Conference Resolution
in Natural Language
Processing (NLP)
Presented By :
Adnan Khan
Group No :14
What is Conference
Resolution?
● Process of identifying
which words or phrases in
a text refer to the same
entity.Key in
understanding
Introduction relationships between
entities (e.g., "John" and
"he").
Why is it important in NLP?
● Critical for tasks like
machine translation,
question answering, and
Types of Reference
Anaphora: Cataphora:
● Refers back to a ● Refers forward to an
previous entity (e.g., entity (e.g., "Before he
"John went to the spoke, John was
store. He bought nervous.").Example:"M
ary loves her dog. She
milk.").
takes it for walks every
day.".
Challenges in Conference
Resolution
Ambiguity in Coreference
Reference across sentences :
● He saw the man with ● Linking pronouns or
the telescope." (Who noun phrases over
has the telescope?) multiple sentences.
Distance
Dependencies :
● Entities referenced far
apart in text
Coreference vs. Anaphora
Conference : Anaphora :
● Any kind of reference ● A specific kind of
between phrases in reference that links
text. back to a previously
mentioned entity.
Example :
● "John met Sarah. They talked for hours."Coreference:
John and Sarah → They
Approaches to Conference
Resolution
Rule-based Methods: Machine Learning
Methods
● Linguistic rules
applied to identify ● Supervised learning with
labeled data
references. ● Example: Using a neural
● Example: "If a network model to predict
pronoun matches a coreference links.
noun in gender and
number, it’s likely a
reference."
Applications of Conference
Resolution
Machine Translation: Chatbots & Virtual
Assistants:
● Ensures correct ● Enables understanding
translation of of ongoing
conversation context.
pronouns and
● Maintaining coherence
references across in summarizing text.
languages.
Modern Techniques
Neural Coreference Pre-trained Language
Models:
Models:
● Models like BERT or GPT
● Based on deep can understand the
context and improve
learning, leveraging
coreference accuracy.
large amounts of
text data.
● Example: BERT-
based coreference
models
Example Use Case
Text:
The president met the prime minister. He later gave a
speech."
Resolution :
He → President or Prime Minister? (Depends on context or
world knowledge).
Model's Output:
Identifying "He" as the president based on the overall
context.
Summary:
● Conference resolution is essential
for making NLP applications
coherent and context-aware.
● Both rule-based and machine
learning methods are used, with
Conclusion the latter showing recent
advances.
Future Directions:
● Improved deep learning models
and integrating world knowledge
for better context understanding.
