Question

Compare the roles of the prepositional phrases in the following two sentences (which differ by only one word). How could a machine be programmed to make such distinctions? The pigpen was built by the barn. The pigpen was built by the farmer.

Short answer

'By the barn' indicates location; 'by the farmer' indicates agency. A machine can use lexical databases and semantic parsing to make such distinctions.

Step-by-step solution

Identify the Prepositional Phrases

In the sentence 'The pigpen was built by the barn,' the prepositional phrase is 'by the barn.' In the sentence 'The pigpen was built by the farmer,' the prepositional phrase is 'by the farmer.' Each phrase starts with the preposition 'by' and gives information about the action 'was built.'

Determine the Function of Each Prepositional Phrase

In 'The pigpen was built by the barn,' the phrase 'by the barn' indicates a location where the pigpen is built, functioning as an adverbial modifier of location. In 'The pigpen was built by the farmer,' the phrase 'by the farmer' indicates the agent (who performed the action), functioning as an adverbial modifier of agency.

Understand the Semantic Roles Through Context

Consider the meanings of 'barn' and 'farmer.' A barn is a structure that cannot perform actions, suggesting it is a location. Conversely, a farmer is a person capable of performing actions like building, suggesting agency. This semantic understanding helps distinguish their grammatical roles.

Program a Machine with Lexical and Contextual Knowledge

A machine could be programmed to distinguish these roles by using a combination of lexical databases (that store typical roles of nouns, like whether 'barn' refers to a location and 'farmer' to an agent) and semantic parsing algorithms that analyze the sentence structure.

Key concepts

Prepositional Phrases

Prepositional phrases are groups of words that include a preposition, its object, and any modifiers of that object. They are crucial elements in sentences as they provide additional details about actions, such as location, manner, or time.
For example, in the sentence 'The pigpen was built by the barn,' the phrase 'by the barn' is a prepositional phrase. Here, 'by' is the preposition, and 'the barn' is the object. Similarly, in 'The pigpen was built by the farmer,' 'by the farmer' functions in the same way but conveys different information.
Understanding prepositional phrases is essential because they can change the meaning of a sentence significantly depending on their role, such as indicating "location" in the first sentence and "agent" in the second.

Semantic Roles

Semantic roles refer to the function that a word or phrase represents in a sentence. These roles help us understand who is doing what to whom, where, and how.
In our examples, while 'by the barn' suggests a location due to the semantic role of the noun 'barn,' 'by the farmer' indicates agency because farmers are typically the actors performing the building. Recognizing these roles requires one to comprehend the meanings and typical actions associated with words.
As such, a noun like 'barn' is inherently different from 'farmer' in terms of potential capabilities—buildings do not act like humans. This understanding is vital in parsing and interpreting sentences correctly.

Contextual Parsing

Contextual parsing involves analyzing subsentential elements with an emphasis on their meaning within a context. This helps to eliminate ambiguity in sentences that may have structurally similar forms but differ in meaning.
Given the sentences 'The pigpen was built by the barn' and 'The pigpen was built by the farmer,' contextual parsing helps to clarify that one implies a location while the other implies an agent responsible for an action.
Contextual parsing involves understanding both the syntactic and semantic layers of language, allowing a more accurate interpretation of sentences by considering word meaning and sentence structure together.

Lexical Databases

Lexical databases are structured repositories of word information, including definitions, synonyms, usage, and even typical semantic roles.
They are powerful tools for natural language processing systems as they provide essential data for understanding and generating human language.
For the sentences discussed, such databases can help a machine distinguish between 'barn' as a location (non-agentive) and 'farmer' as an agent (capable of action). By accessing information about typical roles and meanings stored in lexical databases, machines can better interpret a sentence's intended context and meaning, thus performing tasks more akin to human understanding.