? ? Then why not buy the Swedish chair? ’ (Descriptions of d and c respectively)

• In fact, there is a second problem with Manner. Consider the following formalization: Full Brevity: Never use more than the minimal number of properties required for identification (Dale 1989) An algorithm:

Dale 1989: 1. Check whether 1 property is enough 2. Check whether 2 properties is enough …. Etc. , until success {minimal description is generated} or failure {no description is possible}

Problem: exponential complexity • Worst-case, this algorithm would have to inspect all combinations of properties. n properties combinations. • Recall: one grain of rice on square one; twice as many on any subsequent square. • Some algorithms may be faster, but … • Theoretical result: algorithm must be exponential in the number of properties.

• D&R conclude that Full Brevity cannot be achieved in practice. • They designed an algorithm that only approximates Full Brevity: the Incremental Algorithm.

Incremental Algorithm (informal): • Properties are considered in a fixed order: P= • A property is included if it is ‘useful’: true of target; false of some distractors • Stop when done; so earlier properties have a greater chance of being included. (E. g. , a perceptually salient property) • Therefore called preference order.

• • r = individual to be described P = list of properties, in preference order P is a property L= properties in generated description (Recall: we’re not worried about realization today)

P = < furniture (abcde), desk (ab), chair (cde), Swedish (ac), Italian (bde), dark (ade), light (bc), grey (a), 100£ ({ac}), 150£(bd) , 250£ ({}), wooden ({}), metal (abcde), cotton ({d}) > Domain = {a, b, c, d, e}. Now describe: a = <. . . > d = <. . . > e = <. . . >

P = < furniture (abcde), desk (ab), chair (cde), Swedish (ac), Italian (bde), dark (ade), light (bc), grey (a), 100£ (ac), 200£ (bd), 250£ ({}), wooden ({}), metal (abcde), cotton (d) > Domain = {a, b, c, d, e}. Now describe: a = <desk {ab}, Swedish {ac}> d = <chair, Italian, dark, 200> (Nonminimal) e = <chair, Italian, dark, . . > (Impossible)

[ An aside: shared info will be assumed to be complete and uncontroversial. Consider • Speaker: [[Student]] = {a, b, …} • Hearer: [[Student]] = {a, c, …} Does this make a referable? ]

Incremental Algorithm • It’s a hillclimbing algorithm: ever better approximations of a successful description. • ‘Incremental’ means no backtracking. • Not always the minimal number of properties.

Incremental Algorithm • Logical completeness: A unique description is found in finite time if there exists one. (Given reasonable assumptions, see van Deemter 2002) • Computational complexity: Assume that testing for usefulness takes constant time. Then worst-case time complexity is O(np) where np is the number of properties in P.

Better approximation of Full Brevity (D&R 1995) • Attribute + Value model: Properties grouped together as in original example: Origin: Sweden, Italy, . . . Colour: dark, grey, . . . • Optimization within the set of properties based on the same Attribute

Incremental Algorithm, using Attributes and Values • r = individual to be described • A = list of Attributes, in preference order • Def: = Value i of Attribute j • L= properties in generated description

• Find. Best. Value(r, A): - Find Values of A that are true of r, while removing some distractors (If these don’t exist, go to next Attribute) - Within this set, select the Value that removes the largest number of distractors - If there’s a tie, select the most general one - If there’s still a tie, select an arbitrary one

Example: D = {a, b, c, d, f, g} • Type: furniture (abcd), desk (ab), chair (cd) • Origin: Europe (bdfg), USA (ac), Italy (bd) Describe a: {desk, American} (furniture removes fewer distractors than desk) Describe b: {desk, European} (European is more general than Italian) N. B. This disregards relevance, etc.

P. S. Note the similarity with Van Rooy & Dekker’s semantic of answers: Let A and B be truthful answers to a question, then A is a better answer than B Utility(A) > Utility(B) or Utility(A) = Utility(B) & B A (More about this in the next lecture …)

• Exercise on Logical Completeness: Construct an example where no description is found, although one exists. • Hint: Let Attribute have Values whose extensions overlap.

Example: D = {a, b, c, d, f} • Contains: wood (abe), plastic (acdf) • Colour: grey (ab), yellow (cd) Describe a: {wood, grey, . . . } - Failure (wood removes more distractors than plastic) Compare: Describe a: {plastic, grey} - Success

Complexity of the algorithm nd = nr. of distractors nl = nr. of properties in the description nv = nr. of Values (for all Attributes) Alternative assessment: O(nv) (Worst-case running time) According to D&R: O(nd nl ) (Typical running time)

Minor complication: Head nouns • Another way in which human descriptions are nonminimal – A description needs a Noun, but not all properties are expressed as Nouns – Example: Suppose Colour was the most-preferred Attribute, and target = a

• • • Colours: dark (ade), light (bc), grey (a) Type: furniture (abcde), desk (ab), chair (cde) Origin: Sweden (ac), Italy (bde) Price: 100 (ac), 150 (bd) , 250 ({}) Contains: wood ({}), metal ({abcde}), cotton(d) target = a Describe a: {grey} ‘The grey’ ? (Not in English)

D&R’s repair: • Assume that Values of the Attribute Type can be expressed in a Noun. • After the core algorithm: - check whether Type is represented. - if not, then add the best Value of the Type Attribute to the description

• Versions of Dale and Reiter’s Incremental Algorithm have often been implemented • Still the starting point for many new algorithms. (See later lectures. ) • Worth reading!

Limitations of the algorithm 1. Redundancy does not arise for principled reasons, e. g. , for - marking topic changes, etc. (Corpus work by Pam Jordan et. al. ) - making it easy to find the referent (Experimental work by Paraboni et al. - Next lecture)

Limitations of the algorithm 2. Targets are individual objects, never sets. What changes when target = {a, b, c} ? (Lecture 4) 3. Incremental algorithm uses only conjunctions of atomic properties. No negations, disjunctions, etc. (Lecture 4)

Limitations of the algorithm 4. No relations with other objects, e. g. , ‘the orange on the table’. (Lecture 3) 5. Differences in salience are not taken into account. (Lecture 3) 6. Language realization is disregarded. (Lecture 5)

Discussion: How bad is it for a GRE algorithm to take exponential time? – More complex types of referring expressions problem becomes even harder – Restrict to combinations whose length is less than x problem not exponential. – Example: descriptions containing a most n properties (Full Brevity)

However: – “Mathematicians’ view”: structure of problem shows when no restrictions are put. – What if the input does not conform with these restrictions? (GRE does not control its own input!)

• Compare with Description Logic: - Increasingly complex algorithms … - that tackle larger and larger fragments of logic … - and whose complexity is ‘conservative’ • Question: how do human speakers cope?