WordNet
Can I read the synset or hypernym file twice? No. File I/O is very expensive; read each file only once and store it in an appropriate data structure.
Any advice on how to read and parse the synset and hypernym data files?
Use the readLine() method in our In library to read
in the data one line at a time.
Use the split() method in Java's String library
to divide a line into fields. You can find an example using split() in
Domain.java.
Use either Integer.parseInt()
to convert String id numbers into int values or
Integer.valueOf() to convert into Integer objects.
Which data structure(s) should I use to store the synsets, synset ids, and hypernyms? This part of the assignment is up to you. You must carefully select data structures to achieve the specified performance requirements.
What should sca() return if is there is a tie for the shortest commmon ancestor?
The API does not specify, so you are free to return any shortest common ancestor.
Do I need to store the glosses? No, you won't use them on this assignment.
What is the root synset for the WordNet DAG?
38003,entity,that which is perceived or known or inferred to have its own distinct existence (living or nonliving)
Can a noun appear in more than one synset?
Absolutely. It will appear once for each of the noun’s distinct meanings.
For example, the noun word appears in these 8 synsets:
Do not assume that the number of synsets in which a noun participates is bounded by a constant. The noun35532,discussion give-and-take word,an exchange of views on some topic; "we had a good discussion"; "we had a word or two about it" 56587,news intelligence tidings word,new information about specific and timely events; "they awaited news of the outcome" 59267,parole word word_of_honor,a promise; "he gave his word" 59465,password watchword word parole countersign,a secret word or phrase known only to a restricted group; "he forgot the password" 81575,word,a string of bits stored in computer memory; "large computers use words up to 64 bits long" 81576,word,a verbal command for action; "when I give the word, charge!" 81577,word,a brief statement; "he didn't say a word about it" 81578,word,a unit of language that native speakers can identify; "words are the blocks from which sentences are made"; "he hardly said ten words all morning"
head appears in 33 synsets.
Can a synset consist of exactly one noun? Yes. Moreover, there can be several different synsets that consist of the same noun.
62,Aberdeen,a town in western Washington 63,Aberdeen,a town in northeastern South Dakota 64,Aberdeen,a town in northeastern Maryland 65,Aberdeen,a city in northeastern Scotland on the North Sea
I'm an ontologist and I noticed that your
hypernyms.txt file contains both is-a and
is-instance-of relationships.
Yes, you caught us. This ensures that every noun (except entity)
has a hypernym. Here is an article on the
subtle distinction.
Can I use my own Digraph class? No. You must use Digraph.java. That is the type of the argument to the constructor.
How can I make the data type ShortestCommonAncestor immutable?
You can (and should) save the associated digraph in an instance variable.
However, because Digraph is mutable,
you must first make a defensive copy by calling the copy constructor in Digraph.
Is a vertex considered an ancestor of itself? Yes.
What should ancestor() return if is there is a tie for the shortest common ancestor?
The API does not specify, so you are free to return any shortest common ancestor.
I understand how to compute the length() method
in time proportional to E + V in the worst case
but my lengthSubset()
method takes time proportional to a × b × (E + V),
where a and b are the sizes of the two iterables. How can
I improve it to be proportional to E + V?
The key is use a multi-source version of breadth-first search, as in the
the constructor in BreadthFirstDirectedPaths that accepts an iterable
of sources as an argument (instead of a single source).
Should I construct a new ShortestCommonAncestor object
for each call to sca() and distance()?
No. You need only one ShortestCommonAncestor object per WordNet
object.
The methods sca() and distance() should make
one call to either lengthSubset() or ancestorSubset().
For the extra credit, do length(),
lengthSubset(), ancestor(), and ancestorSubset()
need to take time proportional to
to the number of vertices and edges reachable from the argument vertices
in the worst case? Or, may I use hashing?
You can make standard technical assumptions (such as the uniform hashing assumption).
If you do so, state any assumptions that you make in your readme.txt file.
Should I re-implement breadth-first search to compute shortest common ancestors? Not unless you are attempting the extra credit. For most students, we recommend using BreadthFirstDirectedPaths.java.
Should I re-implement depth-first search to find directed cycles? No. We recommend using either DirectedCycle.java or Topological.java.
Do I need to throw exceptions explicitly with a throw statement?
No, it's fine if they are thrown implicitly, e.g., you can rely on any
method in
Digraph.java
to throw a java.lang.IllegalArgumentException
if passed a vertex argument outside of the prescribed range.
A good API documents the requisite behavior for all possible arguments
but, hopefully, you should not need much extra code to deal with these corner cases.
What should outcast() return if is there is a tie for the outcast?
The API does not specify, so you are free to return any outcast.
My algorithm computes the distance between every pair of nouns. Is that okay? Yes, that's fine.
Some examples. Here are some interesting examples that you can use to test your code.
President (in capitalized form) appears in two different synsets:
13745,President_of_the_United_States President Chief_Executive,the office of the United States .... 13746,President_of_the_United_States United_States_President President Chief_Executive,the person who holds the office .....
municipality has two paths to region:
municipality -> administrative_district -> district -> region municipality -> populated_area -> geographic_area -> region
individual and edible_fruit
have several different paths to their common ancestor
physical_entity:
individual -> organism being -> living_thing animate_thing -> whole unit -> object physical_object -> physical_entity person individual someone somebody mortal soul -> causal_agent cause causal_agency -> physical_entity edible_fruit -> garden_truck -> food solid_food -> solid -> matter -> physical_entity edible_fruit -> fruit -> reproductive_structure -> plant_organ -> plant_part -> natural_object -> unit -> object -> physical_entity
(distance = 23) white_marlin, mileage (distance = 33) Black_Plague, black_marlin (distance = 27) American_water_spaniel, histology (distance = 29) Brown_Swiss, barrel_roll
entity:
Ambrose Saint_Ambrose St._Ambrose
ShortestCommonAncestor,
WordNet, and Outcast.
ShortestCommonAncestor.
First, think carefully about designing a correct and efficient
algorithm for computing the shortest common ancestor.
Consult a staff member if you're unsure.
In addition to the digraph*.txt files,
design small rooted DAGs to test and debug your code.
Modularize by sharing common code.
ShortestCommonAncestor to detect whether a digraph is a rooted DAG.
As defined in the assignment, a digraph is a rooted DAG
if it is acyclic and has one vertex—the root—that is an ancestor of
every other vertex.
synsets.txt
and hypernyms.txt.
Don't worry about storing the data in any data structures yet.
Test that you are parsing the input correctly before proceeding.
WordNet. Divide the constructor into
two (or more) subtasks (private methods).
synsets.txt contains 82,192 synsets, composed from 119,188 nouns.
Do not hardwire either of these numbers;
your program must work for any valid synset file.
Record the number of synsets for use when constructing the underlying digraph
from the hypernyms file.
Digraph. The file hypernyms.txt corresponds to a rooted DAG
with 82,192 vertices and 84,505 edges.
Do not hardwire either of these numbers;
your program must work for any valid hypernym file.
Implement the remaining WordNet methods.
Implement Outcast. This should be relatively straightforward
by calling the appropriate methods from the WordNet data type.
A video is provided for those wishing additional assistance. Warning: the video was made in early 2014 and is somewhat out of date. For example the API has changed.