Third Unit
Third Unit
Movement phenomena refer to structural changes in sentences where certain elements (like
words or phrases) appear to be "moved" from their original positions to form questions or
other constructions.
Basic Observation:
Yes/no questions are similar to their corresponding assertions, except the subject
noun phrase (NP) and the first auxiliary verb are swapped.
Examples:
o Jack is giving Sue a back rub. → Is Jack giving Sue a back rub?
o He will run in the marathon. → Will he run in the marathon?
If there is no auxiliary, insert "do/does/did" (known as do-support):
o John went to the store. → Did John go to the store?
o Henry goes to school every day. → Does Henry go to school every day?
📌 This swapping is called Subject-Auxiliary Inversion.
Wh-questions ask about specific information (e.g., who, what, where, how, why).
To form a wh-question:
1. Replace the questioned element with a wh-phrase.
2. Move the wh-phrase to the beginning of the sentence.
3. Apply subject-auxiliary inversion (unless the subject is being questioned).
Examples:
Assertion: The fat man will put the book in the corner.
o How will the fat man put the book in the corner?
o What will the fat man angrily put in the corner?
o Where will the fat man angrily put the book?
o In what corner will the fat man angrily put the book?
🔁 Still follows the subject-auxiliary inversion rule, and do-support is used if no
auxiliary is present:
o I found a bookcase.
→ Did I find a bookcase?
→ What did I find?
o
Summary Table:
He will run in the marathon next Will he run in the When will he run in the
year. marathon? marathon?
John went to the store. Did John go to the store? Where did John go?
G = (V, T, S, P)
.o Each rule can have only one non-terminal symbol on the left hand side of therule.
🧠
2. Sample CFG and Derivation
Grammar Example:
S → NP VP
NP → Det Noun | NP PP
VP → Verb NP
PP → Pre NP
Det → 'a' | 'the'
Noun → 'cake' | 'child' | 'fork'
Verb → 'ate'
Pre → 'with'
S → NP VP
→ Det Noun VP
→ the Noun VP
→ the child VP
✅ This shows that "the child ate a cake" is generated by the grammar and hence is
grammatically correct.
Goal:
➡️Extend the grammar minimally to support questions, while reusing as much of the
✅ 4. Yes/No Questions
🛠 Mechanism: Subject–Auxiliary Inversion
In yes/no questions, the auxiliary verb moves in front of the subject noun phrase (NP).
📐 Extended Rule:
S[+inv] → (AUX AGR ?a SUBCAT ?v) (NP AGR ?a) (VP VFORM ?v)
Where:
This single rule, when added, enables the grammar to produce valid yes/no questions by just
changing the order of subject and auxiliary. All other grammar components remain
unchanged.
❓ 5. Wh-Questions
These are more complex than yes/no questions because they involve gaps — missing
constituents in the sentence.
🔎 Example:
🧩 Features Needed:
5.2 WH Feature:
Identifies wh-words (who, what, where, etc.) and their corresponding phrase types.
These words serve as fillers for the gaps left in the sentence.
🧮 6. Categories of WH-Words
WH Word Phrase Type Replaced Example
who, whom, what Noun Phrase (NP) Who ate the cake?
where, when Prepositional Phrase (PP) Where did you put the book?
WH Word Phrase Type Replaced Example
1. Introduce GAP feature to non-terminal rules (e.g., NP, PP) to allow missing
elements.
2. Add WH feature to identify wh-phrases and restrict them to appropriate roles.
3. Prevent gap propagation into the subject position in inverted clauses (e.g.,
auxiliary questions).
4. Allow fronted wh-phrases to act as fillers for the gap in the sentence body.
5.
Here’s a brief and clear set of notes on Augmented Transition Networks (ATNs) and hold
mechanisms based on your content:
Traditional FSMs can only handle regular languages, not enough for complex
natural language constructs.
Natural language involves hierarchical and context-sensitive structures.
ATNs are designed to handle context-free and some context-sensitive aspects of
language.
Hold Mechanism
ATNs use a special variable "*" to hold the result of following an arc.
Example:
o If a word like "the" is accepted via a lexical arc, it's stored in "*".
o Then an action like DET := * stores it in the DET register.
o Similarly, AGR := AGR* copies the agreement info from * into the current
network.
Simple NP (Noun Phrase) ATN Example
Arc Tests:
o An arc is followed only if its test passes.
o Example: Arc 2 checks if AGR of the current network and the noun in "*" have
a non-empty intersection.
Push Arcs:
o Used to call sub-networks (e.g., NP).
o The result is stored in "*".
o Example: SUBJ := * assigns the parsed NP to the subject register.
Agreement Check (e.g., Subject-Verb):
o Test compares AGR of SUBJ (noun phrase) with AGR of the verb (in "*").
o Arc proceeds only if their features match (i.e., agree).
Ensures grammatical consistency, such as subject-verb agreement.
Consider an example. A trace of a parse of the sentence "The dog saw Jack"
Here's a concise and structured summary of the new points you've provided:
ATN – Parse Flow Summary
Using threading to fill gaps in data, e.g., during coreference resolution, ellipsis
recovery, or missing data reconstruction.
Using multithreading (in code) to speed up NLP tasks, like tokenization, tagging,
parsing, etc., especially for large corpora.