Coreference Resolution with TensorFlow

Timestamp:
00:00 Introduction
00:34 What is Coreference Resolution?
01:50 What are these expressions called?
02:15 Step 1 Mention Detection
02:24 Named Entities
03:00 POS Tags
03:32 DEP Tags
04:05 Training a Model with Word-Level Features
04:53 Or Extracting Word-Vectors
05:02 Results of Female vs Male Pronoun Mentions
05:58 Creating a recurrent Neutral Network in Tensorflow
07:12 Step 2 Mention Representation
07:23 Classification Outputs loose information
09:30 Step 3 Entity Linking
10:59 Pairwise Ranking
11:25 End-to-end Modelling

La Kopi @ Developers Space is a monthly open mic night for the developer community to learn, connect, and be inspired by each other. Every month, a tech theme is selected and developers submit their topics to be shared with the community.

Name: Ramsha Siddiqui (Machine Learning Engineer, Boomtown)
Topic: Coreference Resolution with TensorFlow

Do your customers reference real-word entities with pronouns, while talking to your chatbot? If yes, you need to add Coreference Resolution to your bot's Natural Language Understanding, in order to extract these references from context. It involves mastering three major components: Mention Detection, Mention Representation and finally Entity Linking.

Learning this skill can also help you understand references in other areas of NLP, such as: Article Search, Automatic Summarization, and Machine Translation. This session includes an example implemented in TensorFlow that demonstrates these three concepts, as well as showing how you can build your very own, contextually-aware chatbot.

For more updates on upcoming events, follow us on social media: