Homework 2 old

Named Entity Recognition

Your task is to implement a named entity recognizer. You are allowed to work in groups of at most 2. Submit your work by Nov. 11th before the class.

• Clone the Emory NLP project.
• Download the benchmark dataset and brown clusters.
• Run NERTrain with the default setting in config_train_ner.xml.
• Improve the named entity recognizer.
• Download the ontology in DBPedia and use it as ambiguity_class in NERFeatureTemplate.
• Ensure the output of all chunks follow the BILOU notation.
• Evaluate the accuracy of your system, precision, recall, and F1, on both the development and evaluation sets.
• Write a report (4-8) pages in the ACL format. Your report must include abstract, introduction, related work, approach, experiments, and conclusion.
• Commit all your work to your Github repository.

Data Format

Only      only      RB   _  O
France    france    NNP  _  U-LOC
and       and       CC   _  O
Britain   britain   NNP  _  U-LOC
backed    back      VBD  _  O
Fischler  fischler  NNP  _  U-PER
's        's        POS  _  O
proposal  proposal  NN   _  O
.         .         .    _  O

Each column represents:

• 0: word-form.
• 1: lemma (predicted).
• 2: POS tag (predicted).
• 3: extra features (blank).
• 4: named entity recognition (gold).

Optimizer

Element	Value
algorithm	perceptron, softmax, adagrad, agagrad-mini-batch, agadelta-mini-batch, agagrad-regression
l1_regularization	L1 regularization, lower-bound (for adagrad*)
learning_rate	Learning rate
max_epochs	Maximum number of epochs
batch_size	Number of sentences used in mini-batch
roll_in	Gold label probability, upper-bound
bias	Bias value

Dictionary

• DBPedia entries with named entities.

Index	Type	DBPedia
0	PERSON	Person, PersonFunction, Mayor, Name
1	NORP	GeopoliticalOrganisation, Legislature, Parliament, PoliticalParty, ReligiousOrganisation, EthnicGroup
2	FACILITY	ArchitecturalStructure, Cemetery, ConcentrationCamp, Garden, HistoricPlace, Mine, Monument, SkiResort, SportFacility, Park, Street
3	ORGANIZATION	GovernmentAgency, Broadcaster, Company, EducationalInstitution, EmployersOrganisation, NonProfitOrganisation, SambaSchool, SportsLeague, SportsTeam, Website
4	GPE	Country, Settlement, State
5	LOCATION	Region, NaturalRegion, HistoricalRegion, Street, Territory, ProtectedArea, SkiArea, Island, NaturalPlace, Continent
6	PRODUCT	Aircraft, Automobile, Locomotive, MilitaryVehicle, Motorcycle, Rocket, Ship, SpaceShuttle, Spacecraft, Train, Device, Drug, Food
7	EVENT	NaturalEvent, Competition, SocietalEvent
8	WORK_OF_ART	Artwork, Cartoon, CollectionOfValuables, Document, Film, Musical, MusicalWork, WrittenWork, TelevisionShow
9	LANGUAGE	Language
10	DATE	TimePeriod
11	MONEY	Currency